- Before 1500 BCE
- 1500 BCE to 500 BCE
- 500 BCE to 500 CE
- Sixth to Tenth Century
- Eleventh to Fourteenth Century
- Fifteenth Century
- Sixteenth Century
- Seventeenth Century
- Eighteenth Century
- Nineteenth Century
- Twentieth Century
- Twenty-first Century
- Geographic Area
- Central America
- Central and North Asia
- East Asia
- North America
- Northern Europe
- South America
- South Asia/South East Asia
- Southern Europe and Mediterranean
- West Asia
- Subject, Genre, Media, Artistic Practice
- African American/African Diaspora
- Ancient Egyptian/Near Eastern Art
- Ancient Greek/Roman Art
- Architectural History/Urbanism/Historic Preservation
- Art Education/Pedagogy/Art Therapy
- Art of the Ancient Americas
- Artistic Practice/Creativity
- Asian American/Asian Diaspora
- Ceramics/Metals/Fiber Arts/Glass
- Colonial and Modern Latin America
- Conceptual Art
- Decorative Arts
- Design History
- Digital Media/New Media/Web-Based Media
- Digital Scholarship/History
- Drawings/Prints/Work on Paper/Artistc Practice
- Fiber Arts and Textiles
- Folk Art/Vernacular Art
- Graphic/Industrial/Object Design
- Indigenous Peoples
- Installation/Environmental Art
- Islamic Art
- Material Culture
- Museum Practice/Museum Studies/Curatorial Studies/Arts Administration
- Native American/First Nations
- Patronage, Art Collecting
- Performance Art/Performance Studies/Public Practice
- Queer/Gay Art
- Sound Art
- Visual Studies
In August 2020, “What Do We Know about the Future of Art History? Part 1” appeared as a special essay in caa.reviews. It explored the history of CAA’s roster of PhD dissertations, beginning with its establishment in 1963 and then delving into the changing circumstances that continue to animate its presentation. The article made the case that this list of art history dissertations constitutes more than just a procedural accounting of projects completed. When analyzed as a data set, the dissertations illuminated unexamined patterns that have occurred within the field in the United States and Canada over the past decades. A series of visualizations distilled some of these findings in a concrete format, but with the author’s full awareness of the limits of this data set—and really any data set—in capturing and conveying the complexity of the stories that we tell about our own relationships to our changing field of study.
That article generated a considerable amount of interest, suggesting that art historians across the discipline are eager to learn more about the shifting shape of the field and are not averse to pursuing the language of descriptive statistics as one possible approach. It also inspired this follow-up essay and another forthcoming one, both commissioned as installments to appear in a series in caa.reviews. All three use the CAA dissertation roster as a salient point of departure and are committed to data-inflected practices of analysis. All are similarly poised at a temporal pivot, with the claim that we must look back to move forward and that understanding our past will help us to sketch the possibilities for the future of our field at a time of great uncertainty.
The present essay, part 2, expands the data set back in time from its previous starting point of 2002 to the year 1980, relying upon the considerable efforts of data gathering undertaken by collaborators at the Visual Resources Centre at Penn State University (PSU).1 Their meticulous work has provided a longer and richer view of these dissertations that offers clarity about the expansion of art history as a field of study. As stewards of this data set, the PSU team is deeply committed to making this resource publicly available, updated, and as accurate as possible, with the understanding that this primary source presents many opportunities for continuing research and further inquiry.2 Additionally, Pepe Karmel has extended the data drawn from the CAA rosters to explore job placements over the past decade. His findings will appear as the third and final essay in this series. Whereas part 1 of this series was written under truly isolated conditions during the tight lockdown that was enacted in the early months of the COVID-19 pandemic in New York, this coauthored sequel has emerged from a lively collaboration, involving art history faculty, librarians, visual resource curators, digital humanities specialists, and graduate and undergraduate students, across several institutions.
