Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
TheaTer as DaTa
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Theater as Data
CompuTaTional Journeys  
inTo TheaTer researCh
Miguel Escobar Varela
University of Michigan Press
ann arbor
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Copyright © 2021 by Miguel Escobar Varela
Some rights reserved
This work is licensed under a Creative Commons Attribution-N onCommercial- NoDerivatives 4.0 
International License. Note to users: A Creative Commons license is only valid when it is applied by 
the person or entity that holds rights to the licensed work. Works may contain components (e.g., 
photographs, illustrations, or quotations) to which the rightsholder in the work cannot apply the 
license. It is ultimately your responsibility to independently evaluate the copyright status of any 
work or component part of a work you use, in light of your intended use. To view a copy of this 
license, visit http://creativecommons.org/licenses/by-nc-nd/4.0/
For questions or permissions, please contact um.press.perms@umich.edu
Published in the United States of America by the
University of Michigan Press
Manufactured in the United States of America
Printed on acid- free paper
First published August 2021
A CIP catalog record for this book is available from the British Library.
Library of Congress Cataloging- in- Publication data has been applied for.
ISBN 978- 0- 472- 07479- 2 (hardcover : alk. paper)
ISBN 978- 0-4 72- 05479- 4 (paper : alk. paper)
ISBN 978- 0-4 72- 12863- 1 (OA)
DOI: https://doi.org/10.3998/mpub.11667458
This open access version made available by the National University of Singapore.
Cover photo: From the performance Pixel by Adrien M & Claire B and CCN de Créteil et du  
Val-de-Marne / Compagnie Käfig—Mourad Merzouki, 2014. Photo © Raoul Lemercier.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Acknowledgments
I would like to thank Yong Li Lan and the rest of the Asian Intercultural 
Digital Archives (AIDA) team at the National University of Singapore 
(NUS) for providing me with a conceptual home from which to embark on 
the study of theater as data, and for helping me understand the problems 
of turning theater into digital artefacts.
Paul Rae, Itty Abraham, Lonce Wyse, Eric Kerr, Maiya Murphy, Felipe 
Cervera, and the late John Richardson all read portions of this book and 
gave me invaluable feedback. At NUS, I would like to thank the Digital 
Cultures reading group at the Faculty of Arts and Social Sciences and the 
Quantitative Reasoning faculty members at the interdisciplinary Univer-
sity Scholars Programme for their conversations, feedback, and support. 
The University Scholars Programme and the Faculty of Arts and Social 
Sciences both provided financial support for making this book available 
through an open access and I am very grateful for this.
While this is not a book about the Indonesian performing arts, the 
traditions of Java do feature in some parts of this book. My understand-
ing of these traditions has been shaped by many people, but I am espe-
cially indebted to Jan Mrázek, Eddy Pursubaryanto, Bima Slamet Raharja, 
Bernard Arps, Kathryn Emerson, Novi Marginingrum, Ki Catur Kuncoro, 
Dewi Ambarwati, Ki Aneng Kiswantoro, Ki Ananto Wicaksono, Matthew 
Cohen, the late Ki Slamet Gundono, and the late Ki Ledjar Soebroto for 
bringing me closer to the performing arts of Java. I would also like to 
thank Imam Maskur for allowing me to use the Kluban data on wayang 
kulit performances. In Singapore I would like to thank The Necessary 
Stage, Centre 42, and The Flying Inkpot for letting me use their data.
The journeys of this book began in earnest when Gea Oswah Fatah 
Parikesit at the Department of Engineering Physics, Gadjah Mada Univer-
sity in Indonesia first invited me to work with him in a collaborative project 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
vi • Acknowledgments
in 2014. It was then that I first saw how theater scholars can have produc-
tive collaborations with scientists. Thank you mas Gea, for your support, 
intellectual generosity, and for your willingness to engage with research-
ers across the disciplinary divide. I am also thankful to Andrew Schauf and 
Luis Hernández- Barraza, with whom I continued this collaborative jour-
ney and learned more about network science and biomechanics.
The global digital humanities community has accompanied and sup-
ported me in this methodological wayfaring. I am especially thankful to 
Clarisse Bardiot, Alex Gil, Isabel Galina, Sean Pue, Padmini Ray Murray, 
David Wrisley, Ernesto Priani, Christof Schöch, Radim Hladík, Frank 
Fischer, and Paul Spence. I’m also very happy to be part of the incipi-
ent DH community in Singapore, and I’m grateful for the conversations 
with, and support from, Kenneth Dean, Lim Beng Choo, Andrea Nanetti, 
Michael Stanley-B aker, Feng Yikang, Sayan Bhattacharyya, and the Digital 
Scholarship team at the NUS Libraries.
It has been an honor to work with the team at University of Michigan 
Press. Thank you to LeAnn Fields for believing in this project.
Last but not least, I wish to thank my family in near and far places. 
Thanks to my parents for being such inspiring and supportive role mod-
els on how to be academic researchers in today’s world. Thanks to Dari 
for her useful feedback on interaction design and visualizations. My wife, 
Yingting, provided me with moral support throughout the writing of this 
book, and helped me better understand how scientists use data in their 
daily work. Sofia is too young to realize this, but she brought enormous 
joy to the process of writing the words that follow.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Contents
Introduction: In Pursuit of Theater’s Digital Traces 1
Part 1: Pre- departure Reflections
 1 Toward a More Nuanced Conversation on Methodology 23
 2 The Roles of Statistics 41
 3 The Roles of Visualizations 57
Part 2: Guided Tours
 4 Words as Data 75
 5 Relationships as Data 94
 6 Motion as Data 116
 7 Location as Data 141
Part 3: Ensuring the Journeys Continue
 8 The Imperative of Open and Sustainable Data 163
 9 The Roles of Software Programming 180
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
viii • Contents
Appendix A: Data Biographies 189
Appendix B: Technical Glossary 193
References 195
Index 219
Digital materials related to this title can be found on the Fulcrum platform 
via the following citable URL: https://doi.org/10.3998/mpub.11667458
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Introduction
In Pursuit of Theater’s Digital Traces
I am on my way to the theater, and I can’t stop generating data. When I 
use public transport, consult a map on my phone, and order a coffee at the 
theater foyer, I am leaving digital breadcrumbs of my activities behind me. 
The data points that have resulted from these activities will be aggregated 
with those generated by thousands of other people. Collectively, these 
datasets will shape the experience of riding buses, consulting maps, and 
ordering coffee for myself and others. From a technological perspective, 
it is doubtlessly exciting that data can be applied to such diverse aspects 
of life. But there are many complex ethical and political corollaries to this 
intense collection and use of data. Data is not only useful for optimiz-
ing bus routes and coffee pricing, it can also be used to increase social 
inequality (O’Neil 2016, 94) and to sway elections (Morgan 2018). We live 
in a world that is made for data and from data, as data shapes us in more 
ways than one. As Ribes and Jackson (2013) note, “we have entered into a 
symbiotic relationship with data—r emaking our material, technological, 
geographical, organizational, and social worlds into the kind of environ-
ments in which data can flourish” (52).
I finish my coffee, switch off my phone, and enter the performance 
space. It seems like I have entered a data-f ree sanctuary and left the world 
of data and its ominous potentials behind. After all, the creative decisions 
that shape performances are most likely not driven by data. While data- 
driven content is becoming increasingly common in films and television 
(Suri and Singh 2018), and data plays a role in the marketing and selec-
tion of performances, I can’t find any documented instance where the 
decisions about what happens on stage were shaped by data, not even in 
commercial productions. There are many performances which are about 
data, such as Of All of the People in All of the World by Stan’s Café (2008), 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
2 •  TheaTer as DaTa
where grains of rice are used to represent population statistics. However, 
creative decisions (lighting, casting, modes of interacting with the audi-
ence) are not decided by data. This is certainly not the case in the intimate, 
experimental productions and cultural performances to which the bulk 
of scholarly attention is devoted to. Likewise, most theater scholarship is 
focused on specific, thoroughly contextualized performance events. The-
ater and performance scholars deploy interpretive methods to study the 
performances that interest us. We don’t usually make use of data— in the 
information science sense— in order to seek answers to our questions or 
to orient our discussions. That is, until recently.
With the advent of the digital humanities (henceforth DH), and the 
pioneering work of several researchers, using data is becoming more 
common in theater scholarship. This does not mean surrendering our 
methods to the nefarious potentials of data, but combining data and criti-
cal insight to offer fresh perspectives. For example, Holledge et al. (2016) 
have used a comprehensive dataset of A Doll’s House performances to offer 
new insights into the global spread of Ibsen’s masterpiece. Using net-
work analysis, Trilcke et al. (2015) have identified previously unreported 
patterns in the interaction of characters in German drama across several 
centuries. Schöch (2017) has identified clear genre differences in the 
words used by classical French playwrights. Caplan (2017) has used net-
work data to show that Yiddish theater performers are far more influen-
tial in popular entertainment that they are usually credited for. Wiesner 
and Stalnaker (2015) were able to estimate conceptual metaphors in dance 
from context- free movement patterns. And Miller (2017) has used Broad-
way data to show that the performance runs of musicals have consistently 
increased in length over time.
These are just some examples of work carried out at the intersection of 
theater and DH, an area with a long history. In an overview of this intersec-
tion, Spence (2013) outlines many projects that were decades in the mak-
ing. In her overview of this area, Leonhardt (2014) describes projects that 
fit into seven categories: library and archival projects, projects on indi-
vidual playwrights or actors, projects on theater architecture, projects on 
dance, larger network projects, applied DH in theater research and educa-
tion/experiential DH projects in performance studies, and projects useful 
for or related to theater and performance research. Caplan (2015) distin-
guishes between four types of projects: digital archives and editions, digi-
tal theatrical environments, digital visualizations, and digital databases. 
And Bay- Cheng (2017) classifies existing projects according to three major 
types of research activities: collection, analysis, and dissemination.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Introduction—In Pursuit of Theater’s Digital Traces • 3
Regardless of how one slices this interdisciplinary cake, one thing is 
certain: there is a great variety of projects carried out at the crossroads of 
theater and DH. This variety will feature in this book, but my main objec-
tive is not to chronicle this growing area. Rather, I focus on a more spe-
cific question: what does it mean to research theater in terms of data? I am 
interested primarily in studying theater as an event rather than as dramatic 
literature (an area that has been extensively studied elsewhere in DH). I 
believe, with Peggy Phelan (1993), that theater “constitutes itself through 
disappearance” (146). As performances blink in and out of existence, they 
leave traces behind. These include program booklets, play schedules, cast 
lists, critical reviews, and videos which can be transformed into digital 
data. Thus, my project is also aligned to what Manovich (2020) calls cul-
tural analytics. While DH is a more expansive term that includes everything 
from the digitization of cultural heritage to the development of metadata 
standards, cultural analytics tends to be more narrowly focused on the 
computational analysis of large cultural datasets. This might well include 
the analysis of textual materials and historical data, but cultural analyt-
ics as practiced by Manovich and his lab since 2005 gives primacy to the 
analysis of contemporary, visual culture at a scale of millions of objects. 
In my projects, I also consider visual and other nontextual dimensions of 
theater, but some of my sources are historical rather than contemporary 
and my datasets are never as extensive as those used by Manovich.
What Is Data?
This book takes a narrow definition of data as “potential information.” 
This comes from information science, a field that assumes data is some-
thing that needs to be automatically processed to be useful (Pomer-
antz 2015). Processed data becomes information. Applied information 
becomes knowledge, and reasoned knowledge becomes wisdom. This 
hierarchy was famously inspired by T. S. Eliot’s play The Rock (1934):
Where is the wisdom we have lost in knowledge?
Where is the knowledge we have lost in information? (7)
This quote is often found in information science books. When theater 
scholars turn to other fields for definitions, we might find some comfort 
in being sent back to look at a theater play. A longer definition of data, as 
understood in this book, is potential information in digital form that needs to be 
processed by a machine for it to become meaningful. But the important question, 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
4 •  TheaTer as DaTa
as Christine Borgman (2015) notes, is not “what are data?” but “when are 
data?” (17). The same digital object could be treated as information or as 
data. Take, for example, a theater script. If one close reads a theater play, 
one is not treating it as data, even if the text is stored in a digital format. 
But if one uses software to calculate the relative frequencies of a word in 
this text, then the same object is treated as data. The examples of data in 
this book are digital texts, videos, motion capture, geographical coordi-
nates, and timestamps. In all cases, software is used to analyze and visual-
ize this data. For example, the videos are not viewed by a human observer, 
but treated as sequences of two-d imensional images made of pixels. The 
distribution of pixels is used to make inferences about the ways actors 
move around the stage.
This narrow definition of data will help bring focus to the arguments 
in this book, but I will not ignore the many problems it poses. For exam-
ple, when data is processed, it is often described as “flowing” from one 
form into another. However, as Jonathan Bollen (2017) notes, “if the met-
aphor applies at all, datasets have the fluidity of plastic; they flow when 
prodded, pushed, or pressed, when the heat is turned up and when placed 
under stress” (619). Data is never fully free of interpretation. As Lisa Gitel-
man (2013) writes: “Data are familiarly ‘collected,’ ‘entered,’ ‘compiled,’ 
‘stored,’ ‘processed,’ ‘mined,’ and ‘interpreted’ [ . . . ] less obvious are the 
ways in which the final term in this sequence— interpretation— haunts its 
predecessors” (3). The history of pre- digital data further illustrates some 
of the problems associated with data. Daniel Rosenberg (2013) argues 
that data has often referred to aspects of an argument taken for granted 
for the sake of discussion or analysis. “The semantic function of data,” 
he tells us, has always been “specifically rhetorical” (8, original emphasis). 
In one of its earlier recorded uses, Euclid differentiates the given quan-
tities of data from the questia of the quantities sought (Rosenberg 2013, 
19). These views evolved during the seventeenth century, when the term 
became more common:
[In] philosophy and natural philosophy, just as in mathematics and 
theology, the term “data” functioned to identify that category of facts 
and principles that were, by agreement, beyond argument. In different 
contexts, such agreement might be based on a concept of self- evident 
truth, as in the case of biblical data, or on simple argumentative con-
venience as in the case of algebra, given X=3, and so forth. The term 
“data” itself implied no ontological claim. In mathematics, theology, 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Introduction—In Pursuit of Theater’s Digital Traces • 5
and every other realm in which the term was used, “data” was some-
thing given by the conventions of argument. Whether these conven-
tions were factual, counter- factual, or arbitrary had no bearing on the 
status of givens as data. (Rosenberg 2013, 20)
The prophets of our own current techno-s cientific era—s cientists, man-
agers, and entrepreneurs— often invite us to treat data as something that 
is beyond argument. A dangerous corollary of this attitude is the belief 
that data is always true or accurate. It is often assumed that data is the 
product of a semi- automatic, or at least systematic process. For example, 
data can be collected by scientific instruments or sensors. But for the sen-
sors to work in a particular way, a series of assumptions about the phe-
nomenon of interest needs to be made: “data need to be imagined as 
data to exist and function as such, and the imagination of data entails an 
interpretive base” (Gitelman 2013, 3 original emphasis). Or, as Borgman 
(2015) puts it “conceptualizing something as data is itself a scholarly act” 
(xviii). Data artist Jer Thorp suggests that data is never inseparable from 
a system where collection, computation, and representation inform each 
other. The following excerpt from his blog vividly captures the impact of 
choice on such a system:
Whenever you look at data—a s a spreadsheet or database view or a 
visualization, you are looking at an artifact of such a system. What this 
diagram doesn’t capture is the immense branching of choice that hap-
pens at each step along the way. As you make each decision— to omit 
a row of data, or to implement a particular database structure or to use 
a specific color palette you are treading down a path through this wild, 
tall grass of possibility. It will be tempting to look back and see your 
trail as the only one that you could have taken, but in reality a slightly 
divergent you who’d made slightly divergent choices might have ended 
up somewhere altogether different. (Thorp 2017, n.p.)
In performance studies, Schechner (2013) famously distinguished 
between “as/is” performance, where certain events are generally thought 
to be performances whereas others (from political events to sporting com-
petitions) could be thought of as performances (32). Following Gitelman, 
Borgman and Thorp, I argue that data never is—n o natural category of 
data exists without preconditions. But many things can be considered 
as data, including theater performances. Perhaps all forms of theater 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
6 •  TheaTer as DaTa
research require us to have an abstract model of what performance is, but 
conventional scholarly writing allows us to refine and revise definitions as 
we go along, through the conceptual process that Kwa (2011) describes as 
a “changing semantic network.” In contrast, data forces us to make defi-
nitions explicit. For example, let’s imagine a project that wants to show 
the average number of actors per performance throughout several decades 
for a given place. For this analysis, we will need an explicit definition of 
actor in order to produce the required dataset.
This definition doesn’t need to encompass all the essential qualities 
of an actor, but it must be narrow and implementable. In other words, it 
should consist of an actionable rule that enables the systematic creation of 
a dataset. For example, a research team might define actors, for the pur-
pose of a project, as the people listed as cast members in program book-
lets. In some cases, this definition will be relatively uncontroversial. But 
other types of productions will be fraught with more intense disagree-
ment. What about a forum theater performance where the audience mem-
bers become spect- actors? Should each audience member be included as 
an actor in the dataset? Or only the resident company actors? Or should 
a separate category be created for spect- actors? All of these options are 
problematic, but building the dataset will force its makers to take a stance 
and decide on definitions of actor. It will be important to make choices 
that will enable the best inferences to be drawn from the data. For exam-
ple, if we are interested in analyzing the careers of professional actors, we 
can leave the spect-a ctors out. Our data will not be a fully accurate model 
of reality, but it will be sufficient to find verifiable patterns, through the 
process that Craig and Greatley-H irsch (2017) call “principled generaliza-
tion” (7).
It is, however, also possible to use data in a way that takes this defini-
tional problem head on. Instead of deciding at the outset whether spect- 
actors should be considered as actors, the goal of a project might be to 
frame this as a question. A data visualization might be used to explore the 
consequences of taking spect-a ctors as actors, showing how different solu-
tions to this definitional problem will yield a completely different number 
of actors. An interactive implementation of this project might enable users 
to reach their own conclusions on the appropriate definition of actors. 
Perhaps the data points in the visualization can be linked to videos, where 
users can then see the way in which specific spect- actors interact with resi-
dent company actors. In this example, data is not used to answer a ques-
tion, but to reframe the question of what it means to be an actor.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Introduction—In Pursuit of Theater’s Digital Traces • 7
This example shows that there are two ways of working with data. One 
brackets off assumptions to find patterns, while the other uses data to 
problematize these assumptions. The former offers new answers, the lat-
ter poses the question in fresh ways. I call the first approach data- driven 
and the second, data- assisted. These two terms form the conceptual back-
bone of the present book.
Data- Driven and Data- Assisted Methodologies
Data- driven methodologies use computers to reason under formal con-
straints. Knowledge is conceived as a rational project, where logic and 
formal representations are used in the pursuit of replicable conclusions 
that disprove previous ideas. In contrast, data- assisted frameworks use 
computers to “imagine in different ways” (Harrell 2013, 79). Data opens 
speculative and subjective avenues for interpretation, which contribute 
to, but don’t necessarily displace, previous interpretations. Data- driven 
methodologies aim at producing consistent explanations that are “hard 
to vary,” as physicist David Detusch (2009) puts it. In turn, data- assisted 
methodologies produce a multiplicity of explanations.
In data-d riven methodologies, we use data to answer questions. We 
create a formal representation of a question and automate a sequence of 
procedures to provide an answer. The criteria for evaluation are defined 
beforehand, and the answer is measured against these criteria. Research-
ers working in this mode know that framing a question in a narrow way 
leaves out many aspects of the object of interest, but they believe that 
there is something to be gained from this reduction. For example, they 
are able to identify features across a different scale (thousands of per-
formances rather than a single instance). This also enables the possibil-
ity of countering our intuitions. This way of working with data brings 
evenness of attention, creating “an opportunity to be surprised: to back 
something other than the sentimental favorite and to reverse consen-
sus views” (Craig and Greatley-H irsch 2017, 8). Using data in this way, 
we keep our intuitions in check, and independently verify deeply held 
beliefs. While the answers aim to provide the best way to characterize 
a phenomenon, they are not final. They can be revised and disproved by 
further research. They also include probabilistic estimates of their accu-
racy in order to provide “measured uncertainty” (Craig and Greatley- 
Hirsch 2017, 3). These estimates can be revised by other researchers as 
more data becomes available.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
8 •  TheaTer as DaTa
In data-a ssisted methodologies, we use data to transform our view of a 
problem. The purpose of framing a theatrical event as data is not to offer a 
clear answer but to augment our capacity to think about such event. Data 
becomes deformance, to use the portmanteau of deformance and perfor-
mance suggested by Lisa Samuels and Jerome McGann (1999) to charac-
terize interpretive transformations in DH. Data, in other words, provides 
a good defamiliarization strategy. Data- assisted methodologies work well 
with theatrical problems that might be fundamentally unquantifiable. But 
in trying to quantify them, as in the spect-a ctors example above, we find 
new ways to address interesting aspects of these problems. According to 
Kramnick (2018), the objective of critical method in the humanities is not 
to give definite answers but to “keep the problem open, poke around its 
edges, ask whether it has been framed in the right way, resituate the con-
versation” (n.p.). This is also what data- assisted methodologies do.
The main distinction between these methodologies is the different 
criteria they require for evaluating conclusions as useful and valid. The 
gold standard of data- driven research is replicable, incremental knowl-
edge. To assess data- driven claims, we should ask how likely the results 
are to be true and whether other independent sources of evidence cor-
roborate these conclusions. In contrast, data- assisted frameworks ask 
to be judged for their generative capacity, their potential to trigger new 
questions and bring forth new perspectives. While in data-d riven meth-
odologies many different procedures and parameters should yield the 
same answer, in data-a ssisted frameworks the same approach should 
yield multiple answers.
Data- assisted methodologies are closer to a constructivist view of 
knowledge, where every piece of data and every conclusion is observer- 
dependent. Data- driven methodologies are closer to a realist epistemol-
ogy. They seek to make intersubjectively verifiable claims, and conceive 
empirical research as a difficult, but worthy goal. Data-d riven approaches 
consider the advantages of thinking in terms of aggregates, and data- 
assisted approaches demand close attention to individual instances. Table 
1 summarizes the comparison made thus far between both approaches.
My term data- assisted is a nod to the DH community, where the term 
“computer- assisted criticism” is common (Siemens 2002; Rockwell and 
Sinclair 2016). Some of the distinctions between data-d riven and data- 
assisted methods have parallels in the epistemological disagreements 
within the DH community. But this is not a book about DH as a field. It 
is a book about what it means to use data to study theater. My starting 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Introduction—In Pursuit of Theater’s Digital Traces • 9
point, though, is that epistemological discussions arise in any field that 
deals with data and that a book on data must be a book about methodol-
ogy. It is disingenuous to think that we can describe data without talking 
about epistemology.
Thus, an epistemological vein runs through this book, as this is a 
meditation on how we come to know things about theater and the roles 
that data might play in this endeavor. For this reason, I have so far been 
characterizing data- driven and data- assisted approaches as methodologies 
rather than methods. A methodology is a framework for estimating the 
pertinence of given methods and for articulating criteria for evaluating 
results as useful, appropriate, or correct. A method is a protocol, a series 
of steps. The same method could be used within data-d riven and data- 
assisted approaches. Take, for example, word frequencies. Within a data- 
driven perspective, one could use word frequencies to identify the most 
likely author of a dramatic text of unknown authorship. But word frequen-
cies could also be used to radically transform the experience of reading a 
text. As Stephen Ramsay (2011) suggests, the font size of each word can 
be altered to represent the relative frequency of that word in the entire 
text. This data- assisted example doesn’t aim to provide a specific answer 
to a clear question, but to enable alternative readings via a computational 
defamiliarization strategy. Both the authorship attribution study and the 
defamiliarizing change in font sizes might use the same statistical meth-
ods for estimating word frequencies.
Statistics are at the heart of many data projects, regardless of whether 
they are conceived within a data- driven or a data- assisted framework. 
The presence of statistics does not in itself determine whether a project is 
Table 1. Comparison of data- assisted and data-d riven methodologies
Data- driven Data- assisted
Reasoning under constraints (bracketing Problematizing assumptions
assumptions)
Answers that are hard to vary Multiplicity of views
Productive reductionism Deformance (productive distortion)
Replicability and falsifiability Adding to a history of interpretation
Estimating the likelihood that an answer  Resituating a question
is correct
Measured uncertainty Fundamentally unquantifiable propositions
Realism Constructivism
Aggregates Individual cases
Disambiguation Exploration of ambiguity
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
10 •  TheaTer as DaTa
aligned with either methodology. Statistics are not always enlisted for the 
pursuit of objectivity, as they can also be used as defamiliarization strate-
gies. There is, however, one key difference: while descriptive statistics are 
used in both approaches, inferential or explanatory statistics belong only 
to data- driven research. The different roles of statistics are described in 
more detail in chapter 2.
Visualizations are also central to many data projects, but the way in 
which they are used varies within both methodologies. Data- driven visual-
izations rely extensively on statistical graphs, using conventions borrowed 
from the scientific community. Data- assisted visualizations require the 
invention of new graphical conventions. Data- assisted visualizations aim 
to resituate conversations, and to communicate how ideological biases 
shape the data. Visualizations used in computational theater research 
can be static or interactive, regardless of which methodology is used. But 
interactivity is particularly useful for data-a ssisted visualizations that aim 
to provide context to each data point, as in the example above, where each 
spect- actors data point is linked to a video-r ecording. The different roles 
of visualizations are described in chapter 3.
When distinguishing between data-d riven and data- assisted meth-
odologies, it doesn’t matter whether a project appears to rely extensively 
on statistics or not, or the extent to which it uses visualizations. The size 
of a dataset doesn’t matter either: both methodologies can be applied 
to very small and very large datasets. What matters is why statistics and 
visualizations are used, and how they support the argument being made. 
The important thing, in other words, is which criteria for evaluation are 
enlisted in the communication of a research project.
There are many differences between data-d riven and data-a ssisted 
methodologies, but they can also be used concurrently within the same 
research project, and both can contribute to a more comprehensive under-
standing of theater history and practice. As noted earlier, computational 
research exists in a spectrum from realism to constructivism. Data- driven 
research tends to be closer to realism and data- assisted to constructivism. 
Some researchers occupy radical positions at the opposite ends of this 
spectrum, affirming either that “we should strive for a science of culture” 
or that “data is always observer-d ependent and socially constructed.” But 
it is possible to carry out research from a more nuanced position in the 
spectrum. I believe, with the proponents of critical realism, that one can be 
a realist about some things and a constructivist about others. Critical real-
ism is a metatheoretical perspective from the philosophy of science that 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Introduction—In Pursuit of Theater’s Digital Traces • 11
is becoming increasingly common in the study of the social world. While 
critical realism is not often invoked in the context of DH, I believe that 
it is consistent with the nuanced ways in which many researchers speak 
about their work. For example, many proponents of data- driven methods 
are quick to recognize that there is a certain subjectivity to their projects. 
The choice of a research topic, for instance, can’t ever be fully derived 
from objective first principles and always betrays a situated way of look-
ing at the world. Likewise, many proponents of what I call data- assisted 
research note that the data used must conform to intersubjectively verifi-
able standards.
In contemporary humanities departments, calling someone a “posi-
tivist” is most certainly an insult. I have witnessed many data research-
ers defending themselves against accusations of positivistic allegiances 
by merely stating that no, their projects are not positivistic, but without 
offering further elaboration. I would like to suggest it is more useful to 
strive for a fine-g rained vocabulary, such as the one that critical realism 
provides. Critical realism is given more attention in chapter 1, but its pre-
suppositions run through the rest of the chapters in this book. For some 
projects, data- driven and data- assisted perspectives might coexist. But 
my goal is not to suggest that every research project must combine data- 
driven and data-a ssisted perspectives. Rather, my goal is to encourage any 
data project to aim towards critical realism. For a data- driven project, this 
means taking history and definitions seriously. For a data-a ssisted proj-
ect, it means aiming to show how playful defamiliarizations and multiple 
perspectives can contribute to a fuller understanding of a complex, shared 
reality.
Some projects will be better suited for a data-d riven perspective, and 
others for a data- assisted one. In my view, data- driven research is only 
useful for research questions that fulfill four conditions:
 1. There is low ambiguity in the definitions of key concepts. Take, 
for example, the analysis of scene structure in theater. In some 
dramatic traditions, it is possible to establish consensus on what 
constitutes a scene. This is the case in Ibsen’s plays. But it would 
be harder to reach consensus in what constitutes scene boundaries 
in the work of other playwrights.
 2. The area of interest can be thought of as discrete features rather 
than continuous processes. It is possible to analyze methodi-
cally the number of productions in which Andre Gregory has 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
12 •  TheaTer as DaTa
been involved as a director. One could create a visualization of his 
creative output through time. But analyzing the aesthetics of one 
of his plays would require interpretive work more suited to data- 
assisted analysis, or other methods.
 3. The data is arithmetically malleable. For example, one could mea-
sure the distance that two touring theater companies covered in a 
year. One could add these measurements and conclude that one 
company traveled twice as far as the second one. But one could not 
say that one company is twice as original as another, even in cases 
where the critical consensus is that one company is indeed more 
original.
 4. Data can be combined into an aggregate. For certain research proj-
ects, it might make sense to take the entire choreographic oeuvre 
of Akram Khan as a unit. But it might not make sense to consider 
migrant theater as a unit. Perhaps what makes migrant theater inter-
esting as a site of study is the diversity of individual approaches.
These four conditions are related, but they are not equivalent to each other. 
One could imagine a research area where there is high ambiguity, but 
researchers can still identify discrete, arithmetically malleable features. 
That is, many researchers identify discrete features but they don’t agree on 
what they are. Let’s take, for example, the plays attributed to Shakespeare. 
There is some disagreement on which plays he wrote, but plays are still 
discrete features that are arithmetically malleable. Different researchers 
can offer competing numbers of plays written by Shakespeare. The analy-
sis of areas such as this, which only meet some of the four criteria, are 
better served by data- assisted methodologies, provided that discrete fea-
tures can still be identified (otherwise there is nothing to be turned into 
data). Cases where none of the criteria can be identified are better served 
by other types of methodologies that don’t use data, such as ethnography 
or practice- based research. This information is summarized in table 2.
I am sure that some people will disagree with my examples above. The-
Table 2. Criteria for different types of research
Criteria Data- driven Data- assisted Other approaches
Low ambiguity Yes Maybe No
Discrete features Yes Yes No
Arithmetic malleability Yes Maybe No
Aggregability Yes Maybe No
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Introduction—In Pursuit of Theater’s Digital Traces • 13
ater and performance studies are, after all, built around sophisticated dis-
agreement (Strine, Long, and Hopkins 1990). Perhaps there are researchers 
who consider the question of Shakespeare authorship settled, scholars 
of Ibsen who could point me to disputed scene boundaries in his plays, 
or dance scholars who believe that Akram Khan’s work is too diverse to 
be considered as an aggregate. My proposed criteria are not meant to be 
taken as inherent properties of a research area, but as articulations of a 
researcher’s perspective. Data-d riven research is useful for someone who 
assumes that the four criteria are met for a given research project. Research-
ers who believe that such assumption would be misleading should bet-
ter use a data- assisted methodology. Assuming that the four criteria are 
met doesn’t mean sweeping problems under the rug, and low ambiguity 
doesn’t mean no ambiguity. A researcher can still acknowledge disagree-
ment in her field, while choosing to pursue a data- driven project. This 
acknowledgment can be communicated together with the results of the 
research. Or the consequences of this disagreement could be further ana-
lyzed through a data- assisted methodology.
In addition to the criteria described above, every data project requires 
good sources of data. The availability of data will always be connected to 
the existence of specific records, and this will limit the kinds of questions 
we can ask. The kinds of data available also signal the existence of spe-
cific histories of representation, and might raise many critical issues that 
should be addressed when using the data. Good data is also thick data. We 
need to know how and why the data was collected, and how it was pro-
cessed. I use the term thick data following Tricia Wang (2016), who extends 
the anthropological concept of thick description to data. Thick data should 
also include thorough descriptions of our doubts and misgivings in the 
collection of data, a chronicle of the decisions made and a description of 
the roads not taken. Data biographies are useful for communicating these 
aspects of data and whenever possible, they should accompany a research 
project (more on this on chapter 8).
Computational Theater Research:  
Becoming Technical, Remaining Critical
In this book, I refer to data- driven and data- assisted methodologies under 
the umbrella term of computational theater research. Computational is a 
fraught term in DH, and some people consider it equivalent to what I call 
data- driven research. In the natural sciences, computational research relies 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
14 •  TheaTer as DaTa
on algorithmic models and simulations rather than experimental and obser-
vational methods. My usage here differs slightly from these other cases.
I refer to computational research as that which requires computa-
tionally intensive processes to transform and analyze data. Some of the 
projects described in the book could theoretically be done by hand cal-
culations. But this would be so time consuming that it would not enable 
the exploratory and iterative processes that characterize these research 
approaches. Thus, I can further refine my definition and describe compu-
tational theater research as the exploratory and iterative use of computers to ana-
lyze digital theater data. I choose the label computational over digital in order 
to distinguish the work described here from the study of digital perfor-
mance (Dixon 2007; Bay- Cheng, Parker-S tarbuck, and Saltz 2015; Chatzi-
christodoulou, Jefferies, and Zerihan 2009; Beyes, Leeker, and Schipper 
2017; Causey 2006) and from other types of work at the confluence of DH 
and theater, such as digital archiving (Bardiot 2015; Carlin and Vaughan 
2015; Sant 2017).
Computational theater research, as defined here, doesn’t include sim-
ulations that enable users to recreate or devise performances in an inter-
active computer environment (Roberts- Smith et al. 2013; Delbridge and 
Tompkins 2009). While I give some attention to dramatic texts, I am pri-
marily interested in the analysis of theater as an event, and I don’t fully 
survey the extensive research on the computational analysis of dramatic 
literature. My project is conceptually similar to what Clarisse Bardiot 
(2017) terms theater analytics, which involves “algorithmic approaches and 
data visualization” (n.p.). However, her approach is somewhat closer to 
what I term data- driven research (Bardiot 2015, 2018) and, for me, com-
putational theater research also includes data- assisted projects.
Another reason for choosing the label computational is to signal the 
importance of computer programming for both data-d riven and data- 
assisted theater research. Some people believe that one must be a pro-
grammer to carry out DH work, a decidedly divisive attitude. I agree with 
the more nuanced position championed by Berry and Fagerjord (2017, 38– 
40). Learning to code is akin to learning to apply a given theoretical frame-
work: knowing how to code equips a scholar with better understanding of 
the tools they use, even if the scholar is not a professional programmer. I 
also believe that being a programmer is not a simple binary matter, since 
there are different levels of expertise.
However, I do believe that those interested in computational theater 
research would benefit from some experience in programming (and ide-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Introduction—In Pursuit of Theater’s Digital Traces • 15
ally some knowledge of statistics and visualization), regardless of whether 
they aim to pursue primarily data-d riven or data-a ssisted work. There are 
epistemological and pragmatic reasons for this. One could perhaps say 
things about Japanese theater without speaking the language (see Leiter 
2012), but having at least some familiarity with the language opens an 
entirely different realm of research avenues to the scholar of Japanese the-
ater. The same is true for programming—a lthough non- programmers can 
say many things about software tools, the nature of the questions that can 
be pursued changes when one is able to follow at least some basic aspects 
of the technical discussions in a field. From a pragmatic point of view, 
we also need theater scholars who can code or the future will not be sus-
tainable (as I argue more fully in chapter 9). If we want to build research 
that is open, durable, and shareable, more of us need to get our hands 
dirty and do some programming. Otherwise, we are at the mercy of finan-
cial resources and tools developed by others. I argue that computational 
research does not need extensive financial resources, but it does require a 
growing community of programmer-s cholars.
This, however, is not a technical handbook. I will describe all techni-
cal procedures in accessible terms, but interested readers can consult the 
appendices (and the web companion) where technical terms are described 
in more detail, and where the data used for the excursions can be down-
loaded by users interested in reproducing my results. For some of the 
projects described in the book, I have also created step- by-s tep Jupyter 
Notebooks, which can be downloaded from the University of Michigan’s 
Fulcrum website. These notebooks are aimed at users who want to learn 
more about the technical procedures described in this book. They include 
explanatory notes, and they are accessible to readers with no program-
ming background who want to peek under the hood and get a sense of the 
technical devices behind some of the conclusions reported in this book. I 
don’t expect all readers of this book to become programmers, but I aim to 
make an argument about why they might consider dipping their toes into 
programming if they haven’t already done so. I also encourage readers 
with more programming experience to reuse my data and find new ways 
of analyzing it beyond what I have initially considered.
Finding a balance between data methods and critical awareness means 
that we need to be simultaneously makers and observers, cultural critics 
and data scientists, programmers and scholars. We need deep under-
standing of the scientific and technological principles of our tools, as 
much as we need a hermeneutic disposition that can unpack their ideolo-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
16 •  TheaTer as DaTa
gies, context, and ethical consequences. A great deal of scholarly attention 
from the humanities has focused on the critical analysis of data produc-
tion and shown how data is always contingent on the financial, historical, 
and cultural contexts in which it is embedded. This type of critical aware-
ness can also refine our data practices, enhancing the ways we collect and 
analyze data. Critical analysis and software programming can go hand in 
hand. As the writers of Data Feminism (D’Ignazio and Klein 2020) argue, 
it is possible to build on critical insights to improve our data methods. 
Many of the contributors to that volume describe themselves as both cul-
tural critics and data scientists. We also need this combination of skills 
to create good scholarship at the intersection of theater studies and com-
putational methodologies. This also means that we need to include not 
only data tables and code in our projects, but also thick descriptions of the 
ways in which the data was collected and analyzed. I believe that a respon-
sible use of data requires us to critically scrutinize the contexts in which 
the data was produced. We need to acknowledge the ethical, social, and 
political complexities of the work we do, and the contexts in which we are 
located. In my own research, I try to demonstrate this by highlighting the 
cultural background of the performances I study and the conditions under 
which the data was collected. In my computational theater research proj-
ects, I ask narrow, highly specific questions concerning theatrical prac-
tices in Indonesia and Singapore. This book is not about my work, but it 
is informed by it.
This Book’s Journeys: A Guided Tour with Excursions
This book is a meditation on methodology, not an analysis of any spe-
cific theater tradition or genre. I will report on many projects carried out 
by research teams around the world—t his is primarily a book about other 
people, about my colleagues and predecessors who have worked at the 
intersection of theater and data in fields ranging from contemporary Por-
tuguese choreography to adaptations of Shakespeare in Asia. The book is 
constructed as a guided tour of this incipient but diverse area, and those 
here represented might not always fully agree with the way I have charac-
terized their work. But my hope is that my proposed vocabulary will lead to 
productive conversations and perhaps inspire alternative conceptual cat-
egories to describe computational research in theater and performance.
In this tour, I systematically apply my proposed vocabulary to char-
acterize a wide range of projects from around the world as data-d riven 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Introduction—In Pursuit of Theater’s Digital Traces • 17
and data- assisted theater research. I complement this guided tour with 
short excursions into my own work on Indonesian and Singaporean the-
ater at the end of each chapter in part 2. These are excursions rather than 
fully fledged case studies that reference work that I have often published in 
more detailed formats elsewhere. The excursions don’t aim to constitute a 
coherent analysis of Singaporean or Indonesian theater, but they serve to 
punctuate the more general guided tour I offer in the book. I don’t include 
a comprehensive analysis of the many issues surrounding theater practice 
in the region where I live, but every time I embark on these excursions, I 
emphasize their histories and contexts. I do this to further a methodologi-
cal tenet: data methodologies are wholly contingent on careful, critical 
adaptation to a given context.
Excursions into Singapore and Indonesia, two neighboring but fairly 
different countries, bring these contextual distinctions into sharp relief. 
Singapore, an island city-s tate with just over five and a half million inhab-
itants is the economic hub of Southeast Asia, and it has developed a 
vibrant contemporary theater culture over the past thirty years. Singapore- 
based theater companies collaborate extensively with foreign partners and 
Singaporean festivals are common stopovers for some of the most influ-
ential theater companies from around the world. The diversity of formats 
and languages found in Singaporean theater attests to the multicultural 
makeup of the city- state, and also to the material wealth that underwrites 
recent theatrical experimentation. Indonesia is Singapore’s largest neigh-
bor (with more than 260 million inhabitants, the fourth most populated 
country in the world) and it is home to a completely different theatrical 
environment. For the past twelve years, I have been primarily interested in 
the traditional performances of Java, which constitute only one of the cul-
tural regions of this vast archipelagic nation. I have conducted data- driven 
and data- assisted work on traditional Javanese puppetry and dance. Some 
of these traditions are at least one thousand years old, but are still the 
focus of active experimentation, and are still performed dozens of times 
every day.
Some readers might think that these excursions are too idiosyncratic 
and too specific to offer generalizable lessons for theater and perfor-
mance scholars working on other areas. For example, one of my projects 
on Javanese dance describes character types, a highly specific convention 
which will not speak directly to specialists of Russian ballet or Argen-
tinean tango. But the methods I describe for the analysis of motion cap-
ture can be adapted to many other movement traditions. Any data proj-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
18 •  TheaTer as DaTa
ect could have been used as a springboard for epistemological reflections 
such as those found in this book. No example is perfect for a book with 
the methodological goals that animate the discussions that follow. A book 
like this one could have also been written by an expert on Arthur Miller 
or Chinese Opera. Their examples, constrained by context and interests, 
would also not generalize to most other contexts. But a reflection on how 
data changes theater can start from any example. The point is not that all 
examples are the same, but rather that any theater example is unique and 
culturally determined. One advantage of using relatively unknown exam-
ples from Singapore and Indonesia is that they will remind every scholar 
that their areas of study, too, are culturally specific.
Ultimately, my goal is to show that computational research has lim-
ited, but important, applications for theater studies. I will try to show that 
pioneering research in this field is scratching the surface of what is pos-
sible, with many tantalizing possibilities just within our reach. I do not 
want to convince every reader to use data and computational methodolo-
gies, but to contribute to the ongoing conversation around epistemology 
and methodologies in theater research. Methodological reflection has 
been at the forefront of theater studies for many years and these conversa-
tions stand to gain from paying some attention to the challenges posed 
by data and computers. While I don’t expect every reader of this book to 
become a computational researcher, I do hope to provide both adherents 
and detractors of these new approaches with more nuanced vocabularies 
with which to carry on their conversations.
This book is divided into three sections: the first delves into the epis-
temological challenges that computational research poses for theater 
studies, the second is a guided tour of key areas of theater amenable to 
computational research, and the last considers the material and ethical 
implications of computational methodologies for theater research.
In the first part, chapter 1 argues for a more textured conversation on 
methodology. I continue the distinctions introduced here. I analyze the 
way methodology is usually conveyed in theater research and pedagogy. I 
then contrast these views with the way methodology is conceptualized in 
the natural sciences. I aim to draw cautionary lessons from the superficial 
application of scientific principles to the study of culture. My goal is to 
enact, by way of critical realism, a more nuanced epistemological attitude 
that doesn’t fit into simplistic distinctions between the humanities and 
the sciences. Chapter 2 describes how both data- driven and data- assisted 
research can make use of statistics. This chapter is not a survey of statis-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Introduction—In Pursuit of Theater’s Digital Traces • 19
tical procedures, but a description of how different statistical methods 
might lead to different types of bias, and what to do about these problems. 
Chapter 3 describes how data- driven and data-a ssisted research can use 
visualizations. This chapter characterizes different approaches to data 
visualization through the work of leading thinkers in statistical graphics, 
data journalism, and DH. I also argue that the conceptual lens of theater 
and performance research could be deployed to highlight the performa-
tive nature of certain data visualizations.
Part 2 is a guided tour of four key areas of theater research that can 
be modeled as data: words (chapter 3), motion (chapter 4), relation-
ships (chapter 5), and locations (chapter 6). In the chapter on words, I 
don’t delve into the digital analysis of dramatic texts but I focus on what 
digital text analysis can bring to the study of program booklets, theater 
reviews, interviews, and other textual sources—a n area of research with 
enormous promise but few precedents. By motion, I refer to the move-
ments of actors and objects on stage, and by relationships I refer to the links 
between fictional characters in a play (copresence in a scene), or between 
people in collaborative networks. In all of these areas, extensive work has 
been already carried out— but I argue that there is much more that could 
be done. Researchers in DH who work on theater and performance are 
not always well known in theater studies circles. For example, Schöch, 
Fischer, and Wiesner are seldom quoted in theater books and articles. In 
my survey, I bring work in theater studies in conversation with work from 
elsewhere in DH, as well as with perspectives from engineering, computer 
science, and statistics.
In each chapter, I first consider what data can mean for the area under 
consideration and describe a variety of methods for the analysis and visu-
alization of this data. Then I describe how these methods can be used 
within data- driven and data- assisted methodologies. I maintain a firm 
distinction between methods and methodologies to show that the same 
procedures can often be used to achieve very different research objectives. 
My overview does not aim to introduce groundbreaking perspectives; I 
have the more modest goal of using a consistent vocabulary to describe 
the potential for data-d riven and data-a ssisted theater research in each of 
these areas. As noted above, each chapter concludes with a short excur-
sion into theatrical practices in Singapore or Indonesia. These are sum-
maries of work that I have carried out, often together with my scien-
tific collaborators Gea Oswah Fatah Parikesit, Andrew Schauf, and Luis 
Hernández- Barraza.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
20 •  TheaTer as DaTa
The third and last part consists of two chapters. Chapter 8 considers 
the conditions for the future of computational research in theater stud-
ies. Both data-d riven and data- assisted methods require consistent ways 
to document, standardize, and preserve data and the software that can 
interpret it. Here, I tackle the challenges that sustainability poses for com-
putational research, from making sure that resources are available in the 
future to thinking of ways to enable other research teams to reuse exist-
ing datasets and to reconstruct their original context. However, I also note 
that making sure every data point is fully conformant with standards for 
reusability and sustainability is futile for a variety of reasons, ranging from 
copyright restrictions to financial limitations. It is much better to empha-
size which areas deserve priority and focus our attention accordingly.
Chapter 9 closes the book with a brief reflection on the ways computa-
tional research can further change the study of theater, and what is needed 
for that to happen. I return to an argument that I made earlier in this intro-
duction. Doing sustainable and meaningful data- driven and data- assisted 
research into theater will benefit from a community of scholars who can 
do some programming. This is not just a technical skill, but a mode of 
thinking about the world. My joys and travails as a programmer have 
deeply shaped the journeys that follow.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
parT 1
Pre- departure Reflections
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
ChapTer 1
Toward a More Nuanced Conversation  
on Methodology
In 2012, Su Wen-C hi was artist- in-r esidence at CERN, the largest scientific 
laboratory in the world. As a media artist and choreographer, she was very 
interested in understanding the ways that scientists work, and how this 
differs from artistic practice. In a workshop on “art and science” in Sin-
gapore several years later, she related her experiences and described, in 
minute ethnographic detail, many fascinating aspects about the lives and 
routines of scientists at CERN (Su 2019). She explained that when deal-
ing with complex problems, scientists are assigned to two groups. If both 
group’s answers are identical, then the answer is considered correct. One 
participant in the workshop noted how this couldn’t be further removed 
from artistic practice. “If we gave a question to two groups of artists,” the 
participant observed, “and they both had the same answer, we wouldn’t 
think that the answer was correct—w e would conclude that the ques-
tion was wrong.” This observation wonderfully captures the differences 
between two ways of producing knowledge: one that values contextual-
ized, unique responses, and one that goes to great lengths to validate and 
corroborate the answers it produces.
Replicability is at the core of a realist view of knowledge, traditionally 
closer to the sciences. Interpretation is the foundation of a constructivist 
approach to knowledge, conventionally closer to the humanities. Data- 
driven research places more emphasis on replicability and data-a ssisted 
research prioritizes interpretation. But thoughtful computational theater 
research requires a sophisticated balance. If data-d riven research does not 
acknowledge the role of interpretation, it will have turned its back on the 
core tenets of theater research. Conversely, if data-a ssisted research has 
no replicable features, then how can data possibly enhance conventional 
theater research?
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research. 23
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
24 •  TheaTer as DaTa
This chapter seeks to enable a nuanced conversation on methodology 
by pursuing three arguments. First, I explain why theater research is pri-
marily interpretive, and why this entails some limitations. Then, I describe 
why replicability is fundamental in the sciences and explain why adapting 
scientific methodologies to the humanities has often been a misguided 
and controversial enterprise. Lastly, I present an overview of critical real-
ism and suggest that this metatheoretical perspective can enable a more 
sophisticated description of the epistemological goals of computational 
theater research.
Theater Research and the Primacy of Interpretation
Theater scholars pursue knowledge through a variety of means, which 
include practice- as- research, ethnography, historical analysis, phenom-
enology, and using different theoretical perspectives for performance 
analysis. In spite of this diversity, most approaches rely on interpre-
tive methodologies which are social, intuitive, situated, and rarely made 
explicit. Within interpretive methodologies, a method can’t be followed to 
the letter, for it must first be transformed by each individual researcher. A 
method, in this perspective, is not a series of steps, but a mode of analysis 
and a perspective on the world. It is hard, for instance, to explicitly state 
what constitutes a good postcolonial analysis. Books and classes that 
teach this approach proceed by example and discussion. Researchers and 
educators often emphasize the why for postcolonial analysis rather than 
the how. There is ample space for disagreement on how to do a good inter-
pretive analysis, and conclusions might differ when the method is carried 
out by a different researcher.
The same is true for methods such as ethnography. Clifford Geertz 
(1973) suggests that ethnography includes some explicit procedures such 
as “transcribing texts, taking genealogies, mapping fields” but that it 
is not these “techniques and received procedures that define the enter-
prise” (6). Ethnography is defined by thick descriptions, a notion Geertz 
took from Gilbert Ryle. Such descriptions constitute a layered analysis 
of actions that consider their situated meanings. Geertz’s example is “a 
wink.” A thin description would gloss a wink as the closing of an eyelid. 
But the same gesture can be understood as provocation, mockery, or even 
rehearsal of mockery— depending on where it is performed and how it is 
perceived. Thick description attends to the complicating factors of con-
text, and describes not only actions, but the surrounding conditions and 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Toward a More Nuanced Conversation on Methodology  • 25
histories that imbue them with specific meanings. Being able to unpack 
these layers of meaning is not a straightforward process. Geertz famously 
said that understanding a culture from the perspective of other people 
is more like “grasping a proverb, catching an allusion, seeing a joke or 
[ . . . ] reading a poem” (Geertz 1974, 70).
The bulk of theater research entails the same kind of epistemic opera-
tion: learning a method is more like grasping a proverb than it is like fol-
lowing an instruction booklet. In some cases, it is hard to distinguish 
method from theory, as the example of postcolonial analysis shows. For 
Soyini Madison, carrying out research means taking a “detour around a 
topic with theory” (Madison 2005, 4, original emphasis). This entails no 
discernible, reproducible steps:
[W]e see and feel theory as something like a collaborator with perfor-
mance, a co- subject however uncomfortably removed from the stabil-
ity of a subject/object relation. In embodied relation to performance, 
theory moves. It is less the primary figure in a new construction of 
performance than it is a reflexive participant in the poiesis of know-
ing, being, and acting that performance initiates. I would thus have to 
call performance- and- theory a project of interanimation: of discern-
ing how many more vital possibilities (for performance, for theory, for 
the world) are wrought by the transactivity of performance and various 
ways of imagining it. (Madison 2005, 2)
Judging the quality of theater research is a task for experts who can appre-
ciate the overall argument of a research piece, and who don’t judge its 
conclusions based on the appropriateness of specific methods for data 
collection and analysis. Writings in theater— and in the humanities more 
generally— are usually not structured according to inflexible, predefined 
formats. The opposite is true of the sciences, where a rigid system dictates 
the structure (literature review, data collection, results, discussion, future 
work). Rigor and intellectual merit are important in theater scholarship, 
but they are not often conveyed or judged according to their compliance to 
inflexible structures amenable to step- wise execution.
Theater scholarship follows conventions, but it is not easy to categori-
cally ascertain what they are, and making them explicit is not a key con-
cern of pedagogy and research practice. We do not generally characterize 
methods as sequences of incontrovertible, executable actions that should 
be strictly reproduced to ensure consistent results. A simplistic counterex-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
26 •  TheaTer as DaTa
ample is the way I prepare my coffee: I weigh and grind the beans, wait for 
a light to go off in the espresso machine, and press a button. This method 
produces consistent coffee, day in and day out, and this is why I like it. 
Many people would be appalled at the thought of doing performance anal-
ysis the way I prepare my coffee. Perhaps a reason for this is that a quality 
that is desirable in coffee—p redictable consistency—i s not desirable in 
performance analysis, as illustrated by Su Wen- Chi’s story.
In graduate school, I used to work as a tutor in a large introductory 
class to theater studies, where one of the student assignments was to 
carry out a performance analysis. Each of the tutors had to grade dozens 
of analyses and one of the most excruciating aspects of the task was the 
distinctly Sisyphean feeling one had when seemingly reading the same 
essay, over and over again. The essays were incredibly similar to one 
another and we took this to mean that they were poorly written. In our 
evaluation, we were looking for creativity and individual expression. Step- 
wise methods are generally not well suited for generating a multiplicity 
of creative and personal responses. They aim exactly at the opposite, at 
producing predictable results through processes that minimize personal 
expression. In the class, we tried to encourage a particular sensitivity in 
the students, enticing them to look at the world of theater in a thoughtful 
way. But as they listened to us, they were aiming to discern protocols: spe-
cific steps to write an essay that can achieve a reasonable grade. Under-
standing what we meant was particularly difficult for science students (as 
a general education class, a large portion of the students came from tech-
nical backgrounds). Part of the cognitive dissonance they experienced is 
due to the fact that reproducible methods (common in the sciences) aim 
to discover things about the world, but methods in the humanities seek 
something else.
In the Routledge Introduction to Theater and Performance Studies, perfor-
mance analysis is described as “a cognitive process that should lead to 
the constitution of meaning” (Fischer-L ichte, Thomasius, and Arjomand 
2014, 54). This wording captures the very essence of interpretive methods: 
they are processes for constituting meaning, and for developing a par-
ticular relationship to performance, rather than for discovering its funda-
mental, unchanging properties. The Routledge Introduction has a section on 
methodologies, comprising three chapters: performance analysis, theater 
historiography, and theorizing theater and performance. Performance 
analysis can only be used when the researcher has accessed the perfor-
mances directly. The book insists that historiographic methods are needed 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Toward a More Nuanced Conversation on Methodology  • 27
to analyze performances from the past, or those seen on video (even when 
they correspond to the very recent past). Performance analysis is then bro-
ken down into semiotics and phenomenology, as two distinct approaches 
that the authors encourage the readers to combine. A semiotic analysis 
could be conducted methodically (in the coffee- making sense). There are 
indeed guides the step-w ise semiotic analysis of theater (Ubersfeld et al. 
1999; Aston and Savona 1991; Pavis 1996). This systematic approach has 
now fallen out of fashion, but as a step-w ise method, semiotic analysis 
could still be taught in an undergraduate- level introductory book. This is 
not what the Routledge Introduction does. Instead of describing the sequen-
tial and executable steps of semiotics, the readers are exposed to the phil-
osophical underpinnings of semiotics. Granted, a strict semiotic analysis 
can be deeply unsatisfying. Although the steps are executable, they are not 
necessarily unambiguous. Perhaps the reason why the writers of intro-
ductory textbooks leave them out is that they feel some epistemological 
unease when dealing with the reductiveness of reproducible methods. I 
suggest there are several reasons for this unease, which I list below:
 1. A stance against positivism and philology: Moving against positivism 
and philology has been an important strategy for theater studies as 
the discipline has found its way throughout the twentieth century 
(Jackson 2004).
 2. The legacy of poststructuralism: This has endowed us both with a sus-
picion of method and a fascination with the vocabulary of method. 
Think, for example, of Derrida (1978) and Foucault (1998), both 
of whom wrote about method in an evocative way. As Bishop and 
Phillips note:
[I]n raising the question of method [ . . . ] one is concerned not 
with what might be called the “technical devices” of the theoretical 
disciplines (philosophy, sociology, psychology, political science, 
etc.) but with a more sustained and comprehensive understand-
ing of the conditions of possibility for disciplines like these, which in 
their foundation and their horizons offer a kind of incomplete heri-
tage for future knowledge. (Bishop and Phillips 2007, 267, original 
emphasis)
This passage signals a poststructuralist approach, where condi-
tions of possibility are the central focus of the inquiry. However, in 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
28 •  theater as data
the pages that follow I will argue that data requires a more sus-
tained engagement with the “technical devices” of methods.
 3. An artistic ethos: Many theater researchers are also artists or at least 
have an artistic sensitivity. They aim to transfer these qualities, or 
at least reflect them, in their academic writing. This is different in 
other fields that also deal with artistic practice. Take, for example, 
cultural policy, where a more dispassionate approach is preferred. 
In theater and performance studies, the ethos of artistic practice 
and a belief in individual expression permeate our own sense of 
what constitutes good and valuable research.
 4. A preference for implicit rules: Theater scholars prefer the language of 
allusion and poetry (see the beautifully written passage from Soyini 
Madison above) and believe there is something prosaic in narrow 
definitions and step- wise sequences of goal- oriented actions.
 5. Comfort with ambiguity: Theater scholars recognize the complexity 
of concepts such as performance, which have no clear definition. As 
mentioned in the introduction, most of us are comfortable in envi-
ronments of sophisticated disagreement. Kagan (2009) argues that 
different disciplines attract different personality types, and that the 
sciences tend to attract people with lower tolerance for ambiguity.
 6.  A principled stance against technocracy and the knowledge economy: Like 
many humanities scholars, we believe that in our world, a super-
ficial veneration of science and quantification has damaged many 
areas of life. Dwight Conquergood (2002) echoes common views 
when he advocates for the importance of seeking knowledge that is 
local, embodied, socially generated and implicit. Replicable meth-
ods are not well aligned to these objectives.
