The Carnegie Museum of Art (CMOA) is a contemporary art museum in Pittsburgh, Pennsylvania. Founded in 1895, the museum calls itself the first museum of contemporary art in the United States, seeking the “Old Masters of tomorrow.” The museum features over 30,000 objects covering a broad range of medium and form, including painting, sculpture, photographs, film, and digital imagery.
Key to CMOA’s aspirations of collecting the Old Masters of tomorrow is understanding contemporary art as a contemporary issue that engages with current social events and conditions. In a note from their director, CMOA asserts that their programming, exhibits, and publications “frequently explore the role of art and artists in confronting key social issues of our time, combining and juxtaposing local and global perspectives.” Their mission statement champions creativity with a global focus, reading in full:
We create experiences that connect people to art, ideas, and one another. At CMOA, we believe creativity is a defining human characteristic to which everyone should have access. CMOA collects, preserves, and presents artworks from around the world to inspire, sustain, and provoke discussion, and to engage and reflect multiple audiences.
This project takes its lead from CMOA's mission statement, examining their accession records to get a sense of how global their collection is. CMOA recorded the nationality of artists within each artwork’s accession record, and was fairly thorough in this effort. In the 4,869 accession records for their contemporary art department, only 11 have artists whose nationality is classified as “undetermined.” While there is no silver bullet for determining the diversity of a museum’s collection, the data visualizations in this project will tease apart aspects of CMOA’s contemporary art collection to get a better understanding of the nationalities and countries represented in their artworks.
You can learn more about this project by reading this poster, which was presented at the Keystone DH 2021 conference.
As part of their 120th anniversary, CMOA released collection records for all of their accessioned works. Each record includes information about the artwork (including title, creation date, and date acquired) as well as information about the artist (including name, nationality, and birthplace).
Since these are working records, the data contributors have a vested interest in maintaining thorough and accurate records. However, human error is very much a possibility within any dataset, and CMOA notes that the dataset may contain incomplete data or errors. Further, the dataset is not a static repository, but a continual project. Art historical research is ongoing at the museum, so CMOA advises downloading the most current version of the dataset to benefit from any updates to the records.
I primarily used Tableau to create data visualizations for this project. I wanted to develop expertise in Tableau, so I used the software to create different types of graphs. Tableau, however, was not a good fit for creating timelines, so I used TimelineJS for the timeline in the Items on View section.
.
With regard to the design of the site itself, I wanted it to be as clear and legible as possible. Mobirise offered an intuitive and simple design. I structured the homepage so that a link to the “About” page is prominently displayed in a bright red button, to encourage users to visit this page first to have more context for the visualizations and the project overall. Otherwise, I gave the four avenues into exploring nationality within the collection more or less equal footing, allowing users to click on what interests them most as opposed to forcing a certain order with a scrolling screen. I believe giving users the freedom to choose what interests them most makes them feel more engaged with the project. Further, the four aspects I explore are not necessarily linear, so forcing an order does little to enhance the understanding of the data visualizations themselves.
While data and data visualizations can be useful for spotting trends, data alone cannot represent an objective truth. Everything about data is subjective—data collection involves a subjective choice as to what you believe is valuable, and can involve errors and oversights. Data visualization software often requires data to fit a certain mold, eliminating nuance and outliers during the data cleaning process. In order to provide more context and transparency, I have included below specific steps and decisions I made when cleaning the data so that I could make legible visualizations.
With regard to the CMOA dataset overall, records were classified into six departments: contemporary art, decorative arts and design, film and video, fine arts, Heinz Architectural Center, and photography. I chose to focus on art that fell under the contemporary art department. From a practical standpoint, winnowing the dataset down from 28,154 records was a necessity, as I knew my data cleaning would involve some detailed record-level work. The contemporary art department has 4,869 records, which was a more reasonable amount for the 10-week period I had to work on the project. As to which department to choose, I selected contemporary art as it felt the more general, and thus inclusive, than some of the other more specialized departments (like film and video or photography). Further, the department felt more in line with CMOA’s point of pride—being the first museum of contemporary art museum in the United States—than some of the other departments, like fine arts.
MAPPING ARTISTS
The dataset provides the nationality of the artists for the artworks. To prepare the data for mapping, I had to change these descriptions to their country name (for example, change “Finnish” to “Finland”). Within OpenRefine, I identified spelling patterns across the nationalities (such as those ending with “ish” or “an”), used chomp GREL statements to trim these endings, and then replaced them with the appropriate endings.
Some artworks had multiple artists, and thus multiple nationalities. I separated the nationality column by separator, and focused on cleaning the nationality of the first artist only. While not ideal, including the nationalities of multiple artists for one piece would have unequally weighted the nationalities, framing the data in a way that equates one artwork by four artists to be the same as four separate artworks. Fortunately, in general multiple people tied to a piece were of the same nationality.
Additionally, the dataset lists both British and English, Scottish, and Welsh as nationalities. I would have liked to have parsed out the British entries to include more specificity; however, Tableau’s mapping function only has United Kingdom as a country, and does not map specifically for England, Scotland, and Wales.
ACQUISITION TRENDS
The “credit line” field in the dataset specifies how the artwork was acquired. The credit lines were very detailed, with names of specific donors. Within OpenRefine, I was able to use text filters on keywords like “gift of” and then run GREL replace statements to clean the data to just four categories: purchase; gift; partial gift, partial purchase; and museum appropriation.