The PSU Data Set
Before launching into a discussion of the findings, we must delve into the data itself, under the premise that art historians ought to grapple with questions about data discovery, structure, and tractability as fundamental concerns in the face of new analytical methods that hinge on computation. In 2019–20, a team of graduate assistants and staff of the Visual Resources Centre at Penn State transcribed the data from the CAA roster, including all completed dissertations, from 1980 to 2018.3 The nature of the roster, which was published first in print and then digitally, required that they manually gather data from The Art Bulletin from 1980 to 2002, using PDF files accessed from JSTOR. Entries dating from 2002 to 2018 were extracted from the caa.reviews platform, also through manual transcription, working entry by entry. By the end of summer 2020, the completed data set, henceforth known as the PSU data set, yielded 6,345 unique entries.4
As acknowledged in part 1, the CAA dissertation roster does not constitute a complete record of all of the dissertations that have been written in the field, due to the fact that it relies on self-reported data from institutions in the United States and Canada. In some years, participating institutions did not submit a roster and, in certain cases, they omitted some of their own eligible dissertations. Furthermore, a manual data collection method is always subject to human error despite the measures taken to ensure consistency and accuracy, which we will elaborate in the paragraphs and endnotes below. No method can be truly comprehensive, but the authors of the PSU data set affirm that they collected the data in good faith and followed best practices.
The first article featured a data set that was harvested computationally from caa.reviews and only included entries from 2002 to 2018.5 By contrast, the PSU data set is much larger, hand-curated, and enhanced. Graduate assistants registered each dissertation as a single entry, according to the standard subfields of the published rosters, with up to three of its associated subject areas.6 In the PSU data set, the information generally appears without modification, with the major exception that graduate assistants added the associated dissertation abstract to each entry. This endeavor hinged on extensive further research, using the ProQuest database, the official repository for theses and dissertations for the Library of Congress, or drawing from institutional electronic theses and dissertations (ETD) databases.7 The added abstracts provide greater substance about the content of the dissertations and thus present a vast catalog of art historical research. With their standard format and associated metadata, these abstracts have the capacity to reveal the emergence of certain research trends, to highlight institutional preoccupations, or to focus on specific scholarly circles. This is a considerable added value of the PSU data set. As an example, the PSU team is currently engaged in a study on the rise of “visual culture” as an area of inquiry within the history of art that can be queried through the abstracts over the past forty years. A team at the Getty Research Institute is using the abstracts to explore approaches to iconography over the same span of time.8 The results of their research, attained by applying techniques of text analysis to the abstracts, will appear in future publications.
This research process revealed certain discrepancies between the entries that appeared in the CAA dissertation roster, ProQuest, and the ETD. As a rule, graduate assistants deferred to the more up-to-date ProQuest or ETD record to override or correct any conflicting information published in the corresponding listing in The Art Bulletin or caa.reviews.9 In particular, many institutions or authors announced completion to CAA in a given year, but the dissertation ended up being filed in another year. By extension, some of the aggregated findings that appear in the following visualizations differ from those in the August 2020 essay, for which the data was not amended aside from basic cleaning and restructuring.10 As a result, the findings of the current essay and the following visualizations significantly enhance and improve upon the results in part 1.
The Growth of the Field
With its wide coverage of 6,345 dissertations, the PSU data set vividly reveals the pace and scope of art history’s rise as a field over the past forty years.11 From 1980 through the mid-1990s, around 60 to 100 dissertations were completed each year. A period of marked and steep growth begins in the late 1990s, bringing the tally to around 200 to 250 dissertations completed annually. Since then, this range has been generally sustained, with 276 as the highest point, attained in the year 2013. Looking forward, this trend line is sure to experience more dips and ebbs, especially considering the fact that some graduate programs have elected to suspend admissions for the academic year 2021–22 as a response to the COVID-19 pandemic. These decisions will likely manifest as a visible drop in the number of dissertations completed over the following years. For that reason, we assert the year 2013 as a peak for the number of dissertations completed, and one that is unlikely to be attained again in the near future. Figure 1 (below) shows the overall reporting of dissertations by year.