An excellent example of the principles outlined above can be seen in 
Research Methods in Theatre and Performance (Kershaw and Nicholson 2011), 
which is aimed at academics and graduate students, primarily those inter-
ested in practice-a s- research. Their aim is described as a “challenge to 
outmoded perceptions that the terms ‘method’ and ‘methodology’ imply 
an attempt to capture, codify and categorize knowledge” (1). The authors 
constantly emphasize the unpredictable nature of theater, and explicitly 
articulate their discomfort with reproducible methods, asking “what are 
methods for but to ruin our experiments?” (15).
The books I’ve quoted thus far don’t constitute a full survey of the 
field. Areas such as cognitive approaches to performance warrant a more 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Toward a More Nuanced Conversation on Methodology  • 29
nuanced take than what this overview permits, as they might not be fully 
described by the characteristics listed above. The books surveyed those 
far are aimed at undergraduate, graduate, and academic researchers but 
they are also indicative of larger patterns. Fischer-L icthe, Arjomand, and 
Mosse provide an introduction to theater studies that is not dogmatic; 
Madison helps us think about theory in provocative and textured ways; 
and Kershaw and Nicholson provide unique insights into the relation-
ship between artistic practice and academic research. Conventional ways 
of approaching methodology in theater are certainly useful, but there are 
three potential problems that demand attention.
First, implicit definitions of methodology limit interdisciplinary exchange. We 
want creative and situated responses, but it is very hard to explain this to 
people outside the humanities. Describing our methods as situated, social 
and intuitive does not diminish their value, it just describes them more 
adequately. I think that a more explicit characterization of interpretive 
methods is crucial for interdisciplinary research and for communicating 
our epistemological projects to society at large.
A second problem is that some aspects of our scholarship would benefit from 
reproducible approaches. Although a reductive strategy at first glance, this 
might help identify previously unsuspected features of theater, which can 
also aid in interpretive endeavors. This applies not only to computational 
research methods but also to standardized surveys such as the ones that 
are commonly used in the social sciences.
Third, our lack of training in reproducible methods prevents us from critically 
assessing the flaws and strengths of relevant scientific research. In reproducible 
research, creativity plays a role in devising new methods for collecting and 
analyzing data, or in identifying which method is relevant. But once the 
design has been made, the collection and analysis of data should proceed 
in a systematic, inflexible way. For these methods to be deemed rigorous, 
there is no scope for creativity in the execution of the methods, only in 
their design.
If you suspect you are sick and go to the doctor, you would expect inter-
pretive flexibility on the part of a good doctor: empathy, creativity, and 
intuition. But you would not want the way your blood samples are col-
lected and analyzed to be creative— you would want this to be conducted 
in a way that is as systematic and consistent as possible. Some aspects of 
medical analysis benefit from situated hermeneutics, while some require 
the rigorous consistency of impersonal, reproducible methods. In fact, 
diagnosis and treatment will work better the more these two areas are dif-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
30 •  TheaTer as DaTa
ferentiated. Might this also be true for theater scholarship? If we are to 
work with data, we need a better way to describe our methodologies and 
their objectives.
The Controversies of Scientific Research in the Humanities
The sciences traditionally favor a very different mode of evaluating evi-
dence. When describing scientific practice, Richard Feynman (1985) 
famously said that “the first principle is that you must not fool yourself— 
and you are the easiest person to fool” (343). To avoid fooling ourselves, 
we need replication: “we’ve learned from experience that the truth will 
come out. Other experimenters will repeat your experiment and find 
out whether you were wrong or right. Nature’s phenomena will agree or 
they’ll disagree with your theory” (342).
Replication is fundamental in the sciences since it is the only way 
to ensure that explanations are likely to be true. Elsewhere, Feynman 
writes: “scientific knowledge is a body of statements of varying degrees of 
certainty— some most unsure, some nearly sure, but none absolutely cer-
tain” (Feynman 1955, 14). Absolute certainty is impossible, but the goal 
of science is to reduce the uncertainty of possible explanations. Science 
aims to achieve this through continued observation and experimentation. 
In another lecture, Feynman describes the process in the following way:
In general we look for a new law by the following process. First we 
guess it. Then we compute the consequences of the guess to see what 
would be implied if this law that we guessed is right. Then we compare 
the result of the computation to nature, with experiment or experience, 
compare it directly with observation, to see if it works. If it disagrees 
with experiment it is wrong. In that simple statement is the key to sci-
ence. It does not make any difference how beautiful your guess is. It 
does not make any difference how smart you are, who made the guess, 
or what his name is— if it disagrees with experiment it is wrong. That 
is all there is to it. (Feynman 2017, 156)
Is a scientific study of culture possible? Here, I agree with Ted Underwood 
(2019b): “Questions that historians and literary critics used to debate 
are increasingly scooped up by quantitative disciplines [ . . . ] Instead of 
saying that the humanities are besieged and giving up ground, we could 
truthfully say that these disciplines are discovering new missions and new 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Toward a More Nuanced Conversation on Methodology  • 31
ways to understand culture” (n.p.). But finding the right way to frame and 
conduct scientific research in the humanities is a difficult matter, as illus-
trated by the controversial work of Jonathan Gottschall. In what follows, I 
dedicate substantial attention to his work, since he presents a persuasive 
argument for a scientific approach to the humanities. But his work also 
provides a cautionary tale on the limitations of such approach. His atten-
tion is focused on literature but his arguments can be extended to other 
areas of the humanities.
Gottschall (2008) argues that the work of literary scholars is only of 
interest to themselves and that literary scholarship has remained irrele-
vant to the world at large. Whereas scientists might sometimes try to write 
for general audiences, this is rarely the case in literary studies. In part, 
this is because of the very way in which literary scholarship is construed. 
For Gottschall, the main problem is that literary scholars have failed to 
produce knowledge that stands the test of time as they “rarely succeed in 
accumulating more reliable and durable knowledge” (7). He identifies 
several reasons for this: a belief in masters who never proved anything, 
a circularity of theory- proof, and stasis in academic departments. He 
also identifies a strong ideological bias, as scholars remain convinced, 
a priori, of the need to fight for just causes and identify them at work in 
literature. He calls this last attitude the “liberationist paradigm” where 
scholars think they can set people free of the workings of power through 
the sheer force of literary criticism. As an antidote, Gottschall suggests 
that scholars should aim to narrow the space of possible explanation. This 
requires a change of theory, attitude, and method, as it would imply that 
literary scholars should submit their ideas to the test of others. Consider 
the following argument for embracing a scientific approach when study-
ing literature:
[T]he praxis of the liberationist era has amounted to endlessly asking 
questions while despairing of more valid answers [ . . . ] The quintes-
sence of the dominant paradigm is, then, constitutional and reflexive 
pessimism about the ability of humans to really know anything [ . . . ] 
But long before Derrida’s neon declarations, the scientists had been 
pretty comfortable with the idea that it is not possible for humans to 
know the truth of something in the sense of its ultimate reality [ . . . ] 
Popper’s concept of falsifiability, which has been a guiding philosoph-
ical principle of scientific investigation for more than a half century, 
is an attempt to grapple with the fact that it is not logically possible 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
32 •  TheaTer as DaTa
to prove any scientific claim by experiment. But science’s response to 
this realization was more reasonable and productive than that of the 
“great generation” of liberationist theorists [ . . . ] We can’t know for 
certain what is true. Science makes no ultimate claims. But through a 
gradual process of rational thinking and falsifying tests, communities 
of scientists can show where the preponderance of evidence lies. This 
is the best that humans can do, and this is no small thing. (Gottschall 
2008, 11)
Like him, I am also sometimes deeply dissatisfied by some aspects of lit-
erary scholarship, but I think that his analysis lacks nuance. While there 
are excesses in the “liberationist paradigm,” there is also value in much 
literary criticism. The most important feature of good literary analysis 
is concerted attention to context and the careful exploration of concepts 
fraught with disagreement. Extrapolating this to theater, I don’t think our 
objective will ever be to limit the possible space of explanation of what per-
formance is. But Gottschall’s insights are useful within limited aspects of 
scholarship. I do think that shrinking the possible space of explanation 
is useful and necessary within certain narrow inquiries, such as those that 
fulfill the criteria for data- driven research (see introduction). To better see 
the pitfalls of Gottschall’s approach, let’s consider the narrower case he 
makes for the application of scientific methods to literature. He advocates 
using evolutionary psychology (EP) to study literature. A substantial part 
of his research aims to show that certain narrative features are indicative 
of cognitive traits that were well suited to the evolution of early humans. 
To his credit, he is not merely applying EP concepts to literary analysis, 
but actually conducting scientific research with several collaborators. One 
problem of this approach, though, is that too many factors have shaped 
human evolution and it is impossible to accurately measure the impact 
of most of them. Scientists have robust and reliable genetic and climato-
logical data which might explain certain aspects of human evolution. But 
we don’t have sufficient data on the early uses of narrative. As Cameron 
(2011) points out in a response to Gottschall, EP research is often faulted 
for its use of circular logic: “why is this trait so common among humans 
now? Because it was adaptive for their ancestors. Why do we think it was 
adaptive? Because it is so common now’” (62).
I think that Gottschall’s larger epistemological argument is correct and 
directly relevant to data-d riven research. But this approach only works on 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Toward a More Nuanced Conversation on Methodology  • 33
narrowly focused questions, which can be answered in terms of the data 
we have. I dream of a world where we could explain the role art has played 
in evolution, or the ways in which it can make societies better. But these 
questions are very difficult to answer given the paucity of evidence and 
the complex interactions of the millions of factors that shape societies. A 
less satisfying, but more feasible approach, is to focus on small questions. 
Taking a cue from Gottschall’s vocabulary, I suggest that data-d riven 
methodologies should also aim to narrow the space of possible questions, not 
only the space of possible explanations. The things I’ve learned from my 
data-d riven research are very limited in scope; for example, I found that 
the frequency of the word “audience” steadily decreased in the writings 
of a group of theater reviewers over a period of twenty years in Singapore, 
and that characters of Indian and Javanese origin are interconnected in 
vastly different ways in Javanese wayang kulit scene structures. These might 
seem as small conclusions, but they can be verified and disproved by other 
researchers. I wish to one day make more substantial discoveries on Indo-
nesian and Singaporean theater, but the steps towards larger discoveries 
are incremental, small questions tightly wrapped around available data.
But leaving that aside, whether Gottschall’s EP papers (or whether my 
own research) constitute solid science or not has no bearing on the valid-
ity of Gottschall’s defense of science as a valid approach to study culture. 
Faulty research examples can’t be used to condemn an approach. One 
could also find examples of superficial postcolonial perspectives and this 
would not in itself invalidate the whole enterprise of postcolonial stud-
ies. Gottschall’s description of the scientific method and what it might do 
for the study of literature is sound and enticing, and relevant to the data- 
driven study of theater. Another line of critique, also espoused by Cam-
eron, is that Gottschall’s scientific analysis stands in fundamental opposi-
tion to literary study:
[C]ritics do not ‘make progress’ towards the ‘true’ interpretation of a 
Keats sonnet; they merely offer different readings. Unlike a new scien-
tific theorem, moreover, a new reading of a text does not automatically 
supersede all previous interpretations. Rather it is of interest to the 
extent that it reveals additional meanings in a text, proposes hitherto 
unnoticed connections between texts, or foregrounds themes in texts 
which resonate with the concerns of a particular moment. (Cameron 
2011, 67)
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
34 •  TheaTer as DaTa
This is a very different kind of argument. The issue at stake here is not 
whether the scientific analysis of literature is possible, or whether Gott-
schall’s own research conforms to the scientific standards he invokes. The 
question here is the purpose of literary study: is it about finding patterns 
in the production of literature or is it about generating situated readings? I 
think that the field is big enough to accommodate both scholarly criticism 
and a science of literature. The same argument can be extended to the-
ater studies. Our discipline aims to enable situated conversations around 
the meaning of theater. But it is also a concerted effort to find verifiable 
patterns in the history and current practice of theater performances. And 
both projects benefit from the study of theater’s digital traces. Data can be 
used to find replicable insights as well as to assist in interpretive, situated 
responses to performances.
Perhaps Gottschall’s mistake is that he militates against all nonscien-
tific perspectives in the study of culture and suggests that only science can 
prove useful. But this is not the only approach, not even when working 
with data. As Ramsay (2007) says: “We would do better to recognize that 
a scientific literary criticism would cease to be criticism.” When analyz-
ing the work of Virginia Woolf, Ramsay says that critics are not trying to 
solve Woolf: “they are trying to ensure that discussion of The Waves con-
tinues into further and further reaches of intellectual depth.” Likewise, 
Rockwell and Sinclair (2016) have created tools for textual analytics which 
they describe as “interactive interpretive toys” (69), rather than as “micro-
scopes revealing the inner structure” (103) of literature. Their goal is 
not to identify objectively verifiable patterns, but to “add to a history of 
interpretation” (103). Some research questions benefit from replicable 
approaches and others do not.
On the topic of replicability, I would like to also address the criticism 
against computational literary research brought forth by Nan Z. Da (2019) 
in a widely circulated— and widely controversial—p aper. Da tried to repli-
cate some DH papers and couldn’t do so. Her analysis on topic modeling 
is particularly sharp. She shows that this procedure might be extremely 
sensitive to the parameters chosen: “When I randomly removed just 1 per-
cent of the original sample, all the topics changed” (2019, 628). We cer-
tainly need more papers that seriously probe the foundations of compu-
tational research and that alert us to methodological blind spots such as 
this. There are different methods we can use to improve our conclusions 
and make them more robust to the problem of parametrization (a point 
I revisit in chapter 2). There are many useful things in Da’s careful atten-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Toward a More Nuanced Conversation on Methodology  • 35
tion to computational errors. The problem is that she then moves on to 
say that such mistakes invalidate the computational study of literature as a 
whole. As Jannidis (2019) notes in his response to her paper, “there is no 
logical way to move from an error in a calculation of a researcher to a gen-
eral statement about the fruitfulness of a research field in general” (n.p.). 
Da’s paper insists that, even if the errors in the calculations were solved, 
computation has nothing to add to the study of literature. This is similar 
to Gottschall’s argument, but in reverse. While Gottschall suggests that 
only a scientific study of culture is of interest, Da claims that science can 
add nothing to the field.
The arguments presented by Da and Gottschall are controversial for 
their totalizing conclusions. In both cases, though, we would be wise to 
recognize the very valid points they are making. Gottschall provides a vivid 
exhortation of what science can do, and Da’s paper should alert us to the 
importance of replication.
Critical Realism: Reconciling Different Epistemologies
We have seen that the natural sciences and the humanities have histori-
cally differed in their epistemic objectives. Data- driven and data- assisted 
methodologies are reconfiguring these faultlines. This has many implica-
tions for how we carry out research projects and also for how we might 
rethink teaching and evaluation practices. Speaking mostly about dif-
ferences between contrasting approaches, as I have done so far in this 
chapter, is useful to orient ourselves in this changing landscape. But it 
has the unfortunate consequence of emphasizing divergence and opposi-
tion. Both interpretive and scientific approaches say things about reality 
in ways that seem at odds with each other. But believing that these two 
modes are fundamentally irreconcilable only makes sense in a zero- sum 
epistemology where we need to choose exclusive allegiances, and where 
we need to be either scientists or humanists, either realists or constructiv-
ists. Such opposition might be encouraged by disciplinary boundaries in 
academic institutions. In the humanities, we are usually taught to be sus-
picious of anything that seems remotely positivistic. In the sciences, we 
are taught to only trust that which can be measured and verified. But per-
haps we can imagine a different kind of engagement with reality, where 
we recognize that some things can be measured whereas others cannot. 
Some areas benefit from systematic experimentation and rigorous repli-
cation practices, while others required the textured explorations of con-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
36 •  TheaTer as DaTa
tested terms that reflect subject- dependent experiences of the world. But 
these different components are sometimes part of the same phenomenon 
and keeping them separate reduces our understanding of complex issues.
Take, for example, climate change. The analysis of climatological pat-
terns provides a good example of an area where a scientific attitude is 
necessary. Consistent, verifiable data has been central to one of the most 
consequential discoveries of our time. As scientists have learned from ana-
lyzing millions of data points, world temperatures are rising at alarming 
rates, and this rise is almost certainly due to human action (Bernstein et 
al. 2008). Reliable data will continue to be indispensable for tracking the 
impact of collective action to combat climate change. But climate change 
will also impact lives in ways that might not always be measurable. Imagi-
native and situated perspectives will be crucial if we want to understand 
how different communities make sense of climate challenges, and if we 
aim to devise creative and fair solutions to these problems.
The biggest challenges of our times will require us to reconcile a belief 
in solid scientific evidence with the recognition that scientific methods 
have been used to oppress and misrepresent people, and that they cannot 
answer every single question. We will need to accurately gauge the kinds 
of questions that science can answer, while complementing them with 
situated and interpretive modes of learning about the world. These are tall 
orders, and this book aims to make a small contribution to a minuscule 
part of the problem: to elucidate what computational research can or can’t 
do for the study of theater.
In order to combine scientific and interpretive approaches, we can 
take a cue from critical realism (henceforth CR), a stance that rejects the 
extreme positions of both positivism and constructivism. The writings of 
Roy Bhaskar (1944– 2014) are associated with CR, but here I refer mostly 
to later formulations, found in the work of Dave Elder- Vass, Ruth Groff, 
Paul Edwards, Frédéric Vandenberghe, and Berth Danermark. Rather than 
a theory or method, CR is a metatheoretical perspective. It is a relatively 
recent perspective and there are differences in the way CR is articulated by 
different theorists— what follows is a select summary of key ideas rather 
than a comprehensive overview.
From a CR perspective, the problem of positivism is that it denies the 
role that power, history, subjectivity, and discourse play in the construc-
tion of knowledge. Researchers in CR agree with constructivists on “the 
political nature of science and are equally skeptical of its truth claims, 
many of which simply represent the current orthodoxy within scientific 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Toward a More Nuanced Conversation on Methodology  • 37
communities” (O’Mahoney and Vincent 2014, 5). In turn, the problem of 
constructivism is that it must reject any claims that the natural or social 
sciences provide a “better” understanding of the world. As long as some-
one holds a particular belief, then we must accept that is “their reality” and 
there is no way to ascertain that other theories might have better explana-
tory power. This conclusion is unreasonable in the natural sciences given 
the demonstrable advances of applied science. But it should also be unten-
able for the study of the social world:
We would hope discourses that girls are naturally bad at science, that 
Western cultures are superior to others [ . . . ] would not be accepted 
solely on the basis that some groups believe these statements to be 
true. (O’Mahoney and Vincent 2014, 6)
Following Bhaskar, critical realists suggest that both constructivism and 
positivism make similar mistakes, as their views are premised on an 
“epistemic fallacy” which conflates ontology (that which exists) and epis-
temology (that which we can know). Positivists reduce ontology to epis-
temology by claiming that only that which is observable exists. For con-
structivists, there is no ontology outside of epistemology, as the world 
only exists through our discourses about it.
In contrast to both perspectives, CR suggest that there is an external 
reality, but that our knowledge of that reality is necessarily subjective, 
contingent, and socially constructed. For CR researchers, this key dis-
tinction between ontology and epistemology leads to a “depth ontology” 
that distinguishes between three realms: the empirical (that which can be 
observed), the actual (that which exists), and the real (the causal mecha-
nisms of reality) (O’Mahoney and Vincent 2014, 9). This means that a veri-
fiable external reality exists, but we can only study it through uncertain 
processes, acknowledging that “different actors will define reality in dif-
ferent ways” (Edwards, Vincent, and O’Mahoney 2014, 321). Thus, CR- 
infused research aims to identify the causal mechanisms present in reality, 
while acknowledging that understandings of the world are socially con-
structed and subjectively experienced.
A CR perspective has clear implications for research; CR- infused 
research must be context- sensitive, as opposed to positivism. But it 
should also pursue the understanding of an objective reality, as opposed 
to constructivism; as a result, CR is committed to both truth and to thick 
explanations (which take politics, history, and subjective experiences into 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
38 •  TheaTer as DaTa
account). As Vandenberghe (2013) notes, CR “enters the ‘science wars’ 
by fighting two fronts” (5). For CR researchers, “useful research is nec-
essarily rich, ‘thick,’ and explanatory as opposed to the ‘thin’ descrip-
tive approaches that positivism necessitates” (O’Mahoney and Vincent 
2014, 4). But it also aims to pursue better explanations of an observer- 
independent reality.
What does this mean for computational theater research? I have 
argued that data- driven methodologies are closer to the sciences and data- 
assisted methodologies are closer to conventional humanities approaches. 
Through CR, we are invited to move beyond this dichotomy; to do science 
without abandoning the presuppositions of the humanities, and to study 
the humanities without abandoning the premises of science.
A humanities-a ware science remains critical about the “facts” on 
which its conclusions rest. We are led by CR to analyze the influence of 
history and power on the construction of facts and categories:
Facts don’t speak for themselves [ . . . ], they are always categorized 
and schematized by one or another theory, philosophy or cosmology 
that is socio-h istorically determined, there is no observation that is not 
an interpretation and no interpretation that does not involve an imagi-
nary representation of reality. (Vandenberghe 2013, 5)
Conversely, scientifically inclined humanities research aims to place 
historically and subjectively situated observations against a larger back-
drop. This means piecing together explanations from different perspec-
tives to elucidate larger, intersubjectively verifiable patterns. Consider, 
for example, what a CR- influenced ethnography would look like. From 
a CR perspective, ethnography can “provide a deeper understanding 
than subjectivism is capable of, one which is able to link the subjective 
understandings of individuals with the structural positions within which 
those individuals are located” (Rees and Gatenby 2014, 135). Extending 
this approach to data- assisted research means that deformances can still be 
understood as indicative of shared, objective structures of culture. How-
ever, this is a point with which many DH theorists would disagree, as 
some people carry out data- assisted research from a purely constructiv-
ist perspective. This is not something I want to disavow. The point I am 
making here is that it is possible to engage in CR-i nspired, data- assisted 
research, in the same way that it is feasible to carry out CR- inspired eth-
nography, without wholly embracing positivism and without fully eschew-
ing constructivism.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Toward a More Nuanced Conversation on Methodology  • 39
A possible criticism of CR is that it is only useful to study the natural 
world, and that there is no objective component in culture. To this, Elder- 
Vass (2012) responds that a purely subjectivist account of culture would 
be incoherent as “it would lack the means to explain how culture can 
acquire the shared quality that makes it culture” (39). Objective culture is 
a product of human agency, but nevertheless “exerts a causal influence of 
its own” once it is produced (41). Culture and beliefs have demonstrable 
agency and create an impact on an observer-i ndependent reality: “the 
tooth fairy is not materially, but ideationally real: the discourse about it 
has real effects, for example, on the bedtime activities of children, even 
though it does not exist” (O’Mahoney and Vincent 2014, 7).
To bring these ideas back to theater, let’s consider the ontological sta-
tus of performances. Performances belong to the realm of actuality, but 
any attempt at grasping their meaning or form is mediated by subjective 
experiences and socially constructed categories. Still, these categories 
have consequences. Let’s imagine an event that is classified as an “experi-
mental mixed-m edia performance.” Each of these terms has contested 
meanings and many people might not agree with how well the label fits 
the specific event. However, if a company uses this label to describe their 
work, this will impact the way it is marketed and funded, and will likely 
change the composition of the audience that will come to see it. We can 
use data-d riven approaches to understand how often this label is used, 
and data- assisted approaches to explore the limitations of this label.
Besides CR, there are other recent perspectives that challenge a binary 
distinction between realism and constructivism. Paul Rae (2018) offers a 
comprehensive survey of “new realisms” and what they mean for under-
standing the fraught connections between theater and reality (7–8 ). Per-
haps other new realisms, as well as CR, can provide a foundation for car-
rying out data-d riven and data-a ssisted research that has learned from 
both hermeneutics and science. Data- driven research should be aware 
of how politics, discourse, and subjectivity shape its methods and con-
clusions. Likewise, data-a ssisted research could contribute to the under-
standing of objective aspects of theater. Both perspectives could comple-
ment each other. One can use a data-a ssisted performative visualization 
to explore how power structures shape the categories of a data-d riven 
project. Conversely, a data- driven analysis could uncover patterns within 
a data-a ssisted project.
But this does not mean that both approaches must always work in tan-
dem. A data- driven project might address its limitations by reflecting on 
the impact of history and power structures in the way the data was defined, 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
40 •  TheaTer as DaTa
collected, and processed. A data-a ssisted project can likewise indicate its 
contribution to the study of shared objective reality. Whether a research 
project requires primarily a data-d riven perspective, a data-a ssisted one, 
or a combination of both depends entirely on the questions it seeks to 
explore. Data- assisted perspectives poke around the edges of a question, 
and data-a ssisted perspectives aim to find the best possible answer to a 
question under formal constraints (see the different criteria outlined in 
the introduction). Projects might also exist in a continuum. A project can 
be said to be driven by data inasmuch as it emphasizes replicability, and 
assisted by data to the extent it draws attention to situated interpretations. 
If we use the foundations of CR to describe the epistemological quests of 
computational theater research, we will be better able to justify why a par-
ticular approach works for a specific question, and how different kinds of 
questions can contribute to a fuller understanding of theater’s history and 
current practice.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
ChapTer 2
The Roles of Statistics
In the majority of the projects reported in this book, at least some of the 
data is quantitative. For example, projects might report the number of 
words in a text, the number of partners in a collaboration network, the 
angular velocity of a dance movement, and the number of performances 
in a given place. But these quantities can be put to very different uses, as 
they can aid in both data- driven and data-a ssisted research. In data- driven 
research, numbers are used to find empirical patterns and convince others 
of the best possible description or explanation of a phenomenon. In data- 
assisted methodologies, numbers are deployed as deformances, in order to 
challenge the assumptions of a question and generate multiple interpreta-
tions that do not supersede each other as more correct or more fitting.
Often the same numerical methods can be used within very differ-
ent epistemological frameworks. Term frequency, inverse document 
frequency (Tf- idf ) is a statistic often used in text mining, and it could be 
adopted for the classification of documents in a textual collection. How-
ever, the same method could also be used for interpretive readings. In his 
study of Woolf ’s The Waves, Ramsay (2007, n.p.) uses tf-i df to give a sense 
of how certain characters’ lexical patterns differ from each other’s. As 
Ramsay notes, this statistic does not aim “to bring the results into closer 
conformity with ‘reality,’ but merely to render the weighting numbers 
more sensible to the analyst.” Thus, quantitative transformations may 
enable new interpretive readings, and they can be enlisted as important 
allies in data- assisted research.
As Underwood (2019a) notes: “Quantitative models are no more objec-
tive than any other historical interpretation; they are just another way to 
grapple with the mystery of the human past, which doesn’t become less 
complex or less perplexing as we back up to take a wider view” (xix). Piper 
(2018) echoes this view: “the literate and the numerate are not agons 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research. 41
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
42 •  TheaTer as DaTa
engaged in a duel. They are two integral components of a more holistic 
understanding of human mentality” (5).
Numbers privilege some modes of representation over others. But the 
same is true for language, and scholars are well aware of this. As Kath-
erine Bode (2012) notes, instead of abandoning language “scholars have 
sought to understand the ways it works and to challenge and critique the 
relations of power it perpetuates.” The same needs to be done with num-
bers, as we seek “to recognise them as a form of representation and, as 
such, to explore how they operate and the ways in which numbers accrue 
authenticity and authority” (12).
In this chapter, I answer Bode’s call to pay attention to the way num-
bers operate as I explore the ways they can be mobilized for data-a ssisted 
and data-d riven research. I describe their potential for bias and insight by 
focusing on statistical methods, which are at the core of the projects I will 
describe later in the book. This chapter looks at a conventional distinc-
tion made in statistics, between descriptive and explanatory data analy-
sis. But what follows is not a series of textbook descriptions. Rather, I 
offer an epistemological reflection on what these two types of statistical 
approaches reveal and obscure, and what they entail for computational 
theater research. I pay specific attention to the potential for bias in each, 
and the steps that can be taken to mitigate bias.
Within descriptive analysis, I refer to procedures that outline the shape 
of the data or which seek to identify outliers and patterns such as trends, 
correlations, similarity, and clusters. Within explanatory analysis, which are 
often called inferential statistics, I refer to methods that aim to explain 
causality in the phenomena under consideration. Descriptive analysis can 
be used in both data-d riven and data- assisted projects. Explanatory sta-
tistical analysis, in contrast, belongs solely to the realm of data- driven 
research.
Descriptive Analysis
Some descriptive methods aim to characterize the general shape of the 
data. This includes calculating measures of central tendency (like the 
arithmetic mean) and measures of dispersion (such as the standard devia-
tion), as well as conveying the distribution of the data (often via graphical 
means such as histograms). Descriptive analysis can also include ways of 
portraying the differences between groups (effect size) and estimating the 
relationship between variables (correlation).
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Roles of Statistics • 43
These operations are often grouped together under exploratory data 
analysis (EDA), a concept first proposed by John Tukey (1977). Rather 
than a specific set of methods, EDA is an agnostic approach to data analy-
sis that aims to identify the shape of the data without imposing assump-
tions about what it must be like. The NIST/SEMATECH e- Handbook of 
Statistical Methods (2003, n.p.) describes EDA as “an attitude/philosophy 
about how a data analysis should be carried out” which is used to achieve 
the following objectives: to uncover the underlying structure of a dataset, 
extract important variables, detect outliers and anomalies, test underlying 
assumptions, develop parsimonious models, and determine optimal fac-
tor settings for further analysis.
Statistician Allen Downey describes a typical data analysis routine as 
consisting of the following steps: (1) importing and cleaning the data, (2) 
single variable explorations such as distributions and summary statistics, 
(3) pair-w ise explorations for possible relations between variables (cor-
relations, linear fits), and (4) multivariate analysis, such as regression 
and control variables for more complex relationships. These exploratory 
procedures are then followed by inferential analysis for estimation and 
hypothesis testing (see the second part of this chapter). Visualizations are 
used throughout the entire process to aid in the analysis and to communi-
cate the outcome of the research. Histograms, boxplots, and violinplots 
are particularly useful for identifying the shape of the distribution (I con-
sider the role of visualizations in more detail in chapter 3).
According to Arnold and Tilton (2019), EDA permeates much of the 
digital humanities, even though it is not often named; it is an iterative, 
agnostic way of discovering the properties of the data. But the fact that 
it is agnostic in a technical sense (i.e., that it doesn’t impose a model a 
priori) doesn’t mean it will necessarily present an unbiased representa-
tion of the data: EDA is only as good as the data that is analyzed. And if 
there are biases in the way the data was collected, EDA will not necessar-
ily reveal this. As Bode (2020) notes, no literary dataset is unbiased. The 
same is true for any theater dataset, or any dataset in the humanities for 
that matter. Bode is particularly concerned with the underrepresentation 
of women writers, which is often the result of systemic bias in many digi-
tal collections. Concerning word usage in a literary collection, EDA can 
offer an agnostic representation, but the systemic imbalances in the data 
collection can’t be removed just by using EDA. Researchers will need to 
critically interrogate the ways the data was collected and pay attention to 
the history of the dataset. If undertaken from the perspective of critical 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
44 •  TheaTer as DaTa
realism, EDA will need to be accompanied by a thorough analysis of the 
cultural and historical conditions under which the data was created.
Besides the problem described by Bode, which we could refer to as 
sampling bias, other features might also make descriptive data analysis 
less objective than it seems. Often, researchers don’t have a direct way of 
measuring a phenomenon they are interested in, and turn to proxies. In 
some cases, the proxy might not be directly related to the phenomenon of 
interest and this can lead to proxy bias. In other words, the availability of 
seemingly neutral or complete datasets proves too tempting for research-
ers, and we might not always question the assumptions that went into the 
creation of the dataset.
Take for example a study of the evolution of novelty in films that uses 
IMDB plot keywords to measure innovation in films (Sreenivasan 2013) over 
the twentieth century. The author uses sound mathematical techniques 
to show a correlation between novelty and revenue: blockbusters tend to 
occupy the midsection of an innovation curve, with films that are either 
too innovative or not innovative enough trailing off at both ends of the 
curve. The main problem with this research project is the assumption 
that the number and variety of tags will indicate novelty. As any film buff 
knows, remakes will often be tagged with completely different tags than 
their original sources, and newer remakes will often have tags that are 
more varied and larger in number than the original films. For example, 
at the time of writing this book, Abre los ojos (dir. Amenábar 1997) has 140 
tags, while the Hollywood remake Vanilla Sky (dir. Crowe 2001) has 176. 
The original Ringu (dir. Nakata 1999) has 154 keywords, while its remake 
The Ring (dir. Verbinski 2002) has 189; the original Ghost in the Shell (dir. 
Oshii 1996) has 222 keywords, and its remake (dir. Sanders 2017) has 
332. It seems unlikely that these newer films are more innovative; perhaps 
the data just shows that newer films will tend to be tagged in more idio-
syncratic ways (or that Hollywood productions are given more tags than 
films produced elsewhere). In other words, films have just become more 
tagged, with newer films being tagged in more granular ways. Likewise, 
when modeling theater as data, it is worth pausing to consider the extent 
to which available data directly measures the phenomenon of interest and 
the extent to which it is a convenient, but misleading proxy.
Besides EDA, other statistical methods can be used to identify pat-
terns in the data, such as clusters and trends. The majority of the proj-
ects I describe in this book—a nd perhaps the majority of DH work—a ims 
at identifying similarity (clusters) and change (trends). For example, in 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Roles of Statistics • 45
chapter 4, I describe a project that identified changes in the words used 
by theater critics over a two-d ecade period. In chapter 6, I show ways to 
compare networks and track their changes over time. And in chapter 7, I 
look at clusters of performances in a given place, and on how that changes 
over time. These projects use methods from computational linguistics, 
network analysis, and geostatistics, as well as some general data min-
ing methods for standardization and dimensionality reduction (reducing 
the number of variables in a large dataset). Elsewhere in DH, research-
ers sometimes use machine learning (ML) techniques rather than classical 
statistics to find clusters and trends. ML researchers distinguish between 
clusters and categories. Generally speaking, clusters are latent structures 
in a dataset discovered through unsupervised ML algorithms. In contrast, 
supervised ML approaches aim to assign objects to predefined categories, 
which are generally determined by human annotators. This book focuses 
on classical statistics rather than ML, and in what follows I refer to clus-
ters as groups identified by classical statistical procedures, rather than by 
ML algorithms.
Clusters and trends are also descriptions of the data, but unlike EDA 
they require us to make assumptions about a dataset, and select param-
eters for what constitutes a trend or a cluster. A cluster is not an inherent 
property of the data as opposed to, say, the arithmetic mean. You can find 
the arithmetic mean of a dataset in an uncontroversial way without mak-
ing any assumptions about what constitutes a meaningful relation between 
items. But a cluster requires a model of what is meaningful. As O’Neil 
(2016) notes, “models are opinions embedded in mathematics” (21).
When identifying trends and clusters, any method is susceptible to 
bias. In some cases, there are technical solutions and certain methods 
can be validated as more useful for given applications. But, as in the case 
of EDA, sometimes bias is ingrained in the way the data was collected. 
This is a particularly important problem for the identification of histori-
cal trends. Many DH projects aim at finding changes over time. But any 
attempt at analyzing historical data should consider chronology bias. This 
type of bias is more often discussed in the medical literature and refers 
to the errors in judgment one makes when comparing evidence from dif-
ferent historical periods (Feinstein 1971, 870). Medical researchers study-
ing the incidence of a disease might conclude that a disease has become 
more prevalent over time, but the reason for an increase in reported cases 
might be that diagnostic methods improved. This is sometimes treated as 
a special kind of confounding bias, where an observed correlation is caused 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
46 •  TheaTer as DaTa
by an agent not directly measured (in this case, this agent would be the 
improvement in diagnostic methods). A close parallel to this situation is 
common in DH. When using historical digitized texts, some words might 
appear to be more common today than in the past, but this might be the 
result of better optical character recognition (OCR) techniques for more 
recent texts. For theater, data might be more readily available for recent 
performances than for those in the past, and this might distort the percep-
tion of trends.
Another problem of historical analysis is that the meanings of con-
cepts might change. An excellent example of this is offered by Rosenberg 
(2013) in his analysis of how the word “data” entered the English lexicon. 
In this study, he compared two different pattern analyses of the usage of 
the word “data,” which had been previously published online. He then 
closely read several early examples that formed the basis of such projects. 
His conclusion is that the patterns available in the published online pieces 
were misleading, since the relative usage of the word “data” needs to take 
a Latinized understanding of data common in the seventeenth century 
into account, and distinguish it from its modern English sense. Historical 
patterns offer tantalizing explanations for cultural change, but it is crucial 
that the data is examined in detail.
To find meaningful clusters in the data, one must acknowledge a per-
vasive problem: clusters will always emerge. This will be true regardless of 
the clustering method: there will always be some pattern to analyze. This 
is a mathematical necessity, a corollary of the Ramsey theory that states 
that a pattern is guaranteed to emerge given enough elements in a set (R. 
L. Graham, Rothschild, and Spencer 1990). Thus, researchers must resist 
the temptation to overinterpret the meaning of a cluster. One will always 
be able to find meaning in any clustering of the data, the way one can 
always find shapes in tea leaves. This is sometimes called clustering illusion 
(Bedek et al. 2018), and in the scientific literature it is often linked to hark-
ing, or hypothesizing after the results are known (Kerr 1998).
For example, imagine a clustering method that shows that the words 
“chair” and “power” are part of the same cluster in a corpus of theater 
reviews. Based on this pattern, we might hypothesize that chairs are seen 
by the theater critics as symbols of power. But if we had chosen another 
clustering mechanism, it is entirely possible that “chair” would form a 
cluster with “contemporary,” and this will send us down an entirely dif-
ferent rabbit hole of possible associations and interpretations. We might 
assume that the first clustering method reveals semantic associations 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Roles of Statistics • 47
between words. A few examples might convince us this is true. But we 
would need to independently calibrate such a method first. We would need 
to run tests on other datasets and confirm whether the method tends to 
reveal words that have semantic associations or not, otherwise the risk of 
deluding ourselves is very high. Alternatively, we could randomly select a 
subsample of sentences from the original dataset and closely inspect them 
to determine the extent to which our initial intuition about the results is 
true. We can also estimate the likelihood that the patterns that we see are 
due to chance. For this, probability values or p-v alues are often used. A 
p- value can be used to estimate the likelihood that the observed pattern is 
due to chance. This is different from the usage of p- values for hypothesis 
testing, which I describe in the next section of this chapter.
As noted in previous chapters, data-d riven research aims at offering 
measured uncertainty. For this reason, it is important that some measure 
of uncertainty is offered. Instead or besides p-v alues, confidence intervals, 
effect sizes, and measures of accuracy (such as the F-s core) can also be 
used (see the excursion in chapter 5 for an example of effect size report-
ing). But it is important to note that, in order to estimate the likelihood 
that a pattern is not random, we need a model that can tell us what would 
happen in this situation under random conditions. This means that we 
need to impose assumptions about what we would see in random situa-
tions and we need to critically inspect those assumptions. For example, 
if one assumes that the data is normally distributed, one would expect 
certain things to appear under random conditions. These would be very 
different if one were to assume an exponential distribution. The general 
audience books and technical papers by Nassim Taleb are good introduc-
tions to these problem (Taleb 2007; Cirillo and Taleb 2016).
Everything we choose to treat as data, and every procedure we per-
form on that data, is the result of specific perspectives on what is wor-
thy of attention. This means that working with data can never be a fully 
objective endeavor. When confronted with this impossibility, we could 
throw our hands in the air and exclaim in despair, “all data approaches 
to the humanities are futile!” But we can also implement a more mea-
sured approach, guided by the principles of critical realism (see chapter 
1). The fact that full objectivity is impossible doesn’t mean we can’t make 
progress towards consensus. If changing the definition of what counts as 
data in a given project leads to entirely different answers, then the onus 
is on the researchers to investigate what are the limitations of the data 
collection procedures used. If changing the settings of a given analytical 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
48 •  TheaTer as DaTa
method yields different results, we need to figure out which are the best 
parameters. Finding foolproof configurations may well be impossible, 
but some parameters are better than others, and we can—a nd should— 
aim for improvement. This is why we need independent methods to verify 
data- driven results and to calibrate our data parameters. If our methods 
and data suggest the existence of a given trend, then we should find other 
approaches that might confirm or complicate these results. This can be 
achieved by fomenting a culture of replication where the goal is to collec-
tively improve our methods rather than to dismiss someone else’s endeav-
ors. Calibration methods, which use alternative approaches to model 
the same phenomena, or to manually verify a subset of computational 
results, should be central to our data research projects. We must also 
develop research environments that encourage the reporting of negative 
findings. Many computational projects will yield no conclusive evidence. 
Rather than overinterpreting vague patterns in the data, we need to be able 
to communicate research that fails to prove an initial hypothesis (this is 
something I attempt in the second excursion of chapter 6).
This short overview has described some potential sources of bias, but 
there are many more types of bias that might influence a computational 
result, and being on the lookout for those is part and parcel of computa-
tional research. However, seeking the best parameters and encouraging 
replication is not the only way in which we can proceed. In data-a ssisted 
research, patterns can be understood as generative devices. The “chair 
power” cluster in the example above might be used as a defamiliarization 
strategy to offer a fresh perspective on the corpus of theater reviews. Then, 
it doesn’t matter if the pattern used as a starting point is “accurate,” but 
the onus is on the researcher to use this defamiliarization to say some-
thing meaningful about the theater reviews under consideration.
Explanatory Analysis
The statistical methods surveyed thus far aim at uncovering structures and 
patterns in the data, and estimating the likelihood that they are not due 
to chance. But this is different from the explanatory statistical techniques 
that aim to test a hypothesis to establish causality. The gold standard, in 
this other scenario, is randomized controlled trials (RCTs). A common 
example is drug discovery, where a drug is administered to a group of 
individuals (the experimental group) and a placebo is given to another 
group (the control group). Cohort sorting should be randomized (i.e., any 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Roles of Statistics • 49
participating individual has equal likelihood of being assigned to a given 
group) to try to minimize confounding factors (some underlying trait that 
would make an individual more likely to be assigned to a given group). 
This procedure was popularized in the early twentieth century when it was 
introduced to precision agriculture by Jerzy Nyman in 1923 (a translation 
of the original article is available in Splawa- Neyman 1990), but it is now 
routinely applied to all areas of science.
A common incarnation of RCTs is A|B testing, which is used in data- 
driven web design. Companies, such as Facebook, make design decisions 
based on experiments. Users around the world are randomly sorted into 
two groups, and they are exposed to almost identical versions of their 
products, which only differ by a single design element (perhaps a particu-
lar shade of blue in a button). Then a specific outcome is measured (per-
haps the number of clicks on an advertisement), and if statistical analysis 
reveals a strategic advantage in one version of the design feature, it is then 
integrated into the product. Most of Facebook’s interface design is based 
on such tests, and the company reportedly runs over a thousand such tri-
als per day (Bakshy, Eckles, and Bernstein 2014).
However, RCTs are often impossible (or unethical) to carry out in the 
humanities and the social sciences. An alternative approach is identify-
ing natural experiments, a strategy more common in social sciences such 
as economics and quantitative sociology. Natural experiments use events 
where people where “naturally” sorted into random groups, in conditions 
which approximate an RCT. In one famous example, researchers were 
interested in evaluating the impact in a person’s life of attending an elite 
high school. In the United States, sometimes entry to elite high schools is 
determined by the grade in a standardized test. The cutoff point depends 
on the number of applicants in a year and is relatively arbitrary. Thus, it 
constitutes a good proxy for the random assignment of similar individuals 
into two groups. Researchers took two groups of people that had attained 
almost the same score in the test: those who barely made it and those who 
barely missed admission (differing by a single point). They assumed that 
both groups were roughly equal in terms of ability but one was admit-
ted to an elite school and one wasn’t, and this provided an ideal natural 
experiment. As the researchers found out, there was no significant differ-
ence in the academic and career achievements between groups in later life 
(Abdulkadiroğlu, Angrist, and Pathak 2014).
Whether experiments are controlled or natural, they are used in sci-
entific research to identify causal explanations for phenomena. Causality 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
50 •  TheaTer as DaTa
cannot be established outside the bounds of experimental conditions: 
researchers might be able to identify strong correlations between obser-
vations, but without experimental designs, it is impossible to assert 
underlying causes. A key tenet of the scientific method is that humans 
are not naturally good at distinguishing between causal relationships 
and spurious correlations. Take, for example, the graphs at the fabu-
lously titled website http://www.spuriouscorrelations.com. One of my 
favorite instances shows a strong correlation between cheese consump-
tion and death by sheet entanglement. This correlation is statistically 
significant, but that does not mean that cheese causes death by sheet 
entanglement (or the other way around). To establish such a causal rela-
tionship we would perhaps need to find a natural experiment (perhaps 
a region where cheese was banned) and see if it affected the number of 
fatal sheet entanglements as compared to a similar region where cheese 
was not banned.
As described above, to test drugs and medical procedures, scientists 
compare a group of animals or humans treated with a particular drug and 
one control group, where no such procedure was applied. The response 
of each individual won’t be exactly the same but it will fall within a prob-
ability distribution. Effect sizes and p-v alues will be used to estimate the 
likelihood that the drug has an effect that is not due to chance. However, 
conventional hypothesis testing is the object of intense controversies and 
it is the result of specific histories. I will consider these controversies and 
histories briefly to show that statistics is not a unified field of practice but 
an area of intense epistemological debates. These discussions, in turn, 
can help us become more attuned to the affordances and limitations of 
statistics for computational theater research.
Hypothesis testing was developed in an earlier environment of less 
intense scientific activity. In the current landscape of frantic scientific 
research and thousands of scientific studies per day, the combination of 
certain social and mathematical conditions might mean that significance 
testing is inadequate to distinguish real results from false ones. The stan-
dard name for this statistical analysis is p- value significance testing. In the 
early twentieth century, when this procedure was developed, the word 
“significant” meant that something was signaled. As the century evolved, 
the word “significant” came to mean “important.” As a result, sometimes 
a significant difference between samples seems to imply an important 
difference. However, when Fisher and others first used this term in the 
context of statistics, they merely wished to emphasize that there was an 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Roles of Statistics • 51
identifiable difference, not necessarily an important or meaningful one 
(Salsburg 2002, 41). This has led to much confusion and unfortunate 
reports, both within academic circles and in the reporting of science to 
general audiences.
As statistics gained a more prominent role in all kinds of research, 
variations of the p-v alue significance testing came to dominate many areas 
of scientific inquiry, particularly in experimental research. Most experi-
mental designs started using p-v alues to estimate the truth of a theory or 
the usefulness of a procedure. The most common variation of this method 
today is the Neyman- Pearson hypothesis testing, where a null hypothesis (no 
effect) is compared against a specific intervention (say, a drug). Although 
Neyman-P earson hypothesis testing is now introduced in almost all sta-
tistical books and is enshrined at the top of scientific practice, its rise 
through the twentieth century was not unchallenged. The antipathy from 
Fisher to Neyman is well known (Salsburg 2002, 52–6 0), and the cur-
rent primacy of a method that bears both their names would have baffled 
them. Neyman himself was not a blind proponent of simplistic hypothesis 
testing. He instead advocated using tests to distinguish between a family 
of distributions, but this more sophisticated approach did not make it into 
the statistics textbooks.
Hypothesis testing now runs rampant and is drawing criticism from 
many quarters. The abuse of the p- value has been exacerbated by the con-
vention that there is a threshold for significance. This threshold is often 
set at 0.05, where results are considered significant when the p- value is 
lower than this threshold (in other words, when there is a 1/20 probability 
of getting a given result when the effect observed is not real, a ridiculously 
high threshold). This convention was set up in the early twentieth century, 
a time of less intense scientific practice. But now it has such a firm place in 
the practice of science that research groups would often waggle their data 
into significance. Some observers have also identified a practice termed 
p- hacking, where scientists try to coerce their data into p- value significance 
(Marasini, Quatto, and Ripamonti 2016; Farcomeni 2017). According to 
these observers, researchers should recognize that a p- value without con-
text or other evidence provides limited information. For example, by itself 
a p- value higher than 0.05 offers only weak evidence that the proposed 
intervention has no actual effect (i.e., that the effect tested for, is false). 
Likewise, a relatively small p- value does not provide incontrovertible evi-
dence for the correctness of a hypothesis; many other hypotheses may 
be equally or more consistent with the observed data. For these reasons, 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
52 •  TheaTer as DaTa
data analysis should not end with the calculation of a p-v alue when other 
approaches are appropriate and feasible.
Some theorists argue that hypothesis testing as developed in the early 
twentieth century should be ditched altogether in favor of Bayesian anal-
ysis (Marasini, Quatto, and Ripamonti 2016, 320). Bayesian approaches 
(named after the seventeenth-c entury reverend Thomas Bayes), thinks of 
probabilities as ways to refine one’s ideas about the world. Bayesian sta-
tistics require analysts to formally express their baseline beliefs (called a 
priori odds) before running an experiment. Calculating these odds effec-
tively is a complex task that was all but impossible before the advent of the 
computer. Currently it is technically more feasible to run Bayesian statisti-
cal analyses, but there are several reasons why this is not yet a dominant 
trend. One is the lack of training, which is connected to the inherent dif-
ficulty of Bayesian statistical analysis. Another one is that frequentist sta-
tistics dominate journals, peer review, and industrial standards.
Beyond Description and Explanation?
This short detour into statistical history, shows that statistics is not a 
monolithic field, but a changing set of practices with many fierce and 
fascinating debates. Questions about the abuse of p- values and debates 
over Bayesian statistics are only scratching the surface. In the age of big 
data, some people have suggested that correlation is sufficient and that 
we don’t need to understand causal mechanisms. A famous example is 
the provocative piece The End of Theory: The Data Deluge Makes the Scientific 
Method Obsolete, written by Chris Anderson (2008), former editor-i n- chief 
of Wired magazine. This oft- cited essay stated that with enough data, 
the numbers speak for themselves. Anderson suggests that correlation 
supersedes causation, and science can advance without coherent mod-
els and unified theories. This point is also echoed by Mayer- Schönberger 
and Cukier’s (2013) influential Big Data: A Revolution that Will Transform 
How We Live, Work, and Think.
However, many scientists have pushed back against this view. Using 
the aforementioned Ramsey theory and other mathematical explanations, 
Calude and Longo (2017) show that most correlations in big data will be 
spurious. Consider a paper on the dangers of relying only on correlation 
for medical diagnostics (Mullainathan and Obermeyer 2017). The authors 
use, by way of example, a project that applied ML to determine if patients 
arriving at the hospital were having a stroke. Using insurance claim data, 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Roles of Statistics • 53
the ML algorithm found four unusual predictors of stroke: accidental 
injury, benign breast lump, colonoscopy, and sinusitis; in addition to 
more conventional predictors, such as cardiovascular disease. Subse-
quent qualitative analysis revealed the reason for the unusual predictors: 
they indicate that someone is a “heavy utilization patient”; in other words, 
someone who is likely to go to the hospital for relatively minor complica-
tions is more likely to have a stroke detected. Evaluating the ML results on 
their own, someone could conclude instead that sinusitis is a risk factor 
for stroke.
I here subscribe to the view that the only way to scientifically investi-
gate causality is by devising experiments. In the humanities, natural or 
controlled experiments are both extremely rare. But if we were interested 
in scientifically studying causation in theater studies, we would need to 
devise experiments. Given the large number of possible factors that 
impact theater performances, it would be very hard to conduct such exper-
iments. This doesn’t mean that we need to abandon the impulse to explain 
things. But perhaps we need to accept that a scientific explanation of theatri-
cal phenomena is a very difficult goal. A scientific approach, of course, is 
not the only road we can take to arrive at explanations. We can also use 
reflection and intuition, as theater scholars have done for a long time, to 
explain theatrical phenomena. Even when working within a data-d riven 
paradigm, it is important to recognize the extent to which our conclusions 
are grounded entirely on data and the extent to which we are extrapolating 
from the data, and speculating beyond what it actually reveals.
If we critically assess our data, and seek estimation and calibration, 
we can find accurate and objective patterns in the data. These observed 
patterns can be replicated and validated by other teams. But validating 
observed patterns is not the same as validating causal explanations. As 
I noted above, outside of experimental conditions, data can’t be used to 
establish causal explanations.
However, a data-d riven methodology can help us estimate the correct 
description of a phenomenon. For example, we might be able to assert 
that the number of theater performances in a given city declined over time, 
within the bounds of a confidence interval. Explaining why the number 
of performances decreased is another matter. We might observe a strong 
correlation with declining population numbers in that city over the same 
period of time. This correlation might be very strong, but it won’t provide 
sufficient grounds for a causal explanation. We might be tempted to think 
that the decrease in the population caused the decrease in the number of 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
54 •  TheaTer as DaTa
performances, but it might have very well been the other way around. The 
decrease in the number of performances might have driven people away 
from this city. Or perhaps both phenomena are the result of the same 
underlying cause, perhaps an economic crisis. An experiment might help 
settle the matter. A randomized controlled trial would be impractical and 
unethical (we would need to forcefully decrease the population in some 
cities and keep it stable in a comparable city to see what happens to the 
theater numbers). But perhaps natural situations could be found that 
approximate those conditions. Even then, cities are complex systems and 
it would be almost impossible to control for every possible confounding 
factor, and untangle true causality from a myriad of contributing causes.
To understand why performance decreased, perhaps the best method 
would be ethnographic: interviews and participant observation of differ-
ent types of practices in the city. I am not suggesting that we abandon 
the pursuit of causal explanations. But merely that we recognize that 
data-d riven methods can’t carry us all the way. I should also stress that 
ethnography and data- assisted methodologies are very different. While 
ethnographers might refer to their notes as data, these notes don’t fit the 
definition of data I use in this book. Also, ethnography cannot be car-
ried out through computational procedures, as opposed to data- assisted 
methodologies which rely on them. To continue the example above, a 
data- assisted approach would consist of developing a performative visual-
ization (see chapter 3) that shows, in a dynamic map, how the city loses its 
color and shape as the number of theater performances go down.
Institutional Bias and Statistics
This chapter has considered many sources of bias. But there is one more 
potential source that demands attention: the institutional conditions that 
arise in fields where statistics become prestigious. As theater moves into 
computational realms, it is important that we consider the challenges that 
statistics pose, and what we can do about them. Biostatistician John Ioan-
nidis (2005) claims that research in many areas is only an “accurate mea-
sure of the prevailing bias” and that most published scientific research is 
false (700). He identifies six trends that decrease the likelihood that the 
findings are true:
 1. The smaller the studies
 2. The smaller the effect sizes
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Roles of Statistics • 55
 3. The greater the number and the lesser the selection of tested 
relationships
 4. The greater the flexibility in designs, definitions, outcomes, and 
analytical modes
 5. The greater the financial and other interests and prejudices
 6. The hotter a scientific field (with more scientific teams involved)
We can find instances of all of these trends in DH. In particular, Ioannidis’s 
trend 3, should give us pause. Digital methods make it almost trivial to 
run many iterations of the same procedure, tweaking parameters slightly 
to get different results. This is the very thing that makes digital research 
possible at a large scale. But the danger is that it becomes extremely easy 
to constantly adjust our methods until we get the results we want. An 
equally important problem is trend 4. If you can always reframe a problem 
in a slightly different way until it yields the results that you are after, you 
can easily trick yourself. This is why data- driven research requires agreed- 
upon definitions on what counts as data, if we want to reach intersubjec-
tively verifiable conclusions. As noted in chapter 1, low ambiguity in the 
definition of data for a given research area is one of the four conditions for 
data- driven research. This is also why we need to find ways to calibrate our 
methods to avoid the clustering illusion and harking, as seen above. We 
should also recognize that DH is a “hot field” with significant financial 
and reputational incentives for finding new applications of digital meth-
ods (trends 5 and 6). But knowing this is a hot field is not a reason to stop 
our work, just to bring extra care to our methods. This is why being aware 
of the limitations of statistics is so important.
The only solution is to use statistical methods in a nuanced way and 
communicate our results with as much humbleness and skepticism as we 
can muster. On their own, the presence of statistics is not an unmistak-
able marker of objectivity. But under certain conditions, statistics can be 
used to help us avoid errors in our reasoning. We can turn again to Ioan-
nidis (2014, 2) for suggestions to improve the likelihood that results are 
true:
• Large- scale collaborative research
• Adoption of a replication culture
• Registration (of studies, protocols, analysis codes, datasets, raw 
data, and results)
• Sharing (of data, protocols, materials, software, and other tools)
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
56 •  TheaTer as DaTa
• Reproducibility practices
• Containment of conflicted sponsors and authors
• More appropriate statistical methods
• Standardization of definitions and analyses
• More stringent thresholds for claiming discoveries or “successes”
• Improvement of study design standards
• Improvements in peer review, reporting, and dissemination of 
research
• Better training in statistical literacy
These are all institutional suggestions. This means that, in order to gen-
erate better computational research we need to rethink the institutional 
environments that can aid in this project. Statistics have many roles to play 
in computational theater research: they can help us describe the data and 
find patterns. These patterns can be used as deformances and provoca-
tive defamiliarization strategies. Or, we can use models to estimate the 
likelihood that these patterns tell us something verifiable about the data. 
But there are also important limits to the usefulness of statistics, and we 
need to consider the many sources of bias in order to counter them. It will 
also be crucial that we report negative findings, rather than try to read too 
much into inconclusive data.
If critical realism informs our usage of statistics, we will be able to 
understand how cultural and historical conditions necessarily shape data 
collection. But we can still carefully use statistics for data-d riven pattern 
discovery. We can also use statistical patterns as launching pads for useful 
data- assisted deformances, which can nonetheless contribute to a fuller 
understanding of theater history and practice.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
ChapTer 3
The Roles of Visualizations
A data visualization is the graphical representation of data and its prop-
erties. In contemporary media environments, data and visualizations are 
almost inseparable from each other. Data could also be described solely by 
numbers and tables or conveyed through other means, such as aural soni-
fications or haptic physicalizations (Jansen et al. 2015; Moere 2008; Hogan 
2015) but these are less commonly used.