Another pathway through the CAA dissertation roster concerns diversity in the field of art history. However, the PSU data set cannot answer this question on its own terms as the roster does not provide any personal information about the dissertation authors, apart from first and last name. For that reason, we inferred the gender of the dissertation authors using methods of prediction that data scientists commonly rely upon for large data sets in which gender is not self-reported. This study draws on the gender package written by Lincoln Mullen, and employs the data from the US Social Security Administration baby names records to predict gender based on first name.12 Needless to say, these methods have some serious limitations and, for that reason, are widely acknowledged as a “blunt tool to study a complex subject.”13 Fundamentally, they hinge on binary conceptions of gender, while assuming sex assigned at birth to be a stable category. More specifically, the package draws on a data set of 91,230 US names, with attention to the frequency of their use over time.14 As such, we can attain a high level of confidence in the results for names that have been used commonly in the United States over the past century. Yet, clearly, many art historians do not fall into that category, particularly those of international origin and those with uncommon first names. If we were to remedy these shortcomings, we would have to enhance the current data set with much more nuanced information reported directly from the authors themselves, which would entail a major effort toward new data collection.
Working within these acknowledged constraints, Figure 2 indicates, perhaps to no surprise, that women have completed a large percentage of art history PhDs in the United States and Canada, even in the early years of the 1980s.15 While the number of male art historians completing dissertations has nearly doubled over the past twenty years, the cadre of newly minted female art history PhDs has tripled in size since 1980. Readers will also note the rise of “undetermined” results, which occurred when the instrument that was used failed to deliver a prediction because the first name was not represented in the database. It is worthwhile to peruse the names that triggered these statistical limitations; for example, Hyewon, Ching, Masumi, Sabine, Zeynep, Fan, and Alla, all names that are uncommon in the Anglo-American context, appeared more than once in the “undetermined” category.16 Rather than manually ascribing gender for the entries where the gender seemed apparent, they were all left as “undetermined,” with the understanding that this category holds interpretational value on its own terms. In 1987, for instance, the “undetermined” names constituted 2 percent of the overall dissertations. By contrast, they composed 16 percent of dissertations completed in 2014, representing the highest proportion during this time period. So, the rise in “undetermined” names points, at least in part, to the increased diversity of the field, with larger numbers of scholars whose names do not have an Anglo-American origin, some of whom may have been international students studying in the United States and Canada.
Fig. 2 Completed Dissertations by Year and Gender, 1980–2018, Based on the CAA Dissertation Roster and the PSU Data Set (created by Nancy Um; click for larger version)
Cameron Blevins and Lincoln Mullen’s analysis of history dissertations serves as a useful point of comparison for the art history data, although their study extends much further back, to the year 1950, and is based on ProQuest records.17 Starting from the year 1980, many more history dissertations were completed each year than art history dissertations. Yet, both areas witnessed steep growth in the 1990s. According to Blevins and Mullen’s study, history dissertations peaked in 1996, compared to art history’s later 2013 peak. While the majority of dissertations in history were completed by men in 1980, the ratio of dissertations written by male and female historians is roughly equal today. As Blevins and Mullen described, writing in 2015, “Female historians have achieved something approaching parity with male historians in terms of how many women and men complete dissertations each year.”18 This comparison highlights significant differences between the composition of the fields of art history and history. A percent stacked chart (Figure 3) demonstrates this difference even more clearly, by presenting the proportion of art history dissertations completed by year broken down by gender. It shows that, in art history, women have completed at least 50 percent of dissertations each year since 1980, and in most years, women completed between 60 and 70 percent of them. While the proportion fluctuates incrementally within this scope, there are no definitive rises or falls in either direction. Again, a comparison with the same breakdown for history illustrates that, by contrast, female historians’ share of completed dissertations rose steadily over this period to reach nearly 50 percent in the past decade.19 In art history, the scales are tipped markedly toward women and have been for the past forty years.