Visualizations offer powerful insight into the data they represent. But 
visualizations can also be intentionally distorted to suit a particular narra-
tive. Unpacking the ethical and political implications of visualizations is 
particularly urgent when data on national economies, public health, and 
environmental destruction are often communicated to specialized com-
munities and to the general public mainly through visualizations. Alberto 
Cairo (2020), a leading thinker and practitioner of data visualization writes 
that “data visualization is a technology—o r set of technologies—a nd, like 
artifacts such as the clock, the compass, the abacus, or the map, it trans-
forms the way we see and relate to reality” (17). As Kennedy and Engebret-
sen (2020) note, visualizations are “cultural artifacts with distinct semi-
otic, aesthetic, and social affordances” (20). Data visualizations privilege 
certain modes of knowledge. For starters, they overemphasize the sense 
of sight, and pose a problem for inclusive design agendas. Understanding 
data visualizations requires a specific kind of literacy, which can lead to 
different types of inequality. As nation- states increasingly make decisions 
based on data, uneven levels of visualization literacy have troubling impli-
cations for democracies (Snaprud and Velazquez 2020). Visualizations 
prioritize correlation over causation, and this can entrench certain types 
of bias (Rettberg 2020, 41–4 2). In chapter 2, I discussed the problem of 
untangling correlation and causation from the point of view of statistics.
In this chapter, I will unpack the implications of data visualizations 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research. 57
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
58 •  TheaTer as DaTa
for computational theater research against the background of theoreti-
cal discussions in statistical graphs, data journalism, and DH. Both data- 
driven and data-a ssisted research rely heavily on visualizations, but they 
use them in different ways.
Different Types of Data Visualization
To better understand how data- driven and data- assisted perspectives 
incorporate visualizations, we can refer to a parallel distinction between 
two communities that use data visualization: scientific and journalism 
communities. This distinction is not absolute, and the borders between 
these communities are porous (Kennedy and Engebretsen 2020). Still, 
there are some glaring differences in the ways both communities approach 
data visualization. The statistical graphs favored by the scientific commu-
nity place a prime in the precise communication of quantitative informa-
tion and include visual representations of statistical features (measures of 
dispersion and outliers). The scientific community tends to rely on stan-
dard graphical representations such as boxplots, scatterplots, and violin-
plots (some of these will be described later and used in chapters 4–7 ).
Statistical graphs are strongly associated with the work of Tukey 
(1977), as they are useful for EDA (as seen in chapter 3), but they have a 
longer history (Friendly and Denis 2001). Boxplots, which were popular-
ized by Tukey were probably first used by Mary Eleanor Spear (1952), and 
scatterplots date back to the early eighteenth century (Friendly and Denis 
2005). The work of Bertin (1983), Cleveland (1985; 1993) and Wilkinson 
(1999) has also been influential to the development of scientific visual-
ization (Friendly and Denis 2001). Throughout their history, statistical 
graphs have generally aimed to be parsimonious. Tufte (1983), one of the 
most famous theorists in this community, says that graphs should aim for 
a low data-i nk ratio. In a digital visualization, this means that every pixel 
should convey an aspect of the data, and any ornamental features should 
be removed. Another key feature of statistical graphs is that they enable 
the comparison across different facets of the data (for example, the treat-
ment given to different samples or differences in populations). This is 
usually achieved through “small-m ultiples”: embedded plots that share 
the same scaling across the x and/or y axis to enable easy visual compari-
sons (for an example, see the pairwise plot in figure 5.7).
While the scientific community uses statistical graphs to communicate 
systematicity and precision, journalists often use visualizations for audi-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Roles of Visualizations • 59
ence engagement. Visualizations in the media rarely indicate confidence 
intervals or measures of dispersion (such as standard deviations or inter-
quartile ranges). Engagement and visual variety are very important, and 
visualizations in the media are often made by artists. Creativity is deployed 
to invent new modes of graphical display, rather than relying on estab-
lished graphical forms such as scatterplots. A common approach is the 
usage of infographics, charts that explore a semantic link between the data 
and their graphical representation, often in ways which are humorous or 
aesthetically innovative. One of the most influential figures in the develop-
ment of infographics is Nigel Holmes, who worked as a designer for TIME 
magazine in the 1970s. In a well- known example, he used bar charts to 
compare medical expenses across different countries, and each country’s 
bar chart was represented as a sick patient lying in a hospital bed.
Tufte (1983) uncharitably described approaches such as this as 
“chartjunk,” as these types of infographics have a high data-i nk ratio and 
the artistic distortions often make comparisons difficult. However, the 
design community has articulated different justifications for the use of 
infographics. Ichikawa (2016), writing from a user experience perspective, 
suggests that effective visualizations must “keep the viewer anchored” 
(n.p.). Visual variation and creativity are not superfluous, but part of an 
engaging experience that can enhance the comprehensibility and utility of 
a visualization aimed at the general audience. The work of Florence Night-
ingale is perhaps one of the most famous precursors of this approach. Her 
1858 coxcomb graph aimed at showing the number of preventable deaths in 
the Crimean War, prompting officials into action. This is a landmark work 
of information graphics, but modern theorists of statistical graphs prefer 
bar charts, or other visualizations that enable straightforward compari-
sons (Gelman and Unwin 2013). For this reason, pie charts are commonly 
despised by the statistical graphs community.
Not all visualizations in the media are infographics. A growing trend 
in the past decade is the development of data- driven interactive stories, 
which have been popularized by the New York Times, the Financial Times, 
and FiveThirtyEight. The objective behind these stories is to enable users 
more direct access to the evidence. Many of these stories are both engag-
ing and statistically sound, so they problematize any simplistic distinction 
between journalistic and scientific visualizations. There are other possible 
ways to distinguish among visualization communities (see Kennedy and 
Engebretsen 2020), but the lines between them are increasingly blurred. 
Still, it is important to identify the goals of different visualizations, even 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
60 •  TheaTer as DaTa
if they can’t be fully mapped to specific communities of practice. Engage-
ment and statistical precision are distinct, if not mutually exclusive objec-
tives. Data- driven theater research usually relies on the established visual-
ization conventions developed by the scientific community. Data- assisted 
visualizations share some features with data- rich reporting in the media, 
but their objective can’t be described merely on the basis of engaging a lay 
audience.
To better characterize the difference between data- driven and data- 
assisted visualizations, I will borrow Thudt et al.’s (2018) distinction 
between exploration and explanation in data visualizations. The aim of 
explanation is to communicate the author’s view, while the aim of explo-
ration is to enable users to find their own story in a dataset. Data-d riven 
visualizations rely on explanatory facets and data-a ssisted visualizations 
rely on exploratory facets. As we have seen, the goal of data-a ssisted 
research is not to settle questions, but to challenge the assumptions of 
a question and find multiple possible answers. Likewise, the exploratory 
facets of a visualization “provide readers with the flexibility to ask a variety 
of questions of the data, to personalize their experience of the story based 
on their own interests, and to view the data from different perspectives” 
(Thudt et al. 2018, 64).
Data-d riven visualizations rely on the graphical conventions of sci-
entific visualizations to estimate the likelihood that an answer is cor-
rect given the data provided, or to indicate the most accurate charac-
terization of the data, describing its structure, or drawing attention 
to trends and clusters (as seen in chapter 2). In contrast data-a ssisted 
visualizations aim to challenge the political, ethic and aesthetic con-
ventions of data visualization tropes. Data- driven analysis can work 
within the constraints of established visualization conventions. But 
data- assisted analysis requires critical interventions to invent new 
forms suitable to its purposes, and it is to these interventions that I 
will now turn to. I want to emphasize, though, that both data-d riven 
and data- assisted research might depend heavily on visualizations, 
or use none at all, as data can also be communicated through tables 
rather than visualizations. Later in this chapter, I will dedicate sub-
stantial attention to the role of interactivity in data- assisted visualiza-
tions. Both data- driven and data- assisted visualizations can make use 
of interactive features. But the goals of data-a ssisted research are par-
ticularly well served by interactive exploratory features.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Roles of Visualizations • 61
Rethinking Visualizations from a Humanities Perspective
“Can graphical means assist the humanities in the project of interpretation?”
—Johanna Drucker
A data-a ssisted visualization doesn’t take the assumptions of visualiza-
tions for granted, and aims to make its own assumptions transparent to 
viewers. The work of Johanna Drucker is a particularly useful conceptual 
guide for this endeavor. Drucker (2014) defines graphesis as “the study of 
visual epistemology” through a “dynamic, subjective process.” Its pur-
pose is to “expose and describe the principles for structuring knowledge 
through graphical form,” and it thus seeks to create methods that are “gen-
erative and iterative, capable of producing new knowledge through the aes-
thetic provocation of graphical expressions.” In other words, the point of 
graphesis is to communicate humanities knowledge through visual means, 
understood as key rhetorical devices in their own right; they are “primary 
methods of analysis” that “create the data, not just display it” (n.p.).
In setting up a theory of graphesis, Drucker argues that graphical 
means of communication have long had a central role to play, but this 
role has often been dismissed in Western traditions of thought. The most 
familiar graphical forms, such as books and letters, have been dramati-
cally overlooked. As Drucker notes, “basic codes for reading are graphi-
cally structured.” This includes conventions for text, footnotes, table of 
contents, and marginalia. She invites us to take these codes seriously and 
give them the attention they deserve: “A margin isn’t an inert space, but a 
field of defining tension between text and page edge, and exerts a graphi-
cal force in relation to other elements on the page” (n.p.). If we give due 
consideration to these seemingly neutral elements, we can mobilize them 
for the reinvention of new graphical languages suited to the purposes 
of humanistic inquiry. She finds useful precedents for such a project in 
the work of several artists, such as Vasiliy Kandinsky, Paul Klee, Bauhaus 
artists László Moholy-N agy and Josef Albers, El Lissitsky, Piet Zwart, Jan 
Tschichold, as well as filmmakers Dziga Vertov and Sergei Eisenstein. 
They all explored the rhetorical force of formal elements such as repeti-
tion, discontinuity and variation. One of Drucker’s most fascinating exam-
ples is the work of Ernst Fraenkel, who presented his analysis of Stéphane 
Mallarme’s Un Coup de Des through very idiosyncratic graphical arguments.
These are things that we can apply to the creation of new visualiza-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
62 •  TheaTer as DaTa
tions for data- assisted research. Drucker urges humanists to not repro-
duce the visual tropes of scientific knowledge, but to deploy visual codes 
to inscribe knowledge as provisional, situated, and observer- dependent. 
Visualizations should “bring the interpretive sensibilities of theoretical 
inquiry to bear on these assumptions while also acknowledging subjectiv-
ity as fundamental to the conception and expression of knowledge” (n.p.). 
A central preoccupation of Drucker’s own work is the graphical represen-
tation of time. She suggests that calendars, for instance, model temporal 
elements rather than represent a “natural” condition of time. In her own 
interventions, Drucker has been interested in modeling “the multi- linear 
(forking paths), heterogeneous (varied in density, rate, and scale), and 
discontinuous (broken, repetitive) temporalities that are part of human 
experience” (n.p.).
Writing elsewhere, Drucker (2011) offers a “call to imaginative action 
and intellectual engagement with the challenge of rethinking digital tools 
for visualization on basic principles of the humanities” (n.p.). To illustrate 
this, she asks us to imagine bar charts that illustrate population changes 
in different fictional nations, which divide people by gender. Visualiza-
tions such as these are common, and therein lie their problems. They are 
so naturalized that they effectively hide the interpretive basis on which 
they are built. Drucker takes apart the categories that underpin this visual-
ization to show there is nothing natural about concepts such as “people,” 
“gender,” and “nation.” In collaboration with visual artist Xárene Eskan-
dar, she offers an alternative mode of representation, where the bar chart 
is reimagined to show tension and indeterminacy in its constitutive cate-
gories. For example, blurred lines show gender ambiguity inside the bars, 
rather than binary distinctions. Floating points between the bars of two 
nations indicate the porosity of a political border and the transient state 
of migrant workers, revealing the instability of the concept of nation. The 
visualization could also indicate culturally specific notions of personhood. 
For this, Drucker imagines a given country where “women only register as 
individuals after coming of reproductive age, thus showing that quantity 
is an effect of cultural conditions, not a self-e vident fact.” Even though the 
bar chart is based on fictional countries, it shows how the humanities can 
do more than merely apply existing visualization frameworks and develop 
graphical approaches that bring interpretation to the fore. Graphical ele-
ments are not discrete bounded entities but “conditional expressions of 
interpretative parameters” (n.p.). Drucker’s article includes other exam-
ples that should be required reading for anyone interested in developing 
data- assisted visualizations.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Roles of Visualizations • 63
Drucker is a proponent of extreme constructivism, for whom interpre-
tation is at the forefront of any knowledge enterprise in the humanities: 
“Nothing in intellectual life is self- evident or self- identical, nothing in 
cultural life is mere fact, and nothing in the phenomenal world gives rise 
to a record or representation except through constructed expressions” 
(n.p.). While working entirely from a constructivist perspective is a poten-
tial avenue for theater researchers interested in data, this is not the only 
possibility. In this book, I champion approaches that combine realism 
and constructivism, by recourse to critical realism (see chapter 1). While 
Drucker herself disavows any traces of observer- independent realism in 
humanities research, her ideas can still be incorporated into a critical real-
ist approach to data visualization. Inspiration for this project can be found 
in feminist data visualization.
In D’Ignazio and Klein’s (2016) formulation, feminist data visualiza-
tion “rejects neither the scientific process nor quantitative ways of know-
ing the world” but aims to “see how all knowledge is situated, how cer-
tain perspectives are excluded from the current knowledge regime” (n.p.). 
Feminist data visualization shows that it is possible to pursue evidence- 
based answers while remaining attentive to the problematic categories on 
which evidence is built. This means that the designers of visualizations 
should rethink binaries, consider edge cases that problematize categories, 
and legitimize embodiment and affect. It will also be crucial for such a 
project to trace each datapoint back to its source. D’Ignazio and Klein also 
stress the importance of using historically and culturally specific modes 
of representation and overturning hierarchies of knowledge transmission 
through participatory design practices. They ask: “What kinds of termi-
nology, symbols, and cultural artifacts have meaning to end users, and 
how can we incorporate those into our designs?” (n.p.). Perhaps we can 
find an answer to their question in the usage of “cultural probes” (Gaver, 
Dunne, and Pacenti 1999), where culturally significant visual conventions 
are incorporated into the graphical language of a visualization. In chapter 
6, I give an example of a kayon plot (figure 6.4), a visualization I developed 
for describing Javanese theater performances, which incorporates statisti-
cal information, has a low data- ink ratio, and is still premised on an inter-
pretive, culturally situated view. This is a possible way in which critical 
realism can inform the development of visualizations that are statistically 
sound, but which also bring interpretation to the fore.
Another possible way of combining statistical information and culturally 
specific information is through Manovich’s (2011) notion of direct visualiza-
tion. According to Manovich, data visualization is conventionally premised 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
64 •  TheaTer as DaTa
on the representation of data through “graphical primitives such as points, 
straight lines, curves and simple geometric shapes to stand in for objects 
and relations between them” (36). Spatial variables such as position, size 
and shape represent key differences in the data and reveal patterns and 
relations. Direct visualization replaces the graphical primitives with actual 
media objects, such as images and video. These representations can be spa-
tially organized according to quantitative measures (such as similarity in 
terms of color palette). This approach works specially well for images, and 
Manovich’s examples include visualizations of magazine covers.
Ethics of Representation
Whether theater researches choose to work entirely within a construc-
tivist perspective, or a critical realist one, a key tenet of data-a ssisted 
perspectives is that visualizations are never neutral. Specific visual 
configurations— colors, shapes, style— have rhetoric implications, and 
are argument- altering, persuasive statements. These choices also have 
ethical dimensions. As Hepworth and Church remind us, visualizations 
are not mere janitorial work, as important decisions made during the pro-
cess of visualizing data can “critically shape historical narratives” (2019, 
n.p.). To highlight the ethical dimensions of data visualizations, they 
critique two projects that map lynchings in the United States. Lynching in 
America (https://lynchinginamerica.eji.org/) was made by Google for the 
Equal Justice Initiative, a mass incarceration not-f or- profit organization. 
In an interactive choropleth map, the shade of each county indicates the 
number of African-A mericans who were lynched between 1877 and 1950. 
Hepworth and Church, while sympathetic to the objectives of this project, 
show that the choice of visualization conventions implies that these mur-
ders were limited to a range of districts in the American South.
The authors contrast this with another project, Map of White Supremacy 
Mob Violence (http://www.monroeworktoday.org/explore/), which they 
describe as more nuanced and comprehensive. This second map focuses 
on the entire United States to depict lynchings and mob violence as a 
country- wide problem. For this they use a point visualization rather than 
a choropleth. State lines are not visible and each incident is represented 
as a single dot. Additional information is provided for each record, which 
includes Native American, Mexican, and Chinese victims, rather than 
only African Americans. In comparing the visual style and the possible 
interactions afforded by each site, Hepworth and Church demonstrate 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Roles of Visualizations • 65
that subtle design differences enact very different arguments. Building 
on these observations, the authors propose a framework of ethical visu-
alization practices that consists of several steps: defining, reviewing, col-
lecting, pruning, describing, surveying, and previsualizing. Each step is 
grouped into phases: pre- data collection (defining, reviewing); data col-
lection and curation (collecting, pruning, describing); and data visualiza-
tion and argumentation (surveying, previsualizing, visualizing, publish-
ing). By showing the context and source of each data point, Map of White 
Supremacy Mob Violence resists the tendency of visualizations to show aggre-
gates rather than individuals. Interactive data- assisted visualizations such 
as this one often try to link the representation of the dataset as a whole 
to the story behind each data point. In doing so, they illustrate the ten-
sion between each unique individual and the categories into which it is 
grouped. As seen above, this is also one of the guiding principles of femi-
nist data visualization.
Thinking about categories and taking apart graphical conventions are 
not the only considerations to ensure an ethical approach to data visual-
ization. As Diakopoulos (2018) notes, computational operations (such 
as algorithmic derivation, filtering, aggregation, and normalization) 
also change the ways the data is presented and can alter the conclusions 
reached. Algorithmic derivation means that data is not directly visualized 
as found in a dataset, but subject first to a computational transformation. 
Normalization is the mathematical operation of dividing a quantity by 
another in order to offer standardization. For example, instead of show-
ing the raw counts of theater venues per city, the number of venues can be 
normalized by population counts (i.e., divided by the population counts). I 
give an example of this in figure 7.1, where I consider changes in the num-
ber of performances over time. Diakopoulos suggests that annotation and 
interactivity are ways in which the impact of computational choices can 
be communicated to the audience. Explaining how the data was obtained 
and transformed is paramount for ethical data visualizations (see Gray et 
al. 2016). For this reason, making the data and the methods available for 
replicability is also important from an ethical perspective, regardless of 
whether a visualization is part of a data-a ssisted or a data- driven project.
Performative Data Visualizations
Scheinfeldt (2012, n.p.) suggests that the humanities are becoming more 
performative (especially DH). His examples are web-a dvertised public 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
66 •  TheaTer as DaTa
events. With the advent of social media, humanities work is done in public 
more and more: “Increasingly digital humanities work is being conceived 
as much as event as product or project.” He concludes that there are per-
formative dimensions which are changing the rules of scholarly commu-
nication: “Performance is a different ball game than publication.” Build-
ing on this insight, Bay-C heng (2017) suggests that history is also being 
increasingly staged, and that data has played a major role in this process. 
History is becoming more performative in the sense that it is increasingly 
performed for a public.
Data visualizations are often publicly available. But there is another 
way in which we can consider them as performative. These digital arti-
facts can be performative in the sense that they enact situated, subjec-
tive responses. Such performative character is less directly linked to the 
public character of a visualization and more closely connected to its dia-
logic, interpretive character. Ramsay (2011) builds explicitly on McGann 
and Samuels’ deformance and also takes inspiration from the Oulipo artists 
to describe how data can be used for playful and performative transfor-
mations. A larger “performative turn” is also discernible in recent think-
ing about visualizations. For example, Parry (2019) proposes the concept 
of enactment to critique visualizations: “Visualizations also act. They do 
things. They are visual-v erbal- numerical enactments.” Visualizations, in 
Parry’s view, enact specific relations, histories, and politics. Bearing this 
in mind when analyzing a digital interface “attends to questions of who, 
under given social, cultural, and political conditions, is allowed to live, 
to speak, and to act— questions, in other words, of who and what gets 
to matter” (n.p.). These principles are not only analytical categories, as 
they can also be mobilized for specific design practices, what Parry terms 
“enactment-i ntensive data visualization,” and which include archival con-
testation, conditional revelation, and fluid interpretation.
Despite these rich discussions, references to performative writing as 
understood in performance studies are largely absent in DH. This is a 
missed opportunity, as the writings of people such as Peggy Phelan, Della 
Pollock, and Soyini Madison, well known to theater and performance 
scholars, are directly relevant to the project of reimagining performative 
visualizations. Della Pollock (1998), characterizes performative writing as 
evocative, metonymic, subjective, nervous, citational, and consequential. 
Finding ways to transpose these characteristics to visualizations provides 
fascinating design challenges and invitations to think differently about 
visualization. What is a nervous data visualization? I understand nervous-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Roles of Visualizations • 67
ness here as a provisional, ever-s hifting relationship to knowledge. For 
Madison (2006), a key characteristic of performative writing is the way it 
interpellates its imagined readers: “Performative writing emphasizes the 
relational dynamic between writer and reader in a spirit of caring about 
the dialogic and communicative quality of the connection” (n.p.). Perfor-
mative visualizations likewise emphasize the relational dynamic between 
designer and user (or “subject,” as Drucker would have it). We can also 
gather data on how people interact with our visualizations and further 
contribute to this dialogue. The fact that digital visualizations can also be 
revised and are essentially open-e nded makes them closer to performance 
in the sense that they are fundamentally unfinished and unfinishable. For 
Madison (2006), “performative writing is evocative because it is a braiding 
of poetry and reportage, imagination and actuality, critical analysis and 
literary pleasure” (n.p.). Performative visualizations, in turn, braid visual 
art, data analysis, data science, and data hermeneutics.
An example can be found in Manovich’s (2013) evocative data essay on 
Vertov’s films:
The presentation is an experiment. Normally an academic article con-
sists from text with a small number of illustrations. Instead, this pre-
sentation is a portfolio of a large number of visualizations, with text 
serving as the commentary. The presentation also does not advance a 
single argument or a concept. Instead, I progressively “zoom” into cin-
ema, exploring alternative ways to visualize media at different zoom 
levels, and noting interesting observations and discoveries. We can 
compare its genre to that of travel writing, where the organizing prin-
ciple is the writer’s movements through space. (Manovich 2013, 45)
Borrowing Manovich’s felicitous metaphor, we can think of performative 
visualizations as more akin to travel writing than to the scientific explora-
tion of landscape. Travel writing, like generative visualization, is aimed at 
suggesting more than what it contains, to signal longing and possibility, 
and perhaps to encourage future travelers to challenge the conventional 
wisdom of known landmarks.
Interactivity
The ethical, interpretive, and performative characteristics of data- assisted 
visualizations can be brought forth in either static or interactive visualiza-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
68 •  TheaTer as DaTa
tions. But interactivity adds specific affordances to data-a ssisted visualiza-
tions. The presence of interactivity per se is not a marker that a visualiza-
tion is performative, or that the project is data- assisted. But there are three 
important objectives that can be enhanced by interactivity: multiple per-
spectives, multiple scales, and thick context.
A key tenet of data-a ssisted approaches is that they should allow mul-
tiple interpretive perspectives. In an interactive visualization, each of these 
perspectives can correspond to a specific pathway. For example, when 
users click on different buttons, the data can be reorganized into different 
interpretive categories. In interaction design, these pathways are some-
times explained with recourse to Aarseth’s (1997) ergodic narratives. An 
ergodic narrative requires non-t rivial effort on the part of the users to tra-
verse a narrative. In the original formulation, any choice in an ergodic nar-
rative forecloses other possible paths. A reader who makes a choice will 
not know what would have been the outcome of treading a different path. 
This is usually not the desired case in interactive data visualizations. Bate-
man et al. (2017) offer a refinement and describe interactive visualizations 
as “ergodic yet immutable” (108). Ergodic yet immutable visualizations 
can present contradictory points of view, aiming to emulate the sophisti-
cated disagreement of theater studies, and the multiplicity of perspectives 
central to data- assisted research. As Hiippala (2020) notes, interactive 
data visualizations “require ergodic work both in the form of exploration 
and composition, a feature which separates [them] from static informa-
tion graphics and non- dynamic data visualizations” (287). In a sense, the 
work of designing data visualizations is always ergodic, as designers must 
choose among different graphical perspectives. But interactive visualiza-
tions extend this ergodic experience to the users.
Moving between different scales is also important for data-a ssisted 
visualizations where attention to individual data points is as important as 
the consideration of data aggregates. Interactive systems are well suited to 
enable shifts between different levels of granularity. When thinking about 
ways to enable changes between scales, it might be useful to consider add-
ing rich- prospect features to a visualization. For Ruecker, Radzikowska 
and Sinclair (2016) a rich-prospect browser enables users to see every item of 
a collection at once while hiding some of its details. Rich-prospect also 
enables users to see how different organizational criteria reveal varying 
connections within a collection (or dataset in this case). Rich- prospect 
browsers were originally developed for the visual display of digital cultural 
heritage and aim to provide “insight into how the collection was under-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Roles of Visualizations • 69
stood by the people who made it” (Ruecker, Radzikowska, and Sinclair 
2016, 176). This feature can be extended to interactive data visualizations 
to reveal how the data was understood by the people who collected or 
organized it, and highlight the situated perspective of these designers.
Lastly, interactivity can be used to provide thick context. Interactive 
visualizations enable users to retrieve the source of specific data points. 
For example, a network diagram might aggregate information on thou-
sands of performers. By clicking on a node in the diagram, the user can 
find additional details about the performer represented by that node, as 
well as a summary of interpretive decisions on how the data was coded, 
or a note alerting the user to missing data (a similar example, for fictional 
characters is presented in chapter 5). For inspiration on how to merge 
visualization with contextual explanations, we can look at Bach et al.’s 
(2018) narrative design patterns for interactive visualizations, which are avail-
able at http://napa-cards.net/. A narrative design pattern is “a low-l evel 
narrative device that serves a specific intent” (Bach et al. 2018, 111). One 
of these patterns, humans- behind-t he- dots is particularly useful for adding 
thick context. The idea behind this pattern is presenting individual sto-
ries in response to users’ clicks on data points. Although Bach et al. are 
writing from a data journalism perspective, this approach can be easily 
extended to data- assisted theater research.
In sum, interactivity can enable a combination of scales, bring together 
multiple perspectives, and add thick context. When thinking about the 
role of interactivity in visualizations, it is important to note that not every 
single element will be interactive, and that it will not be interactive to the 
same extent. According to Thudt et al. (2018), interactivity can be present 
in different degrees as users are only allowed to control certain aspects of 
a visualization: view, focus, and sequence. View means that the user can 
choose what is represented and how it is represented. A high degree of 
interactivity within this aspect would allow users to choose the parameters 
of visual encoding: color, size, and spatial placement. In a less interactive 
version, the users might select what visualizations they want to display 
side by side to see different aspects of the data. Focus refers to the segment 
of a visualization that the user wants to bring to the fore. When there is 
a large amount of data to choose from, users can decide where to look 
by filtering, selecting, zooming, or panning. Lastly, sequence refers to the 
extent to which users can make choices on the progression and temporal 
order of a visualization. Users can be guided through a predefined journey 
through steppers or strollers (in a recent trend called scrolly-t elling). In this 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
70 •  TheaTer as DaTa
case, interactivity enables users to choose the pace of an interactive story. 
But a higher degree of interactivity can be achieved when the users decide 
not just the pace but the next destination of a data journey.
When a visualization enables a high degree of interactivity for all of 
these aspects, it becomes more than just a visualization. Using the con-
ceptual lens of theater studies, we can think of these as intermedial essays. 
The combination of scale, perspectives, and contexts requires not just 
interactivity but the co-p resence of different medial forms: essays, videos, 
sounds, and graphs. Thinking of them as intermedial essays as opposed to 
data stories (a common term in data journalism) highlights their artefac-
tuality and the media specificity of each of its components. As Mee (2018) 
notes when reflecting on the potential of web-b ased theater scholarship, 
“digital platforms allow our scholarship to embody our argument” (8). An 
intermedial essay can exist in between different formats, and this is simi-
lar to the way intermedial performances exist in between media (Chapple 
and Kattenbelt 2006; Bay-C heng et al. 2010). Intermedial performances 
reflexively combine different media, and intermedial scholarship is well 
suited for the reflexive combination of data, textual arguments, and mul-
timedia. For a performance to be intermedial, it must retain some aspects 
of theater, but also include some aspects of other media forms. Likewise, 
intermedial essays might still maintain the conventions of written essays, 
but also include videos and data visualizations.
A key aspect of intermedial essays will be self- reflexivity. Chapple 
and Kattenbelt (2006, 11) argue that “a self- conscious reflexivity that dis-
plays the devices of performance in performance” is an essential feature 
of intermedial performances. In other words, not all performances that 
use media are intermedial, but only those that integrate a certain level of 
self- reflexivity. Likewise, intermedial scholarship should include a level of 
reflexivity with regards to its own mechanisms, and make users aware of 
its own devices. A good example is the embedded “hermeneutic toys” by 
Rockwell and Sinclair (2016) discussed earlier in this book.
One of the most notable theater examples is Erin Mee’s Hearing the 
Music of the Hemispheres, which aims “to offer an alternative, performance- 
driven model for understanding spectatorship” by combining multime-
dia objects and arguments using the Scalar platform (2013, 149). Another 
influential set of intermedial essays is the Hemi Press’s Gesture, a series of 
“evocative digital works that combine multimedia and writing to make 
an original critical intervention in the fields of performance and politics” 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Roles of Visualizations • 71
(Hemispheric Institute 2020). A short excursion that prefigures some 
projects I will describe in part 2 is my intermedial essay “Wayang Kontem-
porer: Innovations in Javanese Wayang Kulit” (Escobar Varela 2015, avail-
able online at http://cwa-web.org/dissertation/wayang-dis/). A short video 
explanation of this project is available as video 3.1, in the web compan-
ion to the present book. Wayang Kontemporer includes videos, essays, and 
interactive visualizations. The different components are organized in two 
major “areas,” which can be described as canvases following Bateman’s 
(2017) semiotic vocabulary for interactive platforms. One canvas is dedi-
cated to essays with embedded videos, and the other to interactive visu-
alizations. As opposed to the majority of the visualizations I will describe 
later in this book, the visualizations in Wayang Kontemporer are entirely pre-
mised on interpretive methodologies (they are the result of fieldwork and 
interviews rather than computational transformations). In that project, I 
analyzed twenty- four performances, which I classified across five dimen-
sions. The explanation for how each dimension was selected is described 
in a series of essays in the essay canvas. A series of radial charts can be 
loaded into the visualization canvas, and the user can overlay them on top of 
each other in order to make comparisons across the performances.
Since the users can choose which performances to compare, this inter-
face affords an interactive view (in Thudt et al.’s terminology). The radial 
chart shows how each performance was classified along the aforemen-
tioned five dimensions. By placing the mouse over the visualization, the 
user can see a short explanation of why the performance was classified in 
such a way. Clicking on the visualization loads a longer essay that explains 
the interpretive decisions behind each visualization in greater detail. This 
is an example of how a visualization can provide thick context. The visu-
alizations in Wayang Kontemporer constantly refer to the essays, and vice 
versa, but it is up to the users to follow such links. There are many pos-
sible pathways between the components of this intermedial essay, and the 
users can choose how to traverse those pathways. In intermedial essays, 
the distinction between visualization and interface collapses, and this 
poses major challenges for the digital sustainability of such projects (see 
chapter 8). But even simpler, static visualizations can achieve the inter-
pretive and performative goals of data-a ssisted visualizations (think of 
Drucker’s bar charts).
This chapter has shown how data- assisted visualizations can be used 
to poke around the edges of a question, while data-d riven visualizations 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
72 •  TheaTer as DaTa
are used to offer the most plausible visual description of a dataset under 
a set of assumptions. These assumptions can in turn be interrogated 
through data- assisted means, visual or otherwise. The chapters in the 
next section show a variety of examples of how data visualization and data 
analysis are used to ask— and answer—a  wide range of questions relevant 
to theater scholarship.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
parT 2
Guided Tours
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
ChapTer 4
Words as Data
The digital analysis of word patterns has the widest array of tools, meth-
ods, and theoretical perspectives of any area covered in this book. Tech-
niques derived from computational linguistics and corpus analysis are 
often used in DH, and have been extensively applied to the study of dra-
matic texts (Craig and Greatley-H irsch 2017). While I survey this type of 
work in some detail, my interest here is to argue that the same methods 
could be applied to study a variety of textual materials related to theater 
practice: advertisements, playbills, reviews, casting calls, production 
notes, academic articles, etc. This is currently an underexplored but prom-
ising area for computational theater research.
Text, as discussed in this chapter, denotes a collection of words rather 
than signaling more expansive definitions which could include diagrams 
and other visual items. Many researchers are interested in applying for-
mal models to answer narrowly construed questions related to authorship 
and other similarly closed questions, whereas other researchers are drawn 
to the speculative potential of computers for the analysis of literary words 
(positions I have characterized as data- driven and data-a ssisted, respec-
tively). Here I will describe four common methods—d imensionality 
reduction, time series analysis, measurement of linguistic differences, 
and topic modelling— and then explore how these methods can be used 
within data- driven and data- assisted methodologies.
Methods for the Analysis and Visualization of Texts
Principal component analysis (PCA) is the most common procedure within 
the family of dimensionality reduction techniques, and it is routinely used 
for authorship attribution, and computational stylometry (Binongo and 
Smith 1999; Eder 2015; Oakes 2014; 2017). When working with textual 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research. 75
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
76 •  TheaTer as DaTa
data, each word in the corpus can be treated as a dimension. Each text can 
be thought of as a series of coordinates in this multi- dimensional space, 
where the value for each dimension corresponds to the number of times 
each word is present in that given text. With hundreds or thousands of 
dimensions it is hard to identify patterns in those numbers. PCA aims to 
make this easier by representing the words in a lower-d imensional space. 
These new dimensions, called principal components, aim to maintain 
most of the variance in the original multi- dimensional space. But hav-
ing less dimensions (typically two) makes it easier to graph and analyze 
patterns in the data. For example, Schöch (2016) has used PCA to identify 
word usage clusters across genres in French drama. Craig and Greatley- 
Hirsch (2017) offer a more comprehensive definition of PCA than the one 
given here, and provide many applications to the study of dramatic texts.
Another common approach is the visualization of word usage changes 
over time. This approach rose to prominence with a paper by Michel et 
al. (2011) that used millions of datapoints from Google Books to identify 
trends in word usage patterns, and interpreted such trends as evidence 
of cultural change. This project relied extensively on line charts (like the 
one in figure 4.1) to visualize the change in word frequencies over time. 
Kulkarni et al. (2015) have developed a more robust set of techniques to 
identify statistically s ignificant trends in word changes that introduces the 
kind of calibration checks I described in chapter 2. For example, they gen-
erated many alternative versions of a Wikipedia corpus where they artifi-
cially introduced word changes to verify that their proposed method could 
detect these shifts. In the excursion below, I describe a simpler statistical 
procedure for distinguishing word patterns over time and apply this to the 
study of a corpus of theater reviews.
A key interest of many digital literary scholars is developing mea-
surements for comparing two texts. Some of these measures have been 
imported from corpus linguistics, such as the type-t oken ratio. To obtain 
this measure, we divide the total number of unique words by the total 
number of words in a text. The resulting number gives a sense of the lexi-
cal richness of a text; the bigger the number, the more varied the vocabu-
lary. Many quantitative measurements have also been specifically devel-
oped for DH purposes, and the most influential is perhaps Burrows’s 
(2002) delta. This and related measures were initially used for authorship 
identification, but are now commonly deployed to trace differences among 
sections of a text or among authors, genres, places, and times (Burrows 
2007; Juola 2018) and they have also been extensively applied to the study 
of dramatic texts (Craig and Greatley-H irsch 2017).
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Words as Data • 77
Topic models can be understood as groups of words that tend to occur 
together. The name might be misleading, as these “topics” might not rep-
resent semantic unities, or “belong to a common abstract theme (such as 
justice or biology)” (Schöch 2017, n.p.). Topic models have been applied 
extensively in digital literary criticism (Jockers 2013; Meeks and Weingart 
2013; Piper 2018). Schöch’s (2017) example is particularly relevant to the-
ater studies, as he constructed a topic model on 391 French plays from 
the classical age and the Enlightenment. Of the topics he found, some are 
thematic and others are related characters, recurring dramatic actions, or 
settings. Some of the findings confirm existing views while others hint at 
new hypotheses. With characteristic DH reserve, Schöch tempers the opti-
mism of his results with a balanced explanation of the technical and inter-
pretive limitations of this study. Although the corpus is of a decent size, 
and of the scale far beyond what a scholar could read in a short time, it’s 
still far from an ideal representation of the period in question. He was able 
to use this corpus, because of excellent efforts to encode French drama in 
free and open text encoding initiative (TEI) formats (more on which in a 
moment). The question of genre, of enormous importance to literary his-
tory, is also well- suited to the possibilities of topic modeling. Thus, here 
we see the happy confluence of available data in reusable formats, relevant 
methods, and questions pertaining to the specific intellectual history of 
French drama.
At this juncture, I want to bring attention to the importance of TEI, a 
series of guidelines for the digital markup of textual phenomena. Markup 
tags are divided into structural components (speaker parts, acts, and 
scenes), renditional aspects (font, size, hue, etc.), logical and semantic 
features (names, dates, addresses), and analytic features (notes and anno-
tations). Consider Hamlet’s soliloquy, as encoded in TEI, taken from the 
Folger Digital Library (Mowat et al. n.d.):
<speaker xml:id="spk- 1762">
<w xml:id="fs- ham- 0271840">HAMLET</w>
</speaker>
<l xml:id="ftln- 1762" n="3.1.64">
<w xml:id="fs-h am-0 271850" n="3.1.64" lemma="to" 
ana="#acp- cs">To</w>
<c> </c>
<w xml:id="fs-h am-0 271870" n="3.1.64" lemma="be" 
ana="#vvi">be</w>
<c> </c>
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
78 •  TheaTer as DaTa
<w xml:id="fs- ham- 0271890" n="3.1.64" lemma="or" 
ana="#cc">or</w>
<c> </c>
<w xml:id="fs- ham- 0271910" n="3.1.64" lemma="not" 
ana="#xx">not</w>
<c> </c>
<w xml:id="fs- ham-0 271930" n="3.1.64" lemma="to" 
ana="#acp- cs">to</w>
<c> </c>
<w xml:id="fs-h am- 0271950" n="3.1.64" lemma="be" 
ana="#vvi">be</w>
<pc xml:id="fs- ham- 0271960" n="3.1.64">— </pc>
<w xml:id="fs- ham- 0271970" n="3.1.64" 
lemma="that" ana="#d">that</w>
<c> </c>
<w xml:id="fs- ham- 0271990" n="3.1.64" lemma="be" 
ana="#vvz">is</w>
<c> </c>
<w xml:id="fs- ham- 0272010" n="3.1.64" lemma="the" 
ana="#d">the</w>
<c> </c>
<w xml:id="fs- ham- 0272030" n="3.1.64" 
lemma="question" ana="#n1">question</w>
<pc xml:id="fs- ham- 0272040" n="3.1.64">:</pc>
</l>
Hamlet is identified as the speaker, within the <speaker> . . . </speaker> 
tags. Each line is marked within the <l> . . . </l> tags. Individual words, 
nested within the line element, are marked within the <w> . . . </w> tags, 
which also provide information on the word’s lemma, via the lemma attri-
bute. This data can then be easily extracted and processed for a wide range 
of research purposes. As a result of large-s cale digitization and encod-
ing projects in literature, many researchers have access to textual data 
encoded in consistent TEI formats, and many software tools have been 
developed specifically to work with TEI- compliant files. Even then, TEI is 
not a panacea, as most researchers need to reformat the data to suit their 
specific needs and there are always omissions and inconsistencies in large 
textual collections. That being said, TEI provides an excellent data model 
and many analyses of dramatic literature build on TEI-c ompliant digital 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Words as Data • 79
text collections. But TEI can also be used to encode other types of materi-
als relevant to theater research which are not necessarily playscripts. For 
example, The Harry Watkins Diary: Digital Edition is a TEI-e ncoded collection 
of diaries written by Harry Watkins (1825–1 894), who regularly recorded 
the plays he saw between 1845 and 1860, and which is a digital compan-
ion to A Player and a Gentleman: The Diary of Harry Watkins, Nineteenth-C entury 
US American Actor (Hughes and Stubbs 2018). Other textual materials rel-
evant to theater research might not be encoded according to consistent 
data models, but many theater reviews and other texts are freely avail-
able online in machine- readable formats, and this is a feature that theater 
scholars could take more advantage of.
In the preceding overview, I have described some examples of dimen-
sionality reduction, time series analysis, measures for linguistic compari-
son, and topic models. But it should be noted that there are many possible 
techniques within each of these approaches. Dimensionality reduction 
can be achieved through the rather traditional PCA method just described, 
but there are many other possibilities. Increasingly, researchers aim to 
arrive at dimensionality reduction through ML, and this approach doesn’t 
require a predefined statistical model (Breiman 2001). Rather, inferences 
are made based on large amounts of data. There are many kinds of ML 
and these techniques can also be applied to time series analysis (Karsdorp 
et al. 2020) and the identification of linguistic clusters. Here, I just want 
to hint at this expansive area with many things to come— more extensive 
discussion of ML for textual analysis research can be found in Piper (2018) 
and Underwood (2019a).
The four major methods just surveyed, and combinations of them, 
can be applied to a wide range of projects and questions. One example 
is the analysis of text reuse: measures for comparing texts and for trac-
ing change over time can be used to study how portions of a text have 
been copied, paraphrased, or cited in other texts. Sometimes text reuse 
is focused on how quotes (e.g., of sacred texts or theater plays) are prop-
agated through time. Another major task of text reuse is alignment: the 
visual representation or description of parts of two or more texts that are 
identical (or at least equivalent). Alignment is particularly important for 
the study of translations, or the study of multiple versions of foundational 
texts. An example from theater is an analysis of thirty-s even different 
German translations of Othello, offered as a test case of the Version Varia-
tion Visualization project (Cheesman et al. 2011). Although text reuse is a 
growing area elsewhere in DH, it has yet to be more consistently applied 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
80 •  TheaTer as DaTa
to the study of theater. It would be fascinating to see to what extent similar 
passages are repeated throughout theater scholarship, publicity materi-
als, or theater criticism. Text reuse is relevant for questions that are less 
literary in nature, such as the study of adaptations for stage performance 
or patterns of textual transmission in oral traditions. Text reuse could be 
of interest to scholars working in theater traditions where improvisation 
based on textual formulas is important.
Procedures of increasing complexity are being developed for the com-
putational analysis of texts. But there is still scope for relatively simpler 
methods. Take for example, Style Inc., where Moretti (2009) analyzes 
seven- thousand titles of British literature. He looks both at the changes in 
the number of words per title and the types of words in the title to explain 
changes in the evolution of novels in the eighteenth and nineteenth cen-
turies. Using simple statistics and visualizations, Moretti discovered that 
titles became progressively shorter, and that they also tended to become 
more abstract in nature. This kind of analysis can also prove inspirational 
for theater studies because of its source of data. Although Moretti’s article 
is ostensibly about literature, it doesn’t focus on the novels themselves, 
but on the registries of their titles. The availability of playbill records 
could constitute a similarly fertile territory for the kinds of analysis pro-
posed by Moretti. There are several playbill repositories being compiled 
at the moment (Prince Lab for Digital Humanities n.d.; UPenn Libraries 
n.d.; NYPL/Zooniverse n.d.) and this might be an area of future growth. 
Of course, the kinds of questions that one could ask of such reposito-
ries are different from the ones warranted by a registry of book titles. In 
the case of novels, we can assume that each one is unique. In contrast, 
when it comes to theater playbills, one would expect that many titles are 
repeated. The reoccurrence of specific titles could be analyzed over time, 
as has been done for the study of the records of the Comédie Française 
Registers Project (York 2017). Compiling and arranging playbill data is a 
complex interpretive endeavor and not a straightforward process of tran-
scription (Vareschi and Burkert 2017). While playbills provide a fascinat-
ing source of data, they are also highly cultural artifacts. Playbills make 
sense within certain production systems but not within others. They are 
commonly employed for commercial, experimental, school, and commu-
nity theater productions in many parts of the world. But traditional theater 
performances (for example, in Southeast Asia) are rarely accompanied by 
playbills.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Words as Data • 81
Data-D riven Text Analysis
A good example of a closed textual question, of the kind favored by data- 
driven analysis is: who is the author of a text with disputed or unknown 
authorship? This has been perhaps one of the most popular questions 
to be explored in digital literary analysis. The history of these questions 
goes back to the early 1950s, and they continue to be hotly debated in DH 
conferences today. An astounding variety of ML and classical statistical 
methods have been applied to study these questions, and all the methods 
seen above can be used for authorship attribution. This is not the space to 
review all possible approaches in full, as my objective here is just to show 
that the question of authorship is fundamentally scientific in the way it is 
pursued. There is a clear hypothesis, in response to explicitly formulated 
questions.
A favorite target of this analysis is Shakespeare. This fascination is 
closely linked to specific intellectual and cultural histories. From today’s 
perspective, the) mystery of Shakespeare’s disputed authorship for certain 
texts is both unknown and important. Authorship in Shakespeare’s time 
was collaborative and the existing records don’t paint a definitive picture 
on authorial contributions. It is impossible to imagine that the kinds of 
questions stirred by Shakespeare’s unconfirmed authorship would have 
much sway in the cultures and contexts not intrigued by individual author-
ship. Perhaps the question of authorship is so tantalizing because it’s both 
not fully answered but still within our grasp— there is data that could, in 
principle, be processed to yield a satisfactory answer.
The data- driven analysis of text can be linked to distant reading, a term 
that was popularized by the Stanford Literary Lab. While distant reading 
sometimes relies on computational techniques, many influential distant 
reading projects rely on systematic reading by human annotators. The 
opposite of distant reading is not “reading,” but a literary history pre-
mised on seemingly haphazard anecdotes. Distant reading is mostly con-
cerned with the systematic analysis of a wide sample of literary texts. In 
that sense it is less directly linked to the history of computational methods 
and closer to the tenets of the social sciences. This point is emphasized by 
Underwood’s (2017) intellectual history of distant reading, which includes 
the work of Janice Radway (1991), whose analysis of romance novels is 
premised on conversations with a sample of readers. Underwood’s own 
work also includes many instances of systematic reading in combination 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
82 •  TheaTer as DaTa
with computational techniques (for example, Underwood 2017), which he 
also labels as distant reading. Tracing similar ideas between the social sci-
ences and distant reading is not the only genealogy possible, and distant 
reading has also been linked to book history (Bode 2012). Distant read-
ing is data- driven in the sense that it is premised on systematically gath-
ered data (even if the methods of collection or analysis are not digital). 
Matthew Jockers’s (2013) macroanalysis is more explicitly modeled on the 
quantitative analysis of digitally collected data. He doesn’t dismiss the 
value of actual reading, but sees macroanalysis as an eminently quantita-
tive approach that seeks to achieve a different objective: “we might think 
about interpretive close readings as corresponding to microeconomics, 
whereas quantitative distant reading corresponds to macroeconomics” 
(25). Macroanalysis, in contrast to macroreading (systematic reading), 
belongs only to the realm of computers. It can be used to explore a variety 
of questions, such as “the historical place of individual texts, authors, and 
genres in relation to a larger literary context” or the ways literary themes 
wax and wane over time (27). Both concepts— macroanalysis and distant 
reading— advocate for looking at literature from afar, and from consider-
ing a larger volume of texts that what is commonly used for conventional 
literary analysis. Macroanalysis and distant reading are also useful for 
the study of a range of theatrical texts, from dramatic literature to critical 
responses.
Data- Assisted Text Analysis
The potential for the speculative, interpretive, and situated analysis of 
words has been articulated in several influential theoretical perspectives. 
Stephen Ramsay (2003; 2007; 2011) has consistently argued that literary 
critics can reproduce their procedures on a computer, an approach he has 
named algorithmic criticism. This approach makes the steps of criticism, 
which are usually hidden, available for scrutiny and reproduction. Fol-
lowing Wittgenstein, he compares literary method to “a ladder that is dis-
carded after one has used it to climb up” (Ramsay 2003, 171). Algorithmic 
criticism is the digital examination of this ladder, as it aims to reproduce 
on a computer the operations that a literary critic would carry out in their 
analysis. Unlike the purveyors of data-d riven analysis, Ramsay is not inter-
ested in settling questions. Rather, his aim is to ensure that discussion of 
relevant literary works continues into grater depths.
Another influential data- assisted approach, which has also been 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Words as Data • 83
alluded to earlier, is Rockwell and Sinclair’s (2016) Hermeneutica, which are 
defined as interactive, interpretive toys that can be embedded into digital 
essays. Rockwell and Sinclair are interested in how interactive visualiza-
tions can “add to a history of interpretation” rather than offer definitive 
answers to closed questions. Their web portal Voyant (http://voyant-tools.
org) enables researchers to easily upload and visualize texts, creating 
interactive charts that can be embedded in other websites. By embedding 
a Voyant visualization in an intermedial essay, an author can exhort read-
ers to explore multiple perspectives, multiple scales, and thick context 
(see chapter 2). An excellent example of these interactive affordances, 
included in Rockwell and Sinclair’s companion website, is the analysis of 
two speeches on race, one by Barack Obama and one by Pastor Jeremiah 
Wright. Readers can find alternative ways to visualize the data, read the 
speeches in full and ask other questions of the same dataset used by the 
authors.
Voyant is a favorite of many researchers and is often used in the context 
of education (I, too, use it for my Intro to Digital Humanities course, as 
well as for the research project described in this chapter’s excursion). The 
ease of use is one of the appeals: the platform automatically creates a set 
of visualizations and statistics. The visualizations and statistics are mostly 
inspired by corpus measurements, such as keywords in context (KWIC), 
collocate analysis, word clouds, and a time series visualization of terms 
in a corpus. The interface is customizable and enables users to select their 
own stop words (i.e., function words such as “the” that should be omit-
ted). New features and tools are constantly being added, and at the time 
of writing this book, the portal has started offering maps and topic mod-
eling. While the portal is excellent for the analysis of English- language 
texts, it’s assumptions about word boundaries delimited by spaces and 
consistent spelling make work difficult in languages where these charac-
teristics are not found, such as in some South Asian languages (Battacha-
ryya 2018). This is by no means a problem endemic to this platform, and 
digital text analysis is plagued by similar problems. Any tool is built on 
cultural assumptions— and this should be taken into account when apply-
ing it to any object of study.
Other platforms are more explicit on the assumptions on which they 
are built. While Voyant Tools relies mostly on automated procedures, 
CATMA (computer-a ssisted text markup and analysis; Meister et al. 2016), 
at http://catma.de enables users to manually or semi- automatically encode 
features of interest, rather than relying on its predefined algorithms. It 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
84 •  TheaTer as DaTa
thus places greater emphasis on interpretation as a provisional and situ-
ated process and also enables collaboration, as different users can create 
their own usernames and collaborate in a project. This can highlight dif-
ferences in interpretation across individuals, contributing to the multi-
plicity of perspectives central to data-a ssisted research. Both Voyant and 
CATMA enable generative and iterative explorations of texts, and can 
help scholars resituate questions of interest, rather than offer definitive 
answers.
As I have argued throughout this book, the same methods can be used 
within different methodologies. Data-a ssisted analysis might rely on delta 
scores, topic models, and time series as much as data- driven analysis, but 
it is premised on a different perspective, one that seeks to ask questions 
in new ways rather than to find better answers to previous questions. For 
example, the inquiries in Literary Detective Work on the Computer (Oakes 2014) 
are framed as closed questions on plagiarism detection, style, authorship, 
and decipherment. These are explicit questions that use straightforward 
methods, even when the answers might be inconclusive (as is the case in 
much scientific work). Is x the author of y? Does genre matter more than 
personal style for the vocabulary choice of a given set of authors? What 
does a given set of graphic characters mean in an unknown language or 
secret code? These results require interpretation, but the quantitative 
methods used (for example, the aforementioned PCA to distinguish words 
that vary among authors) are aimed at trying to establish the answer to the 
question in empirical, replicable, and verifiable means.
This is different from the kinds of examples one finds in the previously 
mentioned Hermeneutica (2016). In the chapter “The Swallow Flies Swiftly 
Through: An Analysis of Humanist,” the authors look at the archive of 
Humanist, the online discussion group led by Willard McCarty that was 
central to the DH community at the time Hermeneutica was written. Rock-
well and Sinclair were interested in analyzing how Humanist changed 
over the years. Several graphs showed changes in lexical choices over 
time (from 1987 to 2008), to indicate that the frequency of phrases such 
as “humanities computing,” “computing in the humanities,” and “digi-
tal humanities” had varied. The latter had been clearly rising in the years 
leading up to 2016. The authors also showed that certain disciplines, such 
as “Visual and Creative Arts” were in an upward trend, whereas “Classics” 
were becoming less often discussed. In their analysis, they used methods 
also espoused by Oakes, but the way the questions were posed is differ-
ent. Rockwell and Sinclair were interested in how the discussion list had 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Words as Data • 85
changed. We find the same exploratory approach and how questions in their 
analysis of Hume’s Dialogues, discursive changes in the Game Studies jour-
nal, and in the analysis of race in the two political speeches mentioned 
earlier. In all cases, the methodology is an exploratory analysis of a given 
corpus. In contrast, in the analyses carried out by Oakes, the methodology 
is less exploratory and less interested in helping answer how questions, 
and more focused in answering whether two things are similar or differ-
ent: are two texts similar enough to warrant the conclusion of identical 
authorship? As Ramsay (2007) notes, there is disagreement in the closed 
questions posed by science, but the assumption is that there is a singu-
lar answer to a given problem. In contrast “literary criticism has no such 
assumption. In the humanities, the fecundity of any particular discussion 
is often judged precisely by the degree to which it offers ramified solutions 
to the problem at hand” (489).
Many researchers, though, weld data- driven inquiries together with 
data- assisted interpretation. Martin Paul Eve (2019), for example, devel-
oped a “computational microscope” to study a single novel: David Mitch-
ell’s Cloud Atlas (2004). There are many reasons this choice of novel is fas-
cinating: differences in editorial intervention between a U.S. and a U.K. 
version, as well as the way the novel crosses genres and historical settings. 
Also, due to copyright restrictions, Eve had to retype the entire novel in its 
different versions, and this made him all the more aware of minute differ-
ences that he could then try to corroborate or expand through computa-
tional techniques.
Theatrical problems come in multiple guises, some of which will 
require closed questions to be settled computationally, whereas others 
will be better served by systems that aid interpretation. Let’s imagine that 
we have a corpus of reviews written by theater critics in a given place (as 
explored in my excursion at the end of this chapter). Different questions 
could be asked about the same corpus. An example of a closed question 
could be: is there a difference in the way experimental and commercial 
productions are described? A question that could be better answered 
through an interpretive analysis would be: how has theater criticism in 
this place changed over time? In both cases, we might be using the same 
data, and some methods could be used to answer both types of questions. 
But the crucial difference is on how the question is framed, and how the 
answers are treated.
Two decades before the publication of this book, a framework for 
using corpus tools to study theater reviews was proposed by Roberts and 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
86 •  TheaTer as DaTa
Woodman (1998). The article described the creation and analysis of a pre-
liminary corpus of British theater reviews. Their database and software 
seem rudimentary from today’s standpoint, but the authors identify an 
area of great promise which has sadly not been taken up by theater schol-
ars, or by corpus linguists. They suggest, for example, that “world” is a 
more frequent term in theater reviews than in common English language 
usage. The prominence of “world” might suggest a “coherent directo-
rial or design concept which is almost always, in the texts used for this 
project, implicitly naturalistic in orientation” (12). Whether this particular 
insight is of relevance or not, it seems clear that theater reviews constitute 
an extraordinary data source that could be easily analyzed through com-
putational techniques. The work of Roberts and Woodman has lingered 
in silence for two decades, and I could not find other research teams fol-
lowing in their footsteps. But perhaps the wider interest in corpus analysis 
spurred by DH will mark a new beginning for the computational explora-
tion of theater reviews.
In this chapter, we’ve seen corpus tools applied to study British theater 
reviews, topic models deployed to study genre in French drama, and text 
reuse techniques mobilized to identify variations in German versions of 
Othello. Given the variety of methods commonly applied to the study of 
textual phenomena, existing theater examples are merely scratching the 
surface of a wide range of possibilities. We can further mobilize this pleth-
ora of techniques to the study of dramatic texts but also extend their reach 
to the analysis of program booklets, tweets, and other media reactions to 
theater performances.
These approaches are particularly useful for texts do not lend them-
selves well to sequential reading because of their disjoint structure. Take, 
for example, marginalia. A series of articles in Leviathan: A Journal of Mel-
ville Studies (2008) used digital text analysis to analyze the marginalia writ-
ten by Melville in the works of Shakespeare, Milton, and Homer. The mar-
ginalia are not beyond the scale where a single reader could easily read 
everything, but their disjoint structure means that it is hard to detect pat-
terns in them. Digital text analysis revealed that Melville annotated Shake-
speare’s work much more than he did other writers, and that the passages 
annotated in Shakespeare tended to contain words with low frequency 
within Shakespeare’s lexicon (Ohge et al. 2018). These two results can 
help draw conjectures about the reading practices, and perhaps the cre-
ative processes of Melville. New insight can similarly be drawn from the 
analysis of production notes or casting calls. It is perhaps in those texts 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Words as Data • 87
in the sidelines that the greatest promise for digital text analysis can be 
found for theater research. We can also apply digital textual methods to 
study performance scholarship itself, as Rachel Fensham (2019) has done 
for dance scholarship. As the most mature area of DH, more digital plat-
forms, step- by- step books and easy to use software packages exist for the 
computational analysis of words than for any of the other area surveyed in 
this book. Our work is cut out for us.