Fig. 3 Proportion of Completed Dissertations by Year and Gender, 1980–2018, Based on the CAA Dissertation Roster and the PSU Data Set (created by Nancy Um; click for larger version)
One wonders how this gender breakdown might be reflected among the ranks of art history faculty and other professions across North America, a topic which would require further study.20 As a promising foray in that direction, Evonne Levy and a team of researchers at the University of Toronto carried out the data-driven project Canada: A History of Art History, which focused on faculty appointments. Levy’s team gathered information from annual university “Calendars,” catalogs that colleges and universities issue to provide information about their academic programs and faculty members. The team also conducted intensive research on individual programs, including interviews with faculty and staff. Their findings indicate that in 1940, 100 percent of the full-time tenure-track or tenured art history faculty in Canada were men. That situation shifted with each passing decade. By 2017, only 36.5 percent were male and the majority, 63.5 percent, were female (Figure 4). The most recent proportion of female to male art history faculty in Canada, with its roughly 2:1 ratio, thus aligns with the composition of new art history PhD holders, as indicated in Figure 3. In order to pursue the same question about the art history faculty in the United States, we would have to engage in a serious undertaking to gather data, both quantitative and qualitative, akin to Levy’s endeavor, but at a much larger scale. The project would also have to address questions about institution type and size, relative ranks of the faculty members, and the ever important issues of race, ethnicity, and contingent labor.
Institutions and Advisers
Part 1 of this series explored doctoral institutions and advisers in order to highlight the major sites that have fostered and mentored new generations of art historians over recent years. The expanded PSU data set confirms those earlier observations in quite stark terms. Figure 5 singles out the forty-four universities at which fifty or more dissertations were completed, which constitutes half of the institutions that were included in the dissertation roster from 1980 to the present. Once again, the Institute of Fine Arts, New York University (IFA/NYU), appears as the most generative site for doctoral work in the United States, presenting a record of 8.6 percent of the overall dissertations completed, equalling 549 dissertations in total.21 It is followed by Columbia, City University of New York (CUNY), Harvard, and Yale. These five high-volume programs, all located in the northeast, clearly play an outsize role in training doctoral students for the field, with over three hundred dissertations each. Comparatively, the majority of PhD-granting institutions have a record of less than one hundred dissertations completed in the same time period.
Based on the expanded PSU data set, 1,281 distinct faculty members advised the 6,345 dissertations listed during this time period.22 Figure 6 singles out only those advisers who supervised twenty or more dissertations, a group that includes forty-four people, labeled with the most recent institutional affiliation for the adviser that appears in the PSU data set. As was suggested in the August 2020 article, the late Linda Nochlin remains the most generative adviser of the group. The PSU data set includes sixty-four dissertations that she supervised from 1984 to 2017, when she passed away. By contrast, ProQuest lists sixty-seven students who completed dissertations under Nochlin’s supervision. Yet, neither list can be taken to be definitive, as both include and exclude certain entries. So, the actual number of students that Nochlin supervised exceeds the PSU data set’s total and likely also ProQuest’s total. Even so, both of these figures, underreported as they may be, are considerably higher than those of the other advisers with large cadres of doctoral students on the list, which are all below forty in total. This comparison is further evidence of the quite extraordinary impact of this legendary scholar on the field.
Fig. 6 Primary Advisers of 20 or More Completed Dissertations, 1980–2018, Labeled with the Most Recent Affiliation Listed, Based on the CAA Dissertation Roster and the PSU Data Set (created by Nancy Um; click for larger version)
When viewed in the format of a list and also the dehumanized bar charts featured above, the relationships between people, namely advisers and their advisees, and their connections to institutions tend to be flattened. Alternate formats of visualization, such as the social network diagram, map ties within and across groups. As an example, Figure 7 delves more deeply into some of the relationships suggested in Figure 6, by singling out those sixty-four students who completed dissertations under Nochlin’s mentorship as they appear in the PSU data set. This diagram presents a view of the adviser-advisee relationship as a pinwheel, linking together various students through their common doctoral adviser, indicated in gray. It also highlights the fact that Nochlin advised doctoral students from three different universities, IFA/NYU, CUNY, and Yale, each differentiated by color, over this time period. Therefore, it underlines quite clearly how an important and generative faculty adviser who taught at more than one institution connected a significant cadre of students across three universities over half a century. It is notable, of course, that these three institutions, IFA/NYU, CUNY, and Yale, are all among the high-volume doctoral institutions identified in Figure 5.