Excursion: The Flying Inkpot
The Flying Inkpot was a volunteer- run website which published theater 
reviews from 1996 to 2015 in Singapore. The bulk of reviews dealt with 
theater and dance, but the project also included some reviews of poetry 
and other art forms in the early years. A branch project dedicated to classi-
cal music still exists (this is excluded from present consideration). All the 
reviews in the site were submitted on a voluntary basis by a core group of 
reviewers. The website itself was also maintained by volunteers, a remark-
able feat for a project of this extent and longevity. This project was particu-
larly relevant in Singapore, where an active theater review culture didn’t 
previously exist within newspapers or other periodicals (but currently 
some newspapers have regular theater criticism columns). The Flying Ink-
pot was the brainchild of Matthew Lyon and Kenneth Kwok. Although the 
project is no longer active, an online archive maintained by Centre 42 pro-
vides access to the reviews for historical research (https://inkpotreviews.
com/). The theater and dance section of this archive includes 1,154 reviews, 
written by 65 reviewers. The reviews include the time and date of the per-
formance reviewed, the reviewer’s name, a rating (in a scale of zero to five 
stars), and a commentary which is typically a few paragraphs long. The 
archive also includes shorter reviews, which are filed under “First Impres-
sions” (these reviews were excluded from the current analysis). The shows 
reviewed include a mixture of local and touring productions. The reviews 
are extensive but, as they relied on the time and interest of reviewers, they 
don’t constitute a comprehensive survey of Singapore- based theater. That 
being said, the 1,064,854 words written by this small army of reviewers 
are in themselves an important cultural object, and one that can be inves-
tigated by historians of Singapore-b ased theater in a number of ways. For 
the analysis below, I excluded the reviews from the first two years, since 
very few were written at that point in time.
Is there any identifiable change in the vocabulary usage of The Flying 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
88 •  TheaTer as DaTa
Inkpot reviewers over time? To answer this question, I first uploaded the 
corpus to Voyant and selected the hundred most common words in the 
corpus, after removing articles (I used the default stopword list from Voy-
ant). I then downloaded the trend data and further analyzed it in Python 
(see appendix A for details and visit the publisher’s website to download 
the data and code). I calculated the Mann-K endall s statistic (Mann 1945; 
Kendall 1975; Gilbert 1987) on the trend data downloaded from Voyant. 
The Mann- Kendall s provides a statistical estimation on whether a trend 
is monotonic (i.e., consistently increasing or decreasing over time). The 
sign indicates the direction of the trend, an s of 103 and −103 are equally 
strong, but the former is increasing and the latter decreasing. The Mann- 
Kendall s has been previously used to detect changes in vocabulary trends 
on Twitter (Malakar et al. 2018). I calculated the Mann-K endall s for the 
top 100 words, and then selected only the strongest trends, provided the 
p- value was lower than 0.05. The top five words with the clearest trends are 
shown in table 4.1 (upward) and 4.2 (downward). They are also graphed 
in figures 4.1 and 4.2, respectively.
I then downloaded the concordances for these words; that is, all the 
sentences where each of these words were used and annotated them by 
hand, pushing the analysis into the data-a ssisted realm. The annotation 
by hand aimed to be systematic, but as a situated and provisional process, 
it offers a different perspective on the monotonic trends and closely exam-
ines assumptions that might be hidden in the statistical analysis. Close 
Table 4.1. Top five words with the strongest markers of an upward trend
Word Mann-K endall s Counts in the corpus p- value
feel 103 549 0.000112
makes 101 460 0.000152
great 87 510 0.001124
light 83 461 0.001897
work 81 1,470 0.002444
Table 4.2. Top five words with the strongest markers of a downward trend
Word Mann- Kendall s Counts in the corpus p- value
woman –8 7 595 0.001124
script – 75 845 0.005064
real – 71 510 0.008015
night – 61 520 0.023047
audience – 57 2,463 0.033909
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Fig. 4.1. Top five words with the strongest markers of an upward trend.
Fig. 4.2. Top five words with the strongest markers of a downward trend.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
90 •  TheaTer as DaTa
annotation of each datapoint incorporates thick context and shifts atten-
tion from the aggregate to the singular sentence. This process revealed 
that, in the upward trend, “work” is most often used as a noun. A the-
ater performance was increasingly referred to as a work, as opposed to a 
“show” or a “performance” (these other words were slightly more com-
mon in the early years of The Flying Inkpot, see figure 4.3). But in 2015, 
the last year, “performance” became again slightly more common than 
“work.” The trends of these other words were not as strong and were not 
captured by the statistical analysis reported above.
When manually disambiguating the multiple senses of “light” (“illu-
mination,” “simple,” “funny”), its trend disappears. It is interesting to 
note that the verb “feel” (which also includes the conjugated forms “feels” 
and “felt,” but not derived nouns such as “feelings”) became a more com-
mon way for the reviewers to express their views. When it comes to the 
decreasing trends, “woman,” “script,” “real,” and “night” are slightly less 
common but they reveal a fascinating trend. One explanation is that pre-
occupation with scripted, nighttime theater that emulates reality dwindled 
over time. An alternative explanation is that, as the group of reviewers 
became more established and diverse, they started looking at different 
types of theater, and describing it with different words. In any case, identi-
fying the reasons for this trend warrants additional research. I also found 
the decreased usage of “woman” surprising, since I previously had the 
impression that feminist readings of performances became more com-
mon over time. I suspect that, as feminist concerns became more com-
mon, a more nuanced vocabulary for the description of gendered expe-
riences was developed and there was less need to directly use the word 
“woman.” However, I have not been able to prove this in terms of the data.
In the close reading of the downward trends, the most surprising 
insight came from reviewing instances of the word “audience” (which 
was overall more common than the other four words with strong down-
ward trends). As I read each of the sentences where this word was used, 
I noticed that most of the times this did not indicate a description of the 
audience (i.e., “the audience laughed”) but rather rhetoric usages (“pro-
voking the audience”), which constitute, in my opinion, an indirect way 
to phrase the reviewer’s own perspectives. I closely read all concordances 
for the word “audience,” and for each sentence indicated whether the 
usage was descriptive or rhetoric. This is a highly interpretive but system-
atic form of reading, that takes every sentence that uses the word “audi-
ence” as an individual case, and then groups them into categories. Other 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Fig. 4.3. “Works” compared to “performance” and “show.”
Fig. 4.4. Changes in the relative frequency of all mentions of “audience,” and trends in purely 
rhetoric usages of “audience.”
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
92 •  TheaTer as DaTa
scholars might disagree with my categories, or with the ways I classified 
individual sentences. For this reason, I am framing this part of my analy-
sis as a data- assisted strategy, and highlighting the situated aspects of my 
interpretive process. The classification of sentences into descriptive and 
rhetoric usages is not necessarily an area of low ambiguity (see the crite-
ria in the introduction). Therefore, interested readers are invited to down-
load my data from the publisher’s website and trace the interpretive move 
behind every data point. They can also reclassify my sentences according 
to other criteria and offer alternative interpretations of this trend.
Given my assumptions, the decrease in the usage of “audience” sug-
gests that as the group of critics that coalesced around The Flying Inkpot 
refined their vocabulary and approach, references to the audience became 
less important as rhetorical devices. Perhaps this indicates that critics 
became more confident of their own voices, encouraged by the prominent 
place that The Flying Inkpot quickly earned among Singapore’s theater cir-
cles. But proving this hypothesis would require additional research as one 
would need to identify other trends in the data that would confirm or dis-
prove this hypothesis. For example, we could find other rhetorical devices 
that also indicate more confident authorial stances, or manually annotate 
a sample of reviews and assign them an “assertiveness” score. It is also 
important to note that the trend in this word usage does not constitute an 
argument about the importance of the audience for Singapore- based the-
ater, but about how critics collectively chose to describe their impressions.
It could be argued that these changes in trends are not particularly spe-
cial. One could imagine that, as the number of reviews increased, many 
terms became less frequent or more frequent. However, looking at general 
trends reveals that the opposite is true. The frequencies of many words 
remained stable across time, as critics came and went, and as the number 
of reviews waxed and waned. Other words have no discernible patterns. 
This is true for terms that have been the focus of intense academic atten-
tion in relation to Singaporean theater (“text,” “stage,” “politics,” “state,” 
“censorship,” “queer,” “postcolonial,” “cultural,” “Chinese,” etc.). Close 
reading would be a better method to study how these concepts were used, 
how their usage responded to specific artistic events, and how they dif-
fer from the ways these terms are used in longer, academic pieces of writ-
ing. I want to emphasize that my results can’t be taken as representative 
of Singapore- based theater as a whole. But they offer a new glimpse into 
the changes in the vocabulary of an influential group of critics. Given the 
importance that The Flying Inkpot had for practitioners over its almost two 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Words as Data • 93
decades of existence, these conclusions still reveal patterns that would be 
easy to miss if one just read all (or a portion) of the reviews.
The approach I demonstrate in this excursion shares some similarities 
with the culturomics project inaugurated by Michel et al. (2011), but that 
study used an entire corpus of digitized texts. It is perhaps closer then to 
the approach of Rockwell and Sinclair, who analyzed changes in the fre-
quencies of words in the archives of Humanist and in the academic journal 
Game Studies. Although the identification of the trends in this excursion is 
driven by data, the results are then examined closely, and the conclusions 
are assisted, rather than purely driven, by data. This book advocates for 
the limited applicability of computational methodologies for studying 
theater performances, and related phenomena. The modest contribu-
tion of this excursion is showing that references to the “audience” in the 
reviews of The Flying Inkpot steadily decreased, and performances tended 
to be increasingly described as “works.” The scope of these observations 
is limited, but they can serve as a stepping stone for more comprehensive 
analysis of a key resource for the history of Singapore-b ased theater at the 
turn of the twenty- first century.
Code and data: Sample code for this chapter is available at  
https://doi.org/10.3998/mpub.11667458.cmp.39.  
The data used can be downloaded from  
https://doi.org/10.3998/mpub.11667458.cmp.26,  
https://doi.org/10.3998/mpub.11667458.cmp.27, and  
https://doi.org/10.3998/mpub.11667458.cmp.28.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
ChapTer 5
Relationships as Data
Theater depends on relationships. A theatrical production is impossible 
without the collaboration and co- presence of different people. By most 
definitions of theater, even the most minimal productions require at least 
a performer and one spectator. “As scholars of a collaborative art form, 
we are always dealing with relational data, for theater artists almost never 
work in isolation,” writes Caplan (2017). Even ritual puppet performances 
in Bali that don’t require a human audience are created by a collective of 
artists. In several dramatic traditions, characters and the relationships 
between them are also important. Relationships in these fictional and col-
laborative spaces can be modeled and analyzed as networks.
To see how a wide range of theater scripts and modes of collaboration 
can be represented and analyzed as networks, a short overview of network 
theory is needed. Following this overview, I will explain how network anal-
ysis can be used for data- driven and data- assisted theater research. This is 
complemented by an excursion into two of my own research projects: an 
analysis of collaborations in the Singaporean theater company The Neces-
sary Stage (TNS) and an analysis of character co- presence in the fictional 
universe of traditional wayang kulit (shadow- puppet theater) in Indonesia. 
This chapter aims to show that the toolkit of network analysis can reveal 
hidden structures in relationships, which has great potential for theater 
research. Data-d riven network analysis aims to show counterintuitive fea-
tures of dramatic literature and artistic collaborations, while data-a ssisted 
network analysis enables a playful defamiliarization of theater history.
Methods for Network Analysis
In theater history, Jacob Levy Moreno is more commonly remembered 
as the father of sociodrama, but he was also the inventor of sociograms: 
Escobar Varela, M9ig4uel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Relationships as Data • 95
visual representations of connections among people. He was one of the 
first researchers to realize that networks could depict social relation-
ships (Moreno 1960). However, the study of networks is older, as the 
first sketches of a theory of networks were proposed by Leonhard Euler 
in 1735, and they were inspired by the Seven Bridges of Königsberg in Saint 
Petersburg. This and many other key moments in the history of network 
theory are masterfully retold by physicist Albert-L ászló Barabási in Linked 
(2002), a nontechnical introduction to the study of networks. In his book, 
Barabási (who is himself part of the history he tells), describes many ways 
in which networks can be analyzed, as well as several practical applica-
tions of the study of networks. His book blends history and context with 
mathematical theory, and it begins, like many an introductory network 
book, by describing the basic components of a network. Networks are 
made of two basic elements: nodes (things that are linked) and edges (the 
links between them). In the study of literary fiction, characters are often 
represented as nodes. But the edges can take a variety of meanings: they 
can represent blood ties between characters, words exchanged between 
them, or co- presence in a scene. Algee- Hewitt (2017) notes that, for the 
analysis of drama, the unit of the network “is not the node, but the edge: it 
measures not characters, but interactions” (752).
Many of the concepts in network analysis come from the sociological 
study of networks (often termed social network analysis or SNA). Quanti-
tative analyses in this field often report measurements such as degree and 
betweenness (Knoke and Yang 2008). Degree indicates the extent to which a 
node is connected to other nodes. The higher the degree, the more con-
nections a node has. For example, Haresh Sharma (the main playwright 
of TNS), is the node with the highest degree. A slightly different measure-
ment is betweenness, which indicates the fraction of all shortest paths 
that pass through a given node. This gives a sense of the importance 
of a node to the network structure. For example, the character Karna in 
the wayang network has a very high betweenness (much higher than his 
degree). This means that although this character is not as well connected 
as others, he is often found in between many characters. Besides offer-
ing useful concepts such as this, SNA has also developed tools that are 
potentially useful for the study of theatrical networks. For example, there 
is abundant SNA literature on the construction of networks through inter-
views or surveys. These are called “egocentric networks” and they stand in 
opposition to “whole-n etworks.” Methods of SNA often estimate general 
properties of networks from just the subsection represented by egocentric 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
96 •  TheaTer as DaTa
ones (Marsden 2005, 8– 29). In computational social science, the purpose 
is estimating the range of variations in a network and determining how 
certain parameters such as gender or education level affect the forma-
tion of networks. A sociological study of theater networks could use this 
approach, but I am unaware of any study that uses such attempt.
In contrast, the study of networks in physics often focuses on general 
properties of the network structure, such as the distribution of degrees, 
the network diameter (the longest path between two nodes in the net-
work), and its clustering coefficient (the average connection between 
two nodes). Many complex networks (from scientific collaborations to 
protein interactions) have three properties: their clustering coefficient is 
very high, the diameter is small, and their degree distributions are expo-
nential (a few nodes have most of the connections, and most nodes have 
very few connections). The first two properties mean the network rep-
resents a “small world,” where even the least connected nodes are only 
a few nodes away from each other. The second property is often called 
“scale- free.” This lack of scale means that all values are expected since 
the distribution of the connections is very uneven (normal distributions, 
in contrast, cluster around an arithmetic mean). These properties have 
been identified in many networks and they are often reported as intel-
lectual curiosities. But these properties can also be used to analyze the 
robustness of a network (say a phone network) in case of accidental 
failure or deliberate attacks. This has many practical applications, such 
as helping engineers devise resilient communication networks or help 
epidemiologists assess the risk of contagion within populations. Later, 
I will describe a project that seeks to understand the meaning of small 
worlds in theater networks.
Of all the areas covered in part 2 of this book, network analysis is 
the one with the most standard set of measurements. Networks, as 
mathematical models, have well-e stablished quantitative properties 
that are straightforward to calculate, as seen above. This is not true 
to the same extent for texts, motion, images, and locations. As Algee- 
Hewitt (2017) notes, networks are “intuitive to grasp and yet mathe-
matically complex” (752). Data-d riven network analysis is well estab-
lished in the sciences and social sciences. But what might one glean 
from network overviews in the humanities? In DH, networks have been 
applied to study relationships between artists (collaborations) and 
interactions between fictional characters. Both of these are directly 
applicable to the study of theater.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Relationships as Data • 97
Network Analysis of Fictional Relationships  
and Artistic Production
Network analysis is an area where scientists have often applied quanti-
tative measurements to datasets from the humanities. Researchers have 
studied, for example, the fictional networks of Greek mythology (Choi 
and Kim 2007) and Marvel superheroes (Alberich, Miro-J ulia, and Ros-
selló 2002). Some studies have also used fictional networks as test cases 
for information segmentation and retrieval (S. B. Park, Oh, and Jo 2012; 
G. M. Park et al. 2013). Some researchers have turned their attention to 
collaborative networks in the arts, and a network analysis of jazz musi-
cians in the early twentieth century found clear evidence of racial segre-
gation (Gleiser and Danon 2003). These papers all present fascinating 
insights from a network science perspective, but they don’t try to con-
textualize their findings within humanities scholarship. Their priorities 
clearly lie elsewhere, and the lack of critical engagement is not a short-
coming in their fields. This is different in the examples that follow, which 
analyze fictional networks in drama, and collaborative networks in theater 
productions against more carefully considered historical and disciplinary 
contexts.
Moretti (2011) focused only on specific plays and his networks are 
therefore much smaller. He also built his networks by hand— the char-
acter nodes were not downloaded from a database or automatically 
extracted from a corpus. He reported some network measurements but 
did not dwell on their geometric and topological features. Still, his close 
analysis of networks reveals counterintuitive features of well-k nown texts. 
For example, his Hamlet network identifies a “region of death,” where 
all characters linked to both Claudius and Hamlet are killed, except for 
Osric and Horatio: “outside that region, no one dies in Hamlet [ . . . ] the 
tragedy is all there” (217). Moretti also found two separate components, 
sub- regions where each node is connected to every other node, which he 
interprets as the separate worlds of the court and the state (Moretti 2013, 
228– 30).
Algee- Hewitt (2017) takes this line of inquiry further, by analyzing 
network properties of 3,568 English dramatic texts written between 1550 
and 1900 retrieved from the ProQuest Literature Online drama corpus. He 
developed his own code to identify the most probable recipient for each 
speech using a rule-b ased system. Speech was represented as directed 
edges that connect speakers to addressees. He then computed several 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
98 •  TheaTer as DaTa
properties for the nodes. He presents a convincing case for taking eigen-
vector centrality (EC) and betweenness centrality (BC) as the most inter-
esting measurements. EC is a measurement that computes the importance 
of a node by calculating its total number of connections, considering the 
relative importance of the nodes to which it is connected. As noted above, 
BC looks at characters that mediate between factions. Algee-H ewitt con-
vincingly demonstrates that BC can be used to estimate protagonism. 
Rather than merely identifying a protagonist, the distribution of BC across 
characters in a play shows how protagonism is shared among characters, 
and how this changed over time, as plays were increasingly “less likely to 
feature a single central character (or cluster of characters) and more likely 
to depend on densely connected communities each featuring prominent 
characters in their own right: they are more likely to look like A Midsummer 
Night’s Dream than Henry V” (Algee-H ewitt 2017, 765). Similarly, BC shows 
what he calls “mediatedness.” Taking both measurements together, he 
identified plays where the protagonist is the mediator and those where 
mediatedness and protagonism are shared by different characters.
This line of analysis—t racking the history of drama through changes 
in network-t heoretical measurements—h as also been systematically 
explored by the Digital Literary Network Analysis (DLINA) group, led by 
Frank Fischer and Peer Trilcke. Over the span of several years, they have 
used different facets of network analysis to study a large corpus of Ger-
man drama. First, they measured how some basic measurements, such as 
network density, changed over time (Trilcke et al. 2015). This made them 
realize that there are consistent markers of genre over the two centuries 
of data which they processed. The average density (the number of actual 
edges divided by the total number of possible edges) for tragedy remained 
constant, as was the case for comedy and libretto (a category where the 
authors include all musical theater).
In another project, these researchers wondered whether theater 
networks were small worlds (Trilcke et al. 2016). As seen earlier in this 
chapter, small worlds are common in nature and in social environments. 
The DLINA group looked for three markers of small worlds in their cor-
pus: a high clustering coefficient, small average path lengths, and scale- 
free degree distributions. They found that only five plays (in a corpus of 
almost five-h undred plays) had small world properties, and this led them 
to pose questions of interest to the study of drama: what does it mean to 
have (or not have) small world properties? Although this is still an open 
question, this line of analysis sets this kind of projects apart from the ones 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Relationships as Data • 99
developed entirely by engineering and scientific teams. For engineers and 
physicists, it is merely interesting to note the presence or absence of cer-
tain quantitative markers in cultural data. But for theater researchers, the 
important questions are what these quantitative features mean for the his-
tory of our field.
On a subsequent project, the DLINA group focused on network dynam-
ics (Fischer et al. 2017). Most of the other network approaches mentioned 
so far “spatialize” the flow of the drama into a single diagram, and thus 
remove the dimension of time. But one could also track the evolution of 
networks as they change. For this purpose, the DLINA group proposed 
a series of time-o riented measurements: all- in index, central-c haracter- entry 
index, final-s cene- size measure, and the drama- change rate. These measure-
ments help identify outliers in drama history, such as extreme changes 
(“provocation”) or extreme uniformity (“boredom”). In the development 
of this project, they were inspired by IntNetViz (Xanthos et al. 2016), an 
interactive tool for visualizing drama as it unfolds.
Yet another of the DLINA group’s projects combined several quantita-
tive measurements to automatically identify the protagonist in a drama. 
They looked at both network measurements (degree, closeness, between-
ness, weighted degree, eigenvector) and word counts (words spoken, 
speech acts, frequency). They found that multidimensionality (i.e., com-
bining different measurements) is the best solution. This project is remi-
niscent of work by another group, which tried to identify the most likely 
romantic couple in a dramatic text (Karsdorp et al. 2015) using a corpus of 
French drama. Also relying on a combination of network and other mea-
sures, they achieved very high accuracy, and were able to identify roman-
tic couples in 80 percent of the cases. But, as they note, their algorithm 
is premised on the assumption that there is one romantic couple in each 
text. A harder problem would be identifying whether a romantic couple is 
to be found in the text (or how many such couples there are).
The DLINA group has announced plans to extend their work to Eng-
lish, French, and Russian dramatic texts and they have developed an 
impressive portal with detailed analysis, network visualizations, and an 
application programming interface (API), to automatically retrieve data 
queries (Trilcke and Fischer n.d.). The tools they propose are excellently 
matched to text-b ased drama, especially when such dramas are encoded 
in TEI formats that are easy to consume and reuse (see chapter 4). DLINA 
is leading the way by combining literary history and criticism with quanti-
tative analysis. If more dramatic texts from around the world were readily 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
100 •  TheaTer as DaTa
available for network analysis, this would open the door for comparative 
analysis across theatrical cultures.
So far, the projects surveyed have focused on dramatic texts. But 
another tantalizing possibility for theater history is the analysis of col-
laborations among theater artists. In an influential paper, Uzzi and Spiro 
(2005) analyzed the small world network of artists involved in Broadway 
musicals from 1945 to 1981. They found that the small world properties of 
the artists’ networks affected their financial and artistic success. To esti-
mate the latter, they used an index created by Suskin (1990), who manually 
assigned a numerical value to 315 productions.
AusStage (AusStage 2013) has amassed an impressive dataset on the-
ater performances in Australia, and theater performances by Australian 
performers around the world. This type of dataset paves the way for new 
theater histories. As Caplan (2017) notes “data- driven theater history, at 
its best, can reveal previously invisible patterns about relationships among 
diverse groups of artists working across languages and cultures” (557). 
Caplan’s own data consists of 290 Yiddish theater artists who worked 
on at least one Vilna Troupe production between 1915 and 1936. Using 
this data, she was able to identify ten previously distinct branches of the 
troupe, a fascinating and previously unreported finding. She suggests that 
this data can also be used to trace how gender influenced the formation 
of friendships and professional networks, or to estimate the influence of 
familiar ties— and of pedagogical figures— on collaborations.
Caplan wonders what collectives of scholars might be able to accom-
plish if they worked together to assemble large datasets on theater col-
laborations. In part, that potential has already been demonstrated by Aus-
Stage, and by projects that build on their data model, such as IbsenStage, 
which is a collectively assembled dataset of performances of Ibsen texts 
around the world. This comprehensive dataset, where the completeness is 
estimated at 60 percent, was the backbone of A Global Doll’s House (Holledge 
et al. 2016). Using network visualizations together with other analytical 
techniques, Holledge and her coauthors showed that the global success of 
the play was linked to genealogies of transmission. One such visualization 
shows how a long chain connects productions from the late nineteenth 
century until the 1990s. In this directed network, edges indicate that at 
least one actor from a given production was involved in another more 
recent performance (more formally, this technique is called a unipartite 
projection). Thus, rather than showing co- presence, as other collaborative 
networks do, this visualization shows movement of actors across different 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Relationships as Data • 101
temporal layers of a production’s history. Their work effectively demon-
strates that it is not only ideas, but specific people, that ushered A Doll’s 
House into global fame. Their network analysis also shows the decisive 
influence of women in the production and promotion of the play—r ather 
than merely in the portrayal of the protagonist—a nother unexpected and 
important finding. Bardiot (2018) also used network analysis to study the 
collaborative networks of Merce Cunningham to show that after 1954, his 
working strategy brought a larger number of artists in collaboration with 
each other, shifting from a “star” to a “spiral” pattern of collaboration.
There are many other possibilities for network analysis in theater, and 
the more datasets are available for reuse, the more likely we will see net-
work analysis in all areas of theater research. For example, one study mod-
eled spectators’ choices to attend a theater show in Belgium as networks, 
and tried to find the probability of people sticking to specific venues, esti-
mating how much knowing another spectator influenced such decisions 
(Agneessens, Roose, and Waege 2004). As theater researchers continue 
to apply network measurements in their work, they can draw inspiration 
from the work of Maximillian Schich, who has systematically used net-
work analysis to study a wide range of topics central to art history. His 
work weds sophisticated quantitative analysis with thorough historical 
analysis (Schich, Lehmann, and Park 2008; Schich et al. 2014, 2017).
Data- Driven and Data- Assisted Network Analysis
The approaches reviewed so far vary in the level of mathematical for-
mality used in the reporting of findings. However, most of them could 
be described as data-d riven projects. This is often made explicit. Caplan 
(2017), for instance says that network analysis and other modes of data 
visualization “can offer an important corrective to our understanding of 
what is central and what is peripheral in theater history” (557). She thus 
urges theater scholars to engage in verifiable, consensus- driven studies of 
theater history premised on data. Holledge et al. categorically situate their 
work within a scientific paradigm. As described above, their interest is to 
study the global history of A Doll’s House. They used network visualizations 
to identify the forces that led to the spread of the play across geographies 
and generations. The authors compare the analysis of the play’s history to 
the scientific study of evolution. Using the zero- force evolutionary law pro-
posed by philosopher Robert Brandon and paleobiologist Daniel McShea, 
they posit that cultural phenomena—s uch as A Doll’s House— tend towards 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
102 •  TheaTer as DaTa
“increasing diversity and complexity” until they find constraints. Natural 
selection is a constraint in biological systems, and “underlying political, 
social, economic, aesthetic, or technological forces” are constraints that 
shape the spread of theater plays (Holledge et al. 2016, 18–1 9).
Some projects also use data- assisted network analysis to enable more 
situated and context-r ich perspectives. Interactive visualizations, like 
Caplan’s digital portal for the Vilna Troupe network (http://vilnatroupe.
com/) and Xanthos et al.’s IntNetViz (https://github.com/maladesimagi-
naires/intnetviz) encourage users to zoom into portions of the data and 
construct their own close analysis of texts and histories, in ways that depart 
from what is purely contained in the data. The interactive features enable 
multiple perspectives, different scales, and thick context (see chapter 3). 
When it comes to aesthetic provocations, the best example is Brecht Beats 
Shakespeare! A Card- Game Introduction to the Network Analysis of European Drama 
(Hechtl et al. 2018). This is a card game set where each card has a network 
visualization and measurements that correspond to different theater plays. 
The players try to win an opponent’s card with a higher value—b ut they 
must agree which network value to use for the game. The game relies on 
networks to produce a playful defamiliarization of theater history. It is 
worth noting that most of these data-a ssisted perspectives are based on 
data- driven projects. Caplan’s visualization is tied to a replicable analysis 
of the Vilna Troupe’s history. Brecht Beats Shakespeare! uses data from Fisher 
et al.’s data on European drama. But inspiration works on both directions: 
data- driven work can inform data- assisted projects and vice versa. Fisher et 
al.’s work, as noted earlier, also derived inspiration from the data- assisted 
IntNetViz for their analysis of network dynamics over time.
First Excursion: Collaboration Dynamics of The Necessary Stage
The first excursion in this chapter shows how network analysis can be 
used to study collaborative relationships among people involved in the-
ater using data from the Singaporean theater company The Necessary 
Stage (TNS). I obtained the data as a spreadsheet (kindly provided by the 
theater company), which listed all the people involved in all their produc-
tions between 1987 to 2015. My research assistant Alyssa Chandra and I 
then went over all the records to standardize the names of the performers. 
Founded in 1987, TNS is one of the most active theater companies in Singa-
pore with over four hundred productions to date. Through their outreach 
programs and sub-c ompanies like Theater for Seniors (TFS) they have 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Relationships as Data • 103
built strong ties to different communities. The most prominent members 
are playwright Haresh Sharma and director Alvin Tan, but the company 
has also served as an incubator for many other companies. Prominent 
artists such as Kok Heng Leun (Dramabox) and Chong Tze Chien (The 
Finger Players), have at some point been involved with TNS, and the com-
pany has actively sought to promote new playwrights and directors. This 
double focus— social involvement and talent incubation— gives the TNS 
network its unique structure. However, it is important to bear in mind that 
the company’s structure has changed over time and so has the meaning of 
“collaboration.” I will describe these two changes before analyzing some 
quantitative features of their collaborative network.
The people involved in the company have changed drastically over 
time, and only Tan has remained since the beginning, but the company 
has retained a distinctive mission. Artists and commentators describe the 
individual performances as part of a longer creative process: “whether 
you like their work or not, its significance outlives every single production 
they stage” (Birch 2004, 58). At the time of writing, TNS commissions an 
average of two performances per year, organizes an annual festival, have 
their own theater venue, and employ a full- time staff of around ten people. 
However, the company started out as a student collective and only eventu-
ally grew into the stable, professional institution it is today. Although they 
now are one of the most respected theater companies in Singapore, their 
political agenda and choice of artistic forms have created trouble for them 
in the past (K. P. Tan 2013).
According to Alvin Tan (2004, 253), in the beginning the artistic com-
mittee had to do administrative and production work to fight the “star-
dom syndrome” but groups were divided and some people left. This 
observation will be specifically important for the network analysis that 
follows. In 1992, only four people were working full time. One of the 
most significant structural changes took place in the year 2000, when they 
moved from their offices in the city center to the Marine Parade neighbor-
hood in the eastern part of the city- state, a change that propelled a dif-
ferent kind of community engagement. In 2002, TNS established a lab to 
encourage more collaborations within group members. The people who 
have worked with TNS—t he nodes in their impressive network—i nclude 
famous actors, producers, academics in different fields, and members of 
parliament. These collaborations have taken many shapes, but they have 
retained a strong social and political objective. Alvin Tan (2004) asserts 
that “for TNS, collaboration is a methodology of resistance, resisting the 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
104 •  TheaTer as DaTa
rationalized mindsets ingrained in cultures of both contemporary urban 
lifestyles and the production structure of the traditional Western model 
theater” (266). Artists affiliated with TNS often reflect on the meaning of 
collaborations, and the term appears extensively in the edited volumes 
they commissioned for their tenth and seventeenth anniversaries (Tan and 
Ng 2004; Krishan 1997).
Haresh Sharma is credited as the playwright of most TNS plays, but his 
writing is often undertaken in collaboration with the performers. Accord-
ing to Tan, the early phase of collaboration at TNS was mostly textual, 
where company members developed a working method based on impro-
visation, writing, revision, and rehearsal as distinct creative stages. Tan 
admits this way of working was still heavily text-c entric and, after the first 
ten years of the company, he wanted to find more adventurous collabora-
tive practices. A wider range of collaborative strategies can be observed in 
several productions. Off Centre (1993), a play about mental illness, required 
extensive research, interviews, improvisation, and writing. The research 
process was even more intensive in productions where the actual per-
formers had personal experiences related to the subject of the plays. For 
example, ’Scuse Me While I Kiss the Sky (1994), was based on the experiences 
of, and performed by, people who had attempted suicide. Similarly, October 
(1996), was also devised with, and performed by a group of elderly people. 
According to David Birch (2004, 61), their most collaborative work is Com-
pletely With/Out Character (1999), the result of working with the late Paddy 
Chew, the first Singaporean to publicly announce his HIV positive status. 
According to Wong (1997), these plays demonstrate that the actors “have 
not only become the subject of both play and interview, they have become 
more pro-a ctive collaborators in determining the shape of the play as their 
experiences, improvisation work and interaction with one another are as 
much part of the process of play-m aking as they are events enacted for the 
stage” (195). This background should be considered when interpreting 
the network analysis below, where people are modeled as nodes and col-
laborations are modeled as undirected edges between every person who 
collaborated in a single play.
Before delving into the analysis of the TNS data, let’s consider the first 
play of a fictional theater troupe by way of example. Let’s say four people 
collaborate in this play, and there is one edge between each of them. There 
are four nodes and six undirected edges. Thus far, the network can be said 
to have a single connected component. If, for a second production, two of the 
original people and two new members are involved, then the shape of the 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Relationships as Data • 105
network will change. The number of nodes will increase from four to six 
and the edges will increase to eleven. But there will still be one single con-
nected component (as the two people involved in both productions are 
linked to the other four who have not collaborated with each other). Now 
let’s imagine a third production, also produced under the banner of the 
same troupe but with three entirely different people. Now the node num-
ber will have increased to nine, and the edge number to twelve. But we will 
observe two connected components, as there is no link between the mak-
ers of the first two productions and the group responsible for the third.
Figure 5.1 shows the staggering increase in the number of connected 
components within the collaborations of TNS. For this analysis, I omit-
ted the names of the directors and playwrights (who are stable across 
many productions and would thus link the network into a giant connected 
component). This visualization shows that TNS was so consistent in its 
recruitment of new people that at some point twelve connected compo-
nents are discernible. But notice also that the line goes up and down. The 
company often involves entirely new people, but then it draws from these 
distinct groups for new productions, bringing the number down. This 
ebb and flow illustrates the principles of collaborations seen in the his-
tory above: the radical pursuit of new collaboration strategies and the con-
stant desire for social engagement by enlisting non-a ctors as performers. 
It is interesting to note that the last recorded number of components (11) 
remained stable since 2002 (when, as seen above, TNS established a lab to 
encourage more collaborations within group members).
We can further drill down into these results and separate the networks 
of people listed in the credits as cast from those listed as part of the pro-
duction crew. Figure 5.2 shows that the number of components for cast 
members at one point soared to 25. In contrast, figure 5.3 shows the more 
closely knit world of the production crews, whose connected components 
have never exceeded 5. This could indicate there are fewer people work-
ing in production roles. But, in fact, as figure 5.4 indicates, there are— 
and have always been— more people working in the production side. The 
number of people (nodes) in this network increased at a far greater pace. 
The explanation is that the nature of the collaborations in the production 
side are different— figure 5.5 shows that edges grow at a greater pace in 
the production network.
These data-d riven results add details to the history of the company, 
and help explain their position in the Singaporean theater landscape. 
However, this bird’s eye overview should be read against a more situated 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Fig. 5.1. The number of connected components over time.
Fig. 5.2. The number of connected components over time for cast members.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Fig. 5.3. The number of connected components over time for production crew members.
Fig. 5.4. The number of nodes over time separated by cast and production crew.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
108 •  TheaTer as DaTa
Fig. 5.5. The number of edges over time separated by cast and production crew.
reading of TNS’s history, which shows that their understanding of collab-
orative work has changed drastically over time.
Second Excursion: Networks of Wayang Narratives
This excursion considers the relationships between characters in a fic-
tional universe. There are different kinds of wayang theater in Indonesia 
and the most notable one is wayang kulit, where shadow puppets made 
of water buffalo hide are controlled by a single puppeteer in all- night per-
formances. Other forms include wayang wong (where human performers 
imitate the leather puppets) and wayang golek (three- dimensional puppets 
made of wood). Most forms derive their narrative materials from the same 
sources, which include the epics Mahabharata and Ramayana. The former 
is often preferred for all- night puppet shows, where stories of feuding 
families, conspiracy, and betrayal are explored through extensive dialogue 
and philosophical explication. The Ramayana is more commonly used in 
dance, as its straightforward plot lends itself nicely to voiceless, visual 
drama (see chapter 6). Wayang theater dates at least to the ninth century 
CE, and the stories can be traced much farther back in South Asia (Esco-
bar Varela 2017). In what follows, I focus on the Javanese versions of way-
ang kulit, and I refer to characters and stories by their Javanese spellings 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Relationships as Data • 109
(which might sound slightly odd to those more familiar with the South 
Asian versions of the stories). Following conventions in English- language 
scholarship on wayang I add an “s” to Javanese nouns to indicate plural 
forms, so I refer to dhalangs as the plural of dhalang (this, however, would 
not be correct in Javanese, which uses different grammatical conventions 
to indicate plural nominal forms).
The Mahabharata narrates the story of two groups of cousins, the Pan-
dawas and the Korawas, who fight for the throne of Astina. The Pandawas 
are the rightful heirs, but the characters of both sides are often depicted 
in nuanced moral tones, offering the dhalangs ample opportunity for 
moral reflection. In Java, the Mahabharata is never performed in full and 
an entire all- night performance concentrates on a single episode. For 
example, a favorite lakon (story) is Dewa Ruci, the spiritual quest of Bima, 
the second Pandawa brother. This episode happens before the Great War 
(Bharatayudha) but does not fit into a specific chronology of the epic. Like 
most of the famous episodes, this is a Javanese addition to the South Asian 
versions. In fact, many key features of the Mahabharata as performed in 
Java are exclusively Javanese inventions. Another example of these addi-
tions is Semar, a hermaphrodite clown-s ervant of divine origin. Semar is 
another name for the Javanese god Ismaya, a remnant from an indigenous 
Javanese religion that predates the arrival of both Hinduism and Islam to 
Java. In the genealogy of wayang characters, Semar is said to be one of 
the brothers of Bhatara Guru (Shiva) and he appears in most stories as a 
clown- servant, offering advice to the protagonists of the story. Regardless 
of the episode, Semar and his sons Petruk, Bagong, and Gareng appear in 
the gara-g ara scene, a comical interlude that takes place in the middle of 
all- night performances (a segment of the performances which itself often 
lasts a couple of hours). These characters are collectively referred to as 
punokawan, or clown- servants.
Working together with physicist Andrew Schauf, I built a network 
model of a set wayang stories (Schauf and Escobar Varela 2018). This 
model considered the numbers of shared scenes in which pairs of char-
acters appear, giving weight values to each edge. Thus, an edge of weight 
twenty between two characters indicates that the corpus contains twenty 
different adegan (scenes) in which both characters appear. The resulting 
network can be explored through a data- assisted interactive visualization 
at the Digital Wayang Encyclopedia’s website at https://villaorlado.github.
io/wayangnetworks/html/canonical (also available as video 5.1 in this 
book’s companion website).
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
110 •  TheaTer as DaTa
In the interactive visualization, the size of a node corresponds to 
its degree (larger nodes have a higher number of edges). An edge (rep-
resented as a line) between two characters indicates that both appear 
together in at least one scene. The thickness of the edges represents 
weight of the degree. By clicking on a node, a user can access several net-
work measurements for the corresponding character. These measure-
ments are explained below:
• Degree: The number of nodes linked to the given node
• Weighted degree: The sum of the weights of all the node’s links
• Closeness centrality: The inverse of the average length of the most 
direct paths between the given node and all other nodes in the 
network
• Betweenness centrality: As seen above, this is a measure of how often 
a node acts as a bridge between other nodes. A high betweenness 
centrality value indicates that the shortest paths between all pairs of 
nodes in the network often pass through the given node.
• Eigenvector centrality: As mentioned earlier, this is a measurement of 
the influence of the node in the network that considers the degrees 
of a node’s neighbors. Nodes with high eigenvector centrality 
tend to be connected with neighbors who are themselves highly 
connected.
The online platform can be used to interactively explore many facets of the 
wayang characters and stories, as it enables situated, data- assisted inter-
pretations. For example, a user can click on a given character’s descrip-
tion and then see which other characters tend to appear in the same scene 
as this character. A user can also trace the different versions of this char-
acter’s stories in South Asia and Java. The extensive notes on each char-
acter point add thick context to each data point. They also enable users 
to shift perspectives from aggregate categories to individual characters. 
The exploratory features (see chapter 3) of the visualization enable users to 
choose where to focus their attention, what to view and which sequence 
to follow.
This dataset can also be used to reach data-d riven conclusions. We can 
combine the network-t heoretical measurements with features that have 
long preoccupied scholars, such as the ways Indian- derived characters 
and characters of local Javanese invention interact in the stories as told 
today. When we considered the origin of the characters (India vs. Java), we 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Relationships as Data • 111
noted that almost half of the characters who appear in the epic are Java-
nese in origin. This seems counterintuitive; when I informally asked peo-
ple familiar with wayang kulit to give an estimate of the percentage of way-
ang characters with Javanese origin, most (including several well- versed 
wayang experts) guessed that only about 20 to 30 percent of the characters 
were Javanese in origin. However, if we look at the weighted degrees of 
the characters, we note that the Indian characters tend to have signifi-
cantly higher values than their Javanese counterparts. This means that 
although almost half of the characters are Javanese in origin, the Indian 
characters tend to appear more often— and tend to appear repeatedly 
alongside the same fellow characters— thus exhibiting stronger connec-
tions within the co-o ccurrence network. This greater prevalence, reflected 
in their weighted degrees, may account for the common perception that 
Javanese characters are fewer in number, even though Javanese and Indian 
characters appear in the same proportion. In some cases, scholars might 
disagree on whether a character is Javanese in origin, or a modified ver-
sion of an Indian character. The Digital Wayang Encyclopedia includes a 
brief discussion on how these decisions were made, and the description of 
each character points to a variety of academic sources. However, readers 
might want to reclassify certain characters and run the analysis again. For 
this purpose, the data can be downloaded from this book’s companion 
website and manually reclassified. The choice of stories might also entail 
bias, as we used a relatively small sample of stories. My hope is that more 
wayang stories and characters will be analyzed in the future following the 
procedures outlined here, in ways that extend or challenge our results.
My comparison of Javanese and Indian characters is summarized in 
table 5.1. I calculated the effect size of weighted degree difference between 
Indian and Javanese characters. Effect sizes are rarely reported in network 
analysis papers in DH, but they are common in other fields (Clemente et 
al. 2015). To estimate effect size, I used Cohen’s (1988) d and obtained 
a value of d = 0.505. This means that there is half a standard deviation 
of difference between both groups’ weighted degrees. In Cohen’s origi-
nal formulation, any effect bigger than 0.1 is considered significant, and 
any value bigger than 0.5 is considered a medium effect. However, more 
recent commentators (Lakens 2013) note that Cohen’s original rules of 
thumb, which categorized effects as small, medium, and large, should 
just be taken as a rough estimate. The importance of an effect size needs 
to be contextualized within the field to which it is applied, and compared 
to other effect sizes reported in the literature. As more theater researchers 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
112 •  TheaTer as DaTa
report effect sizes, it will become easier to gauge how large of an effect is 
0.5 for the study of cultural transmission in theater traditions.
Figure 5.6 visualizes the pairwise comparison between network- 
theoretical measures of Javanese and Indian characters. The measure-
ments considered are weighted degree, betweenness centrality, and eigen-
vector centrality. The usefulness of considering all three measurements 
together is that they all show marked differences between Indian and Java-
nese characters. In this pairwise plot, each measurement is paired against 
the other two, through six scatterplots. This visual presentation follows 
the “small multiples” convention often used in statistical graphs (see 
chapter 3). The variables in the scatterplots are determined by the labels 
of the shared x and y axis. Thus, the scatterplot center- top compares the 
betweenness centrality (x axis) against the weighted degree (y axis). The 
quadrant where a measurement would be compared against itself (such as 
the top left) shows instead a kernel density estimate (KDE) of the distribu-
tions of values for both Indian and Javanese characters. The KDE shows 
how many values (y axis) are present at a given measurement range (x 
axis). This KDE estimate is similar to the violinplots described in chapter 
6 (except that here the kernels are rotated 90 degrees).
Both the scatterplots and the KDEs tell a consistent story: Indian and 
Javanese characters are very different in terms of their network measure-
ments. For every measurement, the Indian characters are spread around 
all possible values, whereas the Javanese characters show two distinct 
peaks: the vast majority of characters are bunched together around lower 
values. But a few characters rank highly for each measurement. These are 
the aforementioned clown-s ervants, who appear in virtually every story 
and have an important role to play as mediators, as they bridge distinct 
factions. The KDE for betweenness centrality shows this dramatically 
important role, with three gray spikes at the far right of the quadrant. 
These spikes indicate that three of the clown- servants have the highest 
betweenness values of the entire dataset (this is also evident in the scat-
terplot). The key role that the clown-s ervants play in wayang is not a new 
insight, but it is strengthened by this data- driven investigation.
Table 5.1. Comparison of Indian and Javanese characters
Percentage of  Average Weighted  
 characters Degree
Indian Characters 52.4% 153.47
Javanese Characters 47.6% 73.36
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Relationships as Data • 113
Fig. 5.6. A pairwise matrix comparing the network- theoretical measurements of Javanese and 
Indian characters.
Elsewhere (Schauf and Escobar Varela 2018), we have generated random 
networks of the same size as this wayang network to estimate the likeli-
hood that the values we see for each character are significant. In another 
paper, I have also considered differences between the network measure-
ments of characters with a stable puppet representation (i.e., those which 
cannot be substituted other puppet), and serambahan or “wildcard” charac-
ters that can be represented by many puppets (Escobar Varela 2019).
The question of the interaction between Javanese and Indian charac-
ters can be answered with data, but the pertinence of the question itself 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
114 •  TheaTer as DaTa
is open to other modes of critical analysis. Many dhalangs and Javanese 
scholars are interested in this issue, as can be seen from wayang diction-
aries published in Indonesia that almost always indicate when Javanese 
characters are also found in the Indian versions of the stories. However, 
this line of inquiry can also be traced to the concerns of colonial-e ra Dutch 
scholars. In a previously quoted paper, I have also given a fuller account 
of how colonial scholarship has shaped this question and how it has 
changed in postcolonial Indonesia (Escobar Varela 2019). This data exper-
iment should not be used to draw definite conclusions about wayang. As 
noted above, the dataset I used is relatively limited and the addition of 
more data will further refine the insights that network analysis can bring 
in to the analysis of Javanese wayang’s history. However, the method of 
using pairplots to visualize differences across types of nodes in a network, 
and the practice of reporting effect sizes, can certainly be applied to other 
types of fictional and collaborative networks.
Conclusions
The networks in both excursions are complex systems with the same struc-
ture: a power- law distribution and small- world properties. This means 
that the same mathematical tools can be applied in both cases, but they 
will be mobilized to very different ends. The TNS network is used to ana-
lyze the history of a theater company. The wayang network reveals patterns 
in the ways characters of different origin interact with each other. The two 
excursions have described quantitative markers against the backdrops of 
history and culture that shaped these networks and their interpretations. 
Both examples look at things that make sense in their contexts. The analy-
sis of collaborative relationships has been central to academic reflections 
on TNS’s history, as is reflected by the edited collections that celebrate and 
critique their work. The analysis of the relationships between Indian and 
Javanese versions of wayang stories has long preoccupied scholars of Java.
Networks can model many things. Here, they model properties that 
are directly relevant to theater scholarship in Singapore and Indone-
sia. Besides these features, many other aspects of theater practice could 
be modeled as networks (international collaborations, funding sources, 
festival circuits, citations in theater scholarship, etc.). If analysis of these 
other areas repeats previous experiences, we might find the markers of 
complexity in most, if not all these networks. But the most promising 
research agenda is to zoom in on the networks, analyzing the way cultural 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Relationships as Data • 115
context determines their structures, and the way their quantitative proper-
ties tell the story of those contexts. The data currently available, and the 
explanatory power of standard measurements, places network analysis as 
the most promising area for the digital study of theater companies and the 
fictional universes of theater plays.
Code and data: Sample code for this chapter is available at  
https://doi.org/10.3998/mpub.11667458.cmp.40.  
The data used can be downloaded from  
https://doi.org/10.3998/mpub.11667458.cmp.29,  
https://doi.org/10.3998/mpub.11667458.cmp.30,  
https://doi.org/10.3998/mpub.11667458.cmp.31,  
https://doi.org/10.3998/mpub.11667458.cmp.32, and  
https://doi.org/10.3998/mpub.11667458.cmp.33.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
ChapTer 6
Motion as Data
The study of motion as data is a fascinating area, and one that holds vast 
promises for our field. As a generalization, motion is more important 
than language for the study of theater-a s-e vent (where I am including 
dance and physical theater). Although there are many genres and tradi-
tions where language is absent or not central, it is difficult to imagine a 
theater performance with no motion. The only example I can think of is 
While We Were Holding It Together (2006, dir. Ivana Muller), where a group of 
actors remained in the same position for the entire performance, as their 
lines were projected behind them. But even then, the subtle eye twitches 
and mouth motions were essential to the meaning of the performance.
In this chapter, I look at how to use digital data to study motion, rather 
than at dance performances that incorporate digital technology (see Bleeker 
2016). In this chapter, I use motion and movement interchangeably for the 
sake of variation. But my interest is analyzing the movement of actors and 
objects on stage. Objects don’t generally move by themselves (unless they 
are powered by electrical or mechanical devices), and in this respect they 
are different from actors. But the movements of both actors and objects 
are computationally tractable and can be modeled as data. The analysis 
of motion is the hardest, but in some ways, most interesting area for the 
computational study of performance. It is also the area that most under-
scores the situated, constructed nature of data. For things such as geo-
graphical coordinates, data-c ollection devices are so embedded in everyday 
life that they can sometimes become transparent and we can forget their 
constructed nature. This is not possible for motion, where the methods of 
data collection are so invasive—a nd so artificial—t hat the cultural specific-
ity and constructedness of the data will remain front and center.
For this reason, the structure of this chapter is somewhat different 
from other chapters in this section, as I first consider the possible ways 
motion can be understood as data. Then I continue the usual structure 
Escobar Varela, M1ig1u6el. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Motion as Data • 117
and survey a range of methods for collecting and transforming this data 
and explain how these methods can be used within data- driven and data- 
assisted methodologies.
What Is Motion Data?
Collecting and processing motion data poses the hardest problem of all 
the areas discussed in this book. There is no standard format for repre-
senting motion (unlike networks for relationships, written words for text, 
and coordinates for maps). There is also no straightforward way to obtain 
motion data. Whereas it is relatively simple to obtain GPS coordinates 
or find the list of people working together in a theater company, obtain-
ing motion data is a whole different issue. Options range from written 
records of choreographies that use dance terms (e.g., demi- plié), to nota-
tion systems (such as Labanotation), and to sophisticated motion capture 
systems. At the time of writing, all available systems are either inaccurate, 
cumbersome to use, or expensive. Since motion data is scarce and it comes 
in a variety of formats, there are no agreed-u pon, widely used methods to 
analyze motion data that would fit the needs of computational theater 
research. For example, quantitative motion analysis is an active subfield 
in engineering, but the level of granularity and types of questions asked 
(e.g., calculating trajectories of pedestrians for autonomous vehicles) are 
not easily mapped to the study of motion in theater and dance. However, 
new software such as Open Pose (Cao et al. 2018) might change this soon, 
as will be discussed later in this chapter.
Motion can be represented in three ways, ordered here in increasing 
level of formalization:
 1. Concepts can belong to a general-p urpose analytical vocabulary such 
as Laban Movement Analysis (LMA) or be specific to a dance tradi-
tion (grande jeté in ballet, trisik in Javanese dance).
 2. Notations are formal vocabularies that include syntactic rules and 
which are usually represented through a series of graphic symbols. 
The systems more commonly used today are Labanotation and 
Benesh Movement Notation (BMN).
 3. Numerical data can be made of points on a 2D or 3D plane (usually 
joint positions, but also weight and muscle activity), or numbers 
that represent speed, effort, or some other magnitude for a given 
motion. Time might also be added to this data.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
118 •  TheaTer as DaTa
This is not the only possible classification of motion representation (for 
an alternative taxonomy see deLahunta and Jennet 2017, 65). The purpose 
of my classification is to explain what is required for the computational 
analysis of motion data. The more formal a language, the more amenable 
it is to automatic transformations, but the harder it is to obtain. The third 
level is required for the computational analysis of dance. Numerical data 
can be obtained directly (from sensors), automatically (from video) or con-
verted from the first and second levels. The following overview describes 
these levels in order of progressive abstraction, which is not always equal 
to their chronological development.
Concepts
Most movement traditions that I am familiar with have a specialized 
vocabulary to describe its motions, and sometimes its aesthetic qualities 
as well. The oldest known description of dance is the Nāṭya Śāstra, attrib-
uted to Bharata Muni, which dates back to at least the second century CE, 
if not much earlier. This text describes karanas, the basic dancing blocks of 
classical dance which comprise posture, gait, and gestures (Pandya 2003). 
Many of the first notation systems to emerge in Europe use words and 
word abbreviations in order to record movement (Guest 1998). These sys-
tems were devised with the intention of documenting and teaching dance, 
and were not as explicitly prescriptive in nature as the Nāṭya Śāstra, but 
they were also premised on specific traditions of dance and social conven-
tions. In spite of their claims to generalizable description, most systems 
were short lived since they could only function within the limited param-
eters of a specific dance practice.
Today, the most widely used conceptual system for the description of 
motion is Laban Movement Analysis (LMA), also known as Laban/Barte-
nieff Movement Analysis. Unlike Labanotation, which will be described 
below, LMA is not used for notation, but for analytical purposes. LMA 
is used in several scientific disciplines, for example to study human- 
computer interaction (Fdili Alaoui et al. 2017), or to establish the con-
nection between movement and emotion in psychology (Tsachor and 
Shafir 2017).
Used by dance researchers to describe dance traditions from different 
parts of the world, LMA facilitates comparative approaches, but it requires 
imposing an external analytical system to describe a dance tradition. In 
other words, it constitutes an etic, as opposed to an emic approach to dance 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Motion as Data • 119
theory (in the latter, the theory is derived directly from a community of 
practice). To generate emic descriptions that still enable comparisons, 
several dance anthropologists have developed comparative guides (Royce 
1977). For example, Kurath’s (1952) Choreographic Questionnaire, takes a for-
malist approach and includes questions on three sections: ground plan, 
body movement, and structure. Later in this chapter, I will describe proj-
ects that aim to map digital motion data to dance concepts.
Notation
The development of dance notation systems independent of natural lan-
guage first emerged in the context of European traditions of dance. In a 
comprehensive historical survey of dance notation systems in the Euro-
pean tradition, Guest (1998) classifies notation systems according to the 
graphical notation medium: words and word abbreviations, track draw-
ings, stick figures, music note systems, and abstract symbol systems. In 
the context of the present overview Guest’s “words and word abbrevia-
tions” correspond to the previous category of “concepts.” I include all 
other systems in the current category. Even though track drawings and 
abstract symbols look very different, they are similar from a data represen-
tation perspective since they require nontextual conventions of encoding. 
If one were to encode natural language conventions, one could make use 
of the many systems for the encoding, representation, and processing of 
text. If, in contrast, one wanted to represent any other system in a com-
puter, a new format would need to be devised for such purpose.
As chronicled by Guest, before the twentieth century, the most influ-
ential systems were those devised by Raoul-A uger Feuillet, Arthur Saint- 
Léon, and Friedrich Albert Zorn. These systems are often referenced in 
historical documents but rarely used today, and I could not find any digi-
tal, machine- readable implementation of these systems. The situation 
is different for Labanotation, also known as Kinetography Laban, which 
was first developed by Rudolf Laban in 1928. The system is cumbersome 
to learn, but widely used today, and many dance traditions have been 
recorded using Labanotation. The Dance Notation Bureau has an exten-
sive catalog of dances encoded in Labanotation, as do many libraries and 
academies.
There have been many efforts over the years to encode Labanotation 
in digital formats, and develop software for this purpose. Nakamura 
and Hachimura have written several papers describing an XML file for-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
120 •  TheaTer as DaTa
mat (similar to the one used for TEI in chapter 4) for Labanotation and a 
software with a graphical user interface for creating notation files inter-
actively, the LabanEditor (Hachimura and Nakamura 2006; Hachimura 
2006). The aim of this program is to generate visual representations of the 
dance automatically, which can be used for the digital reconstruction of 
dances. The authors have worked on Noh theater and other Japanese tra-
ditions with their systems. In their own estimation, the system works well 
for “simple motions,” but is as of yet insufficient to fully document com-
plex dances (Hachimura 2006). Another program, GenLaban is aimed at 
generating Labanotation directly from motion capture data (Choensawat, 
Nakamura, and Hachimura 2015). Although these projects are described 
in great technical detail, the software programs themselves were not avail-
able for download and testing at the time of writing. Labanotator is a com-
mercially available tool for the creation of Labanotation scores. The soft-
ware is common but it does not enable users to export or import files in 
XML formats, which makes comparative analysis difficult. Gábor Misi’s 
Labanatory (2005) is a program that runs on AutoCAD in order to search 
for patterns across Labanotation scores. Misi’s approach does not convert 
the data to XML either, but uses the built- in capabilities of AutoCAD to 
find visual similarities in dance movements. Misi (1983) has used Labano-
tation to carry out formal analyses of male solo Transylvanian dances 
and developed methods for the algebraic representation and analysis of 
Labanotation (Misi 2008). Sadly, this approach has not been followed up 
by others. A widely used algebraic or XML representation would enable 
other teams to reuse data for comparison, retrieval, and analysis.
The other system which is widely used in dance today is Benesh Move-
ment Notation (BMN). It does not have the same level of granularity as 
Labanotation, but it is easier to read as it is visually similar to staff nota-
tion for music. However, the development of digital projects based on 
BMN is a less active area, and I am not aware of any attempts to represent 
BMN in XML or through other digital formats, or to use BMN for quanti-
tative or automatic formalist analyses. However, a program for BMN has 
been developed by the Royal Academy of Dance (Benesh Institute n.d.).
Labanotation is more common in the United States while BMN is more 
common in the United Kingdom. Both have been applied to dance tradi-
tions outside of Europe and North America, but they remain essentially 
bound to specific cultural assumptions about movement. El Raheb and 
Ioannidis (2014) report that different levels of familiarity with dance tradi-
tions produce different scores for the same dance.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Motion as Data • 121
Numerical Data
Numerical data can be acquired directly—f rom markers and sensors—o r 
as the result of processing other kinds of data: words, notations, images, 
and videos. When generating numerical data, there is a trade- off between 
invasiveness and precision. Markers and sensors are expensive and they 
usually need to be attached to the bodies of the performers. This means 
that data can only be gathered in very specific spaces and conditions. 
Obtaining data from other sources is either computationally very difficult, 
or only yields coarse data.
Sensors are electronic components that send signals to a record-
ing device. In biomechanics, sensors are most commonly used to track 
the electrical activity of muscles. This kind of data is rarely used for the 
analysis of dance, but is commonly deployed for the analysis of move-
ment in sports and for the development of therapies for movement dis-
orders (O’Donoghue 2010). Inertial measurement unit (IMU) sensors, 
like the ones commonly found in smartphones, can track the force, angu-
lar velocity, and direction of a movement. A single sensor can’t provide 
enough data for motion capture, but it can be used to track some aspects 
of a given motion, such as its speed and direction. Cuykendall et al. (2016) 
have developed an art project that uses smartphones to gather this kind of 
data, which will be described later in this chapter, in the context of data- 
assisted dance research.
Some companies and research teams are developing motion capture 
technologies based on systems of IMU sensors. But a more common 
solution is to use markers instead. In contrast to a sensor, a marker does 
not have electronic circuity. They only work in rooms fitted with special 
infrared cameras that can calculate the position of the markers in a three- 
dimensional space. Most marker systems come with their own software 
packages that can infer the position and angles of joints based on marker 
data. A system of forty-o ne markers is sufficient for most applications in 
biomechanics. Motion capture (mocap) for film and video games com-
monly requires more granular systems, which typically include full-b ody 
suits, masks, and gloves. Some dance and theater companies have also 
used very detailed mocap graphs in performance. Since my main concern 
is the utilization of mocap for the analysis of movement, I won’t examine 
the extensive literature on the implications of mocap and computer graph-
ics for contemporary performance practice (Kozel 2007; Chan et al. 2010).
The granularity and accuracy of the mocap data determines the kinds 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
122 •  TheaTer as DaTa
of analysis possible, and the reliability of the results. Professional, full- 
body mocap can only be achieved in hi-t ech production studios, where 
the costs are prohibitively expensive for most researchers and artists. 
There are many intermediate options, with portable camera systems 
and different configurations of markers. The most common setup in 
research contexts, both for the biomechanics of sports and the biome-
chanical analysis of dance, is the Vicon plug- in gate model (Duffell, 
Hope, and McGregor 2014). Hachimura and Nakamura used this com-
mercially available optical- type motion capture system to record Japa-
nese classical dance and Noh theater movements (Hachimura 2006; 
Hachimura and Nakamura 2006). One of their primary objectives was 
to use computer graphics to digitally reconstruct intangible heritage, 
for the purpose of preservation and education, and they were thus inter-
ested in the seamless conversion between XML, Labanotation, and 
mocap data. As discussed above, they developed systems for the interac-
tive display of digital animations from Labanotation and for extracting 
Labanotation from mocap systems.
A way to circumvent the financial and technical restrictions of expen-
sive mocap devices is to extract mocap from other sources of data. Naka-
mura’s team has proposed systems for generating data from natural lan-
guage commands, and from Labanotation. Another common approach 
is to obtain data from video. This was traditionally very hard to accom-
plish, but it is becoming easier with software packages such as Open Pose 
(Cao et al. 2018). In the first excursion below, I describe, in more detail, 
a simple technique that Gea O. F. Parikesit and I used to obtain motion 
data from wayang kulit videos. Our measurements are not very granular, 
since we can only estimate speed, but they are sufficiently accurate for very 
specific questions and for video data obtained under controlled environ-
ments, and they could easily be implemented by other research teams.
An intermediate approach between costly mocap systems and 
consumer-e nd video cameras is using commonly available motion cam-
eras such as Microsoft Kinect and IpiSoft, which showed promising 
results in previous studies (Arulampalam, Pierrepont, and Kark 2015). 
Kinect is no longer in development, but motion cameras from Nintendo 
and PlayStation could perhaps be used for dance motion capture in the 
future. Cutting-e dge algorithms can automatically estimate the number 
of persons in a scene, generate their corresponding 3D skeletons, and 
estimate their locations (Elhayek et al. 2018). Video data can also be used 
for gait identification, as a video surveillance system can identify people 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Motion as Data • 123
based on their habitual patterns of walking, even when little data is avail-
able (Balazia and Sojka 2017). This has terrifying social implications for 
privacy, but it could be a boon for the analysis of dance. Still, these algo-
rithms have yet to be applied to dance and physical theater data.
The advent of game- changing technologies might soon radically 
alter the types of research that are possible. As previously noted, Cao et 
al. (2018) have developed Open Pose, an open source software that uses 
a technique called Part Affinity Fields (PAFs) for real-t ime estimation of 
the position of dancers in a video with high accuracy, even when there is 
visual noise in the background of a video. Broadwell (2019) demonstrated 
in a workshop in the 2019 Digital Humanities conference that Open Pose 
could be used to obtain movement data from K-P op dance videos and that 
this data could be potentially used to compare dance styles across a large 
number of videos.
Data- Driven Motion Analysis
Several scientific studies have used biomechanical methods to study 
dance, but they are mostly interested in preventing injuries in professional 
dancers or in using dance as a therapy for specific kinds of trauma or dis-
ease (Koutedakis, Owolabi, and Apostolos 2008; Luksys and Griskevicius 
2016). Other studies have explored the intersection of biomechanics and 
more subjective experiences of dance. A review by Chang et al. (2016) 
looked at the correlation between the subjective perception of beauty and 
the biomechanical markers of skill, such as speed, vigor, and smoothness. 
The review found significant differences between expert and non-e xpert 
judgment, but no general trends in the correlation between aesthetic expe-
rience and biomechanical markers of skill. This is perhaps unsurprising 
to dance scholars who are familiar with the wide range of culturally spe-
cific ideals of dance (see Redding 2019, 62). For example, Javanese dance 
is generally slower than Balinese dance and there is little point in taking 
their different speeds as proxies for skill, appropriateness, or beauty. To 
determine whether a movement is beautiful, well- executed, or appropri-
ate, one needs to be familiar with the specific dance culture where the 
motion is evaluated. A motion that is considered beautiful for one kind of 
dance would be considered ugly if performed in other dances. However, 
the authors of the review, writing from an engineering perspective, con-
fidently hypothesize that the range of responses is due to the lack of clear 
goals in dance. This is different to sports (a much beloved field of biome-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
124 •  TheaTer as DaTa
chanical analysis) where there are usually clearly defined goals that lend 
themselves better to measurement and comparison.
Closer to the objectives of dance studies, Hachimura (2006) proposes a 
method for measuring similarities between dance gestures. The challenge, 
as he explains, is that two gestures perceived as identical by human observ-
ers might look very different in the data. Even when considering simple 
motions such as walking, “there are instances where you start with your 
right foot and instances where you start with your left foot” (Hachimura 
2006, 59). There are also differences in terms of speed, direction, and 
position for different motions that could still be considered to belong to 
the same conceptual category of walking. Thus, he had to develop ways of 
measuring similarity that would be robust to these small variations. For 
this he used a dynamic programming (DP) algorithm which is often used 
in voice and handwriting recognition.
Hachimura also suggests that a minimum convex polyhedron can be 
calculated for different slices of time. An alluring insight hides under 
this technical term. To explain this, let me take a simpler concept first. 
A minimum convex polygon is a two- dimensional shape that encom-
passes a series of points. “Minimum” means that it is the smallest poly-
gon in which no internal angle exceeds 180 degrees and which contains 
all the points. Minimum polygons are used in biological maps to show the 
extent of an animal population, or the area actively defended by an indi-
vidual of that population. The minimum complex polyhedron is a three- 
dimensional version of this concept, or a solid in three dimensions with 
flat polygonal faces, straight edges, and sharp corners. For dance data, 
this is the minimum three-d imensional shape that encompasses all the 
joints of a dancer in a given gesture, as frozen in a moment in time. The 
total volume of this polyhedron can be calculated for any moment in a 
dance, and plotted in a graph as a function of time. This is useful for com-
paring two dances, and serves the purpose of retrieval based on similarity. 
Another interesting possibility, also mentioned by Hachimura, is that the 
principles of LMA, such as space, effort, and shape (which are extensively 
used by dance scholars, as discussed above) can be expressed in terms of 
polyhedra that change through time.
In the case of dance, it is also important to retrieve specific gestures 
of the feet or the hands, rather than whole body motions. Hachimura’s 
example is the okuri motion, which is important for Noh performances. 
Hachimura’s paper describes both methods with broad applicability and 
detailed examples that show engagement with specific dance traditions. 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Motion as Data • 125
Another comprehensive set of dance methods has been proposed by 
Wiesner and her collaborators (2012). They created a database of move-
ments, the ARTeFACT Movement Thesaurus (AMT) which encoded two 
hundred common dance movements into a semantic XML database with 
accompanying motion capture data. Subsequently, they used this data to 
automatically identify, annotate, and retrieve film data (Wiesner 2012). 
Using biomechanical analysis, they were able to achieve high accuracy in 
the automatic classification of these movements (Simpson, Wiesner, and 
Bennett 2014). Acknowledging that dance is more complex than individ-
ual steps, they have also developed systems to identify conceptual meta-
phors in dance (such as “conflict”). They have applied statistical proce-
dures and borrowed principles from corpus linguistics to identify patterns 
in their biomechanical dataset and in verbal descriptions of movements, 
to identify features that correspond to conceptual metaphors.
As the previous projects show, the estimation of difference and sim-
ilarity is central to the data-d riven analysis of dance. MoComp is a tool 
specifically developed for the comparative visualization of mocap data 
(Malmstrom et al. 2016). In their interactive web- based visualizations, 
the authors used streamgraphs to represent changes in angles and angu-
lar velocities over time for a given joint. This project is explicitly premised 
on a scientific understanding of motion. Later, I will describe visualiza-
tions geared towards data- assisted analysis. We will see that many data- 
assisted projects are interested in the affective qualities of movement, but 
this is also important for some data- driven research projects. Li and Pas-
quier (2016) have looked at how different statistical and machine learn-
ing approaches can be used to identify affect in movement, using anno-
tations made by performers and observers as ground truths (an example 
of the calibration practices discussed in chapter 2). Their project deals 
with affective dimensions but aims at consensus- seeking, intersubjec-
tively verifiable observations. The projects discussed in the next section, 
in contrast, aim to enact affective responses through highly interpretive 
visualizations.
Data- Assisted Motion Analysis
In a more recent project, Wiesner and her collaborators (2016) worked 
with the creators of POEM, a mobile platform that can convert move-
ment into whimsical poems, in order to provide a situated classification 
of movement through “embodied visual analytics” (Cuykendall et al. 2016; 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
126 •  TheaTer as DaTa
Cuykendall, Soutar- Rau, and Schiphorst 2016). The collaboration between 
ARTeFACT and POEM builds on the strengths of both projects, combin-
ing the extensive data collection capabilities of POEM with the nuanced 
data model of ARTeFACT (Wiesner et al. 2016). This project helps validate 
and generalize the ARTeFACT data but also shows its potential for creative 
applications. Another example of performative visualization is found in 
Nakamura’s work (2017), where she used lines of different colors to trace 
the movements of fingers, arms, and trunks of Balinese dancers as they 
moved through space. This provides a situated visual interpretation rather 
than replicable conclusions, but it also builds on a data-d riven project, as 
is the case of the ARTeFACT/POEM collaboration.
Interactions between choreographers, engineers, and designers fur-
ther expand the scope of data- assisted movement analysis. These col-
laborations have been thoroughly chronicled by deLahunta (deLahunta 
and Jenett 2017; deLahunta 2017). Of particular significance is Synchronous 
Objects for One Flat Thing, Reproduced by Ohio State University’s Advanced 
Computing Center for the Arts and Design and the Department of Dance 
in collaboration with William Forsythe (Forsythe et al. 2009; Palazzi and 
Shaw 2009). This was perhaps the first collaboration between dancers 
and data scientists to achieve international notoriety in the dance world. 
In parallel with this project, William Forsythe and others started Motion 
Bank, with support from the German Federal Cultural Foundation and 
other sources. The project has generated reusable data and software pack-
ages. A video annotation tool named Piecemaker was developed by The For-
sythe Company member David Kern between 2007 and 2013, and later 
redeveloped as Piecemaker2 (PM2) and PM2GO.
Motion Bank features the work of several choreographers and, even 
though some tools are shared across the platform, each choreographer’s 
portal attends to their specific ideas of dance practice. For example, the 
choreographies of Deborah Hay are not meant to be repeated exactly: “the 
movement may change, but the choreography itself does not change” 
(quoted in deLahunta 2017, n.p.). For Hay’s work, the codification and 
transmission of inflexible dance scores makes no sense. Instead, six per-
formers were invited to perform the same solo dance according to their 
different interpretations for Motion Bank. Each version was recorded 
and processed, and visualizations were invoked to explore these differ-
ences. One such visualization uses minimum convex polygons, the same 
mathematical objects I described earlier. Here, they are instances of what 
I call “performative visualizations” (see chapter 3), and they are used to 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Motion as Data • 127
highlight differences between dance performers as much as to visually 
convey a key principle of Hay’s choreographic philosophy. In the previous 
example, the same quantitative method was used for information- retrieval 
within a data- driven, scientific paradigm. The transferability of the same 
technique across methodologies emphasizes a key theme in this book: 
quantitative methods alone do not signal a specific attitude to knowl-
edge. What matters is the way in which a method is contextualized within 
a methodology— the same method can be used to pursue consensus- 
seeking, replicable conclusions, or situated, performative hermeneutics.
Bermudez et al. (2011) also endeavor to communicate specific choreo-
graphic principles with data. They have developed interactive installations 
to explore the principles of Double Skin/Double Mind, the choreographic 
approach of Emio Greco | PC. Other platforms have also been developed 
for the creative visualization of dance data (see Carlson, Schiphorst, and 
Shaw 2011). But special attention should be paid to the work of Ribeiro, 
Kuffner dos Anjos, and Fernandes (2017), who have used 3D data derived 
from a kinect device to create interpretive visualizations that represent the 
movement and improvisational principles of Portuguese dance creator 
João Fiadeiro’s composition in real time (Composição em tempo real in Por-
tuguese). Working together with him, the authors identified a subset of 
core concepts of Fiadeiro’s method which were explored through several 
performative visualizations. One such principle is suspension, a moment in 
a choreography from where several options are available to a dancer. In 
the context of this dance approach, it is important to think about these 
moments as possible futures. A series of cubes juxtaposed in a 3D space 
were used to emphasize this perspective. In another visualization, they 
wanted to emphasize what happens when a performer leaves the stage 
and another enters. This is described as the cycle of vitality in Fiadeiro’s 
work. This quality is performatively brought out in visualizations where 
the 3D representation of a dancer loses color as they leave the stage and 
another figure regains color as they enter the space and begin a new cycle 
of movement.
This last example highlights that all thinking about dance must be 
situated within specific cultural and analytical contexts. Ribeiro, Kuffner 
dos Anjos, and Fernandes’s work is a felicitous combination of data visu-
alization and contextual sensitivity and can serve as an example for proj-
ects in other dance traditions. Another striking performative visualization 
is EMVIZ (Subyen et al. 2011), which aims to produce visual representa-
tions of the Laban Basic-E fforts. They describe their project as an exam-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
128 •  TheaTer as DaTa
ple of artistic visualizations which combine “interpretive metaphoric map-
pings with aesthetic approaches in representing data from one domain to 
another” (Subyen et al. 2011, 121). They characterize artistic visualizations 
as data mappings that “are interpretive, subjective, and follow a different 
set of conventions than those governing information visualization or sci-
entific visualization.” This echoes my descriptions of performative visual-
izations in chapter 3. Subyen et al’s system has been primarily used in the 
context of art galleries, where dancers wore sensors while performing a 
variety of movements, and real-t ime visualizations were projected behind 
them. It is important to highlight how EMVIZ, as well as the work of the 
teams led by Bermudez and Ribeiro differ from the MoCamp platform we 
encountered in a previous section of this chapter. All of these are compu-
tational transformations of dance data. But while MoCamp aims for the 
readability and systematicity that befit a scientific analysis, the projects 
we have encountered in this subsection enlist an interpretive sensibility 
for the production of data representations, which are in themselves art 
projects.
First Excursion: Calculating the Speed of Puppet Movement  
in a Wayang Video
In what follows, I describe two small research projects, which I carried 
out with my collaborators in order to study the implications of motion 
analysis for the theater traditions of Java. In the first, I collaborated with 
Gea O. F. Parikesit, a physicist who works in image and video processing. 
Here, I summarize our research, and show some new visualizations of 
the data we published earlier. We used Catur Kuncoro’s Wayang Mitologi 
(2012), a sixty- nine- minute video from the Contemporary Wayang Archive 
(CWA, available at http://cwa-web.org/en/WayangMitologi) and, calcu-
lated the speed of the video using difference images. A difference image is 
an image that results from subtracting one image from another. To get a 
better sense of our method, it is important to remember that a video is a 
sequence of digital images, each of which is made of pixels. In turn, each 
pixel contains information for three color channels (red, green, and blue), 
and each channel has a grayvalue, which indicates its brightness. Below is 
a short nontechnical summary of our methods. More comprehensive tech-
nical details can be found in our original publication (Escobar Varela and 
Parikesit 2017).
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Motion as Data • 129
 1. We converted the video to a series of static images.
 2. We computed the difference image for every pair of subsequent 
images.
 3. We calculated the total grayvalue of the pixels in each difference 
image where the value was above a certain threshold. The reason 
for including a threshold is that some of the differences are just 
noise. When recording a black image on a consumer-e nd video 
camera there will be small differences within black image frames. 
Thus, we used two black images in order to estimate the noise 
threshold and then counted every image with a gray value higher 
than such threshold—t hat is, every image that is a reasonable indi-
cator that there was a change from one frame to the next.
 4. We plotted the resulting values in a graph, with the number of 
non- identical pixels as a function of time. Figure 6.1 shows this in 
a slightly different form to the one in the initial publication.
A visual inspection of the patterns in non- identical pixels in differ-
ence images led us to believe that different scenes had different movement 
profiles. In the original paper, we explored these differences by manually 
segmenting the scenes and then using a statistical procedure to iden-
tify differences between them. Here, I use the same segmentation and 
descriptive statistics but present them through new visualizations. The 
main reason for offering new visualizations is that I have presented the 
original results several times after the initial publication, and in that inter-
vening time I have thought of new ways to present the data, which I wish 
to explore here (the original images and analysis can be seen in the afore-
mentioned paper). Part of my interest was to develop a more interpretive 
visualization, which I call the kayon plot, as I will now explain.
For background, let’s consider that Wayang Mitologi is a contemporary 
version of the classical wayang kulit form described in chapter 5. There 
are many differences between classical and contemporary wayang, but 
the most important for present discussion is that contemporary wayang 
is usually shorter: a couple of hours rather than the seven to eight hours of 
the all- night, classical wayang (Escobar Varela 2015). Wayang Mitologi has 
different kinds of scenes: frame scenes, a comic interlude (or gara- gara, a 
fundamental part of a wayang performance), narrative scenes, and normal 
scenes. The comic interlude, narrative sections and the “normal scenes” 
are part of classical wayang conventions, while “frame scenes” are con-
temporary inventions. In the recording of Wayang Mitologi, the scenes 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
130 •  TheaTer as DaTa
Fig. 6.1. Nonidentical pixels in subsequent images as a function of time. Adapted from  
Escobar Varela and Parikesit (2017).
were segmented and labeled by hand. Figure 6.2 compares the mean and 
standard deviation of the number of non-i dentical pixels in subsequent 
images for each scene. A scene with a higher mean will be, on average, 
faster than one with a lower mean. A scene with a higher standard devia-
tion will be less homogeneous than one with a lower one, as it will tend 
to include a wider mixture of slow and fast sequences. Figure 6.2 shows 
very clearly defined clusters. Narrative scenes stand out from the rest in 
terms of both mean and standard deviation, and can be seen in the upper 
right quadrant of the graph. Likewise, the two frame scenes have almost 
overlapping values near the bottom left. The normal scenes and the comic 
interlude occupy the midsection of the graph.
It is interesting to note that the comic scene is very similar to normal 
scenes. But it is also much longer than other scenes, and this information 
is not captured in the visualization in figure 6.2. Another missing aspect 
here is sequence: we don’t know how scenes of different kinds follow one 
another. To include length and sequence information, I combined violin-
plots and bar charts (figure 6.3). The violinplots show the distribution of 
non- identical pixels within each scene, while the bars in the lower half 
show the scene’s duration. A violinplot shows the distribution of values 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Fig. 6.2. The standard deviation and mean of nonidentical pixels for each scene, grouped by 
scene type.
Fig. 6.3. Combined violinplots and bar charts. The violinplots show the distribution of non-
identical pixels within each scene (the mean is indicated by a white circle within each plot). 
The bars show the length of each scene.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
132 •  TheaTer as DaTa
for a given observation. The violinplots in the upper half of this graph are 
cut off at 0 (since a difference image can’t have a negative number of pix-
els). Thus, they look somewhat flatter and less like the musical namesakes 
of more conventional violinplots.
But like other violinplots, the width at each step of the y axis (e.g., 100 
or 4,000) tells you how many such values were found within that range 
(this is similar to the KDE plots described in chapter 5, except that they are 
rotated 90 degrees). A wider contour at the bottom means that the scene 
in question had more lower numbers (i.e., slower parts). Longer shapes 
include higher numbers (i.e., faster parts). The white circle inside each 
shape represents the mean value. Narrative scenes (5 and 7) have both 
similar shapes. They are slender and long—t his means that they have a 
wide distribution of values, but few instances of each value. If you go back 
to figure 6.2 you will see that these narrative scenes (represented there 
as diamonds) are in the upper-r ight section of the graph (another way of 
showing their high mean and high standard deviations). But figure 6.3 
provides the additional information that scenes 5 and 7 are shorter than 
other scenes, as indicated by the bars at the bottom half of the graph.
When exploring this visualization, I was reminded of the shape of the 
kayon, a puppet with a distinctive shape with important functions in way-
ang kulit (figure 6.4). This puppet is used to indicate scene changes and 
forces of nature, and it is also used to open and close a performance (Arps 
2016). The kayon has an important cultural significance in Java, and modi-
fied versions of its shape are used in the logos of many cultural institutions. 
I wanted to use this shape to visually present the information discussed 
thus far. Like all wayang puppets, this one is made of water buffalo hide 
and it has a protruding control stick (traditionally made of bone). Figure 
6.5 represents each scene as a kayon. It is very similar to the previous graph 
(figure 6.3). The upper half is identical, but the y axis of the lower half is 
inverted, and the length of each scene is represented by a line rather than 
by a bar. These simple modifications make each shape somewhat reminis-
cent of a kayon. Let me offer a theoretical justification for the strangeness 
(and perhaps failures) of this graph. This plot doesn’t look like most scien-
tific visualizations and it is an attempt to answer Drucker’s call for human-
istic visualizations (see chapter 3). However, this visualization doesn’t 
fully eschew the principles of statistical graphs either. It includes statisti-
cal information (mean and standard deviation), has a low ink ratio and it 
enables consistent comparisons (the principles advocated by Tufte). This 
visualization is a situated deformance, but is also statistically sound.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Fig. 6.4. A kayon puppet, 
from the front side.
Fig. 6.5. A kayon plot. Similar to figure 6.3 but the y- axis of the lower half is inverted, and the 
length of each scene is represented by a line rather than by a bar.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
134 •  TheaTer as DaTa
Even if other research teams don’t develop their own performative 
visualizations, there is great potential in the usage of difference images 
for the analysis of theater video recordings. Difference image calculation 
is not as sophisticated as some of the other methods described above. 
However, its advantage is its very simplicity: its meaning is clear and it 
is easy to implement. This is also one of the methods proposed by Lev 
Manovich (2013) for the analysis of film. It must be noted, though, that 
the needs of quantitative motion analysis for theater are different from 
those of quantitative film analysis. The theater video recordings I use 
here are those developed for documentary purposes, and they constitute 
digital traces of a performance. In film, the materiality and aesthetics of 
the medium are central concerns. Most of the methods in cinemetrics— 
the quantitative analysis of film— aim to reveal formal aspects of film 
language, such as the number and length of cuts (Salt 1974; Tsivian 
2005; Buckland 2008; Brodbeck 2011). Cuts are arguably one of the key 
aesthetic building blocks in film and patterns in cut usage are strong 
markers of authorial style. In contrast, the best kind of video for the dig-
ital analysis of motion in theater is one where there are no cuts, as the 
one we used here (since a cut would appear as a large number of pixels 
in a difference image).
If one is interested in the movement of actors or dancers, and wants to 
limit the influence of lighting, then the best video is one produced during 
a dry run. This, of course, depends heavily on context and traditions. Way-
ang kulit videos can be recorded from the side of the shadows, with mini-
mal color and depth interference in the analysis of motion (it should be 
noted, though, that the majority of present-d ay spectators watch wayang 
from the front, which is the side where they can see the dhalang animat-
ing the puppets). For many theater forms and genres, an analysis of speed 
can give a quick overview of a recording, and can provide quantitative 
measurements for comparative analysis. The techniques used here could 
be adapted to seek answers that are pertinent to specific traditions of prac-
tice. Copyright still limits the kinds of usage that a video can be submit-
ted to. This might likely change in the future if we implement systems for 
collecting video that enable future analytical uses without compromising 
present-d ay copyright. For example, a theater venue could record all the 
performances from a single camera position and only enable those videos 
to be used twenty or fifty years in the future (or to be only used for research 
rather than for distribution).
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Motion as Data • 135
Second Excursion: Quantitative Analysis of Javanese Dance
For the research project summarize here, I worked with Luis Hernández- 
Barraza, a researcher in the field of bioengineering who works on the 
biomechanical analysis of sports. We borrowed some principles from 
sports biomechanics and tried to identify a question that would be of rel-
evance to Javanese dance, and which could be answered with the tools 
at our disposal. We concentrated on character types, since these are of 
enormous importance to the pedagogy, analysis, and enjoyment of Java-
nese dance. As in many Asian theater traditions, characters in Javanese 
dances are defined along different types, each of which follows conven-
tions of movement, speech, dress, and action. There are many dramas 
and dance- dramas in Java, but we focused on Sendratari, since this genre 
includes character types but no dialogue. Sendratari started in the 1960s 
as a touristic performance genre. It is still a dance that is mostly seen by 
tourists, both foreign and local. However, the Sendratari Ramayana has an 
important place in the city of Yogyakarta (where the present author lived 
for some years). Although some people use the moniker “touristic” to 
denote a lack of authenticity, touristic performances are authentic in their 
own right: they have their histories and politics and are presented for a 
specific audience to achieve a particular purpose (Bruner 2005). Different 
groups take turns performing the Sendratari Ramayana at the Prambanan 
temple in Yogyakarta, and there are many aspects of the performance that 
are open to creative improvisation by the different groups. But the rules 
governing character types are relatively stable, and they are amenable to 
biomechanical comparison. There are many qualitative descriptions of 
the differences between character types, and we used the terms and defi-
nitions proposed by the late R. M. Soedarsono (1983), a leading expert in 
Javanese dance and performance. While we think this list is authoritative, 
it should be noted that there is some disagreement among the dance com-
munity about the names of the character types in the Sendratari Ramayana 
and this is something that future research could investigate further. For 
our project, we focused on main character subtypes for male characters:
• Impur (refined- humble) and kagok- kinantang (refined- proud), two 
subtypes of the alusan (refined) type.
• Kambeng (strong- humble) and kalang- kinantang (strong- proud), two 
subtypes of the gagahan (strong) type.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
136 •  TheaTer as DaTa
Fig. 6.6. The theoretical location of all subtypes except Jatayu along the strong vs. refined, 
and proud vs. humble axes.
Figure 6.6 shows how these subtypes can be classified on the humble vs. 
proud, and refined vs. strong dimensions. We were also interested in the 
character of Jatayu, a mythological bird that has been only recently added 
to Javanese dance and which is not found in Seodarsono’s comprehensive 
list of characters.
We had two questions, each of which was explored in a separate 
paper. First, we wanted to analyze whether there is a difference between 
the refined and the strong characters in terms of common biomechani-
cal measurements. Second, we wanted to understand how the visualiza-
tion of biomechanical data could contribute to an interpretive stylometry 
of dance. We invited a professional dancer to perform the same motion 
(standing up from a kneeling position) as it would befit the different char-
acter subtypes, and we used the Vicon motion capture system, which con-
sists of eight infrared cameras and two force plates, to collect kinetic and 
kinematic data at a sample rate of 100 hertz, full technical details are avail-
able elsewhere (Hernandez-B arraza, Yeow, and Escobar Varela 2019). We 
used the Vicon Nexus 1.8.3 and Polygon 3.5 software for data collection 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Motion as Data • 137
and processing. We distinguished between kinetics and kinematics following 
their technical distinction in biomechanics: kinematics refers to motion 
description whereas kinetics attempts to get at the cause of the motion. 
Thus, we obtained the following kinds of data:
 1. Kinetic data: Ground reaction force (GRF) and joint moments
 2. Kinematic data: joint angles and angular velocities
Our hypothesis was that refined (as opposed to strong) character subtypes 
would exhibit lower values for (1) range of motion (ROM), (2) angular 
velocities, (3) ground reaction force (GRF), and (4) joint moments. We 
used one- way ANOVA, followed by a Holm- Sidak posthoc test, to com-
pare the peak vertical GRF, ROM, angular velocities, and joint moments 
of the character subtypes. All significance levels were set at p = 0.05. We 
found that we could only partially accept this hypothesis for the reasons 
summarized in table 6.1 (full analysis at Hernandez-B arraza, Yeow, and 
Escobar Varela 2019). The difference between the character subtypes is 
more nuanced than what we expected. While these results are not impres-
sive, it is important to report negative findings, and resist the temptation 
to overinterpret the data. Many data- driven projects will find no evidence 
of the hypothesis they sought to prove. The road to replicable and incre-
mental knowledge is built by negative results as much as by awe- inspiring 
breakthroughs.
For the second project, I created a website to explore the data inter-
actively through different visualizations that are linked to animations of 
the dancer’s skeleton. This project can be consulted at https://villaorlado.
github.io/dance/html/ (see video 6.1 in this website’s online companion).
Whereas the results in table 6.1 are data- driven, the website enables 
an interactive, data- assisted analysis. In the website, users can load videos 
Table 6.1. Comparisons of biomechanical markers between refined and  
strong characters
Measurement Expectation Result
ROM Lower values for refined Only true for right knee, and wrist
characters
Angular velocities Lower values for refined Only true for left ankle, right knee, right 
characters shoulder, and right elbow
GRF Lower values for refined Only true for the anterior-p osterior (AP) 
characters component of the right leg
Joint moments Lower values for refined Not true for any joint
characters
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
138 •  TheaTer as DaTa
that correspond to each character subtype. They can then choose any joint 
(e.g., left knee) and see how its angle changes along the x, y, or z axis. This 
change is displayed alongside the video, and it allows a defamiliarized way 
of looking at each character subtype. The page for each character subtype 
also includes photographs of the dance, and notes on the interpretation 
of this movement in the Javanese context. The website also allows users to 
compare the motion of the same joint for different character subtypes. For 
example, users might choose to compare the angles of the left knee for the 
impur and the kambeng character subtypes. Lastly, the ROM for the differ-
ent character subtypes can be compared in interactive graphs. These visu-
alizations add context to the data, as users can see how videos relate to the 
data points. The website also has several exploratory features that enable 
users to choose where to focus their attention, which data to compare and 
how to proceed in an “ergodic, yet immutable” sequence (see chapter 3).
The data used here is limited, but these visualizations are easy to imple-
ment and potentially very useful for the interpretive stylometric analysis of 
dance and of motion in theater. Whereas the scrutiny of character types 
only makes sense within certain traditions, combining animations and 
interactive visualizations is useful for the comparative analysis of a wide 
range of movement techniques, from Butoh to Tango. If movement data 
is extracted from video rather than generated from markers or sensors (as 
detailed earlier in this chapter), then the application of similar techniques 
to movement stylometry could grow at a faster pace. However, these tools 
will always be limited to very specific questions and there are many caveats 
to be considered when analyzing movement, as shall be explained in the 
remainder of this chapter.
Concluding Thoughts on the Application of Quantitative 
Stylometry to Dance
Any quantitative study of culture runs the risk of oversimplifying its object 
of study. Dance is in a constant state of change, and this is particularly 
true for the hectic centers of practice and scholarship in Java. Documen-
tation might give a misleading impression of fixity in what is in real-
ity a dynamic field. This impression might be created in good faith, as 
researchers engage in necessary simplifications in the pursuit of specific 
research questions. However, fixity might also help controversial politi-
cal agendas. In the case of Java, fixity of cultural forms has been linked 
to both the Dutch colonial project and to the repressive Suharto dictator-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Motion as Data • 139
ship (Sears 1996). According to Dyah Larasati, the Suharto Regime (1966– 
1998) systematically prosecuted dancers, while at the same time trying to 
craft an unchanging, idealized version of Javanese dance (Larasati 2013). 
Larasati suggests that these phenomena should be read as the two sides of 
the same coin.
The history of dance is always political, as dancing bodies have often 
been contested sites. I believe that the specific things we tried to mea-
sure in the projects reported in this chapter are not contentious, but this 
doesn’t exempt me and my collaborators from considering the unintended 
political ramifications of our work. The availability of technological tools 
should not be used to the detriment of the historical and ethnographic 
analysis of dance and other movement traditions. A data-d riven analysis 
of motion which is grounded on critical realism should always consider 
the historical and political context of any performance tradition. We also 
shouldn’t forget that dance is an embodied and affective practice, for both 
creators and spectators. When describing the complex, funny, and inter-
sectional dances of Javanese dancer Didik Nini Thowok, Jan Mrázek notes 
that dance can’t always be interpreted, labeled, or described: “laughter, 
tears or an upset stomach will be always a better reaction than an aca-
demic paper” (Mrázek 2005, 279).
Digital data might be an even less appropriate reaction. The dangers 
of oversimplification enhanced by technology are apparent to me when I 
speak about the excursions in this chapter to technical audiences. A com-
mon question is whether my data and results could be used to teach dance 
(or puppetry). My answer in the negative often earns me condescending 
smiles. In engineering disciplines, there is a strong emphasis on real 
world applications, actionable knowledge, and problem solving. In the 
projects I envision in this book the objective is not so much to design 
interventions in the world of performance, but rather to find different 
ways of analyzing culturally situated aspects of theater that are amenable 
to data and computation. My collaborators and I are often told by techni-
cal journal editors and conferences attendees that our research needs to be 
more practical and applied. Scientists are often more open than engineers 
to the pursuit of knowledge to gain understanding of a phenomenon, and 
they tend to be more interested in pursuing questions that might not have 
direct application (however, this sweeping distinction between pure and 
applied research must be taken with a pinch of salt, since it is context- 
dependent and changing rapidly).
I would be worried if the methods presented here were to be used 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
140 •  TheaTer as DaTa
to assess dancers or other performers (for intake admissions to dance 
schools or for hiring into professional troupes). I would also dread the 
possibility that mocap data produced under very specific research condi-
tions is one day considered the canonical or authentic version of a given 
dance. This is a real danger of computational theater scholarship that 
must be addressed in relation to all projects. It is important to document 
the conditions under which data was collected, and the assumptions that 
went into the collection of data. This is as important for data- assisted as it 
is for data- driven research.
Anthropologists of dance and scholars of physical theater are— and 
will continue to be—i nterested in things that motion capture and bio-
mechanical analysis will never capture: the social contexts, meanings, 
and controversies of dance and physical theater. As Balme (2015, 165– 66) 
notes, gender and gendered representations are key concerns in current 
dance scholarship. This is an area perhaps not well captured by motion 
data alone. The more enticing promise of quantitative motion analysis 
is that it can be used to understand dance history in more textured ways. 
Tracing similarities between movements in different historical contexts, 
or from neighboring geographic regions could be used to study the evolu-
tion and spread of dance patterns. As this chapter has laid out, such goals 
will require access to large datasets that are sufficiently standardized and 
shared across research teams. Although several researchers are working 
on related fields, the level of community work this would require is still 
far in the horizon. However, the range of projects chronicled in this chap-
ter show the enormous promise for the computational analysis of dance, 
dance- theater, physical theater, or movement in theater more generally.
Code and data: Sample code for this chapter is available at  
https://doi.org/10.3998/mpub.11667458.cmp.41.  
The data used can be downloaded from  
https://doi.org/10.3998/mpub.11667458.cmp.34 and  
https://doi.org/10.3998/mpub.11667458.cmp.35.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
ChapTer 7
Location as Data
Theater is bound in time and space more than many other artforms and 
more attention should be given to the geographic distribution of theater 
performances. This is an area where data is available in large quantities 
and computational methods are uniquely useful. Films and novels depend 
on specific, place-b ound infrastructures: recording studios, book-s igning 
events, awards ceremonies, etc. But the experience of reading and watch-
ing film is more distributed through time and place than experiencing 
theater events live. The visual arts have their own complex relationships 
to time and place, as artworks are created, sold, exhibited, and stolen in 
specific places. But the ontology of most visual artworks is less strongly 
tied to events that are etched in time and place. This chapter looks at com-
putational methods for the analysis of theater events as spatial and spa-
tiotemporal data. It draws heavily on GIS (geographical information sys-
tems) but considers other relevant geographical tools, including analytical 
methods for combining spatial and temporal analysis. I do not consider 
the 3D reconstruction of historical theater buildings, which is a separate 
area of research (Ioannides et al. 2016; Manzetti 2016).
I believe that spatiotemporal analysis is the area that holds the 
most promise for theater studies—i n chapter 6 I said that the analysis 
of motion is potentially the most interesting, but the technologies are 
much harder to use and there are many more roadblocks. In contrast, 
for geospatial and geotemporal analysis, the needed data is more eas-
ily available, and there are many ways to use this data within different 
types of projects. As I have done in previous chapters, I will first describe 
some general methods for visualization and analysis before explaining 
how they can be used within data-d riven and data-a ssisted methodolo-
gies for spatial theater research.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research. 141
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
142 •  TheaTer as DaTa
Spatial Visualizations
Perhaps the most common spatial visualizations are cartographies. These 
visual artifacts show the location of venues and events as points in a 2D 
map. Even moderate amounts of data can reveal interesting patterns when 
placed on a map and answer a host of interesting questions. Where are the 
contributors to a specific theater conference based? What are the national 
origins of the members of a circus troupe? What are the touring locations 
of commercial troupes and how do they differ from those of troupes that 
depend on festival circuits? Plotting points and regions on maps can con-
stitute useful heuristics to identify inequalities or patterns that warrant 
more investigation.
In many cases cartographic visualizations are sufficient to trigger 
questions and suggest new interpretations of available data. However, as 
Jessop (2008) notes, GIS offers much more than just cartographic possi-
bilities. And GIS might not necessarily be the best option. When dealing 
with historical data, for example, gazetteers (databases that include his-
torical name changes) might be more useful for researchers in the human-
ities. Another common way to visualize geospatial data is to use chorop-
leths. In these maps, areas are shaded according to different quantitative 
values, such as the number of performances in a given country or region. 
The first choropleth map was produced in either 1819 or 1826 by Pierre 
Charles François Dupin, to show rates of illiteracy in different regions of 
France (Friendly 2001). Since then, choropleths have been used for a vari-
ety of purposes but they are less common in the humanities than in the 
natural and social sciences. In the humanities, they are most commonly 
used in linguistics (Liao and Petzold 2010). One problem of choropleths is 
that large geographical areas will have larger shapes, which might suggest 
they are more important. Large but sparsely populated provinces will tend 
to take up a large space in the visualizations. A possible solution is to use 
cartograms, which distort the shape of geographical entities (countries, 
districts, etc.) according to a particular metric or statistic. The resulting 
maps show how much a given area contributes to the total for a given met-
ric in a way that is not distorted by the land area of a province or country. 
They are also rare in DH but they have been used in the analysis of film 
distribution networks (Arrowsmith, Verhoeven, and Davidson 2014).
The difference between cartographies and spatial visualizations is not 
cut and dry. All maps are, in some sense, visualizations of geographical 
data that are the product of specific cultural and epistemological assump-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Location as Data • 143
tions. Yet, it is useful to distinguish between maps that show events as 
points and lines and visualizations that include alternative modes of dis-
playing geographical data. The reason for this distinction is that these 
other modes usually require some type of quantitative transformation, 
even if such transformations are not foregrounded (for example, the color 
scheme of a choropleth is the result of a statistical transformation of the 
data). Choropleths and cartograms require statistical transformations of 
numerical data, but these do not exhaust the possible uses of statistics for 
the analysis of locations.
Geostatistical Methods
A common purpose of statistical procedures, as described in earlier chap-
ters, is to estimate the likelihood that an observed pattern is not a fluke, 
but that it reveals the true existence of a phenomenon. In geostatistics, 
this means that researchers are interested in showing that a geographi-
cal observation is due to an underlying cause, and is not random. Statis-
tical methods can be used to test a given set of observations against the 
assumption of complete spatial randomness (CSR). Several statistical 
tests can be used to distinguish between spatial clustering and CSR, such 
as the Clark- Evans test (Ripley 1979) or a chi- square goodness of fit (Roger-
son 1999). I am not aware of any usage of CSR measurements in computa-
tional theater research, but CSR has been applied to study linguistic data 
(Leino and Hyvönen 2008; Kretzschmar 2013) and tombstone locations 
(Streiter et al. 2012). To imagine a potential application for theater data, 
let’s assume that we had data of people attending theater shows in a par-
ticular venue in a given city. When plotting the addresses of the theatergo-
ers, we might see that they tend to cluster around certain neighborhoods. 
However, this pattern might well be random. To more rigorously validate 
the hypothesis that theatergoers tend to live in certain neighborhoods, a 
Clarke-E vans test could be used to distinguish spatial clustering from CSR 
in the addresses.
Another common object of geostatistical inquiry is to determine 
whether geographical proximity affects a given variable, a feature termed 
spatial autocorrelation. In social science, this often means determining 
whether a country’s value for a certain metric (e.g., “indexes of democ-
racy”) is affected by the values for the same metric in the country’s neigh-
bors (Ward and Gleditsch 2019). The most common statistic used for this 
purpose is Moran’s I (Moran 1948), which estimates the probability that a 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
144 •  TheaTer as DaTa
value for a geographical unit is affected by the average value of neighbor-
ing units. In DH this statistic has been used to analyze linguistic varia-
tions across different places (Asnaghi, Speelman, and Geeraerts 2016). An 
example for theater would be to test whether the presence of certain kinds 
of festivals in a country are affected by the presence of such festivals in 
neighboring countries. The excursion at the end of this chapter offers a 
more detailed example of the usefulness of Moran’s I for theater research.
Data- Driven Location Analysis
Geostatistical estimation methods fit squarely within a data- driven para-
digm, but are uncommon in computational theater research. Cartogra-
phies and other spatial visualizations can be used both for data-d riven and 
data- assisted location analysis.
An excellent example of data- driven cartographic work is the research 
by Holledge and her collaborators (2016). As seen in chapter 5, Holledge 
and her coauthors explicitly invoke a scientific paradigm in their work, and 
aim at answering a close question: what accounts for the global success 
of Et dukkehjem? Holledge and her collaborators use Ibsen’s original Dano- 
Norwegian title, rather than the standard English version, A Doll’s House, to 
refer to an entire body of work that comprises translations, adaptations, 
and productions in film and other media (2016, 4). In their analysis of the 
early global spread of the play before World War I, they show that women 
played a fundamental, and previously underrecognized role. While conven-
tional narratives attribute the success to male backers of Ibsen, Holledge 
et al. convincingly demonstrate that women were the entrepreneurial 
force driving the spread of the play. Their success can be linked to inter-
national movements of women’s emancipation, but also to the routes of 
colonization, migration, and specific biographies. Holledge et al. present 
this argument through detailed historical analysis and through a series of 
cartographic representations. They identified key performances in forty- 
six countries across five continents. To find these key performances, they 
listed all contributors in the previously mentioned IbsenStage database 
who were connected with performances in three or more countries before 
1914, and then used this in conjunction with production records to identify 
twenty- two performer-e ntrepreneurs responsible for the spread of the play 
around the world (Holledge et al. 2016, 29). Their geographical visualiza-
tions show the locations of performances as dots, which are connected via 
straight lings that denote sequential travel patterns.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Location as Data • 145
These routes disappeared with the war in 1914. International touring 
will reappear in the late twentieth century, but then it will be driven by dif-
ferent forces. Rather than land and sea routes, troupes will use air travel 
and subsidized international festivals (Holledge et al. 2016, 64–6 5). In 
another chapter of their book that discusses the role of the state in the 
success of the play, Holledge et al. present a map with all productions of 
Et dukkehjem imported to the Ibsen Stage Festival in Oslo from 1990 to the 
present day, next to a map of the international distribution of Norwegian 
government funds promoting Et dukkehjem as a global play text (98). These 
two maps are reverse mirrors of each other. Both use arrows, but their 
direction is reversed—f rom production places to Oslo in one instance, 
and from Norway to places receiving funding in the other. Taken together, 
these two maps show the close link between funding and international 
exposure. The cartographic displays in Holledge et al. where previously 
presented by Bollen and Holledge (2011) to an audience of cartographers. 
In that previous article, they argue that theater maps show the “impor-
tance of distributional flows through time, across geographical space, and 
between artists from production to production” (226). Although their car-
tographic work is primarily data-d riven, the authors also note the impor-
tance of complementing visualizations with in-d epth social, historical, 
and political analysis. Thus, their research occupies a nuanced position in 
the spectrum between data- driven and data- assisted work. Their work is 
data- driven inasmuch as it answers closed questions with verifiable data, 
but it is data-a ssisted when it zooms in and out of different scales, switch-
ing between interpretive historical analysis and the distant vision afforded 
by the maps.
As part of a project that examines the convergence of people, theater 
venues, and media, Circuit: Mapping Theatre Performances in Victoria (Tombe 
et al. 2017) used data from AusStage to visualize connections between 
touring productions among theater venues in Victoria, Australia. These 
connections were overlaid in interactive choropleths that also include sta-
tistics on the ethnic diversity of the cities where the venues are located.
Other types of geographical visualizations and geostatistical analyses 
are less common in theater research. Thus, I turn briefly to film studies 
in this overview to show projects that could provide inspiration for the 
data- driven analysis of theater locations. Arrowsmith, Verhoeven, and 
Davidson (2014) explore different modes of visualizing data from cinema 
venues. Besides choropleths, their work includes circos plots and carto-
grams. Circos plots are a kind of chord diagram which is sometimes also 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
146 •  TheaTer as DaTa
used for network data (such as Caplan’s interactive visualization of the 
Vilna Troupe, discussed in chapter 5). Circos plots show interrelations 
between data in a matrix, and that makes them popular choices for net-
work analysis and for indicating movement across geographical locations. 
Arrowsmith et al. use them to show the influence of different distributors 
in the movement of Greek films across venues in Australia. They also used 
cartograms (see above) to display the different number of cinema venues 
in different Australian provinces from 1948 to 1971.
Quantification is sometimes used for creating choropleths and other 
geographic visualizations, but the explicit description of geostatistical 
estimation procedures is not common in DH. A notable exception is Ver-
hoeven and Arrowsmith (2013), who use Markov chain analysis to test 
whether the distribution of Greek films from different distributors fol-
lows statistically discernible pathways through different exhibition venues 
in Australia. Markov chains are often used for analyzing the probability 
distributions of a sequence of events, such as the pathways that are of 
interest to Verhoeven and Arrowsmith. Markov chain analysis is a complex 
method which is sometimes used for other areas of DH, such as author-
ship attribution (Khmelev and Tweedie 2001) or social network analy-
sis (Warren et al. 2016). This procedure could also be applied to theater 
research, for example to analyze the touring circuits of theater companies.
Data- Assisted Location Analysis
The spatial humanities have led to the conceptualization of many theoreti-
cal positions that support the usage of GIS and mapping within interpre-
tive epistemologies: “little g” GIS, deep mapping, and thick maps. There is 
some overlap between these concepts, but taking them in turn will help 
clarify the possibilities for the data- assisted analysis of theater locations.
Bodenhamer (2013) sees a major challenge in the clash between the 
epistemology of cartographic GIS and the interest of historians. Like most 
humanists, historians are interested in what he terms “extractive scholar-
ship”: scholars constantly shift perspectives “in the pursuit of the fullest 
possible understanding of heritage and culture” (5). This means that, tra-
ditionally, narratives are used to construct and present arguments. Narra-
tives enable the “interweaving of evidentiary threads, each of which can be 
qualified, highlighted or subdued through a variety of literary devices”. In 
contrast, GIS privileges a world that values “authority, definition, and cer-
tainty over complexity, ambiguity, multiplicity, and contingency,” the very 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Location as Data • 147
things that narrative enables and that historians value (7). Bodenhamer’s 
suggestion is to construct “deep maps,” GIS artifacts that are interwoven 
with critical commentary to enable the narrativization and contextualiza-
tion of spatial data.
Klenotic (2011) has proposed “little g” GIS, as opposed to “big G” GIS, 
to signal that researchers need not use the full range of technical options 
available in GIS, but that there is value in “partial, self- taught, bottom-
 up applications of GIS” (59). Even basic, easy-t o- use tools are powerful 
enough to trigger interesting questions, and visualize data in ways that 
problematize previous assumptions about spatial knowledge. Klenotic’s 
approach is modeled after Knigge and Cope’s (2016) notion of grounded 
visualization, which embraces messy, exploratory, and iterative visualiza-
tion processes that are amenable to multiple perspectives. Klenotic sug-
gests that applying the notions of grounded visualization to GIS means 
developing maps and visualizations with piecemeal iterations that require 
constant reanalysis of the data. Klenotic’s view embraces a progressive 
and recursive ethos of praxis. “Deep” and “thick” maps introduce other 
theoretical dimensions. Thick mapping is the name of the project by Todd 
Presner and his collaborators. The name is a direct allusion to Geertz’s 
thick description. This approach acknowledges the social construction of 
maps as unstable, culturally specific objects that “make claims and har-
bor ideals, hopes, desires, biases, prejudices, and violences” (Presner, 
Shepard, and Kawano 2014, 15). Considering maps as contingent invites 
makers of GIS devices to treat them as useful but transient representa-
tions that don’t correspond neatly to an external reality. This enables an 
interpretive usage of GIS that tells complex, multilayered stories.
Deep maps also aim to produce multilayered descriptions of places, 
by integrating data and stories from different sources. Although this 
term activates similar references to thick mapping, the term reveals a dif-
ferent intellectual heritage. Deep maps show inherent contradictions by 
reflexively including different kinds of media, in ways that draw inspira-
tion from critical geography (Bodenhamer, Corrigan, and Harris 2013). 
Deep maps aim to use cartographic and geospatial conventions while also 
drawing attention to the problematic, positivistic assumptions of GIS and 
related technologies.
There are some nuanced distinctions among the concepts just 
reviewed—“ little g” GIS emphasizes process, thick maps highlight 
the constructedness of maps, and deep maps stress the multiplicity of 
sources—b ut all of them ultimately provide epistemological frameworks 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
148 •  TheaTer as DaTa
for using GIS technology within data-a ssisted theater research. An exam-
ple is the work of Bench and Elswit (2017), who are interested in using 
databases and maps to study the international tours of dance companies. 
They argue that tours are central to studying the global nature of dance, 
and that this endeavor requires a combination of historical research and 
digital methods. This combination triggers many fascinating questions 
and lines of inquiry, some of which I summarize below:
 1. How do key choreographic works become canonical, even when 
other works were almost as commonly performed (for example, in 
relation to Pavlova’s “Dying Swan,” 583)?
 2. Are aesthetic inclinations reflected in different modes of touring? 
For example, are the routes and repertoires of American and Rus-
sian modernisms substantially distinct from each other (587)?
 3. What does the data on performers reveal? Even basic data such as 
nationality has the potential to become a window into historically 
c ontingent classifications. For example, in a 1941 American Cara-
van Tour, two German- born dancers were classified as stateless 
(582).
 4. How do different geopolitical events affect touring? The most 
interesting examples are not necessarily big conflicts, but smaller 
scale wars such as the Ecuadorian- Peruvian war of 1941 which 
disrupted the travels of the American Ballet Caravan (590).
Bench and Elswit explore these avenues of inquiry and present them 
together with different cartographies of travel, where tours are repre-
sented as point- to- point lines, which are also offered through online 
visualizations.
On the surface, these visualizations might appear similar to the work 
of Holledge et al., but I believe that Bench and Elswit are more interested 
in using maps to find new ways of thinking about touring rather than 
to answer closed questions. The questions listed above are not entirely 
answerable in terms of the data, but they could not have been formulated 
without systematic, careful data collection. Bench and Elswit adamantly 
highlight the pitfalls of reducing the politics of touring to lines and points 
in a coordinate system, and seek instead “scholarly modalities” that move 
between stories and data in “dynamic spatial histories of movement” 
(575). They argue for scalability, moving between the close and the dis-
tant, the single data point and the aggregate. Movement across scales, as 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Location as Data • 149
I have argued in this book, is a key marker of data- assisted research. To 
achieve this movement when working with maps, Bench and Elswit stress 
that it is important to never use tools for their own sake, but to critically 
deploy them to rethink questions relevant to dance history. They right-
fully note that a choropleth would be ill- suited to their needs, as they are 
more interested in travel as process rather than in the number of events 
in a given region. In chapter 3, I have also noted how choropleths can be 
misleading, by referencing two projects that map lynchings in the United 
States (Hepworth and Church 2019).
One fascinating perspective for interactive, data- assisted location 
analysis is to use maps as interfaces for multiform materials to enable 
alternative modes of scholarly communication. Interactivity is particularly 
important for bringing depth and thickness to maps. Maps as interfaces 
enable their makers to chart multiple, contradictory voices. In so doing, 
they critically reimagine common technologies in modes that transcend 
conventional mapping tropes. Maps as interfaces are prime examples of 
the intermedial scholarship I described in chapter 3.
While not specifically a theater project, The Digital Literary Atlas of 
Ireland (Travis and Breen 2017) is an excellent example of an interac-
tive, deep map that deals with the lives of some playwrights. The project 
includes maps with geographical information for fourteen writers as well 
as several themed maps: Emigration, House- Island and Provincial Town, Dublin 
Bricolage, and Northern Impressions. The maps are also linked to pages with 
more detailed information about each of the writers. Although this proj-
ect has comparatively little data, its strength is the depth and context of 
its resources, which can be used to examine different aspects of the writ-
ers’ lifepaths. For example, Travis (2015) shows how the Atlas can be used 
to explore Beckett’s connection to landscape, in ways that are deforma-
tive and ergodic (I have discussed both concepts in chapter 3). Designing 
digital platforms that enable ergodic experiences, and which are expressly 
defined as deformances, is particularly important for work on writers such 
as Beckett, who challenged many assumptions about narrative in his 
work. It would be a contradiction to impose a closed realist space as the 
sole mode to access Beckett’s life and work. The Atlas aims to allow “bri-
colage by interactively juxtaposing different scales of time, space, text and 
image” (Travis 2015, 223). This approach helps resituate cartographic 
questions as provisional and subjective, and to reconsider maps as gen-
erative objects that trigger new interpretations.
Earlier I described the film venue visualizations of Arrowsmith, Ver-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
150 •  TheaTer as DaTa
hoeven, and Davidson (2015) as examples of data- driven work. However, 
they have also invented the petal plots, which address the unique challenges 
of their datasets and which are closer to the playful transformations 
required by data- assisted visualization. The dataset that gave rise to these 
diagrams showed multiple attributes of cinemas in Melbourne. The data 
included the lifespan of a cinema, transformations in seating capacity, and 
location. The petal plot is able to represent all these different dimensions 
at once. Each cinema is visualized as a curved line. It’s location along a 
circular axis indicates its direction from the city center of Melbourne. The 
beginning and end points indicate the years in which the cinema began 
and ended operation. The curve of the line shows the change of cinema 
capacity—a  convex curve indicates that a given cinema first increased and 
then decreased its seating capacity. These diagrams might not be readily 
applicable to other kinds of data. But the lesson here is that sometimes we 
should develop visualizations that are specific to our problems rather than 
merely copying visual tropes from scientific disciplines. This is one of the 
most effective examples I know of a performative visualization that echoes 
Drucker’s call for developing new humanistic visual tropes. Petal plots 
effectively deploy new visual metaphors but they also preserve the preci-
sion and readability of statistical graphs. I was directly inspired by their 
work for developing the kayon plots I describe in chapter 6. Davidson and 
collaborators convincingly show that the nuance of their historical data 
would have been lost had they employed other kinds of visualizations.
Excursion: The Geographies of Wayang Kulit
Wayang kulit performances (which were described first in chapter 5, and 
also mentioned in chapter 6) are very common in Java. My Facebook feed 
and several WhatsApp groups that I belong to are full of announcements 
of upcoming performances and discussion of recent performance events. 
It is surprising that such an old form continues to exist to this day and 
age. Wayang is at least one thousand years old, and a performance is full 
of allusions to a rich oral literature. Arps (2016) says that part of the plea-
sure of watching a wayang show is “distinctly philological,” as a wayang 
aficionado delights in catching references to old performances, or to dif-
ferent variations of the musical repertoire associated with specific artists 
(28). A traditional wayang show lasts all night (seven to eight hours). And 
even if people are free to come and go— and most spectators typically 
leave before the show concludes— wayang invites us to rethink temporal-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Location as Data • 151
ity in ways that are not fully compatible with a modern concept of time, 
linked to ideas such as efficiency and opportunity costs. Jan Mrázek (2019) 
quotes his wayang kulit teacher saying that the nights lasted longer in the 
past (56).
Wayang, in other words, is not for the faint of heart. And it also does 
not seem particularly compatible with the demands of contemporary, 
pressed- for- time urban living. Wayang aficionados in Java often long for 
the past and are anxious about the future, and they often remark, in con-
versation, that wayang was more common in the past. Anxiety about the 
future of wayang is perhaps as old as the written record of the form itself. 