Fig. 7 Linda Nochlin and Her Students, Based on the CAA Dissertation Roster, 1980–2018, and the PSU Data Set (created by Nancy Um with visNetwork; click for interactive version, where you can scroll to zoom in or out/click and drag to rearrange the nodes)
This set of relationships is extended in Figure 8, with the inclusion of the doctoral students of two of Nochlin’s students, drawn from the PSU data set. This results in two additional pinwheels, which radiate from Nochlin’s own, in addition to the appearance of a fourth institution, Binghamton University, highlighted in gold. In the cases of these two added advisers, Patricia Mainardi and Tom McDonough, there are also discrepancies between the PSU data set and ProQuest lists, so their student cohorts are only partially represented.23 Even so, this graphic format suggests the reach of Nochlin’s training, which clearly extended far beyond the three main institutions of her affiliation. More radiating pinwheels could be added to amplify these generational registers with other institutions and programs. Figure 8 also attributes to Nochlin a high “betweenness centrality,” an index that refers to the number of times in which a node, or a person in this case, serves as a bridge between others.
Fig. 8 Linda Nochlin, Her Students, and Some of Her Students’ Students, Based on the CAA Dissertation Roster, 1980–2018, and the PSU Data Set (created by Nancy Um with visNetwork; click for interactive version, where you can scroll to zoom in or out/click and drag to rearrange the nodes)
The interpersonal ties that appear in Figures 7 and 8 constitute the fundamental fabric of art historical discourse as carried out through exchanges that operate on a human level. Yet these ties, which are both intellectual and social, appear in print in uneven ways, often relegated to the interstices of academic writing, such as the acknowledgments in a book’s opening pages or tucked away in our more personalized footnotes. There is only one scholarly format that enshrines these relationships, the Festschrift, which is unique in that it declares implicit social underpinnings as the basis for its table of contents. But clearly there is great interest in preserving these connections and seeking to understand their enduring impact within the academy. For instance, the Dictionary of Art Historians sketches the biographies of major figures in the field, providing their bibliographies and tracing connections to other art historians, both collaborators and students, although it focuses only on Western art. On another level, the crowd-sourced effort The Academic Family Tree maps adviser-advisee relationships across disciplines, thereby capturing the multiplicative character of these cross-generational relationships in a publicly accessible record.24
Rosters, Repositories, and the Limits of Institutional Memory
While it has been long overlooked, the CAA dissertation roster is a valuable primary source that sheds light on aspects of the history of art history as a discipline. With its chronological expansion and the enhancement of the added abstracts, the new PSU data set will be of continuing value to the field. The PSU team is dedicated to updating this resource as new rosters become available and is exploring mechanisms to address some of the data set’s errors and absences, many of which are based on original reporting problems. The team is also currently in discussion with CAA about ways to ensure the roster’s long-term maintenance in a machine-readable format, along with public access to it. This particular project of data curation and preservation should be acknowledged as a service to the discipline and, in fact, an open alternative to commercial indexing services such as ProQuest, which controls access to vast amounts of data about scholarly production. Indeed, the fundamental premise is that we, as scholars, should make information about our own work publicly available in order to circumvent the need to rely upon those commercial entities that ultimately seek to monetize access to the same body of information.25 So, there is more at stake here than a mere set of quantitative exercises.
As the researchers at PSU delved into these dissertation entries, several questions emerged that required investigation outside of the limits of the roster. Namely, they sought to track down dissertation advisers who had long left their positions in a given department or had retired or passed away. They also endeavored to locate dissertations that were reported but seemingly unfiled or those authors whose names had changed and were thus difficult to track. So, the work of compiling the data set was not at all mechanical, and in fact it inspired deep reflection about how to capture the past of our programs and, by extension, how we may keep a durable record of the people who have animated them. Levy’s Canadian study also responded to this particular concern. Her team endeavored to record departmental histories by featuring a Department Fact Sheet and a Faculty Info Sheet for almost every program in the country.26 These valuable fact sheets, composed in close consultation with each department, trace the development of degree programs over time, mark historical milestones, and even record past departmental debates. The faculty info sheets serve as important records, containing lists of all tenured and tenure-track faculty that appeared in the annual university calendars, thus likewise tracking historical changes.
On the other hand, for most institutions in the United States, departmental websites serve as the main repositories for program identities. Their contents are rarely standardized and many are not kept up to date.27 Moreover, they generally do not preserve historical data. For instance, some departments include the profiles of emeritus faculty members or memorial pages for those who have passed away, but rarely do they include the names of those who departed the program for another institution or the many contingent faculty members who take up a large share of teaching and advising on short-term contracts. As one very useful but quite unique example, the University of Chicago features a robust history of its art history department since 1890 on its website, including a list of former faculty members dating back to the department’s founding. For some, but not all, short biographical sketches and lists of publications were compiled by past students. The site provides a listing of dissertations completed in the department, but only since 2005.28 By contrast, the list of placements of its past doctoral students extends back to the 1990s. There are, of course, various ways to capture these institutional histories and legacies, although there are no codified standards for how to do so across the discipline or across universities. Fragments of a department’s history are more frequently excavated from the dusty files (or now hard drives) of individual faculty and staff, or simply committed to the memories of its longest-standing members. This essay and its earlier counterpart serve as prompts to inspire a discussion about how (and if) we wish to write these histories and which formats might be best suited to their long-term preservation, as art history barrels forward to the next stage in its development as a discipline.
Finally, we offer these provocations and findings with a certain amount of urgency. Institutions of higher education sit at a crossroads, as we have lived through a full year of COVID-19 and now anticipate the ways the pandemic will continue to affect our lives. We are all sensitive to the momentous changes that are occuring due to increasingly constrained resources and larger shifts within our own institutions. They include the possibility of program closures, the loss of faculty and staff, the continuing casualization of the academic workforce, and diminished support for our students, in addition to a belated reckoning with structural racism, social injustice, and increasing inequality. Humanities departments are vulnerable to such precarities and realignments, even while their scholars are ideally poised to provide much-needed understanding about this changing social landscape and its implications. Institutions will make weighty decisions based on the analysis of quantitative data, such as enrollments, budgets, outcomes, and placements. Art historians need to think deeply about how we can tell the story of our field, communicating through numbers as well as through words and mobilizing the data about our past to envision a promising future for the discipline and its next generations.
Penn State University
1 Elizabeth Mansfield, head of art history at Penn State University, spearheaded this project to compile and analyze data from the CAA roster, beginning in late 2019. The authors wish to thank Mansfield for her encouragement and generous support. They also thank Aaron Hyman, caa.reviews Editorial Board member, who provided the spark for this continuing project along with valuable feedback; Julie Nelson Davis, caa.reviews editor-in-chief, for her continuing enthusiasm for studies drawn from the CAA dissertation roster; and Evonne Levy, who kindly consulted about the data from Canada, in addition to Tom McDonough and Pepe Karmel, who both read the final draft and offered thoughtful comments.
2 Catherine D. Adams and Carolyn J. Lucarelli, Art History Dissertations and Abstracts from North American Institutions (University Park, PA: Penn State Libraries Open Publishing, 2021/ongoing), https://doi.org/10.26209/ahd.
3 The Penn State team included Catherine Adams, assistant curator of visual resources; Jennifer Glissman, MA student; Emily Hagen, PhD candidate; Claire Heidenreich, PhD candidate; Carolyn Lucarelli, curator of visual resources; and Kyle Marini, MA student. Adams and Lucarelli were essential collaborators on this project. Additionally, Heather Froehlich, literary informatics librarian, provided valuable guidance, feedback, and training throughout. While this essay was being finalized, caa.reviews published the dissertations that were completed in 2019 and 2020. These recent additions have not yet been incorporated into the PSU data set, and thus were not included in this article, although the PSU data set is currently being updated.