Thow (2018) has traced how gamelan music (which is an integral part of 
wayang) has been said to be disappearing and degrading for at least a 
century. In these nostalgic expressions, geography plays a role too. Those 
afflicted with wayang nostalgia say that people in the cities don’t have time 
to enjoy wayang anymore, and that wayang is eminently a rural form (see 
Mrázek 2019, 98–1 07 for an excellent treatment of nostalgia and wayang). 
But is it really a rural form? This is where the data comes in.
Since the middle of 2015, Imam Maskur and his organization have 
been systematically collecting information on wayang kulit performances 
all over Java, Indonesia, which is then posted to the website https://
kluban.net at the beginning of every month. For this he mobilizes a large 
group of volunteers, managers, and artists who mainly communicate over 
social media. Their main aim is to let people know which performances 
are taking place in the month to follow, but at the end of each month they 
also correct the records, as many performances get canceled and others 
are only confirmed a few days before the actual event. My research assis-
tant Wejo Seno Nugroho ran a survey over one year in order to verify the 
accuracy of randomly selected performances and locations. We were sur-
prised to conclude that the website is ~97% accurate and captures ~85% 
of all performances. However, this is just a preliminary assessment. The 
performance records are most likely biased toward self-r eporting by 
slightly more famous artists, and very local performances in remote vil-
lages might go unreported. Thus, the results that follow should be taken 
with a pinch of salt. Since the data has only been collected since 2015, it is 
hard to draw historical comparisons. Previous data were collected in the 
1960s and 1970s but they were based on surveys and it is hard to estimate 
how accurate they were. But even if the historical question is not acces-
sible to us, we can use the data from the website to ask whether wayang is 
today primarily a rural form.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
152 •  TheaTer as DaTa
Fig. 7.1. Scatter plots with regression lines for average number of performances per month 
compared to population density, land area, and total population per regency.
The distinction between urban and rural areas in Java is not hard and 
fast, so I have refined the question: are performances more common in 
less densely populated areas? Truisms spoken by those longing for the 
past would have us believe so. But the data from Kluban shows that there 
is no correlation between population density and the number of perfor-
mances, as shown in figure 7.1. As the unit of analysis, I used data on 
each regency (kabupaten in Indonesian, similar to a municipality). I used 
the census data from 2016, since this was the year for which data was 
available for all regencies from official Indonesian sources (Badan Pusat 
Statistik 2019). Therefore, I also used the data for performances in 2016 
for this calculation. As I will show later, the number of performances has 
remained stable over the years under consideration, so it is reasonable to 
extrapolate from 2016 data.
As figure 7.1 shows, there is no correlation between the average num-
ber of performances in a regency and the regency’s size, population den-
sity, or total population. So, there must be something else explaining why 
some regencies have more performances than others. Visualizing the dis-
tribution of performances in a map seems to suggest that areas with more 
performances tend to cluster together. Figure 7.2 is a choropleth of the 
regencies in Java colored according to the total number of performances. 
The colors are assigned based on the decile in which each regency is 
found. A decile is a division of the data into ten bins, where each bin holds 
an incremental 10 percent of the data. Thus, the regencies with the darkest 
hue are the ones in the first decile (from the first to the tenth percentile).
The map shows that the places with most performances are located 
at the peripheries of Central Java, in provinces that border West Java and 
East Java. As we saw earlier, Moran’s I is the kind of statistical measure 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Location as Data • 153
Fig. 7.2. A choropleth map of Java, with districts colored by average number of performances 
per month (grouped into deciles).
that can be used to further characterize this type of geographical distribu-
tions. Moran’s I tells us how much a variable (number of performances) is 
explained by its “spatial lag,” or the number of performances in neighbor-
ing regions. Using the built in Moran’s I function of the PySAL package 
(Rey and Anselin 2010) I obtained a Moran’s I value of 0.173 with a p- value 
of 0.016 for my data. This means there is a reasonably strong spatial auto-
correlation in the data. Understanding the meaning of Moran’s I is clearer 
when it is presented in visual form. Figure 7.3 shows a scatter plot of 
each provinces’ total number of performances (x axis) against the average 
value of its neighbors, also known as its spatial lag (y axis). Neighbors 
are defined here as those regions which border the regency in question. 
Using a nomenclature borrowed from chess, this is called queen contigu-
ity (as opposed to bishop contiguity). All values are visualized in a scatter-
plot. The regression line indicates the correlation between the number of 
performances and its spatial lag, and Moran’s I can be understood as the 
slope of this line (which is 0.173 in this case).
To calculate the p-v alue of Moran’s I, or the likelihood that this result 
is merely the product of chance, the most common procedure is a boot-
strapping method, where random permutations are generated of the same 
data (in this case 999 permutations were created). Then we can count how 
many of those permutations have a slope of 0.173 or larger, which in this 
case is less than 1.6% of all cases (more formally a calculation derived 
from permutations is called a pseudo p-v alue). If researchers aim to repli-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
154 •  TheaTer as DaTa
Fig. 7.3. A scatterplot of each regency’s total number of performances (x axis) against the  
average value of its neighbors, also known as its spatial lag (y axis).
cate this result, they will get slightly different p-v alues, since the permuta-
tions will be different each time. But they should still be reasonably close 
to the numbers reported here. Figure 7.4 shows the distribution of these 
results as a shaded curve (a KDE, the same technique described in chap-
ter 6). The dotted line shows the actual observed slope of 0.173. Although 
not huge, the p- value increases our confidence that the results we are wit-
nessing are not a random effect, but that they reveal the presence of a real 
underlying phenomenon.
Thus, we can assert that regencies in Java are more likely to have more 
performances if their neighbors also have more performances. But this 
is still a very general observation. To delve deeper, we can see the areas 
where this effect is the most pronounced. For this we can calculate the 
local Moran’s I, which gives us a sense of the “hot areas” where the spa-
tial lag is the most pronounced. I used PySAL for this purpose again and 
found that the provinces with the highest local Moran’s I are: Cilacap, 
Banyumas, Kebumen, and Ciamis. These regions might not mean much 
to the reader. And this might be true even if they are scholars of Javanese 
performing arts, as these are regions which are rarely the focus of aca-
demic attention. The bulk of local and foreign scholarship focuses on the 
Central Javanese court cities of Surakarta and Yogyakarta (where I have 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Location as Data • 155
Fig. 7.4. The results of 999 random permutations of the data. The shaded area shows the ker-
nel density estimate of these results, and the vertical dashed line shows the actual observed 
Moran’s I of 0.173.
also done most of my research). But the Kluban data shows that scholars 
could extend their attention to other places. I have decided, upon look-
ing at these charts, to embark on further ethnographic research in these 
other regions. I do not yet have enough information to say why there are 
more performances there. But this data-d riven analysis has shown a blind 
spot of current scholarship. This is similar to what Miller (2017) suggests 
with his analysis of underrepresented American dramas in theater source-
books. Looking at the actual data reveals a trend different from what 
scholarship alone suggests.
As noted earlier, the data was obtained from https://kluban.net. How-
ever, the data is not presented there in a machine- readable format, so I 
had to scrape it using a common Python library (as other procedures, this 
is described in more detail in appendix A, and the curated dataset is avail-
able for download in this book’s companion site). I then wrote Python 
routines to check the data for accuracy and to detect spelling errors. For 
each performance, the website states the date, the name of the performer 
(dhalang, see chapter 6), and the location. Sometimes the location is very 
precise and includes a specific address, but this is only true for less than 
40 percent of all records, so I had to resort to a slightly less granular cat-
egory (regency) for comparisons. Other conclusions could be drawn if I 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
156 •  TheaTer as DaTa
had access to the specific coordinates of the event for each of the 11,255 
records. But here, as is often the case when working with data, there is a 
trade- off between accuracy and granularity. A researcher faces decisions 
and must aim to communicate them to others when the research is made 
public. One advantage of using regencies is that there are population sta-
tistics available for analysis. The number of performances per month per 
regency, as well as the population statistics can be easily traced to other 
sources, and they help move this project in the direction of replicable 
research (this also means that the conclusions reached here might be later 
challenged by other researchers).
Other types of information are absent or inconsistent in my wayang 
dataset. For example, most records don’t state the reason for a perfor-
mance (marriage, institutional celebration, village cleansing, etc.). This is 
something I would wish to know more about, but my dataset alone won’t 
shed light into this issue. The data does include the name of the dhalangs, 
but this also carries problems. Many dhalangs have similar names, and 
name disambiguation is very tricky except in the case of the most famous 
performers. I identified between 4,000 and 5,000 possible distinct names 
in the dataset, but further disambiguation would require other kinds of 
research, beyond purely computational techniques as I would most cer-
tainly need to interview people directly. Thus, I have avoided looking at 
questions about who performs where, tantalizing as they are. But there is 
another variable in the data which is accurate and precise: the dates.
When I visualized the dates as a time series, I was struck by the regular-
ity of their patterns. Figure 7.5 shows the total number of performances 
per month for 2016, 2017, and 2018. There are less performances in the 
fasting month of Ramadan (since only ritual performances are allowed 
then). This lunar month moves from one year to the next, but in the data 
graphed in the figure it fell in May and June, as can be seen from the dip in 
the performance numbers. The peaks are connected to the anniversary of 
the Indonesian Independence (August 17), the first month in the Islamic 
New Year (which also moves, but fell around September and October in 
the years when the data was collected) and the anniversary of the inscrip-
tion of wayang into UNESCO’s representative list of intangible heritage 
(the date itself is November 7, but the preceding week usually sees a 
surge in performances). This data can also be combined with geographi-
cal information. Even though the total number of monthly performances 
is constant over the years, some regencies see more performances than 
others in specific months. To visualize this, I created a choropleth map 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Location as Data • 157
Fig. 7.5. Performance totals per month across three years: 2016, 2017, and 2018.
for each month, and then joined the results into an animation that can be 
consulted in the book’s companion website (video 7.1).
This data-a ssisted animation, more performative and evocative than 
the static choropleth, shows Java as a “beating heart,” as the highest con-
centration of performances flows back and forth between the court cit-
ies of central Java and the eastern regencies (the ones in the local Moran 
hotspot mentioned earlier). This analysis of wayang kulit performance 
data is highly specific, but it shows a way of working with large amounts 
of location data, and how it can be analyzed through different kinds of 
visualizations and statistics. Similar research is certainly possible for 
many other theatrical forms around the world and I think it is likely that 
an explosion in this type of research will be seen in the near future.
Further Considerations: Time and Place
In this chapter I have spoken mostly about geospatial rather than geotem-
poral data, even if time features into many of the projects described earlier. 
All the pioneering projects in theater studies mentioned above accounted 
for time in one way or another. Bollen and Holledge (2011) superimposed 
networks into their cartographic displays to show the temporal sequence 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
158 •  TheaTer as DaTa
in which performances followed one another. Bench and Elswit (2017) 
represented tours as point to point lines that can be animated in the online 
versions of their maps, to show the order in which the tours were carried 
out. This consistent attention to time is certainly linked to the nature of 
theater, and should attune us to the possibilities for geotemporal analy-
sis in the future. Projects in other areas of DH can also provide inspira-
tion. Starting Out from 23.5°N (Academia Sinica Center for Digital Cultures 
n.d.) explores the life and work of Taiwanese artist Chen Cheng-P o (陳澄
波, 1895–1 947) as he moved between China, Taiwan, and Japan in the first 
half of the twentieth century. This project combines maps-a s- interfaces, 
interpretive resources, controlled-v ocabularies, and interactive timelines 
to explore the life of an influential figure through deep maps. This and 
many other GIS projects have been affected by a change in the terms and 
conditions of the Google Maps API which took effect on June 11, 2018 
(Google Developers n.d.), a time when I was doing research for the pres-
ent book. This should raise the alarm bells of depending on commercial 
providers for critical DH architecture, a theme further explored in the 
final part of this book. There are other, more robust platforms that offer 
some degree of independence from commercial providers, such as Neat-
line (Nowviskie et al. 2013). It is also possible to build custom interfaces 
with Leaflet (Open Source Community [2010] 2018) using only data from 
OpenStreetMap (OpenStreetMap Community 2004).
In this book, I have argued that data- driven and data- assisted meth-
odologies can coexist to address different aspects of the same project. In 
the computational analysis of location data, we can seek statistical rigor 
to answer discrete questions, which can in turn be expanded and contex-
tualized in complex narrative displays that don’t take mapping tropes for 
granted, but that examine them as culturally situated objects. Location 
data can be prodded and deformed, in ways that situate observers front 
and center and which guide nuanced interpretive perspectives. But this 
data can, in turn, be used to test patterns. We can readily imagine complex 
hermeneutic circles, where the results of statistical patterns guide further 
situated analysis, and where deformances suggest new empirical investiga-
tions. These possibilities are contingent on the potential of our work to be 
accessible and reusable in the future. As hinted in the discussion above in 
relation to Google Maps, we need to pay attention not only to epistemo-
logical possibilities but also to the technical and institutional conditions 
that can limit or enable the future of computational theater research. This 
is the subject of the next, and last, part of this book.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Location as Data • 159
Code and data: Sample code for this chapter is available at  
https://doi.org/10.3998/mpub.11667458.cmp.42.  
The data used can be downloaded from  
https://doi.org/10.3998/mpub.11667458.cmp.36,  
https://doi.org/10.3998/mpub.11667458.cmp.37, and  
https://doi.org/10.3998/mpub.11667458.cmp.38.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
parT 3
Ensuring the Journeys Continue
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
ChapTer 8
The Imperative of Open and Sustainable Data
The Australian Performing Arts Database (AusStage) is one of the most 
ambitious performing arts databases ever undertaken. It aims to capture 
all live theater performances for the entire history of Australia, as well as 
performances by Australian artists around the world (AusStage 2013). The 
project is remarkable also for its clear and robust data model and the fact 
that the entire dataset is openly available. Researchers can download and 
reuse the data as long as they provide proper acknowledgment. The prog-
ress of data-d riven and data-a ssisted theater research will depend on the 
well-p lanned openness of projects such as this, and it will be underwritten 
by institutional infrastructures that can keep these kinds of projects avail-
able for future generations of scholars.
The Comédie Française Registers Project is also extraordinary in 
terms of its scope and openness, as it enables users to download and 
reuse data on the performances of the Comédie Française from 1680 
until 1791 (Biet et al. 2015). There are many other carefully planned data-
base projects around the world, such as IbsenStage (modeled explicitly 
in the footsteps of AusStage), Records of Early English Drama (Black et 
al. 2017), 19th Century Acts (Gonzalez et al. n.d.), the Digital Yiddish 
Theater Project (Baker et al. 2019), Reseña Histórica del Teatro en México 
2.0–2 .1 (Historical Theater Reviews in Mexico 2.0–2 .1) (Franco 2020), 
Base de Datos de Comedias Mencionadas en la Documentación Teatral 1540–1 700 
(Database of plays mentioned in theater records: 1540–1 700), (Ferrer 
Valls 2019), and the Cuban Theater Digital Archive (Manzor, Rimkus, 
and Ogihara 2013). Some projects integrate textual sources with other 
media, such as the Map of Early Modern London (Jenstad 2011), which 
combines maps and textual annotation. Other projects, such as the 
Hemispheric Institute’s (2008) Digital Video Library, the Asian Shake-
speare Intercultural Archive (Yong et al. 2015), and the Digital Dance 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research. 163
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
164 •  TheaTer as DaTa
Archives (Fensham 2016) contain extensive video material, interactive 
visualizations, and copious scholarly annotations.
In many of these projects, data is bound up in other kinds of artifacts, 
as it is used for visualizations or it constitutes the backbone of archival 
projects (and it is not always readily available for download). Different 
projects also vary in their policies for sustainability and reusability, and 
in how clearly they display such information. A comprehensive survey of 
such policies is available elsewhere (Escobar Varela and Lee 2018). Proj-
ects have good reasons for sharing data in different ways. In some cases, 
this is limited by copyright restrictions or other institutional constraints. 
Sometimes sustainability policies are not necessarily explicitly stated. This 
is understandable, as many projects are new and struggling to find finan-
cial and technical support to ensure their subsistence. But it poses serious 
problems for the continued existence of computational theater research.
The bulk of this book deals with methodology and epistemology. Yet, 
the implementation of methods depends on material and institutional 
foundations, and it is to these that I now turn my attention to. I will con-
sider three aspects of data: modeling, sharing, and preservation. I will 
dedicate some attention to each of them in isolation, but my larger aim 
is to show how these aspects are deeply interconnected. The way data is 
modeled limits and constraints how it is to be shared and preserved. Shar-
ing without preservation is meaningless. And preserving things that won’t 
be shared (at least eventually) would be preposterous.
Data Models and Metadata
Imagine an assiduous theater goer that keeps a notebook with reflec-
tions on every theater show she watches. The notes contain some factual 
information— name of the performers, time, maybe even ticket price— 
but these are interspersed with her thoughts on the performances. Now 
let’s imagine that she wanted, after several years of note-t aking, to turn 
this notebook into a digital database. For this, she would need to explicitly 
formalize the kinds of things she is capturing in her notebook. She thus 
designs a spreadsheet with the following column names: creative team, 
title, venue, and comments. Each performance’s data will be entered in a 
separate row. This formalization is a data model, albeit a relatively simple 
one. To see where it falls short, let’s imagine that she wanted to compare 
her data with someone else’s. This other person included two categories— 
performers and director— to describe the information that the first the-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Imperative of Open and Sustainable Data • 165
atergoer included within the single category of “creative team.” If they 
wanted to make their data more directly comparable, both people could 
settle on a shared model. But let’s say that both projects are too far along 
their development and that standardization is impractical. Each could 
continue with their own system, but establish ways to map one onto the 
other. For example, the theatergoers could note down that “creative team” 
in system A includes “director” and “performers” in system B, and that 
they are all instances of a larger category of data called “people,” which 
have properties such as dates of birth and names. In technical parlance, 
this means that the two systems would be interoperable. The systems will 
remain distinct but formal rules would map one to the other. These rules 
can also be said to be part of a more general data model.
In both situations—s tandardization and interoperability—o ne could 
decide to use an existing data model or to develop a new one. The latter 
strategy would add more nuance, but would require time and effort. More 
importantly, it would need to be adopted by other people who might not 
be easily persuaded to change to a bespoke, little- used system that is not 
theirs. Using an existing data model reduces cost and effort, but necessar-
ily forgoes some nuance and granularity. In some of the excursions in this 
book, I have described Javanese wayang kulit performances. The dhalang 
is the creative leader of a wayang show, and he or she speaks all character 
parts, animates the puppets and cues in the musicians. One could describe 
the dhalang as a director or a puppeteer, but I think that both terms fail to 
capture the actual role of a dhalang in a wayang kulit show. To describe a 
wayang performance, I could use a theater specific data model, or I could 
just refer to a more general model such as Dublin Core (DC), a widely used, 
small set of vocabulary terms that can be deployed to describe resources. 
The core element set consists of 15 terms (description, format, identifier, 
language, publisher, relation, rights, source, subject, title, and type), and 
several dozen properties, classes, datatypes, and vocabulary encoding 
schemes. Using DC, I could specify the role of a dhalang as “creator.” This 
is indeed what I did in the Contemporary Wayang Archive (CWA). Table 
8.1 shows an example of a wayang kulit performance described according 
to the DC elements in the CWA.
The elements modeled by DC are more formally called metadata ele-
ments. Metadata is data about data. In the introduction, I used Pomer-
antz’s definition of data as a potentially informative object. For him, meta-
data is “a statement about a potentially informative object” (Pomerantz 
2015). A good data model includes some metadata. Without it, a model 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
166 •  TheaTer as DaTa
would be much less useful. Continuing the example above, let’s consider 
an entry in our imaginary theatergoer’s diary for Robert Lepage’s Far Side of 
the Moon (2002). The director’s name (Robert Lepage) is a data point. But 
it does not belong to the same conceptual category as Far Side of the Moon, 
even though both are text strings. A human perusing the database (at 
least one familiar with conventions of play titles and Francophone nam-
ing conventions) would know which is the director and which is the title. 
But a machine would be unable to do this without some formal mecha-
nism. Metadata provides such a mechanism, as it describes which data 
point belongs to which category. For these descriptions to be maximally 
useful, the relationships between different types of data statements need 
to be explicitly defined. One option is to use a formal language such as 
the Resource Description Framework (RDF). In the RDF, relationships 
between data and data categories are made through subject- predicate- 
object statements, known as triples. These statements indicate relation-
ships between different data elements. An example of the previous record 
for Enthus Susmono’s Dewa Ruci (2008) that uses RDF conventions is 
available in XML format at http://cwa-web.org/en/metadata/DewaRuci.
xml. As seen in chapter 4, XML is the same language used to encode TEI 
files. The excerpt below gives a sense of how it is structured:
<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-
rdf-syntax-ns#" xmlns:dc="http://purl.org/dc/ 
elements/1.1/">
<rdf:Description rdf:about="http://dublincore.org/">
<dc:title>Dewa Ruci</dc:title>
<dc:identifier>http://cwa-web.org/en/DewaRuci</
dc:identifier>
<dc:date>2008</dc:date>
<dc:publisher>Contemporary Wayang Archive</
dc:publisher>
<dc:creator>Enthus Susmono</dc:creator>
<dc:language>id</dc:language>
</rdf:Description>
</rdf:RDF>
Metadata operates at different levels of granularity. In the CWA, the 
metadata describes a performance. But the performance itself is obviously 
not part of the archive, which includes only a video recording of the per-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Imperative of Open and Sustainable Data • 167
Table 8.1. A DC metadata record for Enthus Susmono’s Dewa Ruci (2008)
Title Dewa Ruci
Identifier http://cwa-web.org/en/DewaRuci
Data 2008
Publisher Contemporary Wayang Archive
Creator Enthus Susmono
Language ID
Description In this performance, Bima is on a spiritual quest to find the meaning of life. 
His teacher Durna tries to trick him by telling him the answer will be found 
at the bottom of the ocean. Bima dutifully follows, defeats a dragon that 
lives in the ocean and finds a miniature version of himself, Dewa Ruci, from 
whom he receives a lesson on the spiritual meaning of life.
formance. A more complex metadata standard could have been used, such 
as the CIDOC-C RM (the Conceptual Reference Model of the Comité Inter-
national pour la Documentation [International Council for Documentation]). 
The CIDOC-C RM is often used to capture information on intangible cul-
tural heritage, and has been proposed for documenting theater perfor-
mances (Pendón Martínez and Bueno de la Fuente 2017). The CIDOC- 
CRM could be used to describe multiple instances of a performance and to 
indicate how multiple video recordings refer to them. But even this is not 
extremely detailed. We could imagine using an even more granular data 
model where the content of each frame of each video is described in detail. 
But there would need to be a category for each thing in the video. One can 
easily imagine how quickly this would run out of control— and Borges’ 
(1946) short story “Del Rigor en la Ciencia” (On Exactitude in Science) 
comes to mind. The characters in this story are so obsessed with creating 
a map that fully represents reality— every bird, every leaf on a tree— that 
the resulting map ends up being so comprehensive that it overlaps with 
reality, and people don’t know any more whether they are living in the 
map or in the reality it seeks to represent.
The limitations of any metadata model for the humanities are obvious. 
Calling the dhalang a creator misses the cultural-s pecific aspects of a dha-
lang’s role in performance, as he or she is someone who is often reinter-
preting oral traditions rather than making entirely new works. Here I am 
advocating for using metadata models for the sake of sustainability, but it 
is important to note that there are serious limitations with these systems. 
Brown and Simpson (2013) note that standards limit the ability to make 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
168 •  TheaTer as DaTa
the nuanced statements that characterize humanities research. They use 
the example of a “writer” (Michael Field), who is not an actual person but 
the pen name used by the authorial collaboration between two women in 
the Victorian era (Katherine Harris Bradley and Edith Emma Cooper). The 
nuances of their collaborations can’t be captured by existing data models 
and, as Brown and Simpson note, this is not an extreme case, but a com-
mon example of the kind of careful attention to specificity that the human-
ities require.
When choosing data models and metadata standards, there is always 
a tradeoff. Flanders and Jannidis (2015) suggest different strategies for 
dealing with this tradeoff. When it comes to the purpose of a model, they 
distinguish between curation-d riven and research-d riven data modelers. 
The first group wants to achieve models that are as generalizable as pos-
sible, finding the most widely applicable common ground. The second 
group is interested in formalizing very specific research ideas for spe-
cific purposes and narrow domains. The CWA follows a curation-d riven 
approach. My objective in relying on a widely used, though less granular 
standard, is to make it easy for people to find and cite these resources, as 
the DC records can be easily ingested by library systems. But I might one 
day want to develop a research-d riven model, one that is aimed specifi-
cally at video analysis. I might then wish to manually annotate the video 
recordings and describe the actions that take place in the videos. The 
same action can be described in a wide variety of ways. Whichever model 
I end up developing and using will betray specific prejudices and prefer-
ences and will reflect my interpretation of what matters in a performance. 
As Jannidis and Flanders (2015) note, most researchers in DH know that 
models are social constructs rather than representations of objective real-
ity, but this doesn’t mean all models are equally good. Thus, Jannidis and 
Flanders suggest that models can be assessed in terms of persuasiveness, 
intellectual elegance, or strategic value. Unlike natural objects, digital 
artifacts “are created with a purpose by identifiable agents and they have 
a history which is part of their identity” (235). We thus need to represent 
not only the history of the artifact, but also the history of the ways in which 
it has been described and contextualized— and there is always some degree of 
uncertainty in these histories.
Bollen (2017), who has been instrumental in the development of 
AusStage, analyzed twelve influential, large-s cale theater databases and 
found that they tend to coalesce around five aspects: places, people, com-
panies, performances, and works. Models that include these aspects are 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Imperative of Open and Sustainable Data • 169
fittingly described as persuasive, intellectually elegant, and strategic, 
using the terms proposed by Jannidis and Flanders. But Bollen is also 
quick to highlight cases where the proposed categories are too coarse or 
too value- laden to describe border cases, such as touring productions. 
Large- scale projects would do well to adopt the model proposed by Bol-
len and to pay attention to cases that test the limitations of such model. 
But what about smaller datasets? In many cases, we might not want to 
add metadata to each individual data point. For example, in my analy-
sis of the geographic distribution of performances in Java described in 
chapter 7, I scrapped data from http://www.kluban.net and then put it 
into tables. The tables are consistently formatted but the data model is 
not formalized into a standard such as RDF, and I don’t provide explicit 
metadata for each data point.
When gathering this data, I had very specific questions I wanted to 
answer, and these didn’t require granular metadata or formal statements. 
I am interested in sharing this data with other people who might want to 
use them for their own analysis and I imagine their objectives would be 
one of the following: to verify my results, to use my data as an example 
in their teaching, to ask other questions of my data, or to combine it with 
other datasets. In each case, they will likely apply transformations to the 
data. So, rather than serve it via a semantic server which is hard to main-
tain, I will just offer the data for download in its entirety, in a simple and 
sustainable CSV format. I will add a description of how I obtained and 
transformed the data, but this will be a data biography, rather than a more 
formal, machine-a ctionable description. The data biography is a concept 
I borrow from Simon Eliot’s (2002, 289) “biography of a data source,” 
which I came to know by way of Bode (2012, 14). I will add metadata that 
is less granular, and which describes the entire dataset as a single object 
to facilitate citation, and the data will be maintained in this book publish-
er’s website. This is the strategy followed by most recent DH books (Piper 
2018; Eve 2019; Underwood 2019a; Mullaney et al. 2019).
Projects such as AusStage have many times more data than the exam-
ple above, and their objective is to support a more varied range of research 
agendas. Thus, it makes sense for them to structure their data according 
to more formal data models. There is no one-s ize fits all solution. Perhaps 
the distinction introduced by Jannidis and Flanders, between research- 
driven and curation-d riven data modelers, is best thought as a spectrum of 
possibilities. Each data team needs to think of what type of data model is 
practicable, and which best suits its needs. I agree with Toni Sant’s (2014) 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
170 •  TheaTer as DaTa
admonition that “performance scholars need to make use of best practices 
in information science more consistently,” but we also need to do what is 
technically and institutionally feasible. The ways in which data are to be 
shared are perhaps the most important considerations when choosing a 
data model, and it is to these that I now turn to.
Sharing, Reusing, Citing
Many projects require the creation of their own data. But a vibrant compu-
tational research environment is one where not everyone needs to produce 
all of their data, and research teams can reuse previously existing datasets. 
This is what literary studies, probably the most successful area of DH, have 
achieved through a shared model for literary data. This model is premised 
on the guidelines of the TEI that I discussed in chapter 4. These guidelines 
enable researchers to add structural markup to digital texts. For example, 
theater texts can be encoded in ways that specify the names of speak-
ers, the content of their speeches, and the division of plays according to 
scenes and acts. This is the format that Trilcke and his collaborators have 
used to study the networks of German dramatic texts (as seen in chapter 
5). Based on available TEI data, they could easily extract information on 
the interactions between speakers in a scene. They didn’t have to manu-
ally encode texts, as they were able to reuse extensive datasets that use the 
same format. The TEI format also includes metadata on each text, such as 
the author and year of publication. This can be easily extracted from the 
digitized texts and used to make comparisons across time. However, even 
standardized data needs to be transformed to suit specific research agen-
das. Even when their data was TEI-c onformant, Trilcke and his collabora-
tors spent a considerable amount of time cleaning their data (Trilcke et 
al. 2015). But their project would have been all but impossible without an 
open repository of texts in reasonably standard formats that they could tap 
into. The way in which network researchers use the TEI format to study 
theatrical texts is only one example of what this shared format enables. 
The widely used Stylo in R package (Eder, Rybicki, and Kestemont 2016) 
and Voyant can also ingest TEI- conformant texts.
Sharing data will be important for the future of computational the-
ater research for several reasons: to allow others to verify our results, to 
enable other researchers to combine our data with their own datasets and 
ask new questions, and, equally important, for use in training courses. 
In ideal conditions, systems should be in place to enable other people 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Imperative of Open and Sustainable Data • 171
to always access all the data. There need to be clear policies about how 
the data is to be reused and the data must be described in ways that other 
people can understand. Lastly, the data must conform to formats that oth-
ers can understand and reuse. The common names for these principles 
are: findability, accessibility, interoperability, and reusability, also known 
as the FAIR guiding principles for data management and stewardship 
(Wilkinson et al. 2016).
Christine Borgman (2009) wrote an often-c ited “call to action” based 
on her keynote at the 2009 Digital Humanities conference. In it, she 
encouraged digital humanists to think critically about their data practices 
so that a robust infrastructure for data scholarship can be developed. Sci-
ence, she says, offers inspiration—b ut it is important to design a schol-
arly information infrastructure that caters specifically to humanities 
concerns. For her, this infrastructure encompasses technology, services, 
practices, and policies. She identifies six factors for comparison between 
the humanities and the sciences, “selected for their implications for the 
future of digital scholarship in the humanities”: publication practices, 
data, research methods, collaboration, incentives, and learning (n.p.). Of 
particular relevance to the present discussion is her analysis of the reasons 
why people don’t share data: “(1) faculty get more rewards for publish-
ing papers and books than for releasing data; (2) the effort of individuals 
to document their data for use by others is much greater than the effort 
required to document them only for use by themselves and their research 
team; (3) data and sources offer a competitive advantage and are essen-
tial to establishing the priority of claims; and (4) data are often viewed as 
one’s own intellectual property to be controlled.” The problem of copy-
right is indeed very complex in the humanities as researchers typically do 
not own the data they work with. But in some cases, research teams can 
make provisions to share some data in ways that don’t violate intellectual 
property laws.
A 2016 report on the state of Open Access in DH for the EU Digital 
Research Infrastructure for the Arts and Humanities (DARIAH) follows 
along similar lines and identifies several ways in which humanists can 
develop an ecosystem for data citation (Buddenbohm et al. 2016, hence-
forth “OA Report”). In their view, the data management plan (DMP), 
should be viewed as a key instrument for ensuring that data can be shared. 
The DMP is a formal document, proposed by the UK’s Digital Curation 
Center (Rusbridge et al. 2005), where a researcher describes how the data 
is to be used and shared and what metadata will be added (an example is 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
172 •  TheaTer as DaTa
available at https://dmponline.dcc.ac.uk/). The DMP should be a “living” 
document where changes are made through time. The authors of the OA 
Report also cite eight principles for encouraging data reuse from the Joint 
Declaration of Data Citation Principles issued by the Data Citation Synthe-
sis Group in 2014. These principles are: importance, credit and attribu-
tion (whenever claims are made on data, the data should be cited); unique 
identification; persistence; specificity and variability; interoperability; and 
flexibility. It is important that access to the datasets themselves (in CSV or 
another simple, machine-r eadable format) is always possible.
The OA Report highlights the key role that data citation might play 
in a scholarly ecosystem where data is openly shared and identify several 
things that prevent citation from happening. Some of these barriers are 
data-r elated, such as the diversity of formats, fragmentation, and miss-
ing organization. Other hindrances are technical, such as data that is not 
clearly separated from other digital artefacts and is, for example, inter-
spersed with other aspects of a PDF file. There are also legal barriers such 
as privacy issues and copyright. A final category of barriers include things 
such as funding and its impact on sustainability, the difficulty in ascertain-
ing provenance, as well as problems related to versioning and granularity. 
The authors of the OA Report identify the European Holocaust Research 
Infrastructure (https://www.ehri-project.eu/) as an excellent case study 
and mention several data repositories where individual researchers can 
deposit humanities data: Re3Data.org, Harvard Dataverse, DataDryad.
org, Zenodo, Figshare, and Mendeley. At the time of writing, many teams 
working on computational humanities projects are also depositing their 
code and data at the Open Science Framework (https://osf.io/). The right 
incentives must be in place for people to share and cite data, and this must 
contribute to the prestige of academic careers if this is to be a sustainable 
model. Theater journals, conferences, and university departments have an 
enormous role to play in implementing a culture where these principles 
can guide academic practice.
Should All Digital Projects Be Preserved?  
A Multitiered Approach to Digital Sustainability
Many interactive visualizations require specific infrastructures to stay 
alive, and this is particularly difficult for the case of intermedial essays 
(see chapter 3). For both the preservation of data and the preservation of 
intermedial essays, a multitiered approach will come in handy. Sustain-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Imperative of Open and Sustainable Data • 173
ability seems like a reasonable aim for all data and interfaces. But perhaps 
not every aspect of every single project should— or can— be maintained. 
While long- term preservation is a major challenge, it is also a choice. Not 
all projects might aim to endure for centuries and there is value in what 
I will call bloom-a nd- fade experiments in digital publishing, which can 
uncover new modes of scholarly expression without having to worry about 
the technical infrastructure required for long-t erm preservation. Perfor-
mance’s essence has often been theorized in relation to its ephemerality— 
performance studies is thus uniquely positioned to conceptualize the 
contribution of digital projects that are only meant to exist temporarily. 
However, if the aim of a project is to become a data archive, then its cre-
ators must devise systems for supporting its continued existence into 
future centuries, as books and libraries have succeeded in doing. The 
designers of digital projects might opt for a multitiered approach, where 
certain digital objects are marked for long- term preservation and others 
are developed solely for their more immediate, ephemeral value. This sec-
tion enables the strategic construction of this multitiered approach by 
bringing together conceptual insights from performance studies and best 
practices from digital curation and information science. I will place espe-
cial emphasis on how projects with limited financial resources can achieve 
such goals.
In another short story, “Funes el memorioso” (Funes the memorious), 
Borges (1942) imagined a man who could remember everything. It soon 
transpires that this is a curse rather than a blessing. Being unable to forget 
things, Borges’ character was also unable to make abstractions and to iso-
late details from their contexts. To think, Borges concludes, is to forget. 
One of the challenges of digital scholarship today is that we need to con-
stantly consider the preservation of our materials. But this is not entirely 
a new problem, as the long-t erm preservation of the cultural record has 
always been central to any endeavor in the humanities. The humanities 
could be defined as the intergenerational custodianship and commentary 
of the cultural record. This is true for areas as different from each other as 
architecture and literature. What to preserve, and how to preserve things 
are open questions. But the centrality of preservation is an assumption 
that underpins all endeavors of the humanities. Even scholars who ana-
lyze present phenomena do so in relation to the past. It is important to 
note that there is a cultural dimension to preservation. Not all intellectual 
traditions are interested in preserving things in the same way. There are 
different ways of keeping the past alive, as Western societies place a pre-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
174 •  TheaTer as DaTa
mium on the preservation of artifacts which are ideally kept in conditions 
that work against the passage of time (temperature-c ontrolled, hermetic 
environments, for example). Museums and archives are very important in 
this context, and the intellectual traditions of the West use these artifacts 
to reconstruct specific histories. Questions about how and when things 
came to be the way they are today are considered central within these 
intellectual traditions. Indonesia, to give a contrasting example, is a place 
where keeping oral traditions alive is perhaps more important than the 
preservation of physical artifacts and the precise adjudication of prove-
nance. Both attitudes, however, signal a preoccupation with the intergen-
erational preservation of cultural memory.
Memory institutions devoted to preservation have evolved over many 
centuries and many generations. As records become increasingly digital, 
the previous institutional structures for preservation need major over-
hauls. Within a paper-b ased model of scholarly publishing, scholars don’t 
need to spend much time thinking about how their materials will be pre-
served. Or about which materials will be preserved. Well-d eveloped infra-
structures and procedures are in place. Journals, publishers, and librar-
ies employ highly  trained professionals and dedicate substantial financial 
resources to these purposes. Digital preservation still works better when 
preserving digital facsimiles of print materials. But interactive data visu-
alizations and intermedial essays require interactive and multimedia 
platforms. Since there is no true- and- tested method of digital preserva-
tion for interactive online scholarship, many scholars have set up their 
own archives and portals. This is a key area of DH experimentation and, 
in many cases, initiatives are not driven by institutions but by individual 
researchers who must spend significant time and effort thinking about 
how to preserve their materials. One challenge is defining what to pre-
serve— a significant problem in an era where generating data is much eas-
ier than storing it in sustainable formats. As Borges’ short story suggests, 
perfect memory is a hindrance to thought. It is also impossible from a 
practical point of view. Bloom- and- fade projects enable a nimbler, more 
adaptive development strategy that eventually settles on the data, the data 
models, and the visualizations and interfaces worth keeping for posterity.
Modifying digital media is almost too easy. Its inherent instability is 
linked to a series of principles identified by Lev Manovich (2000): numeri-
cal representation, modularity, automation, variability, and transcoding. 
These are well known, so I will only describe them briefly to show how they 
make digital media easy to change and hard to preserve. Numerical rep-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Imperative of Open and Sustainable Data • 175
resentation is the principle from which the others stem. In current com-
puter architectures everything is represented by binary numbers (images, 
texts, videos, etc.) and it is very easy to apply mathematical transforma-
tions to these numbers. Any operation on a file (search, delete, change 
color) consists of a series of mathematical transformation applied to these 
series of numbers. The second characteristic, modularity, means that all 
new media files and systems are made of independent entities that can be 
easily modified. An image is made of many pixels and one can change the 
color of a single pixel without affecting the others. Likewise, a web page 
is made of many modular components, such as texts and images. It is very 
easy to change an image without affecting the integrity of the whole web-
page. This is obviously impossible to do with print materials.
Automation, the third principle, means that any mathematical opera-
tion can be repeated in the same way with no intervention of the user. 
Changing thousands of images or reverting them back to their previous 
state is almost effortless. The fourth principle is variability. Digital media 
is so easy to change and replace, that it is difficult for media creators to 
decide when a product has reached a finished stage and digital files can 
“exist in different, potentially infinite versions” (Manovich 2000, 36). 
Transcoding, the last principle, is the translation of media from one for-
mat to another. It is easy to convert an image file to a lower resolution, or 
to export a video as a series of still images. Transcoding and variability 
mean that files can exist in an endless state of flux. These two last char-
acteristics are of course underpinned by the numerical representation 
and modularity of digital files, as well as by the ease with which transfor-
mations can be automated. Manovich argues that the enormous creative 
potential of digital media can be explained by these five principles. But 
they also account for the difficulty of preserving digital media. The very 
instability that is so enticing for creative purposes is what makes preserva-
tion hard.
In addition to these characteristics of digital media, certain institu-
tional aspects of DH work contribute to constant change (Escobar Varela 
2016):
Funding: Funding for research projects around the world tends to focus 
on short cycles (3–5  years). Often, priority is given to new proj-
ects. For projects that are new versions of older ones, significant 
improvements need to be made. Usually funding is not given for 
keeping projects alive, but to revamp design features. These prob-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
176 •  TheaTer as DaTa
lems, and a series of potential solutions, are further explored in the 
Endings Project (The Endings Project Team 2019).
Teams change: Often the people responsible for a project change, bring-
ing with them new expertise and ideas for new features in a project
Practice-b ased approach: Digital projects often begin without a clear sense 
of the final product, which means that teams often must feel their 
way forward, experimenting as they go along.
Constant user feedback: Obtaining user feedback on software products 
and web portals is not difficult. This is both a blessing and a curse. 
Feedback certainly makes products better, but it provides an end-
less supply of new ideas. Preservation requires keeping a product 
in a finished state for posterity and the essential variability of new 
media, coupled with the best practices of iterative design—w hich 
keep new users in mind—c onspires against this goal.
This situation is compounded by the physical constraints of storage 
media (Bollacker 2010). As a rule of thumb, data written into smaller 
physical entities is easier to rewrite but it has a shorter lifespan. For exam-
ple, it is easier to amend a message saved on a solid state drive than to 
change a message carved out in a stone surface. The stone does not lend 
itself easily to continuous rewriting, which makes each successive inscrip-
tion cumbersome. The upside, though, is that stone inscriptions from 
previous centuries have endured to our day, often in harsh environmental 
conditions. Solid state drives can hold the same information as thousands 
of stone inscriptions, but they won’t survive for more than a dozen years. 
Digital data requires constant migration to newer storage media, and 
managing this imposes manpower constraints.
To complicate matters further, it is not sufficient to maintain physical 
storage media. To read files from the past, it is crucial that we can run the 
software that can process those files, and this is not always a straightfor-
ward process. In other words, preservation won’t happen automatically. 
People will need to design and enforce action plans to preserve digital 
data and infrastructure. Molloy (2014) suggests that theater makers are 
insufficiently familiar with the needs of digital data preservation and don’t 
usually have a preservation plan in place. Initiatives such as the Long Now 
Foundation are trying to develop best practices (Bollacker 2010). I have 
thus far argued that it is impossible to save everything— constant change 
is easy and preservation is hard. But preservation is not always desirable, 
since it requires enormous costs in terms of money, infrastructure and 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Imperative of Open and Sustainable Data • 177
effort. As mentioned above, there is strategic value in doing faster, nim-
bler projects. I suggested that teams can better identify what their long- 
term goals are if they first work on bloom-a nd- fade projects that are not 
meant to last for a long time. When they have done so, they will be in a 
better position to preserve the key aspects of their work in the long term. 
In addition to this, computational research teams can implement the fol-
lowing suggestions:
Experiment widely: One of the best things that a DH team can do is carry 
out little experiments to figure out what they want to achieve in 
terms of preservation. Bollacker (2010) suggests that since no one 
knows what will be useful in the future, it is important to experi-
ment with as many formats and combinations of formats as budget 
and conditions allow for:
Because we also can’t predict the future to know the best data- 
representation choices, we try to do as nature does. We can copy 
our digital data into as many different media, formats, and encod-
ings as possible and hope that some survive. This is our best shot 
at ensuring that at least some formats and some things will still be 
available in the future. (110)
Be informed: It is sometimes hard to fathom how volatile and frag-
ile current storage media and infrastructures are. I’m always sur-
prised when speaking to people who overestimate the durability of 
platforms, formats, and physical storage media. There are many 
resources that people can consult to understand better how media 
works. All the documents from the Long Now Foundation are use-
ful reading for anyone interested in DH and digital preservation. 
Minimal computing also applies here (a concept I will revisit in chap-
ter 9). It is important to choose minimal architectures since they 
tend to be easier to maintain, and more compliant with existing 
standards.
Arrange things into baskets: Based on small-s cale experiments and on 
awareness of the technical and social constraints of long-t erm pres-
ervation, research teams can make decisions about what aspects of 
their projects should be marked for long- term preservation and 
which are meant to exist as relatively ephemeral digital projects. 
There are some aspects of a project that are essential and that must 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
178 •  TheaTer as DaTa
be preserved even if everything else is lost. I think of this as the 
“preservation core” of a project. Having grown up in an earthquake- 
prone city, I learned to keep key documents in a sealed plastic bag 
by the door of my house. In case of an emergency, I could take these 
documents and run. The same principle applies to the design of a 
digital preservation core. A second basket will include digital arti-
facts that can be kept in case we have a bit more time to run away 
before the building (or the funding) comes crumbling down. 
This is similar to what the Digital Curation Centre (DCC) recom-
mends (Higgins 2008). We often don’t want to believe that disas-
ter will strike and when it does, we are caught off guard. I know 
of one project where the digital architecture collapsed overnight, 
when there was a mandatory update from the base system they were 
using. Since there was no multitiered approach to preservation 
it was very difficult for the team to recover functionality quickly. 
More importantly, this imperils the long-t erm preservation of this 
project. Data marked for long-t erm preservation should be stored 
according to explicit data models and ideally made openly acces-
sible. Both preservation and openness depend on each other. Open 
sharing requires sustainability practices. But in turn, resources that 
are openly shared are more likely to be preserved. This is one of 
the core principles of LOCKSS or “Lots of Copies Keeps Stuff Safe” 
(Maniatis et al. 2005).
When planning projects, we also need to plan for our absence 
or retirement. This is of course hard for academics to do (as it is 
hard for everyone) but it is a must if digital projects are to endure 
into the future. For example, the key data of the CWA are the perfor-
mance recordings and the translations and notes. They constitute 
the innermost preservation basket, or digital preservation core. For 
this reason, they are kept in the most standard possible files. The 
translations, transcripts and notes are stored in vanilla text files 
(UTF- 8 encoding) and the videos in raw formats. On the next bas-
ket we have the documentation and source code on GitHub. Given 
that interactive websites are hard to preserve, it is important to take 
video snapshots of interactive features for future reference. This 
is what I have done with the interactive visualizations described in 
this book. The website companion includes videos that show how 
these visualizations are meant to work, in case the online versions 
of those visualizations stop working in the future due to any of the 
problems described earlier in this chapter.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Imperative of Open and Sustainable Data • 179
This last part of this chapter might read as a prescriptive instruction man-
ual, but its message underpins the entire project of Theater as Data. Ear-
lier in the book, I discussed mostly epistemological and methodological 
concerns. We cannot, however, think about epistemology without paying 
attention to the material conditions on which data survives or perishes. 
Matters of digital preservation are more than just technical issues. They 
point to the essence of the humanities. If we want to have viable projects, 
careful attention to the conditions that encourage and limit preservation 
is essential. On a practical level, the issues discussed here will help us 
think about how to strategically construct a multitiered approach that val-
ues both ephemerality and long- term preservation. The double focus on 
ephemerality and long- term cultural transmission is, after all, a defining 
feature of theater.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
ChapTer 9
The Roles of Software Programming
Programming is more than a technical skill, it is a mode of looking at the 
world. Engaging with programming has pedagogical, ethical, and institu-
tional implications for computational theater research. In this short, con-
cluding chapter, I will untangle some of these implications by suggesting 
that programming can best be thought as a range of creative and critical 
endeavors. I believe that programming will be useful to anyone attempt-
ing to work with theater data, whether they are interested in the data- 
driven quest to answer specific questions, or in the provisional and situ-
ated polyvocality of data-a ssisted research. My objective is not to say that 
those who are not programmers should refrain from doing computational 
research. Given the range of available software and online tools, many 
people can carry out computational theater research without knowing 
how to program. People who are not programmers have and will continue 
to make fundamental contributions to computational theater research. 
My objective in this chapter is not to defend programming knowledge as a 
precondition for working with data. Rather, I want to show how program-
ming can help us think about theater in new ways, and to encourage more 
people to tinker with the art of programming.
Learning to program is not a binary choice. Programming skills 
are located along a continuum, and it is difficult to ascertain a single 
moment where someone becomes a programmer. But embarking on the 
road towards the acquisition of programming skills will make our work 
richer and deeper as it will expand our potential to decide how data is 
collected, processed, and presented. Both data- driven and data- assisted 
research can be carried out by deploying existing software, by tweaking 
existing packages, or by writing bespoke code. An interesting feature of 
online platforms, such as Voyant, is that they encourage playfulness and 
reward their users with the aesthetic joy of seeing texts transformed into 
Escobar Varela, M1ig8u0el. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Roles of Software Programming • 181
tables and visualizations which can then be tinkered with. A more direct 
version of this experience is available to users who write their own code. 
Several DH books take readers through step-b y- step instructions of how 
to write code for text analysis (for example, Jockers 2014). An intermedi-
ate step between using platforms such as Voyant and writing code directly 
can be found in certain programming packages that have a graphical user 
interface (GUI) which can be run directly by novice users but also modi-
fied by users with more programming experience. For working with text, 
the most famous example of this approach is the Stylometry in R pack-
age (Eder, Rybicki, and Kestemont 2016). Many current projects depend 
on libraries built in the R and Python programming languages, which 
are often easy to use. The code I have used for projects reported in this 
book was mostly written in Python, and I made extensive use of libraries 
developed by other people for the analysis and visualization of data, as can 
be seen in the online companion to this book. These libraries are open 
source, which means others can reuse and extend them. In this way, pro-
gramming, like theater, becomes a social activity that builds on collective 
effort and shared ideals.
Even a rudimentary knowledge of programming will enable us, theater 
scholars, to better understand our projects, and to be more attentive to 
latent assumptions and blind spots in our methodologies. The more we 
increase our programming proficiency, the more resilient we will be as a 
community that doesn’t need to outsource decision making to other dis-
ciplines and to commercial companies. There is, of course, a danger of 
romanticizing self- sufficiency. No one can carry out the entire process on 
their own. You can’t build a computer from scratch and write every piece 
of software that runs on it. But an incursion into the making of tools will 
sensitize us to things at stake, and to kinds of labor which are often invis-
ible. A central theme running through this book is that every method is 
value- laden. When choosing a method, or a source of data, there is no 
value-f ree option, as every choice represents a set of preferences and prej-
udices. If we just choose “default” options in a software package or online 
portal, that means someone else has made the choice for us. If we want to 
carry out nuanced data-d riven and data-a ssisted research, we need to be 
able to own our assumptions. This is easier to do if we rethink software 
programming as a mode of critical thinking.
For example, thinking as programmers will enable us to remain atten-
tive to cultural specificity. If we can trace the assumptions of the algorithms 
we use, we will be better able to reflect on what gets misrepresented and 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
182 •  TheaTer as DaTa
craft tools better suited to our objectives. Many named entity recognition 
algorithms are premised on the idea that personal names are made of a 
first name, an optional middle name, and a family name (expressed in that 
order). But Chinese names often begin with the family name, followed by 
a bisyllabic first name (where each syllable might be separated by a space 
when written in Roman characters or hyphenated). Mexican names (like 
mine) are based on Spanish naming conventions where a first name is 
followed by two family names (one from the father’s and one from the 
mother’s side); many Indonesians have only one name, or two first names 
but no hereditary family name. These are just some examples, but nam-
ing conventions vary extensively around the world. This is not often recog-
nized by the designers of systems for name recognition which underpin, 
among other things, academic citation databases. If we can code, we can 
try to circumvent uneven practices in the work that we do. We might not 
be able to change the ways of big corporate systems, but we should be able 
to intervene in the systems that we need for data research in the humani-
ties. We need open platforms that others can inspect and adapt, but we 
also need people who are able to do this. People who know how to pro-
gram are more likely to demand rightful access to the source code. And 
open source code encourages people to program. Software programming 
will also better enable us to develop performative, data-a ssisted visualiza-
tions that challenge conventional visual tropes, and to craft intermedial 
essays in ways that are not predefined by other people, but that stay close 
to the topics and intellectual dispositions that interest us.
Learning to code expands a scholar’s horizon in a way akin to learning 
another language. One could perhaps say insightful things about French 
theater without speaking French. But even a tentative incursion into the 
French language will enable a scholar to grasp subtle differences, ask bet-
ter questions, and trace a more nuanced history of prejudices and intel-
lectual positions. Code is also quickly becoming a lingua franca, and the 
imperative to learn to think as programmers has gained urgency in a word 
where software shapes so many aspects of life. As Kitchin and Dodge 
(2014, 1) note:
It is very difficult to avoid the effects—t he work— of software in the 
world [ . . . ] because of the difference it makes to the constitution and 
practices of everyday life. Indeed, to varying degrees, software con-
ditions our very existence. Living beyond the mediation of software 
means being apart from collective life.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Roles of Software Programming • 183
These remarks have implications for pedagogy, and for what we expect 
of our students. Undergraduate students starting their theater studies 
degrees at the time of writing are digital natives, or so the conventional 
thinking goes. But are they really native to this world, if they are only 
users, rather than makers? If they can’t rewrite the code that underpins 
their reality, they don’t have full citizenship in this digital world. As Rush-
koff (2011, 12) puts it, in equally provocative terms:
In the emerging, highly programmed landscape ahead, you will either 
create the software or you will be the software. It’s really that simple: 
program, or be programmed. Choose the former, and you gain access 
to the control panel of civilization. Choose the latter, and it could be 
the last real choice you get to make.
This passage, which I often use in my introductory DH classes is also 
quoted by Ramsay (2012), who reflects on how to teach programming to 
humanities students. He makes a crucial point when he says that teaching 
students to program also teaches them to think, and that programming 
teaches humanities thinking as much as computational thinking. While 
there is a strong appeal in teaching humanities students to code in order 
to improve their chances in the job market, I agree with Ramsay’s point 
that we need to teach programming skills to make sure that humanities 
thinking can continue into a digital age in our own terms, rather than 
merely on the conditions of dominant software ideologies.
When I teach my students to program, I also use another quote from 
Ramsay, in this case to set them at ease. Many feel inadequate when learn-
ing to program and they see these type of work as alien to their core inter-
ests. But via Ramsay (2014), I reassure them that the key ingredients of 
programming are curiosity and an open mind:
One thing is certain: Being good at mathematics in no way guarantees 
that one will be good at programming (or vice versa). My own (admit-
tedly anecdotal) experience as a teacher suggests that being musical, 
enjoying games and puzzles, being a tinkerer, loving to cook, and 
being a good long- form writer are far better predictors of success. A 
very high tolerance for frustration helps as well. (n.p.)
There is a craft, an art even, to programming. And this point has been 
perhaps best captured in the words of Paul Graham (2003), who com-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
184 •  TheaTer as DaTa
pares hackers to artists. For him, as for many others, “hacking” refers to 
an approach to programming premised on trial- and- error rather than on 
engineering models:
The fact that hackers learn to hack by doing is another sign of how dif-
ferent hacking is from the sciences. Scientists don’t learn science by 
doing it, but by doing labs and problem sets. Scientists start out doing 
work that’s perfect, in the sense that they’re just trying to reproduce 
work someone else has already done for them. Eventually, they get to 
the point where they can do original work. Whereas hackers, from the 
start, are doing original work; it’s just very bad. So hackers start origi-
nal, and get good, and scientists start good, and get original. (n.p.)
Programming is closer to the modes of working of theater makers and 
researchers than people conventionally assume. And even learning a little 
bit of programming, learning to playfully hack and adapt our tools can 
help us reimagine our relationship to our technologies and to our theater 
traditions. By learning to program—o r rather, by discovering program-
ming through creative tinkering—w e can also choose to represent things 
differently, away from hegemonic paradigms of representation. Vikram 
Chandra, the renowned Indian novelist, worked for a long time as a soft-
ware programmer, although this is a little known fact of his biography. 
Like me (and many others in DH), he did not receive formal training in 
software engineering but learned by endless creative tinkering. Chandra 
(2014) vividly describes the pleasure of experimentation as the driving 
force in his learning process:
The work of making software gave me a little jolt of joy each time a 
piece of code worked; when something wasn’t working, when the 
problem resisted and made me rotate the contours of the conundrum 
in my mind, the world fell away, my body vanished, time receded. And 
three or five hours later, when the pieces of the problem came together 
just so and clicked into a solution, I surfed a swelling wave of endor-
phins. [ . . . ] Even after you are long past your first “Hello, world!” 
there is an infinity of things to learn, you are still a child, and— if you 
aren’t burned out by software delivery deadlines and management- 
mandated all- nighters— coding is still play. You can slam this pleasure 
spike into your veins again and again, and you want more, and more, 
and more. (18– 19)
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Roles of Software Programming • 185
What we need is this playfulness, this tinkering that characterizes hack-
ing. As Mark Olson (2013) has suggested:
A hacker ethos is a way of feeling your way forward, through trial and 
error, up to and perhaps beyond the limits of your expertise, in order 
to make something, perhaps even something new. It is provisional, 
sometimes ludic, and involves a willingness to transgress boundaries, 
to practice where you don’t belong. (238)
Besides the ludic pleasure that programming can bring, programming 
also frees up financial resources and enhances our imagination. These 
two related points are more cogently articulated by the proponents of 
minimal computing from GO::DH (Global Outlook Digital Humanities, a 
special interest group within the Association for Digital Humanities Orga-
nizations), from whom I have borrowed the concept:
We use “minimal computing” to refer to computing done under some 
set of significant constraints of hardware, software, education, net-
work capacity, power, or other factors. Minimal computing includes 
both the maintenance, refurbishing, and use of machines to do DH 
work out of necessity along with the use of new streamlined comput-
ing hardware like the Raspberry Pi or the Arduino micro controller to 
do DH work by choice. This dichotomy of choice vs. necessity focuses 
attention on computing that is decidedly not high-p erformance. By 
operating at this intersection between choice and necessity minimal 
computing forces important concepts and practices within the DH 
community to the fore. (Minimal Computing Working Group 2015, 
n.p.)