4 The authors of this essay assert that it is the responsibility of scholars to be explicit about how they obtained their data and the choices that they made in processing, cleaning, and analyzing it, following Catherine D’Ignazio and Lauren F. Klein’s directive that all aspects of data-oriented labor and decision-making should be documented and accessible. But, it is also acknowledged that the readers of this essay may be less engaged with these procedures. For this reason, the details of data processing and cleaning are generally relegated to the lengthy endnotes that follow. See D’Ignazio and Klein, Data Feminism (Cambridge, MA: MIT Press, 2020), 173–201.
5 Data set and scripts are available on GitHub: https://github.com/nancyum/caa.
6 The standard fields include: year, author last name, author first name, title, adviser(s), institution, and the CAA subject area categories, currently broken down along the lines of chronology, geographic area, and subject, genre, media, and artistic practice. In the roster, the dissertations are organized by year and further subdivided according to subject area. Within these subject areas, dissertations appear alphabetically by author. Dissertations with more than one subject area categorization appear in the online roster multiple times. The PSU team culled the duplicates from the data set, such that a single dissertation entry may have associated with it more than one subject area category, numbered from one to three, which reflects the order in which they appeared on the roster and is not indicative of any preference or priority. As described in the August 2020 essay, the subject area categories changed considerably according to the expanding interests and shifting approaches to art history across the roster’s sixty-year span (Nancy Um, “What Do We Know about the Future of Art History? Part 1,” caa.reviews, August 18, 2020, https://doi.org/10.3202/caa.reviews.2020.74). Each of the four assistants maintained a separate spreadsheet devoted to dissertations completed within a ten-year span, from the 1980s through the 2010s, which were then combined.
7 Institutional repositories for electronic theses and dissertations (ETDs) are maintained by individual institutions and are not searchable via a single database. To find abstracts not recorded by ProQuest, the Penn State team identified entries with missing abstracts, sorted by institution, located the institutional ETD, and manually queried these databases for each dissertation. The source of the abstract text, if not found in ProQuest or not found at all, was recorded in a separate notes field. Of the current 6,345 entries, 256 remain without abstracts, some due to institutional embargo. See Alexandra Gold, “The Great Embargo Debate,” Inside Higher Ed, September 20, 2018, https://www.insidehighered.com/blogs/gradhacker/great-embargo-debate.
8 Colleagues at the Getty Research Institute, Emily Pugh and Tracy Stuber, have also kindly consulted on this project’s next steps and future potential.
9 Modifications were made most commonly in the year and title fields, and sometimes in the author fields, due to inconsistent naming conventions, specifically in regard to the use of initials, middle names, and diacritics. Dissertations that had different titles in the ProQuest/ETD databases were noted in a separate field.
10 Upon completion of the data entry, the PSU team cleaned the data using OpenRefine. Obvious spelling errors, leading and trailing spaces, corrupted diacritics, and special characters were corrected. Care was taken to identify duplicates and transcription errors, corrected against information recorded by ProQuest. Institution and adviser fields, which were often reported inconsistently, were reconciled to achieve consistency with the conventions of part 1 of this series. Adviser names, including those with hyphenated first names, were standardized in the format “Initial.Initial. Last Name.” Institution names appear in their shortest, most common form; for example, Washington University in St. Louis was reconciled to Washington University, and University of Maryland was reconciled to Maryland. Reconciliation of the institution field reduced the number of unique institutions from 211 to 88 in the PSU data set, which indicates the high level of variation in institutional naming conventions that appeared in the CAA dissertation roster over this period. The data was restructured for analysis, which was carried out using various packages with the R programming language.
11 The PSU data set includes entries that predate 1980, when the roster first began to appear in The Art Bulletin. Those entries, which were inconsistently recorded, were not included in the aggregated results that appear in the following figures, as they would skew the appearance of change over time. For information about the scope and changing shape of the CAA dissertation roster and its move from Art Journal to The Art Bulletin in 1980, see Um, “What Do We Know about the Future of Art History?, Part 1.”