The interface of choice and necessity is articulated in different ways by the 
most active members of the Working Group. Alex Gil (2015) suggests that 
scholars around the world (including librarians and students) need to ask 
themselves what counts as sufficient, and ask themselves what is their spe-
cific goal. He urges us to remember that the most important objective of 
our efforts should be the “renewal, dissemination and preservation of the 
scholarly record.” Doing this in a sophisticated, ethical, and sustainable 
way requires that we learn “to produce, disseminate and preserve digital 
scholarship ourselves, without the help we can’t get, even as we fight to build 
the infrastructures we need at the intersection of and beyond our librar-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
186 •  TheaTer as DaTa
ies and schools” (n.p., original emphasis). The projects I advocate for and 
participate in are a response to this call for action. Like Gil, I also endeavor 
with my collaborators to do things ourselves, while also building long- 
lasting infrastructure. Jentery Sayers (2016) responds to Gil in another 
short position piece saying that minimal DH should be “not only what we 
need but what we want” (n.p.). Sayer’s crucial question is: “How might 
minimal computing increase our shared capacities to think or imagine, 
and not just our individual capacities to work or produce?” (n.p.). The 
question of minimal infrastructures is thus not purely practical, but car-
ries with it an important disposition towards our code and machines.
Inspired by these views, I imagine data-d riven and data-a ssisted the-
ater projects built with the minimum possible financial investment and 
built in such a way that minimal conservation efforts are needed for their 
long-t erm sustainability. This is at odds with the way many DH projects 
are done, which require large financial investments, but allocate little 
resources (or thought) to long- term sustainability. An expensive project 
might be harder to maintain than one that requires modest investment, 
but this also depends on where the investment is made. This might seem 
counterintuitive to people who assume that keeping projects alive is very 
expensive. A good data project, in my view, is like a bicycle—i t won’t be fast 
but it will be reliable, requiring much less maintenance than an expensive 
car. Even if you have the money to buy a car, it might be a smarter strategy 
to buy a bicycle or a series of them and distribute the money throughout 
the years, making sure the bike- enabled transportation is available in the 
future. Digital sustainability (see chapter 8) provides an important justi-
fication for the allocation of financial and other resources, including the 
substantial amount of time and effort that goes into the development of 
such projects. But sustainability should be a primary concern for intellec-
tual reasons, as the traditions of the humanities—i ncluding theater and 
performance studies— are comparative and historical.
It is, however, difficult to talk about DH work without considering 
money (Liu 2012). The presence of large financial resources for DH fund-
ing has been a cause for both celebration and concern. In a time when 
funding for the humanities is decreasing across the Western world, DH 
has enabled academic positions, grants, and research centers to flour-
ish. But this situation has also been assessed negatively by those who fear 
that the presence of large budgets has co-o pted humanities interests into 
subservience to capital, technoscience, and managerism at universities 
(Berry and Fagerjord 2017, 10). Another reason why we should be vigilant 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
The Roles of Software Programming • 187
of the impact of financial resources is that they tend to be disproportion-
ately appointed to performance traditions from rich countries and to well- 
known projects. My fear is that theater projects with no data materials will 
become even less well known. But it is precisely here where I see a shim-
mer of hope. Data scholarship has a greater potential to break through 
the gatekeepers of academia than other kinds of projects. Smartly planned 
data projects from around the world might disrupt knowledge distribu-
tion channels if resources are allocated in a way that ensures sustainabil-
ity. My own work is mostly funded by generous research grants in Singa-
pore, but I also work with cash- starved institutions in Indonesia and have 
developed several projects with minimal financial resources.
As Gil, Sayers, and other members of the Minimal Computing WG 
suggest, technical knowledge is required in order to ensure that minimal 
possible investment and sustainability are possible. Think of the bicycle 
again: if you know how to repair it yourself, the costs of maintenance will 
be much cheaper. Or, at least you will know when you are overpaying for 
a service. Financial constraints are conventionally seen as limitations for 
DH, especially in parts of the world where funding is harder to come by. 
But there is a way of doing sustainable, highly technical work with mini-
mal investments: learning to code and relying on open infrastructures. 
This strikes a particular chord with my colleagues in Indonesia, where 
funding is scarce. However, this should be a goal even for well-f unded 
projects in rich institutions, for the reasons discussed above. Funding is 
often seen as one of the greatest impediment, or enablers, of computa-
tional research. However, computational theater research can be carried 
out by people who understand the foundations of their tools, with the 
minimum possible financial investments and it can be built to last.
Even as we learn to code, we should still be able to engage in produc-
tive research dialogues. Good work will certainly come out of collabora-
tions with colleagues in technical and scientific fields. But these collab-
orations will be all the more meaningful if all partners understand each 
other’s tools and disciplinary histories. In carrying out research for proj-
ects reported in this book, I have been particularly lucky to collaborate 
with Gea O. F. Parikesit and Andrew Schauf. Both of them are physicists 
with an extraordinary awareness of the histories of the humanities and the 
complexities of Indonesian arts. When theater scholars collaborate with 
scientists and programmers, the communication will be more meaning-
ful as both sides learn to understand different paradigms and approaches.
If we, as theater scholars want to engage with data, we need to learn 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
188 •  TheaTer as DaTa
to craft our software for the variety of reasons I have argued above: to 
enhance our way of thinking critically about methodology, to create cultur-
ally  sensitive work, to gain control of our tools, and to deepen our collabo-
rations and to make work that lasts. Embedding programming into our 
academies will no doubt be tricky. But as Bench and Elswit (2017) remark, 
who better than theater scholars to think about new modes of scholarship 
that are collaborative, creative, and practice-b ased? The paradigm shift in 
theater studies that enabled the rise of practice-b ased research might pro-
vide the right model for computational work. We have already rethought 
the place of practice in the making of new knowledge, it is now time to 
consider the ways in which programming can further expand our method-
ological imaginations.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Appendix A
Data Biographies
All the code for the excursions in the book was implemented in Python 
3.7.6, using the Pandas package, version 1.0.1. All images were generated 
with Matplotlib 3.1.3 (Hunter 2007) and Seaborn version 0.10.0.
Chapter 4
The theater reviews were downloaded from https://inkpotreviews.com/
archive.html in January 2019. The reviews for each year were grouped into 
separate text files, which omitted the names of the reviewers and titles of 
the reviews. These files were uploaded to Voyant (http://voyant-tools.org), 
where I used built- in functions to calculate the trends over time and the 
concordances. I then downloaded these files for processing in Python.
4.1_trendOverTime.csv https://doi.org/10.3998/mpub.11667458.cmp.26
This file was downloaded from Voyant, and then analyzed in Python. To 
calculate the Mann-K endall statistics I used the pymannkendall package 
version 1.4.1 (Hussain and Mahmud 2019) for Python.
4.2_audienceKWIC.csv https://doi.org/10.3998/mpub.11667458.cmp.27
This file is the concordance for sentences containing the KWIC concor-
dance for the word audience, which I then manually categorized into 
descriptive and rhetoric mentions.
4.3_audienceTrends.csv https://doi.org/10.3998/mpub.11667458.cmp.28
This file includes the results of manual analysis of the audience 
concordance.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research. 189
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
190 • Appendix A
Chapter 5
The file with the list of cast members and roles was provided by The Nec-
essary Stage (TNS) in January 2016. My research assistant Alysha Chandra 
and I verified all names and roles and standardized the spelling of names 
in the files for consistency. Based on this data, the network was built using 
NetworkX version 2.4 (Developers 2010). The number of nodes, edges, 
and components were calculated using the built-i n functions of this 
library. The files below include data on the component size, as well as the 
number of nodes and edges over time.
5.1_TNScomponents.csv https://doi.org/10.3998/mpub.11667458.cmp.29
This file includes information on all people involved in the production.
5.2_TNScast.csv https://doi.org/10.3998/mpub.11667458.cmp.30
This file includes information on people labeled as cast members in the 
file provided by TNS.
5.3_TNSproductionCrew.csv https://doi.org/10.3998/mpub.11667458.cmp.31
This file includes information on people labeled as production crew mem-
bers in the file provided by TNS.
The interactive visualizations for the Digital Wayang Encyclopedia are 
available at https://villaorlado.github.io/wayangnetworks/html/ and 
in video 5.1 (https://doi.org/10.3998/mpub.11667458.cmp.23) in this 
book’s web companion. The data was obtained from several published 
wayang encyclopedias (Hardjowirogo 1948; Sudibyoprono et al. 1991; 
Sudjarwo, Sumari, and Wiyono 2010; Purwadi 2013; Solichin et al. 2017). 
The data was verified by my research assistants Losheini Ravindran and 
Yosephine Novi Marginingrum.
5.4_wayangNodeInfo.csv https://doi.org/10.3998/mpub.11667458.cmp.32
This resource includes information on each character.
5.5_wayangNetwork.gephi https://doi.org/10.3998/mpub.11667458.cmp.33
This is the Gephi file of the full network.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Appendix A • 191
Chapter 6
The original Wayang Mitologi (Catur Kuncoro, 2012) video recording is 
available at http://cwa-web.org/en/WayangMitologi
6.1_mitologiDifferenceImages.csv https://doi.org/10.3998/mpub.11667458.cmp.34
This resource includes the number of pixels in each difference image. The 
video was processed by Gea Oswah Fatah Parikesit using Scilab and the 
Scilab Image and Video Processing Toolbox (SIVP). In the video record-
ing of Wayang Mitologi, the noise level is 10 greyvalues, so we only con-
sider pixels with greyvalues higher than 10 as the non- static pixels. This 
was checked by recording two dark images and measuring the greyvalues 
of the supposedly dark pixels. The video is 69 minutes long. We obtained 
1,000 images per minute, and then sampled one out of each 150 difference 
images, which corresponds to ~0.1 min. The file indicates how each of the 
460 difference images corresponds to the 69,652 frame numbers and the 
69 minutes.
This file also includes scene segmentation information, which I manu-
ally added.
For the Sendratari Ramayana dance, Luis Hernández- Barraza and I 
used a motion- capture system (Vicon MX, Oxford Metrics, Oxford, UK), 
that consists of seven infrared cameras to collect kinematic data at a sam-
ple rate of 100 hertz. Forty-o ne reflective markers (14 mm diameter) were 
attached to the dancer following the full body Plug- In- Gait Marker model 
(Vicon, Oxford Metrics, Oxford, UK) to facilitate capture of the dancer’s 
motion. Two embedded force plates (AMTI, Watertown, Massachusetts, 
USA) were used to obtain GRF data at a sampling rate of 1,000 hertz. The 
force plates were synchronized to the motion capture system, and both 
were calibrated according to the manufacturer’s recommendations before 
the dance data was recorded.
6.2_danceSubtypesAngleData.csv https://doi.org/10.3998/mpub.11667458.cmp.35
The material is this resource was obtained and processed by Luis 
Hernández-B arraza. The columns are named according to the character 
subtype, the joint and the plane (x, y, or z). Each row corresponds to the 
measurement of an angle (1 observation per ms).
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
192 • Appendix A
Chapter 7
The data was obtained from https://kluban.net. I used the Python Beau-
tifulSoup version 4.8.2 package to scrap the website and then format the 
results into a CSV file. The geographical regions were assigned via a rule 
based system. All cases where the rules could not sufficiently disambigu-
ate the region, were manually resolved. I built a custom Python script to 
achieve both operations (rule-b ased and manual disambiguation). The 
population statistics were obtained from Badan Pusat Statistik (Indone-
sian Center for Statistics), and they correspond to 2016 estimates. The 
maps were generated with Geopandas version 0.8.1. Moran’s I was calcu-
lated using PySAL version 1.14.4.
7.1_placeAndDate.csv https://doi.org/10.3998/mpub.11667458.cmp.36
This file includes the date and location of each performance. The date is 
meant to be used as a Python datetime object. The time information can 
be discarded, as only the dates are relevant. Time was assigned at 00:00 
for consistency, but most performances begin at 21:00 and end around 
04:00 the next day. The date in the file corresponds to the beginning of the 
performance.
7.2_performancesAndPopulation.csv https://doi.org/10.3998/mpub.11667458.cmp.37
This resource includes aggregate counts per regency as well as statistics 
on the size, population and population density of each regency.
7.3_performancesPerRegency.csv https://doi.org/10.3998/mpub.11667458.cmp.38
This resource includes aggregate counts per regency.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Appendix B
Technical Glossary
Betweenness centrality (network analysis): This is a measure of how often 
a node acts as a bridge between other nodes. A high betweenness cen-
trality indicates that the shortest paths between all pairs of nodes in the 
network often pass through the given node.
Cartogram: A geographic visualization which distorts the shape of geo-
graphical entities (countries, districts, etc.) according to a particular 
metric or statistic. The resulting maps show how much a given area 
contributes to the total for a given metric in a way that is not propor-
tional to the actual land area of a province or country.
Choropleth: A geographic visualization where areas (districts, counties, 
countries) are colored according to different quantitative values.
Circos plots: A type of chord diagrams that show interrelations between 
data in a matrix. Each data point is represented as a dot in the circle’s 
circumference, which is connected by lines drawn across the area of 
the circle.
Closeness centrality (network analysis): The inverse of the average length 
of the most direct paths between the given node and all other nodes in 
the network.
Cohen’s d: A measure of the difference between groups that can be used 
to estimate the magnitude of a phenomenon (effect size). Cohen’s d is 
the difference between two means (one for each group) divided by the 
standard deviation for the data.
Complete spatial randomness (CSR): A distribution of points in a geo-
graphical location that conforms with a random model’s prediction.
Degree (network analysis): In directed networks, the degree is the num-
ber of edges of a given node. In undirected networks, we can distin-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research. 193
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
194 • Appendix B
guish between in- degree (edges with the given node as target) and out- 
degree (edges with the given node as source).
Difference image: An image that results from subtracting one image from 
another. This is used to compare subsequent images extracted from 
videos. The number of pixels in an image difference can be used to 
estimate the amount of movement in a video.
Eigenvector centrality (network analysis): A measurement of the influ-
ence of the node in the network that considers the degrees of a node’s 
neighbors. Nodes with high eigenvector centrality tend to be con-
nected with neighbors who are themselves highly connected.
Kernel density estimation (KDE): A non- parametric (i.e., that doesn’t 
impose a model) method for visually estimating the probability density 
function of a variable.
Mann- Kendall s: A measure for the strength of a monotonic trend (that 
is, one that increases or decreases consistently). A negative value 
means that the trend is decreasing, while a positive value means it is 
increasing.
Minimum convex polyhedron: A solid in three dimensions with flat polyg-
onal faces, straight edges and sharp corners.
Moran’s I: A score used to determine whether statistical proximity affects 
a given variable (spatial autocorrelation). Moran’s I estimates the prob-
ability that a value for a geographical unit is affected by the average 
value of neighboring units.
Network density: The number of actual edges in a network divided by the 
maximum total number of edges.
Network dynamics: The study of networks as they change over time.
Weighted degree (network analysis): The sum of the weights of all the 
node’s links.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
References
Aarseth, Espen J. 1997. Cybertext: Perspectives on Ergodic Literature. UK edition. Baltimore: 
Johns Hopkins University Press.
Abdulkadiroğlu, Atila, Joshua Angrist, and Parag Pathak. 2014. “The Elite Illusion: 
Achievement Effects at Boston and New York Exam Schools.” Econometrica 82 (1): 
137– 96.
Academia Sinica Center for Digital Cultures. n.d. “Starting Out from 23.5°N: Chen 
Cheng- Po.” Accessed December 18, 2018. http://chenchengpo.asdc.sinica.edu.tw 
/about_en
Agneessens, Filip, Henk Roose, and Hans Waege. 2004. “Choices of Theatre Events: 
P* Models for Affiliation Networks with Attributes.” Metodoloski Zvezki 1 (2): 419.
Alberich, Ricardo, Joe Miro- Julia, and Francesc Rosselló. 2002. “Marvel Universe 
Looks Almost like a Real Social Network.” ArXiv Preprint Cond-M at/0202174.
Algee- Hewitt, Mark. 2017. “Distributed Character: Quantitative Models of the English 
Stage, 1550– 1900.” New Literary History 48 (4): 751– 82. https://doi.org/10.1353/nlh 
.2017.0038
Angwin, Julia, and Jeff Larson. 2016. “Machine Bias.” Text/html. ProPublica. May 23, 
2016. https://www.propublica.org/article/machine-bias-risk-assessments-in-crimi 
nal-sentencing
Arnold, Taylor, and Lauren Tilton. 2019. “New Data? The Role of Statistics in DH.” 
In Debates in the Digital Humanities 2019, by Matthew K. Gold and Lauren F. Klein. 
Debates in the Digital Humanities. Minneapolis: University of Minnesota Press. 
https://dhdebates.gc.cuny.edu/read/untitled-f2acf72c-a469-49d8-be35-67f9ac1e3 
a60/section/a2a6a192-f04a-4082-afaa-97c76a75b21c#ch24
Arps, Bernard. 2016. Tall Tree, Nest of the Wind: The Javanese Shadow-P lay Dewa Ruci Per-
formed by Ki Anom Soeroto: A Study in Performance Philology. Singapore: NUS Press.
Arrowsmith, Colin, Deb Verhoeven, and Alwyn Davidson. 2014. “Exhibiting the Exhib-
itors: Spatial Visualization for Heterogeneous Cinema Venue Data.” Cartographic 
Journal 51 (4): 301– 12. https://doi.org/10.1179/1743277414Y.0000000096
Arulampalam, J., J. Pierrepont, and L. Kark. 2015. “Markerless Motion Capture: Valid-
ity of Microsoft Kinect Cameras and IPisoft.” Gait & Posture, 24th Annual Meeting 
of ESMAC 2015 Abstracts, 42 (September): S76. https://doi.org/10.1016/j.gaitpost 
.2015.06.141
Ashworth, Peter D., and Man Cheung Chung, eds. 2006. Phenomenology and Psychologi-
cal Science: Historical and Philosophical Perspectives. History, Philosophy, Psychology. 
New York: Springer.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research. 195
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
196 • References
Asnaghi, Costanza, Dirk Speelman, and Dirk Geeraerts. 2016. “Geographical Patterns 
of Formality Variation in Written Standard California English.” Literary and Linguis-
tic Computing 31 (2): 244– 63. https://doi.org/10.1093/llc/fqu060
Aston, Elaine, and George Savona. 1991. Theatre as Sign- System. Routledge.
AusStage. 2013. “AusStage.” 2018 2013. https://www.ausstage.edu.au/
Bach, Benjamin, Moritz Stefaner, Jeremy Boy, Steven Drucker, Lyn Bartram, Jo Wood, 
Paolo Ciuccarelli, Yuri Engelhardt, Ulrike Köppen, and Barbara Tversky. 2018. 
“Narrative Design Patterns for Data- Driven Storytelling.” In Data- Driven Storytelling, 
edited by Nathalie Henry Riche, Christophe Hurter, Nicholas Diakopoulos, and 
Sheelagh Carpendale, 59–8 3. New York: A. K. Peters/CRC Press. https://doi.org/10 
.1201/9781315281575-3
Badan Pusat Statistik. 2019. “Badan Pusat Statistik.” 2019. https://www.bps.go.id/
Baker, Zachary, Joel Berkowitz, Sonia Gollance, Debra Caplan, Barbara Henry, Faith 
Jones, C. Tova Markenson, et al. 2019. “Digital Yiddish Theatre Project.” Text/
html. Digital Yiddish Theatre Project. March 23. https://yiddishstage.org/about
Bakshy, Eytan, Dean Eckles, and Michael S. Bernstein. 2014. “Designing and Deploy-
ing Online Field Experiments.” In Proceedings of the 23rd International Conference on 
World Wide Web, 283– 92. ACM.
Balazia, M., and P. Sojka. 2017. “You Are How You Walk: Uncooperative MoCap Gait 
Identification for Video Surveillance with Incomplete and Noisy Data.” In 2017 
IEEE International Joint Conference on Biometrics (IJCB), 208–1 5. https://doi.org/10.1109 
/BTAS.2017.8272700
Balme, Christopher. 2015. The Cambridge Introduction to Theatre Studies. Cambridge: Cam-
bridge University Press.
Barabási, Albert- László. 2002. Linked. Cambridge: Perseus.
Bardiot, Clarisse. 2015a. “Arts de La Scène et Big Data. Retracer et Analyser Le Proces-
sus de Création d’un Spectacle Grâce à La Visualisation de Données.” In Le Numéri-
que à l’ère de l’Internet Des Objects: De l’hypertexte à l’hyperobjet. Paris.
Bardiot, Clarisse. 2015b. “Rekall: An Environment for Notation/Annotation/Denota-
tion.” Performance Research 20 (6): 82– 86. https://doi.org/10.1080/13528165.2015.11 
11058
Bardiot, Clarisse. 2017. “Arts de La Scène et Culture Analytics.” Revue d’historiographie 
Du Théâtre 4.
Bardiot, Clarisse. 2018. “Measuring Merce Cunningham: A Theatre Analytics 
Research.” In Bridges/Puentes. Mexico City: ADHO.
Bardzell, Jeffrey, and Shaowen Bardzell. 2013. “What Is Critical about Critical Design?” 
In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 3297– 
3306. ACM.
Bateman, John A. 2017. “Triangulating Transmediality: A Multimodal Semiotic Frame-
work Relating Media, Modes and Genres.” Discourse, Context & Media 20 (Decem-
ber): 160– 74. https://doi.org/10.1016/j.dcm.2017.06.009
Battacharyya, Sayan. 2018. “Non-n ormative Data from The Global South and Epis-
temically Produced Invisibility in Computationally Mediated Inquiry— DH2018.” 
In Proceedings of the 2018 Digital Humanities Conference. Mexico City. https://dh2018.ad 
ho.org/en/non-normative-data-from-the-global-south-and-epistemically-produc 
ed-invisibility-in-computationally-mediated-inquiry/
Bay- Cheng, Sarah. 2017. “Digital Historiography and Performance.” Theatre Journal 68 
(4): 507– 27. https://doi.org/10.1353/tj.2016.0104
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
References • 197
Bay- Cheng, Sarah, Robin Nelson, Andy Lavender, and Robin Nelson. 2010. Mapping 
Intermediality in Performance. Amsterdam: Amsterdam University Press.
Bay- Cheng, Sarah, Jennifer Parker- Starbuck, and David Z. Saltz. 2015. Performance and 
Media. Ann Arbor: University of Michigan.
Bedek, Michael A., Alexander Nussbaumer, Luca Huszar, and Dietrich Albert. 2018. 
“Methods for Discovering Cognitive Biases in a Visual Analytics Environment.” In 
Cognitive Biases in Visualizations, edited by Geoffrey Ellis, 61–7 3. Cham: Springer.
Bench, Harmony, and Kate Elswit. 2017. “Mapping Movement on the Move: Dance 
Touring and Digital Methods.” Theatre Journal 68 (4): 575– 96. https://doi.org/10.13 
53/tj.2016.0107
Benesh Institute. n.d. “Benesh Notation Editor—B enesh Notation Editor Help Cen-
tre.” Accessed December 9, 2018. http://www.ok-edesign.com/Benesh/BNBNE 
_Whatisbne.html
Bermudez, Bertha, Scott Delahunta, Hoogenboom Marijke, Ziegler Chris, Frédéric 
Bevilacqua, Sarah Fdili Alaoui, and Barbara Meneses Gutierrez. 2011. “The Double 
Skin/Double Mind Interactive Installation.”
Bernstein, Lenny, Peter Bosch, Osvaldo Canziani, Zhenlin Chen, Renate Christ, Ogun-
lade Davidson, William Hare, Saleemul Huq, David Karoly, and Vladimir Kattsov. 
2008. Climate Change 2007: Synthesis Report: An Assessment of the Intergovernmental Panel 
on Climate Change. IPCC.
Berry, David M., and Anders Fagerjord. 2017. Digital Humanities: Knowledge and Critique 
in a Digital Age. Cambridge: Polity Press.
Bertin, Jacques. 1983. Semiology of Graphics; Diagrams Networks Maps. Madison: Univer-
sity of Wisconsin Press.
Best, Michael, and Janelle Jenstad. 2013. The Internet Shakespeare Editions. University of 
Victoria. http://internetshakespeare.uvic.ca/
Beyes, Timon, Martina Leeker, and Imanuel Schipper, eds. 2017. Performing the Digi-
tal: Performativity and Performance Studies in Digital Cultures. Digital Society. Bielefeld: 
Transcript Verlag.
Bhaskar, Roy, and Mervyn Hartwig. 2010. The Formation of Critical Realism: A Personal Per-
spective. Ontological Explorations. Routledge.
Biet, Christian, Pierre Frantz, Florence Filippi, Georges Forestier, Sylvaine Guyot, Sara 
Harvey, Tiphaine Karsenti, Sophie Marchand, Jeffrey Ravel, and Agathe Sanjuan. 
2015. “The Comédie- Française Registers Project— Comédie Française Registers 
Project.” 2015. https://www.cfregisters.org/en/
Binongo, J. N. G., and M. W. A. Smith. 1999. “The Application of Principal Component 
Analysis to Stylometry.” Literary and Linguistic Computing 14 (4): 445– 66. https://doi 
.org/10.1093/llc/14.4.445
Birch, David. 2004. “Celebrating the Ordinary in Singapore in Extraordinary Ways: The 
Cultural Politics of The Necessary Stage’s Collaborative Theatre.” In Ask Not: The 
Necessary Stage in Singapore Theatre, edited by Chong Kee Tan and Tisa Ng, 266–9 0. 
Singapore: Times Editions.
Bishop, Ryan, and John Phillips. 2007. “Of Method.” Theory, Culture & Society 24 (7– 8): 
264– 75. https://doi.org/10.1177/0263276407084796
Black, Carolyn, John Craig, James Cummings, Matthew Davies, Alexandra Gillespie, 
Peter Greenfield, Diane Jakacki, et al. 2017. “REED Online.” 2017. https://ereed.li 
brary.utoronto.ca/
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
198 • References
Bleeker, Maaike. 2016. Transmission in Motion: The Technologizing of Dance. Oxon and New 
York: Routledge.
Bode, Katherine. 2012. Reading by Numbers: Recalibrating the Literary Field. London and 
New York: Anthem Press. http://site.ebrary.com/id/10595451
Bode, Katherine. 2020. “Why You Can’t Model Away Bias.” Modern Language Quarterly 
81 (1).
Bodenhamer, David J. 2013. “Beyond GIS: Geospatial Technologies and the Future of 
History.” In History and GIS: Epistemologies, Considerations and Reflections, edited by 
Alexander von Lünen and Charles Travis, 1– 13. Dordrecht: Springer Netherlands. 
https://doi.org/10.1007/978-94-007-5009-8_1
Bodenhamer, David, John Corrigan, and Trevor M. Harris. 2013. “Deep Mapping and 
the Spatial Humanities.” International Journal of Humanities and Arts Computing 7. 
http://diginole.lib.fsu.edu/islandora/object/fsu%3A209941/
Bollacker, Kurt D. 2010. “Avoiding a Digital Dark Age.” American Scientist 98 (2): 106– 10.
Bollen, Jonathan. 2017. “Data Models for Theatre Research: People, Places, and Per-
formance.” Theatre Journal 68 (4): 615– 32. https://doi.org/10.1353/tj.2016.0109
Bollen, Jonathan, and Julie Holledge. 2011. “Hidden Dramas: Cartographic Revela-
tions in the World of Theatre Studies.” Cartographic Journal 48 (4): 226– 36. https:// 
doi.org/10.1179/1743277411Y.0000000026
Borgman, Christine L. 2009. “The Digital Future Is Now: A Call to Action for the 
Humanities.” Digital Humanities Quarterly 3 (4). http://www.digitalhumanities.org 
/dhq/vol/3/4/000077/000077.html
Borgman, Christine L. 2015. Big Data, Little Data, No Data. Cambridge: MIT Press.
Bowker, Geoffrey C., and Susan Leigh Star. 1999. Sorting Things Out. Inside Technology. 
Cambridge: MIT Press.
Breiman, Leo. 2001. “Statistical Modeling: The Two Cultures (with Comments and a 
Rejoinder by the Author).” Statistical Science 16 (3): 199–2 31.
Broadwell, Peter M. 2019. “Automated Movement and Choreography Analysis of Video 
Data via Deep Learning Pose Detection.” 2019. http://broadwell.github.io/DH2019 
_Movement_Choreography.html
Broadwell, Peter M., and Timothy R. Tangherlini. 2017. “GhostScope: Conceptual 
Mapping of Supernatural Phenomena in a Large Folklore Corpus.” In Maths Meets 
Myths: Quantitative Approaches to Ancient Narratives, edited by Ralph Kenna, Máirín 
MacCarron, and Pádraig MacCarron, 131–5 7. Understanding Complex Systems. 
Cham: Springer International. https://doi.org/10.1007/978-3-319-39445-9_8
Brodbeck, Frederic. 2011. “CINEMETRICS— Film Data Visualization.” 2011. http://cin 
emetrics.fredericbrodbeck.de/
Brown, Susan, and John Simpson. 2013. “The Curious Identity of Michael Field and Its 
Implications for Humanities Research with the Semantic Web.” 2013 IEEE Interna-
tional Conference on Big Data, 77– 85. Silicon Valley, CA, October 6– 9. https://doi.org 
/10.1109/BigData.2013.6691674
Bruner, Edward M. 2005. Culture on Tour: Ethnographies of Travel. Chicago: University of 
Chicago Press.
Buckland, Warren. 2008. “What Does the Statistical Style Analysis of Film Involve? A 
Review of Moving into Pictures. More on Film History, Style, and Analysis.” Literary 
and Linguistic Computing 23 (2): 219– 30. https://doi.org/10.1093/llc/fqm046
Buddenbohm, Stefan, Nathanael Cretin, Elly Dijk, Bertrand Gaiffe, Maaike De Jong, 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
References • 199
Nathalie Le Tellier- Becquart, and Jean- Luc Minel. 2016. “State of the Art Report on 
Open Access Publishing of Research Data in the Humanities.” DARIAH.
Bulger, Monica, Greg Taylor, and Ralph Schroeder. n.d. “Data- Driven Business Mod-
els: Challenges and Opportunities of Big Data,” 74.
Burrows, John. 2002. “‘Delta’: A Measure of Stylistic Difference and a Guide to Likely 
Authorship.” Literary and Linguistic Computing 17 (3): 267– 87. https://doi.org/10.10 
93/llc/17.3.267
Burrows, John. 2007. “All the Way Through: Testing for Authorship in Different Fre-
quency Strata.” Literary and Linguistic Computing 22 (1): 27–4 7. https://doi.org/10.10 
93/llc/fqi067
Busa, Roberto. 2005. “Index Thomisticus.” http://www.corpusthomisticum.org/it/in 
dex.age
Cairo, Alberto. 2020. “Foreword.” In Data Visualization in Society, edited by Martin 
Engebretsen and Helen Kennedy, 17–1 8. Amsterdam: Amsterdam University Press. 
https://doi.org/10.2307/j.ctvzgb8c7.6
Calude, Cristian S., and Giuseppe Longo. 2017. “The Deluge of Spurious Correlations 
in Big Data.” Foundations of Science 22 (3): 595–6 12. https://doi.org/10.1007/s10699 
-016-9489-4
Calvert, Thomas W. 1986. “Toward a Language for Human Movement.” Computers and 
the Humanities 20 (1): 35–4 3. https://doi.org/10.1007/BF02393462
Cameron, Deborah. 2011. “Evolution, Science and the Study of Literature: A Critical 
Response:” Language and Literature, February. https://doi.org/10.1177/09639470103 
91126
Cao, Zhe, Gines Hidalgo, Tomas Simon, Shih- En Wei, and Yaser Sheikh. 2018. “Open-
Pose: Realtime Multi- Person 2D Pose Estimation Using Part Affinity Fields.” ArXiv: 
1812.08008 [Cs], December. http://arxiv.org/abs/1812.08008
Caplan, Debra. 2015. “Notes from the Frontier: Digital Scholarship and the Future of 
Theatre Studies.” Theatre Journal 67 (2): 347– 59. https://doi.org/10.1353/tj.2015 
.0059
Caplan, Debra. 2017. “Reassessing Obscurity: The Case for Big Data in Theatre His-
tory.” Theatre Journal 68 (4): 555–7 3. https://doi.org/10.1353/tj.2016.0106
Carlin, David, and Laurene Vaughan, eds. 2015. Performing Digital: Multiple Perspectives 
on a Living Archive. First ed. Farnham, Surrey, England; Burlington, VT: Routledge.
Carlson, K., T. Schiphorst, and C. Shaw. 2011. “ActionPlot: A Visualization Tool for 
Contemporary Dance Analysis.” In Proceedings of the International Symposium on Com-
putational Aesthetics in Graphics, Visualization, and Imaging, 113–2 0. CAe ’11. New York: 
ACM. https://doi.org/10.1145/2030441.2030466
Causey, Matthew. 2006. Theatre and Performance in Digital Culture. Vol. 5. Routledge 
Advances in Theatre and Performance Studies. London: Routledge.
Chan, Jacky C. P., Howard Leung, Jeff K. T. Tang, and Taku Komura. 2010. “A Virtual 
Reality Dance Training System Using Motion Capture Technology.” IEEE Transac-
tions on Learning Technologies 4 (2): 187– 95.
Chandra, Vikram. 2014. Geek Sublime: The Beauty of Code, the Code of Beauty. Minneapolis: 
Graywolf Press.
Chang, Michael, Mark Halaki, Roger Adams, Stephen Cobley, Kwee-Y um Lee, and 
Nicholas O’Dwyer. 2016. “An Exploration of the Perception of Dance and Its Rela-
tion to Biomechanical Motion: A Systematic Review and Narrative Synthesis.” Jour-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
200 • References
nal of Dance Medicine & Science 20 (3): 127–3 6. https://doi.org/10.12678/1089-313X 
.20.3.127
Chapple, Freda, and Chiel Kattenbelt. 2006. Intermediality in Theatre and Performance. 
Themes in Theatre: Collective Approaches to Theatre and Performance. Asterdam: 
Rodopi.
Chatzichristodoulou, Maria, Janis Jefferies, and Rachel Zerihan. 2009. Interfaces of Per-
formance. Digital Research in the Arts and Humanities. Farnham and Burlington: 
Ashgate.
Cheesman, Tom, Kevin Flanagan, Zhao Geng, and Sebastian Sadowski. 2011. “Version 
Variation Visualization.” http://www.delightedbeauty.org/vvv
Choensawat, Worawat, Minako Nakamura, and Kozaburo Hachimura. 2015. 
“GenLaban: A Tool for Generating Labanotation from Motion Cap-
ture Data.” Multimedia Tools and Applications 74 (23): 10823–4 6. https://
doi.org/10.1007/s11042-014-2209-6
Choi, Yeon- Mu, and Hyun- Joo Kim. 2007. “A Directed Network of Greek and Roman 
Mythology.” Physica A: Statistical Mechanics and Its Applications 382 (2): 665–7 1.
Cirillo, Pasquale, and Nassim Nicholas Taleb. 2016. “Expected Shortfall Estimation for 
Apparently Infinite- Mean Models of Operational Risk.” Quantitative Finance 16 (10): 
1485– 94. https://doi.org/10.1080/14697688.2016.1162908
Clemente, Filipe Manuel, Fernando Manuel Lourenço Martins, Dimitris Kalamaras, P. 
Del Wong, and Rui Sousa Mendes. 2015. “General Network Analysis of National 
Soccer Teams in FIFA World Cup 2014.” International Journal of Performance Analysis in 
Sport 15 (1): 80–9 6. https://doi.org/10.1080/24748668.2015.11868778
Cleveland, William. 1985. The Elements of Graphing Data. Monterey, CA: Wadsworth 
Advanced Books and Software.
Cleveland, William. 1993. Visualizing Data. Summit, NJ: Hobart Press.
Cohen, Jacob. 1988. Statistical Power Analysis for the Behavioral Sciences. Hillsdale, NJ: Law-
rence Erlbaum Associates.
Conquergood, Dwight. 2002. “Performance Studies: Interventions and Radical 
Research.” TDR/The Drama Review 46 (2): 145–5 6. https://doi.org/10.1162/1054204 
02320980550
Craig, Hugh, and Brett Greatley- Hirsch. 2017. Style, Computers, and Early Modern Drama: 
Beyond Authorship. Cambridge: Cambridge University Press.
Craig, Hugh, and Arthur F. Kinney. 2009. Shakespeare, Computers, and the Mystery of 
Authorship. Cambridge: Cambridge University Press.
Cuykendall, Shannon, Ethan Soutar-R au, and Thecla Schiphorst. 2016. “POEME: A 
Poetry Engine Powered by Your Movement.” In Proceedings of the TEI ’16: Tenth Inter-
national Conference on Tangible, Embedded, and Embodied Interaction, 635–4 0. TEI ’16. 
New York: ACM. https://doi.org/10.1145/2839462.2856339
Cuykendall, Shannon, Ethan Soutar- Rau, Thecla Schiphorst, and Steve DiPaola. 2016. 
“If Words Could Dance: Moving from Body to Data Through Kinesthetic Evalua-
tion.” In Proceedings of the 2016 ACM Conference on Designing Interactive Systems, 234– 38. 
DIS ’16. New York: ACM. https://doi.org/10.1145/2901790.2901822
D’Ignazio, Catherine, and Lauren F. Klein. 2016. “Feminist Data Visualization.” In 
Workshop on Visualization for the Digital Humanities (VIS4DH). Baltimore, MD: IEEE.
D’Ignazio, Catherine, and Lauren F. Klein. 2020. Data Feminism. Cambridge, MA: MIT 
Press.
Da, Nan Z. 2019. “The Computational Case against Computational Literary Studies.” 
Critical Inquiry 45 (3): 601– 39. https://doi.org/10.1086/702594
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
References • 201
Diakopoulos, Nicholas. 2018. “Ethics in Data-D riven Visual Storytelling.” In Data- 
Driven Storytelling, edited by Nathalie Henry Riche, Christophe Hurter, Nicholas 
Diakopoulos, and Sheelagh Carpendale, 59–8 3. New York: A. K. Peters/CRC Press. 
https://doi.org/10.1201/9781315281575-10
Davidson, Alwyn, Deb Verhoeven, and Colin Arrowsmith. 2015. “Petal Diagrams: A 
New Technique for Mapping Historical Change in the Film Industry.” International 
Journal of Humanities and Arts Computing 9 (2): 142– 63. https://doi.org/10.3366/ijhac 
.2015.0146
DeLahunta, Scott. 2017. “Dance Becoming Data: Part One Software for Dancers.” Com-
putational Culture, no. 6 (November). http://computationalculture.net/dance-beco 
ming-data-part-one-software-for-dancers/
DeLahunta, Scott, and Florian Jenett. 2017. “Making Digital Choreographic Objects 
Interrelate. A Focus on Coding Practices.” In Performing the Digital. Performativity 
and Performance Studies in Digital Cultures, by Martina Leeker, Imanuel Schipper, and 
Timon Beyes. Bielefeld: transcript. http://dx.doi.org/10.25969/mediarep/2079
Delbridge, Matthew, and Joanne Tompkins. 2009. “Using Virtual Reality Modelling 
in Cultural Management, Archiving and Research.” In EVA London 2009, 260– 69. 
London.
Derrida, Jacques. 1978. Writing and Difference. Chicago: University of Chicago Press.
Deutsch, David. 2009. A New Way to Explain Explanation. TEDGlobal 2009. https://www 
.ted.com/talks/david_deutsch_a_new_way_to_explain_explanation
Developers, NetworkX. 2010. “NetworkX.” Networkx. Lanl. Gov.
Dixon, Steve. 2007. Digital Performance: A History of New Media in Theater, Dance, Perfor-
mance Art, and Installation. Cambridge, MA: MIT Press.
Downey, Allen. 2014. Think Stats: Exploratory Data Analysis in Python. Open Textbook 
Library. Needham: Green Tea Press.
Drucker, Johanna. 2009. Speculative Computing: Basic Principles and Essential Distinctions. 
Chicago: University of Chicago Press. http://chicago.universitypressscholarship 
.com/view/10.7208/chicago/9780226165097.001.0001/upso-9780226165073-
chap ter-2
Drucker, Johanna. 2011. “Humanities Approaches to Graphical Display.” Digital 
Humanities Quarterly 5 (1).
Drucker, Johanna. 2013. “Diagrammatic Writing.” New Formations: A Journal of Culture/
Theory/Politics 78 (1): 83– 101.
Drucker, Johanna. 2014. “Graphesis.” Paj: The Journal of the Initiative for Digital Humani-
ties, Media, and Culture 2 (1).
Duffell, Lynsey D., Natalie Hope, and Alison H. McGregor. 2014. “Comparison of 
Kinematic and Kinetic Parameters Calculated Using a Cluster-B ased Model and 
Vicon’s Plug- in Gait.” Proceedings of the Institution of Mechanical Engineers, Part H: Jour-
nal of Engineering in Medicine 228 (2): 206– 10.
Dunne, Anthony, and Fiona Raby. 2013. Speculative Everything: Design, Fiction, and Social 
Dreaming. Cambridge, MA: MIT Press. http://muse.jhu.edu/book/28148
Eder, Maciej. 2015. “Does Size Matter? Authorship Attribution, Small Samples, Big 
Problem.” Literary and Linguistic Computing 30 (2): 167– 82.
Eder, Maciej, Jan Rybicki, and Mike Kestemont. 2016. “Stylometry with R: A Package 
for Computational Text Analysis.” R Journal 8 (1): 107–2 1. https://doi.org/10.1093 
/llc/fqt066
Edwards, Paul K., Steve Vincent, and Joe O’Mahoney. 2014. “Concluding Comments.” 
In Studying Organizations Using Critical Realism: A Practical Guide, edited by Paul 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
202 • References
Edwads, Joe O’Mahoney, and Steve Vincent. Oxford: Oxford University Press. http://
www.oxfordscholarship.com/view/10.1093/acprof:oso/9780199665525.001 
.0001/acprof-9780199665525-chapter-17
Eijk, Gwen van. 2017. “Socioeconomic Marginality in Sentencing: The Built-i n Bias in 
Risk Assessment Tools and the Reproduction of Social Inequality.” Punishment & 
Society 19 (4): 463– 81. https://doi.org/10.1177/1462474516666282
El Raheb, Katerina, and Yannis E. Ioannidis. 2014. “From Dance Notation to Concep-
tual Models: A Multilayer Approach.” In MOCO. https://doi.org/10.1145/2617995 
.2618000
Elder- Vass, Dave. 2012. The Reality of Social Construction. Cambridge.
Elhayek, Ahmed, Onorina Kovalenko, Pramod Murthy, Jameel Malik, and Didier 
Stricker. 2018. “Fully Automatic Multi-P erson Human Motion Capture for VR 
Applications.” In Virtual Reality and Augmented Reality, edited by Patrick Bourdot, 
Sue Cobb, Victoria Interrante, Hirokazu Kato, and Didier Stricker, 28– 47. Lecture 
Notes in Computer Science. Springer International.
Eliot, Simon. 2002. “Very Necessary but Not Quite Sufficient: A Personal View of 
Quantitative Analysis in Book History.” Book History 5: 283– 93. https://www.jstor 
.org/stable/30228195
Eliot, T. S. 1934. The Rock. London: Faber and Faber.
Endings Project Team, The. 2019. “The Endings Project.” 2019. https://projectendings 
.github.io/
Escobar Varela, Miguel. 2015. “Wayang Kontemporer: Innovations in Javanese Wayang 
Kulit.” Singapore: National University of Singapore. http://cwa-web.org/dissertati 
on/wayang-dis/index.php
Escobar Varela, Miguel. 2016. “The Archive as Repertoire: Transience and Sustainabil-
ity in Digital Archives.” DHQ: Digital Humanities Quarterly 10 (4).
Escobar Varela, Miguel. 2017. “From Copper-P late Inscriptions to Interactive Web-
sites: Documenting Javanese Wayang Theatre.” In Documenting Performance: The 
Context and Processes of Digital Curation and Archiving, 203– 14. London and New York: 
Bloomsbury Methuen Drama.
Escobar Varela, Miguel. 2019. “Towards a Digital, Data-D riven Wayang Kulit Encyclo-
pedia.” Indonesia and the Malay World 47 (137): 23–4 6. https://doi.org/10.1080/1363 
9811.2019.1553382
Escobar Varela, Miguel, and Nala H. Lee. 2018. “Language Documentation: A Reference 
Point for Theatre and Performance Archives?” International Journal of Performance Arts 
and Digital Media 14 (1): 17– 33. https://doi.org/10.1080/14794713.2018.14 53242
Escobar Varela, Miguel, and Gea Oswah Fatah Parikesit. 2017. “A Quantitative Close 
Analysis of a Theatre Video Recording.” Digital Scholarship in the Humanities 32 (2): 
276– 83. https://doi.org/10.1093/llc/fqv069
Eve, Martin Paul. 2019. Close Reading with Computers: Textual Scholarship, Computational 
Formalism, and David Mitchell’s Cloud Atlas. Stanford: Stanford University Press.
Farcomeni, Alessio. 2017. “Contribution to the Discussion of the Paper by Stefan 
Wellek: ‘A Critical Evaluation of the Current p- Value Controversy.’” Biometrical Jour-
nal 59 (5): 880– 81. https://doi.org/10.1002/bimj.201700053
Fdili Alaoui, Sarah, Jules Françoise, Thecla Schiphorst, Karen Studd, and Frédéric Bev-
ilacqua. 2017. “Seeing, Sensing and Recognizing Laban Movement Qualities.” In 
Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems, 4009– 20. 
ACM.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
References • 203
Feinstein, Alvan R. 1971. “XI. Sources of ‘Chronology Bias’ in Cohort Statistics.” Clini-
cal Pharmacology & Therapeutics 12 (5): 864– 79.
Fensham, Rachel. 2016. “Searching Movement’s History: Digital Dance Archives.” In 
Transmission in Motion: The Technologizing of Dance, edited by Maaike Bleeker, 94–1 03. 
Oxon and New York: Routledge.
Fensham, Rachel. 2019. “Research Methods and Problems.” In The Bloomsbury Compan-
ion to Dance Studies, edited by Sherril Dodds, 35. London: Bloomsbury.
Ferrer Valls, Teresa. 2019. “Base de Datos de Comedias Mencionadas En La Documen-
tación Teatral 1540– 1700.” 2019. http://catcom.uv.es/consulta/
Feynman, Richard. 1955. “The Value of Science.” Engineering and Science 19 (3): 13–1 5.
Feynman, Richard. 1985. “Surely You’re Joking, Mr. Feynman!”: Adventures of a Curious Char-
acter. Edited by Ralph Leighton and Edward Hutchings. New York: W. W. Norton.
Feynman, Richard. 2017. The Character of Physical Law. Edited by Alan Sleath. Cam-
bridge, MA: MIT Press.
Fischer, Frank, Matthias Göbel, Dario Kampkaspar, Christopher Kittel, and Peer 
Trilcke. 2017. “Network Dynamics, Plot Analysis: Approaching the Progres-
sive Structuration of Literary Texts.” In Digital Humanities 2017 Conference Abstracts. 
Montréal.
Fischer, Frank, Peer Trilcke, Carsten Milling, and Daniil Skorinkin. 2018. “To Catch 
A Protagonist: Quantitative Dominance Relations In German- Language Drama 
(1730– 1930).” In Digital Humanities 2018 Conference Abstracts. Mexico City.
Fischer- Lichte, Erika, Ramona Thomasius, and Minou Arjomand. 2014. The Routledge 
Introduction to Theatre and Performance Studies. Routledge.
Flanders, Julia, and Fotis Jannidis. 2015. “Data Modeling.” In A New Companion to Digi-
tal Humanities, by Susan Schreibman, Ray Siemens, and John Unsworth, 229–3 7. 
John Wiley & Sons. https://doi.org/10.1002/9781118680605.ch16
Forsythe, William, Maria Palazzi, Norah Zuniga Shaw, and Scott deLahunta. 2009. 
“Synchronous Objects for One Flat Thing, Reproduced.” Website Installation or 
On Line Resource. Columbus, Ohio.
Foucault, Michel, James D. Faubion, and Robert Hurley. 1998. Aesthetics, Method, and 
Epistemology. Vol. 2. New York: New Press. http://www.yorku.ca/rajagopa/documen 
ts/Topic-conceptualthemes.docx
Franco, Israel. 2020. Reseña Histórica Del Teatro En México 2.0– 2.1. Sistema de Información de 
La Crítica Teatral (Historical Theater Reviews in Mexico 2.0– 2.1, Theater Review Information 
System). Accessed May 21, 2020. http://criticateatral2021.org/
Friendly, Michael, and Daniel J. Denis. 2001. “Milestones in the History of Thematic 
Cartography, Statistical Graphics, and Data Visualization.” http://datavis.ca/mile 
stones/
Friendly, Michael, and Daniel Denis. 2005. “The Early Origins and Development of the 
Scatterplot.” Journal of the History of the Behavioral Sciences 41 (2): 103–3 0. https://doi 
.org/10.1002/jhbs.20078
Gaver, Bill, Tonny Dunne, and Elena Pacenti. 1999. “Cultural Probes.” Interactions 6 (1). 
https://interactions.acm.org/archive/view/jan.-feb.-1999/design-cultural-probes1
Geertz, Clifford. 1973. The Interpretation of Cultures: Selected Essays. New York: Basic Books.
Geertz, Clifford. 1974. “‘From the Native’s Point of View’: On the Nature of Anthro-
pological Understanding.” Bulletin of the American Academy of Arts and Sciences 28 (1): 
26– 45. https://doi.org/10.2307/3822971
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
204 • References
Gelman, Andrew, and Antony Unwin. 2013. “Infovis and Statistical Graphics: Differ-
ent Goals, Different Looks.” Journal of Computational and Graphical Statistics 22 (1): 
2– 28. https://doi.org/10.1080/10618600.2012.761137
Gil, Alex. 2015. “The User, the Learner and the Machines We Make · Minimal Comput-
ing.” 2015. http://go-dh.github.io/mincomp/thoughts/2015/05/21/user-vs-learner/
Gilbert, Richard O. 1987. Statistical Methods for Environmental Pollution Monitoring. New 
York: Van Nostrand Reinhold.
Gitelman, Lisa. 2013. “Raw Data” Is an Oxymoron. Infrastructures Series. Cambridge, 
MA: MIT Press.
Gleiser, Pablo M, and Leon Danon. 2003. “Community Structure in Jazz.” Advances in 
Complex Systems 6 (4): 565– 73.
Gonzalez, Anita, Clara McClenon, Katy Robinson, Allyson Mackay, Stacey Bishop, 
Amana Kazkazi, Justin Joque, et al. n.d. “19th Century Acts.” Accessed March 23, 
2019. http://19thcenturyacts.com/
Google Developers. n.d. “Important Updates | Google Maps Platform.” Accessed 
December 18, 2018. https://developers.google.com/maps/billing/important-upd 
ates
Gorski, Philip S. 2013. “‘What Is Critical Realism? And Why Should You Care?’” Con-
temporary Sociology 42 (5): 658– 70. https://doi.org/10.1177/0094306113499533
Gottschall, Jonathan. 2008. Literature, Science, and a New Humanities. First Edition. New 
York: Palgrave Macmillan.
Graham, Paul. 2003. “Hackers and Painters.” http://www.paulgraham.com/hp.html
Graham, Ronald L., Bruce L. Rothschild, and Joel H. Spencer. 1990. Ramsey Theory. 
Wiley- Interscience Series in Discrete Mathematics and Optimization. Wiley.
Gray, Jonathan, Liliana Bounegru, Stefania Milan, and Paolo Ciuccarelli. 2016. “Ways 
of Seeing Data: Toward a Critical Literacy for Data Visualizations as Research 
Objects and Research Devices.” In Innovative Methods in Media and Communication 
Research, edited by Sebastian Kubitschko and Anne Kaun, 227– 51. Cham: Springer 
International Publishing. https://doi.org/10.1007/978-3-319-40700-5_12
Groff, Ruth. 2004. Critical Realism, Post-P ositivism, and the Possibility of Knowledge. Vol. 11. 
Routledge Studies in Critical Realism. London and New York: Routledge.
Guest, Ann Hutchinson. 1998. Choreo-G raphics: A Comparison of Dance Notation Systems 
from the Fifteenth Century to the Present. Amsterdam: Psychology Press.
Hachimura, Kozaburo. 2006. “Digital Archiving of Dancing.” Review of the National Cen-
ter for Digitization 8: 51– 66.
Hachimura, Kozaburo, and Minako Nakamura. 2006. “An XML Representation of Lab-
notation, LabanXML, and Its Implementation on the Notation Editor LabanEdi-
tor2.” Преглед НЦД 9: 47– 51.
Hajian, Sara, Francesco Bonchi, and Carlos Castillo. 2016. “Algorithmic Bias.” In Pro-
ceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data 
Mining, 2125– 26. KDD ’16. New York: ACM. https://doi.org/10.1145/2939672.294 
5386
Hallam, Julia, and Les Roberts. 2013. Locating the Moving Image: New Approaches to Film 
and Place. Bloomington: Indiana University Press. http://muse.jhu.edu/book/27089
Hardjowirogo. 1948. Sejarah Wayang Purwa. Jakarta: Balai Pustaka.
Harrell, D. Fox. 2013. Phantasmal Media: An Approach to Imagination, Computation, and 
Expression. Cambridge, MA: MIT Press.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
References • 205
Hechtl, Angelika, Frank Fischer, Anika Schultz, Christopher Kittel, Elisa Beshero- 
Bondar, Steffen Martus, Peer Trilcke, Jana Wolf, Ingo Börner, and Daniil Skorinkin. 
2018. “Brecht Beats Shakespeare! A Card- Game Intervention Revolving Around 
the Network Analysis of European Drama.” In Digital Humanities 2018 Conference 
Abstracts, 595– 96. Mexico City: ADHO.
Hemispheric Institute. 2008. “Hemispheric Institute Digital Video Library.” 2008. 
www.hemisphericinstitute.org/eng/hidvl/
Hemispheric Institute. 2020. “Gesture— HemiPress.” 2020. https://hemi.press/gest ure/
Hepworth, Katherine, and Christopher Church. 2019. “Racism in the Machine: Visu-
alization Ethics in Digital Humanities Projects.” Digital Humanities Quarterly 12 (4).
Hernandez- Barraza, Luis, Chen- Hua Yeow, and Miguel Escobar Varela. 2019. “The 
Biomechanics of Character Types in Javanese Dance.” Journal of Dance Medicine & 
Science 23 (3): 104– 11.
Higgins, Sarah. 2008. “The DCC Curation Lifecycle Model.” International Journal of Digi-
tal Curation 3 (1).
Hiippala, Tuomo. 2020. “A Multimodal Perspective on Data Visualization.” In Data 
Visualization in Society, edited by Martin Engebretsen and Helen Kennedy, 277–9 4. 
Amsterdam: Amsterdam University Press. https://doi.org/10.2307/j.ctvzgb8c7.23
Hirsch, Brett D., and Janelle Jenstad. 2016. “Beyond the Text: Digital Editions and Per-
formance.” Shakespeare Bulletin 34 (1): 107.
Hogan, Trevor. 2015. “Tangible Data, a Phenomenology of Human- Data Relations.” 
In Proceedings of the Ninth International Conference on Tangible, Embedded, and Embodied 
Interaction, 425– 28. TEI ’15. New York: ACM. https://doi.org/10.1145/2677199.269 
1601
Holledge, Julie, Jonathan Bollen, Frode Helland, and Joanne Tompkins. 2016. A Global 
Doll’s House: Ibsen and Distant Visions. Palgrave Studies in Performance and Technol-
ogy. London: Palgrave Macmillan.
Hughes, Amy E., and Naomi J. Stubbs. 2018. A Player and a Gentleman: The Diary of Harry 
Watkins, Nineteenth- Century U.S. American Actor. Ann Arbor: University of Michigan 
Press.
Hunter, John D. 2007. “Matplotlib: A 2D Graphics Environment.” Computing in Science 
& Engineering 9 (3): 90.
Hussain, Md. Manjurul, and Ishtiak Mahmud. 2019. “PyMannKendall: A Python Pack-
age for Non Parametric Mann Kendall Family of Trend Tests.” Journal of Open Source 
Software 4 (39): 1556. https://doi.org/10.21105/joss.01556
Ichikawa, Tomoko. 2016. “Visualization as Experience.” Digital Studies/Le Champ 
Numérique 5 (3). https://doi.org/10.16995/dscn.32
Ioannides, Marinos, Eleanor Fink, Antonia Moropoulou, Monika Hagedorn-S aupe, 
Antonella Fresa, Gunnar Liestøl, Vlatka Rajcic, and Pierre Grussenmeyer. 2016. 
Digital Heritage. Progress in Cultural Heritage: Documentation, Preservation, and Protection: 
6th International Conference, EuroMed 2016, Nicosia, Cyprus, October 31— November 5, 
2016, Proceedings. Cham, Switzerland: Springer.
Ioannidis, John P. A. 2005. “Why Most Published Research Findings Are False.” PLoS 
Medicine 2 (8). https://doi.org/10.1371/journal.pmed.0020124
Ioannidis, John P. A. 2014. “How to Make More Published Research True.” PLOS Medi-
cine 11 (10): e1001747. https://doi.org/10.1371/journal.pmed.1001747
Jackson, Shannon. 2004. Professing Performance: Theatre in the Academy from Philology to 
Performativity. Cambridge: Cambridge University Press.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
206 • References
Jannidis, Fotis. 2019. “On the Perceived Complexity of Literature. A Response to Nan 
Z. Da.” Journal of Cultural Analytics Blog (blog). June 17, 2019. https://culturalanal 
ytics.org/2019/06/on-the-perceived-complexity-of-literature-a-response-to-nan 
-z-da/
Jansen, Yvonne, Pierre Dragicevic, Petra Isenberg, Jason Alexander, Abhijit Karnik, 
Johan Kildal, Sriram Subramanian, and Kasper Hornbæk. 2015. “Opportunities 
and Challenges for Data Physicalization.” In Proceedings of the 33rd Annual ACM Con-
ference on Human Factors in Computing Systems, 3227– 36. CHI ’15. New York: ACM. 
https://doi.org/10.1145/2702123.2702180
Jenstad, Janelle. 2011. “Using Early Modern Maps in Literary Studies: Views and Cave-
ats from London.” In GeoHumanities: Art, History, Text at the Edge of Place, by Michael 
Dear, Jim Ketchum, Sarah Luria, and Doug Richardson. London and New York: 
Routledge. https://doi.org/10.4324/9780203839270-22
Jessop, Martyn. 2008. “The Inhibition of Geographical Information in Digital Human-
ities Scholarship.” Literary and Linguistic Computing 23 (1): 39– 50. https://doi.org/10 
.1093/llc/fqm041
Jockers, Matthew Lee. 2013. Macroanalysis. Topics in the Digital Humanities. Balti-
more: University of Illinois Press.