12 Lincoln Mullen, “Gender: Predict Gender from Names Using Historical Data,” R package version 0.5.2, https://github.com/lmullen/gender.
13 Cameron Blevins and Lincoln Mullen, “Jane, John . . . Leslie? A Historical Method for Algorithmic Gender Prediction,” Digital Humanities Quarterly 9, no. 3 (2015), http://www.digitalhumanities.org/dhq/vol/9/3/000223/000223.html.
14 Birth dates were determined using thirty-four as the average age of dissertation completion.
15 The authors offer their thanks to Esol Cho, a PhD candidate in political science at Binghamton University, who wrote the original scripts and helped to troubleshoot the analysis using the gender package. Cho also ran the same data using <a href=“https://genderize.io/>genderize.io”, another package for predicting gender in large data sets. Genderize.io generated fewer “undetermined” entries per year, but it resulted in a comparable breakdown in the proportion of male and female authors.
16 Names that include a hyphen, such as Anne-Marie, were split so that only the first part was included, in order to achieve a lower rate of “undetermined” findings. The “undetermined” group also includes all of those who use an initial as a first name.
17 “Dissertations about History by Gender, 1950–2012,” graph from Blevins and Mullen, “Jane, John . . . Leslie?” It should be noted that the ProQuest database of doctoral dissertations can be searched and mined on the basis of subject fields, but not by the discipline of record of the author. So, as with the CAA dissertation roster, one cannot gain a full picture of dissertations completed in the field through ProQuest, even though that database has been used to check, rectify, and enhance the entries that appear in the PSU data set.
18 Blevins and Mullen, “Jane, John . . . Leslie?”
19 “Proportion of Dissertations about History by Gender, 1950–2012,” graph from Blevins and Mullen, “Jane, John . . . Leslie?”
20 Pepe Karmel’s forthcoming study, to be published as part 3 of this essay series, will look at the career diversity of those who hold doctoral degrees in art history, including positions in museums, libraries, archives, commercial galleries, and other areas.
21 The preliminary results from the August 2020 essay can be viewed as comparison, including the whole group and only high-volume institutions; see Um, “What Do We Know about the Future of Art History?, Part 1.”
22 Some dissertations on the roster named more than one adviser; up to three were recorded in the PSU data set for each entry. In each case, the first name listed was taken as the primary adviser.
23 In both of these cases, as well, the filtered lists from the PSU data set and from ProQuest differ, with certain exclusions and errors in each. Neither source provides a definitive list of Mainardi’s or McDonough’s doctoral advisees.
24 The Academic Family Tree project relies on user-generated content and is divided by discipline. See “About the Academic Family Tree,” https://academictree.org/about.php.
25 ProQuest is actively leveraging its database for research purposes. See “ProQuest’s TDM Studio™ Service Transforms Text and Data Mining with Efficiency, Flexibility and Power,” ProQuest News Release, January 24, 2020, https://about.proquest.com/about/news/2020/ProQuests-TDM-Studio-Service-Transforms-Text-and-Data-Mining.html.
26 The linked example Department Fact Sheet and Faculty Info Sheet are for the University of British Columbia. To see additional sheets, consult the program page for each institution at https://arthistoryincanada.ca/faculty/.
27 CAA’s Graduate Programs in Art History and Graduate Programs in the Visual Arts directories, formerly published annually, may serve as another resource about departmental histories, although as of 2017 they are no longer being updated.
28 An examination of the list of dissertations on the University of Chicago website indicates some of the problems of reporting and record keeping that have been mentioned amply above in regard to the CAA dissertation roster. Although its entries matched up with the PSU data set for certain years, in others there were as many as six entries that were unreported to CAA, and thus did not appear in the PSU data set. Some dates of completion differed from those that were sent to CAA. In 2011 no dissertations were listed in the PSU data set, although the website includes five, one of which appeared in the PSU data set but under a different year.