Jockers, Matthew Lee. 2014. Text Analysis with R for Students of Literature. Quantitative 
Methods in the Humanities and Social Sciences. Cham: Springer.
Juola, Patrick. 2018. “Large-S cale Accuracy Benchmark Results for Juola’s Authorship 
Verification Protocols.” In Digital Humanities 2018, DH 2018, Book of Abstracts, El Cole-
gio de México, UNAM, and RedHD, Mexico City, Mexico, June 26–2 9, 2018, 411. https:// 
dh2018.adho.org/en/large-scale-accuracy-benchmark-results-for-juolas-authorsh 
ip-verification-protocols/
Kaeppler, Adrienne L. 1978. “Dance in Anthropological Perspective.” Annual Review of 
Anthropology 7: 31.
Kagan, Jerome. 2009. The Three Cultures. Cambridge University Press.
Karsdorp, F., M. Kestemont, C. Schöch, and A. P. J. van den Bosch. 2015. “The Love 
Equation: Computational Modeling of Romantic Relationships in French Classical 
Drama.” In Proceedings of the 6th Workshop on Computational Models of Narrative (CMN- 
2015), Atlanta, GA. https://repository.ubn.ru.nl/handle/2066/142767
Karsdorp, Folgert, Enrique Manjavacas, Lauren Fonteyn, and Mike Kestemont. 2020. 
“Classifying Evolutionary Forces in Language Change Using Neural Networks.” 
Evolutionary Human Sciences 2. https://doi.org/10.1017/ehs.2020.52
Kealiinohomoku, Joann. 1974. “Dance Culture as a Microcosm of Holistic Culture.” 
In New Dimensions in Dance Research: Anthropology and Dance (The American Indians), 99– 
106. Tucson: University of Arizona, Committee on Research in Dance.
Kendall, Maurice. 1975. Multivariate Analysis. London: Charles Griffin.
Kennedy, Helen, and Martin Engebretsen. 2020. “Introduction:” In Data Visualization 
in Society, edited by Helen Kennedy and Martin Engebretsen, 19–3 2. Amsterdam: 
Amsterdam University Press. https://doi.org/10.2307/j.ctvzgb8c7.7
Kerr, Norbert L. 1998. “HARKing: Hypothesizing after the Results Are Known.” Person-
ality and Social Psychology Review 2 (3): 196– 217.
Kershaw, Baz, and Helen Nicholson. 2011. Research Methods in Theatre and Performance. 
Edinburgh: Edinburgh University Press.
Khmelev, Dmitri V., and Fiona J. Tweedie. 2001. “Using Markov Chains for Identifica-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
References • 207
tion of Writer.” Literary and Linguistic Computing 16 (3): 299–3 07. https://doi.org/10 
.1093/llc/16.3.299
Kitchin, Rob, and Martin Dodge. 2014. Code/Space: Software and Everyday Life. Cam-
bridge, MA: MIT Press.
Klenotic, Jeffrey. 2011. “Putting Cinema History on the Map: Using GIS to Explore the 
Spatiality of Cinema.” In Explorations in New Cinema History: Approaches and Case Stud-
ies, by Richard Maltby, Daniel Biltereyst, and Philippe Meers, 58– 84. Oxford.
Knigge, LaDona, and Meghan Cope. 2016. “Grounded Visualization: Integrating the 
Analysis of Qualitative and Quantitative Data through Grounded Theory and Visu-
alization.” Environment and Planning A, July. https://doi.org/10.1068/a37327
Knoke, David, and Song Yang. 2008. Social Network Analysis. Vol. 154. Quantitative 
Applications in the Social Sciences. Los Angeles: SAGE.
Koene, Ansgar. 2017. “Algorithmic Bias: Addressing Growing Concerns [Leading 
Edge].” IEEE Technology and Society Magazine 36 (2): 31– 32. https://doi.org/10.1109 
/MTS.2017.2697080
Koutedakis, Yiannis, Emmanuel O. Owolabi, and Margo Apostolos. 2008. “Dance Bio-
mechanics: A Tool for Controlling Health, Fitness, and Training.” Journal of Dance 
Medicine & Science: Official Publication of the International Association for Dance Medicine & 
Science 12 (3): 83– 90.
Kozel, Susan. 2007. Closer: Performance, Technologies, Phenomenology. Cambridge, MA: 
MIT Press.
Kramnick, Jonathan. 2018. “The Interdisciplinary Delusion.” The Chronicle of Higher Edu-
cation, October 11, 2018. https://www.chronicle.com/article/The-Interdisciplinary 
-Delusion/244772
Kretzschmar, W. A. 2013. “GIS for Language and Literary Study.” In Literary Studies in 
the Digital Age: An Evolving Anthology, edited by R. Siemens and K. Price. New York: 
Modern Language Association. http://dx.doi.org/10.1632/lsda.2013.0
Krishan, Sanjay, ed. 1997. 9 Lives, 10 Year of Singapore Theatre 1987–1 997. Singapore: First 
Printers.
Kulkarni, Vivek, Rami Al-R fou, Bryan Perozzi, and Steven Skiena. 2015. “Statistically 
Significant Detection of Linguistic Change.” In Proceedings of the 24th International 
Conference on World Wide Web, 625– 35. International World Wide Web Conferences 
Steering Committee.
Kurath, Gertrude P. 1952. “A Choreographic Questionnaire.” Midwest Folklore 2 (1): 
53– 55.
Kwa, Chunglin. 2011. Styles of Knowing: A New History of Science from Ancient Times to the 
Present. University of Pittsburgh Press. https://muse-jhu-edu.libproxy1.nus.edu.sg 
/book/1988
Lakens, Daniel. 2013. “Calculating and Reporting Effect Sizes to Facilitate Cumula-
tive Science: A Practical Primer for t- Tests and ANOVAs.” Frontiers in Psychology 4. 
https://doi.org/10.3389/fpsyg.2013.00863
Larasati, Rachmi Diyah. 2013. The Dance That Makes You Vanish: Cultural Reconstruction in 
Post- Genocide Indonesia. Difference Incorporated. Minneapolis: University of Minne-
sota Press.
Larson, Jeff, and Julia Angwin. 2016. “How We Analyzed the COMPAS Recidivism 
Algorithm.” Text/html. ProPublica. May 23, 2016. https://www.propublica.org/ar 
ticle/how-we-analyzed-the-compas-recidivism-algorithm
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
208 • References
Leino, Antti, and Saara Hyvönen. 2008. “Comparison of Component Models in Ana-
lysing the Distribution of Dialectal Features.” International Journal of Humanities and 
Arts Computing 2 (1–2 ): 173– 87. https://doi.org/10.3366/E1753854809000378
Leiter, Samuel L. 2012. “Is the Onnagata Necessary?” Asian Theatre Journal 29 (1): 112– 
21. https://doi.org/10.1353/atj.2012.0028
Leonhardt, Nic. 2014. “Digital Humanities and the Performing Arts: Building Com-
munities, Creating Knowledge.” In Keynote Address, SIBMAS/TLA Conference, New York 
City. Vol. 12.
Li, William, and Philippe Pasquier. 2016. “Automatic Affect Classification of Human 
Motion Capture Sequences in the Valence- Arousal Model.” In Proceedings of the 3rd 
International Symposium on Movement and Computing, 15:1–1 5:8. MOCO ’16. New 
York: ACM. https://doi.org/10.1145/2948910.2948936
Liao, Han-t eng, and Thomas Petzold. 2010. “Analysing Geo-L inguistic Dynamics of 
the World Wide Web: The Use of Cartograms and Network Analysis to Understand 
Linguistic Development in Wikipedia.” Cultural Science 3 (2).
Lindinger, Christian, David Labbe, Philippe Pollien, Andreas Rytz, Marcel A. Juillerat, 
Chahan Yeretzian, and Imre Blank. 2008. “When Machine Tastes Coffee: Instru-
mental Approach To Predict the Sensory Profile of Espresso Coffee.” Analytical 
Chemistry 80 (5): 1574– 81. https://doi.org/10.1021/ac702196z
Liu, Alan. 2012. “The State of the Digital Humanities: A Report and a Critique.” Arts 
and Humanities in Higher Education 11 (1–2 ): 8–4 1. https://doi.org/10.1177/14740222 
11427364
Luksys, Donatas, and Julius Griskevicius. 2016. “Quantitative Assessment of Dance 
Therapy Influence On the Parkinson’s Disease Patients’ Lower Limb Biomechan-
ics.” Mokslas 8 (6): 583–8 6. http://dx.doi.org.libproxy1.nus.edu.sg/10.3846/mla.20 
16.978
Madison, D. Soyini. 2005. The SAGE Handbook of Performance Studies. SAGE Publications.
Malakar, Sourav, Saptarsi Goswami, and Amlan Chakrabarti. 2018. “An Online Trend 
Detection Strategy for Twitter Using Mann– Kendall Non- Parametric Test.” In 
Industry Interactive Innovations in Science, Engineering and Technology, edited by Swapan 
Bhattacharyya, Sabyasachi Sen, Meghamala Dutta, Papun Biswas, and Himadri 
Chattopadhyay, 185– 93. Singapore: Springer Singapore.
Malmstrom, Carl, Yaying Zhang, Philippe Pasquier, Thecla Schiphorst, and Lyn Bar-
tram. 2016. “MoComp: A Tool for Comparative Visualization Between Takes of 
Motion Capture Data.” In Proceedings of the 3rd International Symposium on Movement 
and Computing, 11:1– 11:8. MOCO ’16. New York: ACM. https://doi.org/10.1145/294 
8910.2948932
Maniatis, Petros, Mema Roussopoulos, Thomas J. Giuli, David SH Rosenthal, Mary 
Baker, and Mary Baker. 2005. “The LOCKSS Peer- to- Peer Digital Preservation Sys-
tem.” ACM Transactions on Computer Systems (TOCS) 23 (1): 2–5 0.
Mann, Henry B. 1945. “Nonparametric Tests against Trend.” Econometrica: Journal of the 
Econometric Society, 245– 59.
Manovich, Lev. 2000. The Language of New Media. Leonardo. Cambridge, MA: MIT Press.
Manovich, Lev. 2011. “What Is Visualisation?” Visual Studies 26 (1): 36– 49. https://doi 
.org/10.1080/1472586X.2011.548488
Manovich, Lev. 2013. “Visualizing Vertov.” Russian Journal of Communication 5 (1): 44–5 5.
Manovich, Lev. 2020. Cultural Analytics. Cambridge, MA: MIT Press.
Manzetti, Maria Cristina. 2016. “3D Visibility Analysis as a Tool to Validate Ancient 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
References • 209
Theatre Reconstructions: The Case of the Large Roman Theatre of Gortyn.” Virtual 
Archaeology Review 7 (15): 36–4 3. https://doi.org/10.4995/var.2016.5922
Manzor, Lillian, Kyle Rimkus, and Mitsunori Ogihara. 2013. “Cuban Theater Digital 
Archive: A Multimodal Platform for Theater Documentation and Research.” In 
International Conference on Information Technologies for Performing Arts, Media Access, and 
Entertainment, 138– 50. Springer.
Marasini, Donata, Piero Quatto, and Enrico Ripamonti. 2016. “The Use of P- Values in 
Applied Research: Interpretation and New Trends.” Statistica; Bologna 76 (4): 315– 
25. http://dx.doi.org.libproxy1.nus.edu.sg/10.6092/issn.1973-2201/6439
Marsden, Peter. 2005. “Recent Developments in Network Measurement.” In Models 
and Methods in Social Network Analysis, edited by Peter J. Carrington, John Scott, and 
Stanley Wasserman. Vol. 28. Structural Analysis in the Social Sciences. Cambridge: 
Cambridge University Press.
Mayer-S chönberger, Viktor, and Kenneth Cukier. 2013. Big Data: A Revolution That Will 
Transform How We Live, Work, and Think. Boston: Houghton Mifflin Harcourt.
Mee, Erin B. 2013. “Hearing the Music of the Hemispheres.” TDR: The Drama Review 57 
(3): 148–5 0.
Mee, Erin B. 2018. “Born- Digital Scholarship.” TDR: The Drama Review 62 (3): 8– 9.
Meeks, Elijah, and Scott Weingart. 2013. “The Digital Humanities Contribution to 
Topic Modeling.” Journal of Digital Humanities, April 9. http://journalofdigitalhuma 
nities.org/2-1/dh-contribution-to-topic-modeling/
Meister J. C, Petris, M, Gius, E, and Jacke, J. 2016. “CATMA 5.0 [Software for Text 
Annotation and Analysis.” http://catma.de/
Michel, Jean- Baptiste, Yuan Kui Shen, Aviva P. Aiden, Adrian Veres, Matthew K. Gray, 
Joseph P. Pickett, Dale Hoiberg, et al. 2011. “Quantitative Analysis of Culture Using 
Millions of Digitized Books.” Science 331 (6014): 176–8 2. https://doi.org/10.1126/sc 
ience.1199644
Miller, Derek. 2017. “Average Broadway.” Theatre Journal 68 (4): 529– 53. https://doi.org 
/10.1353/tj.2016.0105
Minimal Computing Working Group. 2015. “About · Minimal Computing.” 2015. 
http://go-dh.github.io/mincomp/about/
Misi, Gábor. 1983. “Formal Methods in Form Analysis of Transylvanian Male Solo 
Dances.” Dance Studies 7: 21– 56.
Misi, Gábor. 2005. “Labanatory.” 2005. http://www.labanatory.com/eng/software 
.html
Misi, Gábor. 2008. “An Algebraic Representation of Labanotation for Retrieval and 
Other Operations.” In Proceedings of the 25th Biennial Conference of ICKL, Mexico City, 
143– 60.
Moere, A. V. 2008. “Beyond the Tyranny of the Pixel: Exploring the Physicality of Infor-
mation Visualization.” In 2008 12th International Conference Information Visualisation, 
469– 74. https://doi.org/10.1109/IV.2008.84
Molloy, Laura. 2014. “Digital Curation Skills in the Performing Arts—a n Investiga-
tion of Practitioner Awareness and Knowledge of Digital Object Management and 
Preservation.” International Journal of Performance Arts and Digital Media 10 (1): 7– 20. 
https://doi.org/10.1080/14794713.2014.912496
Moran, Patrick A. P. 1948. “The Interpretation of Statistical Maps.” Journal of the Royal 
Statistical Society. Series B (Methodological) 10 (2): 243–5 1.
Moreno, J. L. 1960. The Sociometry Reader. Glencoe: Free Press.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
210 • References
Moretti, Franco. 2000. “The Slaughterhouse of Literature.” Modern Language Quarterly 
61 (1): 207.
Moretti, Franco. 2007. Graphs, Maps, Trees. Verso.
Moretti, Franco. 2009. “Style, Inc.: Reflections on Seven Thousand Titles (British Nov-
els, 1740– 1850).” Critical Inquiry 36 (1): 134.
Moretti, Franco. 2011. “Network Theory, Plot Analysis.” New Left Review, no. 68 
(March): 80.
Morgan, Susan. 2018. “Fake News, Disinformation, Manipulation and Online Tactics 
to Undermine Democracy.” Journal of Cyber Policy 3 (1): 39– 43. https://doi.org/10.10 
80/23738871.2018.1462395
Mowat, Barbara, Paul Werstine, Michael Poston, and Rebecca Niles, eds. n.d. Hamlet. 
Washington, DC: Folger Shakespeare Library. Accessed 13 April 2020. www.folge 
rdigitaltexts.org
Mrázek, Jan. 2005. “Masks and Selves in Contemporary Java: The Dances of Didik Nini 
Thowok.” Journal of Southeast Asian Studies 36 (2): 249–7 9. https://doi.org/10.1017 
/S0022463405000160
Mrázek, Jan. 2019. Wayang and Its Doubles: Javanese Puppet Theatre, Television and the Inter-
net. Singapore: NUS Press.
Mullainathan, Sendhil, and Ziad Obermeyer. 2017. “Does Machine Learning Automate 
Moral Hazard and Error?” American Economic Review 107 (5): 476–8 0. https://doi.org 
/10.1257/aer.p20171084
Mullaney, Thomas Shawn, Christian Henriot, Jeffrey Snyder-R einke, David William 
McClure, and Glen Worthey. 2019. The Chinese Deathscape: Grave Reform in Modern 
China. Stanford: Stanford University Press.
Nakamura, M. 2017. “The Postures and Movements of Balinese Dance.” In 2017 Inter-
national Conference on Culture and Computing (Culture and Computing), 63–6 4. https://doi 
.org/10.1109/Culture.and.Computing.2017.37
NIST/SEMATECH. 2003. “NIST/SEMATECH e-H andbook of Statistical Methods.” 
2003. https://www.itl.nist.gov/div898/handbook/
Nowviskie, Bethany, David McClure, Wayne Graham, Adam Soroka, Jeremy Boggs, 
and Eric Rochester. 2013. “Geo- Temporal Interpretation of Archival Collections 
with Neatline.” Literary and Linguistic Computing 28 (4): 692– 99. https://doi.org/10 
.1093/llc/fqt043
NYPL/Zooniverse. n.d. “Ensemble@Yale.” Accessed August 3, 2018. http://ensemble 
.yale.edu/#/about
O’Donoghue, Peter. 2010. Research Methods for Sports Performance Analysis. Oxon and New 
York: Routledge.
O’Mahoney, Joe, and Steve Vincent. 2014. “Critical Realism as an Empirical Project.” In 
Studying Organizations Using Critical Realism: A Practical Guide, edited by Paul Edwads, 
Joe O’Mahoney, and Steve Vincent. Oxford: Oxford University Press. http://www 
.oxfordscholarship.com/view/10.1093/acprof:oso/9780199665525.001.0001/acpr 
of-9780199665525-chapter-1
O’Neil, Cathy. 2016. Weapons of Math Destruction. New York: Crown.
Oakes, Michael P. 2014. Literary Detective Work on the Computer. Amsterdam and Philadel-
phia: John Benjamins Publishing.
Oakes, Michael P. 2017. “Computer Stylometry of C. S. Lewis’s The Dark Tower and 
Related Texts.” Digital Scholarship in the Humanities. https://doi.org/10.1093/llc/fq 
x043
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
References • 211
Ohge, Christopher, Steven Olsen-S mith, Elisa Barney Smith, and Adam Brimhall. 
2018. “At the Axis of Reality: Melville’s Marginalia in The Dramatic Works of Wil-
liam Shakespeare.” Leviathan 20 (2): 37– 67.
Olson, Mark J. V. 2013. “Hacking the Humanities: Twenty- First- Century Literacies and 
the ‘Becoming-O ther’ of the Humanities.” In Humanities in the Twenty-F irst Century: 
Beyond Utility and Markets, edited by Eleonora Belfiore and Anna Upchurch, 237– 
50. London: Palgrave Macmillan. https://doi.org/10.1057/9781137361356_13
Open Source Community. (2010) 2018. JavaScript Library for Mobile-F riendly Interactive 
Maps: Leaflet/Leaflet (Version 1.3.4). JavaScript. Leaflet. https://github.com/Leaflet 
/Leaflet
OpenStreetMap Community. 2004. “OpenStreetMap.” https://www.openstreetmap 
.org/
Palazzi, Maria, and Norah Zuniga Shaw. 2009. “Synchronous Objects for One Flat 
Thing, Reproduced.” In SIGGRAPH: International Conference and Exhibition on Computer 
Graphics and Interactive Techniques, 2. New Orleans: ACM.
Pandya, Ami. 2003. “Notation System in Indian Classical Dance— Bharatanatyam.” 
PhD diss., India: Maharaja Sayajirao University of Baroda (India). http://search.pro 
quest.com/docview/1771297342/citation/1957BE1310BF4463PQ/1
Park, Gyeong-M i, Sung- Hwan Kim, Hye-R yeon Hwang, and Hwan- Gue Cho. 2013. 
“Complex System Analysis of Social Networks Extracted from Literary Fictions.” 
International Journal of Machine Learning and Computing 3 (1): 107.
Park, Seung-B o, Kyeong-J in Oh, and Geun-S ik Jo. 2012. “Social Network Analysis in a 
Movie Using Character-N et.” Multimedia Tools and Applications 59 (2): 601– 27.
Parry, Kyle. 2019. “Reading for Enactment: A Performative Approach to Digital Schol-
arship and Data Visualization.” In Debates in the Digital Humanities 2019, by Matthew 
K. Gold and Lauren F. Klein. Debates in the Digital Humanities. Minneapolis: Uni-
versity of Minnesota Press. https://dhdebates.gc.cuny.edu/read/untitled-f2acf72c 
-a469-49d8-be35-67f9ac1e3a60/section/a2a6a192-f04a-4082-afaa-97c76a75b21c 
#ch24
Pavis, Patrice. 1996. The Intercultural Performance Reader. Routledge.
Peer, Willie van, Jèmeljan Hakemulder, and Sonia Zyngier. 2007. Muses and Measures. 
Newcastle: Cambridge Scholars.
Pendón Martínez, Alberto, and Gema Bueno de la Fuente. 2017. “Description Mod-
els for Documenting Performance.” In Documenting Performance: The Context and Pro-
cesses of Digital Curation and Archiving, by Toni Sant, 29–4 6. London and New York: 
Bloomsbury.
Phelan, Peggy. 1993. Unmarked: The Politics of Performance. London and New York: 
Routledge.
Piper, Andrew. 2018. Enumerations: Data and Literary Study. Chicago: University of Chi-
cago Press.
Pollock, Della. 1998. “Performing Writing.” In The Ends of Performance, by Peggy Phelan 
and Jill Lane, 73– 103. London and New York: New York University Press.
Pomerantz, Jeffrey. 2015. Metadata. The MIT Press Essential Knowledge Series. Cam-
bridge, MA: MIT Press.
Presner, Todd Samuel, David Shepard, and Yoh Kawano. 2014. HyperCities. MetaLAB-
projects. Cambridge, MA: Harvard University Press.
Prince Lab for Digital Humanities. n.d. “English Playbills | Price Lab for Digital 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
212 • References
Humanities.” Accessed August 3, 2018. https://pricelab.sas.upenn.edu/projects 
/english-playbills
Purwadi. 2013. Mengenal Gambar Tokoh Wayang Purwa Dan Keterangannya. Surakarta: 
Cendrawasih.
Radway, Janice A. 1991. Reading the Romance: Women, Patriarchy, and Popular Literature. 
Revised edition. Chapel Hill: University of North Carolina Press.
Rae, Paul. 2018. Real Theatre: Essays in Experience. Theatre and Performance Theory. Cam-
bridge: Cambridge University Press. https://doi.org/10.1017/9781316890752
Ramsay, Stephen. 2003. “Toward an Algorithmic Criticism.” Literary and Linguistic Com-
puting: Journal of the Association for Literary and Linguistic Computing 18 (2): 167.
Ramsay, Stephen. 2007. “Algorithmic Criticism.” In A Companion to Digital Literary Stud-
ies, by Raymond George Siemens and Susan Schreibman. Vol. 50. Blackwell Com-
panions to Literature and Culture. Malden, MA: Blackwell.
Ramsay, Stephen. 2011. Reading Machines: Toward an Algorithmic Criticism. Champaign: 
University of Illinois Press. http://www.jstor.org/stable/10.5406/j.ctt1xcmrr
Ramsay, Stephen. 2012. “Programming with Humanists: Reflections on Raising an 
Army of Hacker- Scholars in the Digital Humanities.” In Digital Humanities Peda-
gogy, edited by Brett D. Hirsch, 3:227–4 0. Practices, Principles and Politics. Open 
Book Publishers. www.jstor.org/stable/j.ctt5vjtt3.14
Ramsay, Stephen. 2014. “ENGL 4/878 FAQ.” 2014. http://jetson.unl.edu/syllabi/2014 
/fall/dh/html/01_overview.html
Redding, Emma. 2019. “The Expanding Possibilities of Dance Science.” In The Rout-
ledge Companion to Dance Studies, by Helen Thomas and Stacey Prickett, 56–6 7. Lon-
don: Routledge.
Rees, Chris, and Mark Gatenby. 2014. “Critical Realism and Ethnography.” In Study-
ing Organizations Using Critical Realism: A Practical Guide, edited by P. K. Edwards, Joe 
O’Mahoney, and Steve Vincent. First ed. Oxford: Oxford University Press.
Rettberg, Jill Walker. 2020. “Ways of Knowing with Data Visualizations.” In Data 
Visualization in Society, edited by Martin Engebretsen and Helen Kennedy, 35– 48. 
Amsterdam: Amsterdam University Press. https://doi.org/10.2307/j.ctvzgb8c7.8
Rey, Sergio J., and Luc Anselin. 2010. “PySAL: A Python Library of Spatial Analytical 
Methods.” In Handbook of Applied Spatial Analysis, 175– 93. New York: Springer.
Ribeiro, Claudia, Rafael Kuffner dos Anjos, and Carla Fernandes. 2017. “Capturing 
and Documenting Creative Processes in Contemporary Dance.” In Proceedings of the 
4th International Conference on Movement Computing, 7:1–7 :7. MOCO ’17. New York: 
ACM. https://doi.org/10.1145/3077981.3078041
Ribes, David, and Steven Jackson. 2013. “Data Bite Man: THe Work of Sustaining a 
Long- Term Study.” In “Raw Data” Is an Oxymoron, by Lisa Gitelman, 147–6 6. Infra-
structures Series. Cambridge, MA: MIT Press.
Ripley, B. D. 1979. “Tests of `Randomness’ for Spatial Point Patterns.” Journal of the 
Royal Statistical Society. Series B (Methodological) 41 (3): 368–7 4.
Roberts, David, and Lance Woodman. 1998. “A Corpus Linguistics Study of the The-
atre Review: First Steps.” Studies in Theatre Production 18 (1): 6–2 8. https://doi.org/10 
.1080/13575341.1998.10806987
Roberts, Les. 2015. “Navigating the ‘Archive City’: Digital Spatial Humanities and 
Archival Film Practice.” Convergence 21 (1): 100– 115. https://doi.org/10.1177/13548 
56514560310
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
References • 213
Roberts, Les. n.d. “Liverpool— City in Film.” Accessed December 14, 2018. https:// 
www.google.com/maps/d/viewer?mid=1f8v4bF2fBwk2Z5tjtmWq2QnQOlg
Roberts- Smith, Jennifer, Shawn DeSouza- Coelho, Teresa M. Dobson, Sandra Gabri-
ele, Omar Rodriguez- Arenas, Stan Ruecker, Stéfan Sinclair, et al. 2013. “Visualiz-
ing Theatrical Text: From Watching the Script to the Simulated Environment for 
Theatre (SET).” Digital Humanities Quarterly 7 (3).
Rockwell, Geoffrey, and Stéfan Sinclair. 2016. Hermeneutica: Computer- Assisted Interpreta-
tion in the Humanities. Cambridge: MIT Press. http://www.jstor.org/stable/j.ctt1c0 
gm6h
Rogerson, Peter A. 1999. “The Detection of Clusters Using a Spatial Version of the Chi- 
Square Goodness- of- Fit Statistic.” Geographical Analysis 31 (1): 130–4 7. https://doi 
.org/10.1111/gean.1999.31.1.130
Rosenberg, Daniel. 2013. “Data before the Fact.” In “Raw Data” Is an Oxymoron, by Lisa 
Gitelman, 15–4 0. Infrastructures Series. Cambridge: MIT Press.
Royce, Anya Peterson. 1977. The Anthropology of Dance. Bloomington: Indiana University 
Press.
Rozik, Eli. 1999. “The Corporeality of the Actor’s Body: The Boundaries of Theatre and 
the Limitations of Semiotic Methodology.” Theatre Research International 24 (2): 198– 
211. https://doi.org/10.1017/S0307883300020824
Ruecker, Stan, Milena Radzikowska, and Stefan Sinclair. 2016. Visual Interface Design for 
Digital Cultural Heritage: A Guide to Rich- Prospect Browsing. Farnham and Burlington: 
Ashgate.
Rusbridge, Chris, Peter Burnhill, Seamus Ross, Peter Buneman, David Giaretta, Liz 
Lyon, and Malcolm Atkinson. 2005. “The Digital Curation Centre: A Vision for 
Digital Curation.” In 2005 IEEE International Symposium on Mass Storage Systems and 
Technology, 31– 41. Baltimore, MD: IEEE.
Rushkoff, Douglas, and Leland Purvis. 2011. Program or Be Programmed: Ten Commands for 
a Digital Age. Berkeley, CA: Counterpoint.
Salsburg, David. 2002. The Lady Tasting Tea. New York: W. H. Freeman / Owl Book.
Salt, Barry. 1974. “Statistical Style Analysis of Motion Pictures.” Film Quarterly 28 (1): 
13– 22. https://doi.org/10.2307/1211438
Samuels, Lisa, and Jerome McGann. 1999. “Deformance and Interpretation.” New Lit-
erary History: A Journal of Theory and Interpretation 30 (1): 25.
Sant, Toni. 2014. “Interdisciplinary Approaches to Documenting Performance.” Inter-
national Journal of Performance Arts and Digital Media 10 (1): 3– 6. https://doi.org/10.10 
80/14794713.2014.912495
Sant, Toni. 2017. Documenting Performance: The Context and Processes of Digital Curation and 
Archiving. London and New York: Bloomsbury.
Santos Unamuno, Enrique. 2017. “GIS and Telescopic Reading: Between Spatial and 
Digital Humanities.” Neohelicon: Acta Comparationis Litterarum Universarum 44 (1): 65.
Sayers, Jentery. 2016. “Minimal Definitions · Minimal Computing.” 2016. http://go-dh 
.github.io/mincomp/thoughts/2016/10/02/minimal-definitions/
Schauf, Andrew Johnathan, and Miguel Escobar Varela. 2018. “Searching for Hidden 
Bridges in Co- Occurrence Networks from Javanese Wayang Kulit.” Journal of His-
torical Network Research 2 (1): 26– 52. https://doi.org/10.25517/jhnr.v2i1.42
Schechner, Richard. 2013. Performance Studies. Edited by Sara Brady. London and New 
York: Routledge.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
214 • References
Scheinfeldt, Tom. 2012. “Game Change: Digital Technology and Performative Human-
ities.” Found History. February 15, 2012. http://foundhistory.org/2012/02/game-ch 
ange-digital-technology-and-performative-humanities/
Schich, Maximilian, Christian Huemer, Piotr Adamczyk, Lev Manovich, and Yang-Y u 
Liu. 2017. “Network Dimensions in the Getty Provenance Index.” ArXiv Preprint 
ArXiv:1706.02804.
Schich, Maximilian, Sune Lehmann, and Juyong Park. 2008. “Dissecting the Canon: 
Visual Subject Co- Popularity Networks in Art Research.” 5th European Conference 
on Complex Systems, September 3, Jerusalem. https://archiv.ub.uni-heidelberg.de 
/artdok/711/
Schich, Maximilian, Chaoming Song, Yong- Yeol Ahn, Alexander Mirsky, Mauro Mar-
tino, Albert- László Barabási, and Dirk Helbing. 2014. “A Network Framework of 
Cultural History.” Science 345 (6196): 558– 62. https://doi.org/10.1126/science.124 
0064
Schmicking, Daniel, and Shaun Gallagher, eds. 2010. Handbook of Phenomenology and 
Cognitive Science. Dordrecht: Springer Netherlands. https://doi.org/10.1007/978-90 
-481-2646-0
Schöch, Christof. 2016. “Principal Component Analysis for Literary Genre Stylistics.” 
Billet. The Dragonfly’s Gaze (blog). 2016. https://dragonfly.hypotheses.org/472
Schöch, Christof. 2017. “Topic Modeling Genre: An Exploration of French Classical 
and Enlightenment Drama.” Digital Humanities Quarterly 11 (2).
Sears, Laurie Jo. 1996. Shadows of Empire: Colonial Discourse and Javanese Tales. Durham: 
Duke University Press.
Septi Rito Tombe. 2017. University of Melborne Digital Studio, and Social Media and 
Performance Impact Survey. “Circuit: Mapping Theatre Performances in Victoria.” 
https://circuit.unimelb.edu.au/
Siemens, Raymond G. 2002. “A New Computer- Assisted Literary Criticism?” Computers 
and the Humanities 36 (3): 259– 67.
Simpson, Travis T., Susan Wiesner, and Bradford C. Bennett. 2014. “Dance Recogni-
tion System Using Lower Body Movement.” Journal of Applied Biomechanics 30 (1): 
147–5 3. https://doi.org/10.1123/jab.2012-0248
Snaprud, Mikael, and Andrea Velazquez. 2020. “Accessibility of Data Visualizations:” 
In Data Visualization in Society, edited by Martin Engebretsen and Helen Kennedy, 
111–2 6. Amsterdam: Amsterdam University Press. https://doi.org/10.2307/j.ctvzg 
b8c7.13
Soedarsono. 1983. “Wayang Wong in the Yogyakarta Kraton: History, Ritual Aspects, 
Literary Aspects, and Characterization.” Ann Arbor: University of Michigan.
Solichin, Suyanto H., Sumari, Undung Wiyono, and Sri Purwanto. 2017. Ensiklopedi 
Wayang Indonesia. Jakarta: Mitra Sarana Edukasi.
Spear, Mary Eleanor. 1952. Charting Statistics. New York: McGraw- Hill.
Spence, Paul. 2013. “Teatro clásico y humanidades digitales: el cruce entre método, 
proceso y nuevas tecnologías.” Teatro de Palabras 7 (9): 31.
Splawa-N eyman, Jerzy. 1990. “On the Application of Probability Theory to Agricultural 
Experiments. Essay on Principles. Section 9.” Translated by Dorota M. Dabrowska 
and T. P. Speed. Statistical Science, 465– 72.
Sreenivasan, Sameet. 2013. “Quantitative Analysis of the Evolution of Novelty in Cin-
ema through Crowdsourced Keywords.” Scientific Reports 3: 2758.
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
References • 215
Streiter, Oliver, Yoann Goudin, Chun (Jimmy) Huang, and Ann Meifang Lin. 2012. 
“Matching Digital Tombstone Documentation to Unearthed Census Data: Sur-
veying Taiwan’s Family Names, Ethnicities and Homelands.” International Journal 
of Humanities and Arts Computing 6 (1– 2): 57– 70. https://doi.org/10.3366/ijhac.2012 
.0038
Strine, Mary S., Beverly Whitaker Long, and Mary Frances Hopkins. 1990. “Research in 
Interpretation and Performance Studies: Trends, Issues, Priorities.” Speech Commu-
nication: Essays to Commemorate the 75th Anniversary of the Speech Communication Associa-
tion, 181– 204.
Su, Wen- Chi. 2019. “Dramaturgy and Technology.” In Dramaturgy and the Human Condi-
tion. Singapore.
Subyen, Pattarawut, Diego Maranan, Thecla Schiphorst, Philippe Pasquier, and Lyn 
Bartram. 2011. “EMVIZ: The Poetics of Movement Quality Visualization.” In Pro-
ceedings of the International Symposium on Computational Aesthetics in Graphics, Visualiza-
tion, and Imaging, 121–2 8. CAe ’11. New York: ACM. https://doi.org/10.1145/20304 
41.2030467
Sudibyoprono, R. Rio, Suwandono, Dhanisworo, and Mujiyono. 1991. Ensiklopedi Way-
ang Purwa. Jakarta: Balai Pustaka.
Sudjarwo, Heru S., Sumari, and Undung Wiyono. 2010. Rupa Dan Karakter Wayang 
Purwa. Jakarta: Kakilangit Kencana.
Suiter, Ted. 2013. “Why ‘Hacking’?” In Hacking the Academy: New Approaches to Scholarship 
and Teaching from Digital Humanities, by Daniel J. Cohen and Tom Scheinfeldt. Ann 
Arbor: University of Michigan Press.
Suri, M., and S. N. Singh. 2018. “The Role of Big Data in the Media and Entertainment 
Industry.” In 2018 4th International Conference on Computational Intelligence Communica-
tion Technology (CICT), 1– 5. https://doi.org/10.1109/CIACT.2018.8480281
Suskin, Steven. 1990. Opening Night on Broadway: A Critical Quotebook of the Golden Era of 
the Musical Theatre, Oklahoma! (1943) to Fiddler on the Roof (1964). New York: Schirmer 
Trade Books.
Taleb, Nassim. 2007. The Black Swan: The Impact of the Highly Improbable. London: Allen 
Lane.
Tan, Alvin. 2004. “A Necessary Practice.” In Ask Not: The Necessary Stage in Singapore The-
atre, edited by Chong Kee Tan and Tisa Ng, 266– 90. Singapore: Times Editions.
Tan, Chong Kee, and Tisa Ng, eds. 2004. Ask Not: The Necessary Stage in Singapore Theatre. 
Singapore: Times Editions.
Tan, Kenneth Paul. 2013. “Forum Theater in Singapore: Resistance, Containment, and 
Commodification in an Advanced Industrial Society.” Positions: East Asia Cultures Cri-
tique 21 (2).
TextGrid Consortium. 2006. “TextGrid: A Virtual Research Environment for the 
Humanities.” 2014 2006. textgrid.de.
Thorp, Jer. 2017. “You Say Data, I Say System.” Hacker Noon, July 13. https://hackernoon 
.com/you-say-data-i-say-system-54e84aa7a421
Thudt, Alice, Jagoda Walny, Theresia Gschwandtner, Jason Dykes, and John Stasko. 
2018. “Exploration and Explanation in Data- Driven Storytelling.” In Data- Driven 
Storytelling, edited by Nathalie Henry Riche, Christophe Hurter, Nicholas Dia-
kopoulos, and Sheelagh Carpendale, 59–8 3. New York: A. K. Peters/CRC Press. 
https://doi.org/10.1201/9781315281575-3
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
216 • References
Törnberg, Anton. 2019. “Abstractions on Steroids: A Critical Realist Approach to Com-
puter Simulations.” Journal for the Theory of Social Behaviour 49 (1): 127–4 3. https:// 
doi.org/10.1111/jtsb.12194
Travis, Charles. 2015. “Acts of Perception: Samuel Becket, Time, Space and the Digital 
Literary Atlas of Ireland, 1922– 1949.” International Journal of Humanities and Arts Com-
puting 9 (2): 219– 41. https://doi.org/10.3366/ijhac.2015.0150
Travis, Charles, and Richard Breen. 2017. “Digital Literary Atlas of Ireland.” 2017. 
https://uploads.knightlab.com/storymapjs/dfe287c1e4ed312fe68fb5e395ee7d57 
/digital-literary-atlas-of-ireland-time-map/index.html
Trilcke, Peer, and Frank Fischer. n.d. “Dracor.Org.” Accessed September 5, 2018. 
https://dracor.org/
Trilcke, Peer, Frank Fischer, Mathias Göbel, and Dario Kampkaspar. 2015. “Comedy 
vs. Tragedy: Network Values by Genre.” DLINA: Digital Linterary Analysis (blog). 
https://dlina.github.io/Network-Values-by-Genre/
Trilcke, Peer, Frank Fischer, Matthias Göbel, and Dario Kampkaspar. 2016. “Theatre 
Plays as ‘Small Worlds’? Network Data on the History and Typology of German 
Drama, 1730– 1930.” In Digital Humanities 2016 Conference Abstracts. Krakow, Poland.
Trilcke, Peer, Frank Fischer, and Dario Kampkaspar. 2015. “Digital Network Analysis 
of Dramatic Texts.” In Digital Humanities 2015 Conference Abstracts. Sydney, Australia.
Tsachor, Rachelle P., and Tal Shafir. 2017. “A Somatic Movement Approach to Foster-
ing Emotional Resiliency through Laban Movement Analysis.” Frontiers in Human 
Neuroscience 11 (September). https://doi.org/10.3389/fnhum.2017.00410
Tufte, Edward R. 1983. The Visual Display of Quantitative Information. Vol. 2. Cheshire, CT: 
Graphics Press.
Tukey, John W. 1977. Exploratory Data Analysis. Vol. 2. Reading, MA: Addison- Wesley.
Ubersfeld, Anne, Frank Collins, Paul Perron, and Patrick Debbeche. 1999. Reading The-
atre. Toronto Studies in Semiotics. University of Toronto Press.
UiO Universitetet i Oslo. n.d. “IbsenStage.” Accessed September 5, 2018. https://ibse 
nstage.hf.uio.no/
Ulmer, Gregory L. 1994. Heuretics: The Logic of Invention. Baltimore, MD: Johns Hopkins 
University Press.
Underwood, Ted. 2017. “A Genealogy of Distant Reading.” DHQ: Digital Humanities 
Quarterly 11 (2).
Underwood, Ted. 2019a. Distant Horizons: Digital Evidence and Literary Change. Chicago: 
University of Chicago Press.
Underwood, Ted. 2019b. “Dear Humanists: Fear Not the Digital Revolution.” Chronicle 
of Higher Education, March 27. https://www.chronicle.com/article/Dear-Humanists 
-Fear-Not-the/245987
UPenn Libraries. n.d. “19th Century Playbills, 1803– 1939.” Accessed August 3, 2018. 
http://dla.library.upenn.edu/dla/pacscl/detail.html?id=PACSCL_FLP_FLPTHCPLA 
YBILL
Uzzi, Brian, and Jarrett Spiro. 2005. “Collaboration and Creativity: The Small World 
Problem.” American Journal of Sociology 111 (2): 447–5 04. https://doi.org/10.1086/43 
2782
Vandenberghe, Frédéric. 2013. What’s Critical About Critical Realism? Essays in Reconstructive 
Social Theory. Routledge. https://doi.org/10.4324/9780203798508
Vareschi, Mark, and Mattie Burkert. 2017. “Archives, Numbers, Meaning: The 
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
References • 217
Eighteenth- Century Playbill at Scale.” Theatre Journal 68 (4): 597– 613. https://doi 
.org/10.1353/tj.2016.0108
Verhoeven, Deb, and Colin Arrowsmith. 2013. “Mapping the Ill-D isciplined? Spatial 
Analyses and Historical Change in the Postwar Film Industry.” In Locating the Mov-
ing Image: New Approaches to Film and Place, by Julia Hallam and Les Roberts, 106– 29. 
Bloomington: Indiana University Press. http://muse.jhu.edu/book/27089
Wang, Tricia. 2016. “Why Big Data Needs Thick Data.” Ethnography Matters (blog). Jan-
uary 20, 2016. https://medium.com/ethnography-matters/why-big-data-needs-thi 
ck-data-b4b3e75e3d7
Ward, Michael Don, and Kristian Skrede Gleditsch. 2019. Spatial Regression Models. Vol. 
155. Quantitative Applications in the Social Sciences. Thousand Oaks: SAGE.
Warren, Christopher N., Daniel Shore, Jessica Otis, Lawrence Wang, Mike Finegold, 
and Cosma Shalizi. 2016. “Six Degrees of Francis Bacon: A Statistical Method for 
Reconstructing Large Historical Social Networks.” Digital Humanities Quarterly 10 
(3).
Thow, Xin Wei. 2018. “‘All the Good Musicians Are Dead’: A Sense of Decline in Java-
nese Gamelan.” MA thesis, National University of Singapore.
Wiesner, Susan. 2012. “ARTeFACT Project Summary.” Charlottesville: University of 
Virginia. http://avillage.web.virginia.edu/ARTeFACT/default.asp
Wiesner, Susan, Bennett Bradford, and Rommie Stalnaker. 2012. “The ARTeFACT 
Movement Thesaurus: Toward an Open-S ource Tool to Mine Movement-D erived 
Data.” In Digital Humanities 2012. Hamburg, 413–1 4. See https://dh-abstracts.libra 
ry.cmu.edu/works/1503
Wiesner, Susan, Shannon Cuykendall, Ethan Soutar- Rau, Rommie L. Stalnaker, The-
cla Schiphorst, and Karen Bradley. 2016. “Schrifttanz: Written Dance/Movement 
Poems.” In Digital Humanities 2016: Conference Abstracts, 402–4 . Kraków: Jagiellonian 
University and Pedagogical University, Kraków.
Wiesner, Susan, and Rommie L. Stalnaker. 2015. “Representing Conflict through 
Dance: Using Quantitative Methods to Study Choreographic Time, Stage Space, 
and the Body in Motion.” In With(out) Trace: Inter-D isciplinary Investigations into Time, 
Space and the Body, by S. Dwyer, R. Franks, and R. Green. Oxford: Inter- Disciplinary 
Press.
Wilkinson, Leland. 1999. The Grammar of Graphics. Berlin: Springer.
Wilkinson, Mark D., Michel Dumontier, IJsbrand Jan Aalbersberg, Gabrielle Appleton, 
Myles Axton, Arie Baak, Niklas Blomberg, et al. 2016. “The FAIR Guiding Prin-
ciples for Scientific Data Management and Stewardship.” Scientific Data 3 (March): 
160018. https://doi.org/10.1038/sdata.2016.18
Wong, Audrey. 1997. “Collaboration in the Work of The Necessary Stage.” In 9 Lives, 
10 Year of Singapore Theatre 1987– 1997, edited by Sanjay Krishan, 266–9 0. Singapore: 
First Printers.
Xanthos, Aris, Isaac Pante, Yannick Rochat, and Martin Grandjean. 2016a. “Visualis-
ing the Dynamics of Character Networks.” In Digital Humanities 2016: Conference 
Abstracts, 417–1 9.
Xanthos, Aris, Isaac Pante, Martin Grandjean, and Yannick Rochat. 2016b. “About 
IntNetViz.” 2016. https://maladesimaginaires.github.io/intnetviz/about.html
Yong, Li Lan, Eng Hui Alvin Lim, Ken Takiguchi, Chee Keng Lee, Hyon- u Lee, Ha- 
young Hwang, Michiko Suematsu, and Kaori Kobayashi. 2015. “Asian Shake-
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
218 • References
speare Intercultural Archive (A| S| I| A). National University of Singapore.” English, 
Chinese, Japanese. Accessed 1. http://a-s-i-a-web.org
York, Christopher. 2017. “Exploratory Data Analysis for the Digital Humanities: The 
Comédie- Française Registers Project Analytics Tool.” English Studies 98 (5): 459– 
82. https://doi.org/10.1080/0013838X.2017.1332024
Yuri Tsivian. 2005. “Cinemetrics— About.” http://www.cinemetrics.lv/index.php
Zahavi, Dan. 2013. The Oxford Handbook of Contemporary Phenomenology. Oxford: Oxford 
University Press.
Zillner, Sonja, Margrit Gelautz, and Markus Kallinger. 2002. “‘The Right Move’—A  
Concept for a Video- Based Choreography Tool.” In. Graz, Austria. http://eprints 
.cs.univie.ac.at/1148/
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Index
accuracy, 5– 7, 32, 36, 47– 48, 53–5 4, 60, Caplan, Debra, 2, 94, 100– 102, 146
99, 117, 121– 23, 125, 151, 155–5 6 CATMA (computer-a ssisted text markup 
actor, 100, 103– 5, 116, 134. See also and analysis), 83–8 4
performer causality / causal relationships, 37– 39, 
algorithmic criticism, 82. see also Ramsay, 42, 48–5 4
Stephen chance: probability and statistics, 47– 50, 
ambiguity, 9, 11– 13, 27–2 8, 55, 62, 90, 153
92, 146, 156, 192 character types: in Javanese dance, 
annotation, 65, 77–7 8, 90, 125– 26, 135– 38
163– 64 characters: dramatic, 77, 94– 99,  
artistic collaborations: as object of study, 108– 14
17, 19, 81, 84, 94, 96– 97, 100– 114 chartjunk, 59
AusStage, 100, 145, 163, 168–6 9 choropleth, 64, 142–4 3, 145–4 6, 149, 
authorship attribution, 75–7 6, 81, 84– 85, 152, 153, 156–5 7
146 CIDOC- CRM, 167. See also data model
classification, 41, 92, 118, 125, 148
Bardiot, Clarisse, 14, 101 clusters: statistics, 42, 44– 48, 55, 60, 76, 
Bay- Cheng, Sarah, 14, 66, 70 79, 130, 143, 152
Bayesian: statistics, 52 Comédie Française Registers Project, 
Bench, Harmony, and Kate Elswitt, 148– 80, 163
49, 158, 188 complete spatial randomness (CSR), 143
bias, 10, 19, 31, 42–4 8 consensus, 7, 11– 12, 47, 101, 125, 127
institutional bias, 54– 56 constructivism / constructivist, 8– 10, 23, 
proxy bias, 44, 49 35– 39, 63–6 4
biomechanics, 123– 125, 135–4 0 copyright, 20, 85, 134, 164, 171– 72
bloom- and- fade publishing, 173– 77 corpus: linguistics, 48, 75– 77, 83– 88, 
Borges, Jorge Luis, 167, 173–7 4 93, 97– 99, 109, 125
Borgman, Christine, 4–5 , 171 critical realism, 10– 11, 18, 24, 35–4 0, 43– 
boxplot, 43,58 44, 56, 63, 139
culturally specific / cultural specificity, 
calibration (of methods), 47– 48, 53, 55, 18, 62– 63, 116, 123, 132, 139, 147, 
76, 125, 191 181, 188, 208
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research. 219
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
220 • Index
dance, 87, 108, 116– 28,135– 140, 148– 49, ephemerality, 173, 177, 179
163 ergodic: platforms, 68, 138, 149
dance notation systems, 119– 20. See also ethics / ethical implications, 64–6 7
Labanotation ethnography, 12, 24, 38, 54
data exploratory data analysis (EDA), 43– 45, 
data biography; biography of a data- 58
set, 13, 169
data collection, 25, 43, 47, 56, 65, 116, feminist data visualization, 63, 65
126, 148 Feynman, Richard, 30
data management plan (DMP), 171– 72 Fiadeiro, João, 127
data model, 78– 79, 100, 126, 163– 70, film studies / cinema, 67, 134, 141–4 9
178 (see also CIDOC- CRM;  Fischer, Frank. See Digital Literary Net-
Dublin Core) work Analysis (DLINA)
data repository, 80, 170, 172 (see also The Flying Inkpot, 87–9 3
digital archives) funding (financial resources; money), 
deep mapping, 146 15–1 6, 20, 55, 114, 122, 145, 172– 75, 
defamiliarization, 8– 11, 48, 56, 94, 102, 178, 185– 87
138
difference image, 128– 29, 132, 134 genre: written drama and literature, 76– 
digital archives, 2, 84, 87, 93, 128, 163, 77, 82– 86, 98
165–6 7. See also data repository geographical information systems (GIS), 
digital curation, 65, 168– 69, 171, 173, 178 141–4 8, 158
digital humanities (DH), 2– 3, 8, 11, Gottschall, Jonathan, 31– 35
13–1 4, 18–1 9, 34, 38, 43–4 6, 55, 58, graphical conventions, 10, 60, 65
65– 66, 75– 79, 81, 83– 84, 86– 87, 96, 
111, 144, 146, 158, 168– 71, 174– 75, Hachimura, Kozaburo, 119– 20, 122, 124
177, 181–8 7 hacking / hackers, 184– 85
Digital Literary Network Analysis Hamlet, 77– 78, 97. See also Shakespeare
(DLINA), 98– 99 Hay, Deborah, 97
dimensionality reduction, 45, 75–7 6, hermeneutics, 15, 29, 39, 67, 70, 127, 158
79. See also principal component 
analysis (PCA) IbsenStage, 100, 144– 45, 163. See also A 
director: theater, 12, 86, 103, 105, 164– 66 Doll’s House
dispersion: statistics, 42, 58– 59 Indonesia. See wayang, Sendratari
distant reading, 81– 82 interactivity, 10, 60, 65, 67– 72, 149
A Doll’s House, 2, 100– 102, 144–4 5. See also intermedial essays, 70–7 2, 83, 149, 172, 
IbsenStage 178, 182
Drucker, Johanna, 61–6 3, 67, 71, 132, 150 interoperability, 165, 171– 72
Dublin Core, 165– 68, 168. See also data interpretive approach, 2, 5, 8, 12, 24, 26, 
model 29, 34– 36, 41, 62– 63, 66– 71, 77, 80, 
82–8 3, 85, 90, 92, 125, 127– 29, 136, 
effect size, 42, 50, 54, 74, 111– 14 138, 145– 47, 158
Elswitt, Kate. See Bench, Harmony, and iterative research, 14, 43, 61, 84, 147,  
Kate Elswitt 176
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
Index • 221
Jannidis, Fotis, 35, 168– 69 Parikesit, Gea Oswah Fatah, 122, 128– 30, 
Jockers, Matthew, 77, 82, 181, 206 187
PCA. See principal component analysis 
kernel density estimation (KDE), 112, (PCA)
132, 154, 194 performance studies, 5, 13, 26, 28, 66, 
173, 186. See also Schechner, Richard
Labanotation, 117– 20, 122. See also dance performative (data) visualization, 39, 54, 
literary scholarship / literary analysis, 65–6 7, 126– 28, 134, 150
30–3 4, 43, 75– 77, 80– 85, 95 performer, 2, 69, 94, 100, 102, 104– 5, 
108, 121, 125– 27, 140, 144, 148, 155– 
machine learning (ML), 45, 47, 49, 52, 56, 164–6 5. See also actor
79, 81, 125 permutations (bootstrapping method), 
Mann- Kendall, 88, 101. See also trends 153– 55
Manovich, Lev, 3, 63– 64, 67, 134,  Piper, Andrew, 41, 77, 79, 169
174– 75 playbills, 75, 80
markers: motion capture, 121– 22 political aspects
metadata, 164– 71. See also data of dance, 138– 39, 148
methods and methodology, 9– 10 of research, 1, 16, 27, 36, 57, 60, 66
minimal computing, 177, 185–8 7 of theater, 102– 3, 145
minimum convex polyhedron; minimum positivism, 27, 36– 38
convex polygon, 124, 126 postructuralist; poststructuralism, 27
mocap (motion capture), 17, 117, 120– 22, principal component analysis (PCA), 
125, 136, 140 75– 76, 79, 84. See also dimensional-
models / modelling, 6, 14, 41, 43– 48, 52, ity reduction
56, 75, 79, 94, 96, 100–1 01, 104, 114, program booklets, 3, 6, 19, 25, 86
116, 126, 164– 70, 184
Moran’s I, 143–4 4, 152– 55, 157 Ramsay, Stephen, 9, 34, 41, 46, 66, 82, 
Motion Bank, 126 85, 183
Mrázek, Jan, 139, 151 Ramsey theory: mathematics, 46, 52
randomized controlled trials; controlled 
natural experiments, 126 trials, A|B testing, 48, 49, 54
negative findings, 48, 56, 137 randomness; random conditions, 47–4 9, 
network theoretical measurements, 95, 54, 113, 143, 153–5 4
98–9 9, 110, 112–1 3, 193 replicability / reproducibility, 7– 9, 15, 
23– 29, 28, 30, 34– 35, 40, 48, 53, 
open access, 171– 72 55– 56, 65, 82, 84, 102, 126– 28, 137, 
Open Science Framework (OSF), 172 153, 156
open source, 181– 82 rich- prospect, 68
Rockwell, Geoffrey, and Stéfan Sinclair, 
p- value, 47, 50–5 2, 88, 153– 54 8, 34, 70, 83– 84, 93
pairwise plot, 112–1 3
parameters / parametrization, 8, 34, 45, Schauf, Andrew, 19, 109, 113, 187
48, 55 Schechner, Richard, 5
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution
222 • Index
Schöch, Christof, 19, 76– 77 thick data, 13
scholarship: theatre, 16, 25, 29, 30– 32, thick description, 13, 16, 24, 147
70, 72, 80, 87, 97, 114, 138, 140, time series analysis, 75, 79, 83– 84, 107– 
146, 149, 155, 171, 174, 185–8 8 8, 156
scientific: approaches to research, 10, 15, topic modelling, 34, 75, 77, 79, 83– 84, 
18– 19, 23–2 4, 29, 30– 35, 46, 49–5 5, 86
58– 62, 81, 84, 101, 118, 123, 125, training; education, 29, 52, 170, 183
127– 28, 144, 187 trend detection, 42, 44– 46, 48, 60, 
semiotics, semiotic, 27, 57, 71 76, 84, 88– 92, 155. See also 
Sendratari: Javanese dance, 135– 38 Mann- Kendall
sensors: motion capture, 5, 118, 121, 128, Trilke, Peer. See Digital Literary Network 
138 Analysis (DLINA)
Shakespeare, 12–1 3, 81, 86, 102, 163. See 
also Hamlet uncertainty / doubt, 7, 9, 30, 47, 168
Sinclair, Stéfan, 68. See also Rockwell, Underwood, Ted, 30, 41, 81– 82
Geoffrey, and Stéfan Sinclair
Singapore, 16– 19, 23, 33, 87–9 3, 102– 8. Verhoeven, Deb, 142, 145– 46
See also The Flying Inkpot; The Nec- video, 27, 70– 71, 109, 118, 121– 23, 126, 
essary Stage (TNS) 128–2 9, 134, 137– 38, 163– 64, 166– 
social network analysis (SNA), 95– 68, 175, 178
96. See also network theoretical violinplot, 43, 58, 112, 130–3 2
measurements Voyant Tools, 83–8 4, 88, 170, 181. See 
spectator, 101, 134, 139, 150 also Rockwell, Geoffrey, and Stéfan 
stylometry, 75, 136, 138– 40, 181 Sinclair
TEI, 77– 79, 99, 120, 166, 170 wayang, 33, 71, 94– 95, 108–1 4, 122, 128– 
text reuse, 79– 80, 86 34, 150–5 7, 165–6 7
textual differences: measures of textual Wiesner, Susan, 2, 19, 125– 26
difference, 76 word frequencies, 9, 76, 88– 93
The Necessary Stage (TNS), 94– 95, 
102– 8 XML, 77–7 8, 119– 20, 122, 125, 166
thickness
thick context, 67–7 1, 83, 90, 102, 110
Escobar Varela, Miguel. Theater As Data: Computational Journeys Into Theater Research.
E-book, Ann Arbor, MI: University of Michigan Press, 2021, https://doi.org/10.3998/mpub.11667458.
Downloaded on behalf of Unknown Institution