The dataset also provided creation and acquisition dates for the artworks as a text string data type. Within OpenRefine I transformed these entries into the date data type, and then ran a GREL formula to subtract the creation date from the acquisition date and return a year amount.
Because American artworks comprise over 54% of the dataset, I decided to break these visualizations down by American and non-American art. While it would have been interesting to plot every nationality to see more detail and nuance, too many nationalities have too few artworks to make those graphs readable. Of the 56 nationalities recorded, 21 of them have 3 or less artworks. For the visualization of the time between an artwork’s creation and acquisition, then, plotting each country would have yielded dozens of small squiggles that do not clearly show purchasing trends. The graphs with American art versus art from outside the United States are able to tell a clearer story.
Some date entries had circa, or a range of years. In order to map dates, I had to normalize them, so when there was a range I took the average of the dates. I put a 20-year max on taking the average of date ranges. Ranges that exceeded 20 years were too broad to be meaningful; for example, a handful had creation dates like 1916-1956 or 20th century. I excluded artworks with these overly broad and general date ranges, as I would not be able to map them and there were relatively few (approximately 25 of the 4,689 entries).
ITEMS ON VIEW
The dataset provides the physical location of each artwork, specifying whether it is in a specific gallery, the museum lobby, the museum grounds, or not on view. I standardized these in OpenRefine, to be either “on view” or “not on view” to create the visualizations.
The dataset did not specify how long the piece has been in a location, so it is possible that the pieces move frequently or stay in their location for years. It would have been nice to have this information to situate the data more. Further, one of the categories was “off-site.” The dataset does not have a data dictionary, so it is not clear if the piece was off-site for restoration or off-site as part of a loan to another museum. Either way, I chose to classify artworks with this classification as “not on view,” because CMOA was not themselves putting that piece on view.
For the timeline, it was tricky to identify the nationalities of the 32 artists and artist collectives that participated in the Carnegie International, 57th Edition. The other artists in the timeline were either in the CMOA dataset, or their place of residence and birth were the same. This was not always the case with the artists from the Carnegie International. Accordingly, I referred to the documentation that CMOA provided about the event, which listed all of the "national affiliations by residence and birth." Of the 32 artists from the event, 20 lived in the United States and 26 nations were represented in country of birth. Recording all of these nationalities differs from how artist nationality was recorded within the dataset, as only one nationality was chosen. However, as the dataset did not clarify how nationality was chosen (specifically, if residence was favored over place of birth, and how residences in different countries were weighted), the best option was to include the nationalities as described by CMOA in their event programming.
THE AMBIGUITY OF NATIONALITY
The dataset provides the birthplace of the artists as one long text string (for example, “Seattle (King County, Washington, United States)”). Within OpenRefine, I clustered these birthplaces to standardize any variants. Then, I used GREL replace statements to transform the birthplaces into a string that used a comma as a delimiter (for example, “Seattle,Washington,United States”). Since Tableau maps through a hierarchy—city, state, country—I parsed these values out by separating the columns by the comma delimiter.
Often times the dataset only recorded the home country of artists. Despite research, there were six artists for whom I could not track down their specific birthplace. Either this information was not available online or their town was not appearing on a map. Since there were so few artists out of the 1,304 total that had this issue, I omitted them from the visualization.
“About.” n.d. Carnegie Museum of Art. Accessed December 2, 2019. https://cmoa.org/about/.
“Art & Architecture Thesaurus Full Record Display (Getty Research).” n.d. Accessed December 3, 2019. http://www.getty.edu/vow/AATFullDisplay?find=appropriation&logic=AND¬e=&english=N&prev_page=1&subjectid=300311629.
Burleigh, Paula. “The Politics of Museum Joy.” Art Journal 78, no. 2 (March 2019): 125–28. https://doi.org/10.1080/00043249.2019.1626164.
Carnegie Museum of Art. (2015) 2019. “Cmoa/Collection,” September. https://doi.org/10.5281/zenodo.35013.
Drucker, Johanna. “Graphical Approaches to the Digital Humanities.” Digital Humanities Quarterly 5, no. 1 (2011).
“Exhibitions.” n.d. Carnegie Museum of Art. Accessed December 5, 2019. https://cmoa.org/exhibitions/.
Groskopf, Christopher. “The Quartz Guide to Bad Data.” Quartz. December 15, 2015. https://qz.com/572338/the-quartz-guide-to-bad-data/.
“History.” n.d. Carnegie Museum of Art. Accessed December 2, 2019. https://cmoa.org/about/history/.
McNamee, Donald. 1983. "Abstract Painting and Sculpture in America 1927-1944." The Structurist 0 (23): 102.
“Participants.” Carnegie International 2018. Accessed December 9, 2019. https://2018.carnegieinternational.org/participants/.
Ramos, E. Carmen. “The Latino Presence in American Art.” American Art 26, no. 2 (2012): 7–13.
Rawson, Katie, and Trevor Muñoz. “Against Cleaning.” Curating Menus, July 6, 2016. http://curatingmenus.org/articles/against-cleaning/.
“Past Exhibitions.” n.d. Carnegie Museum of Art. Accessed December 5, 2019. https://cmoa.org/exhibitions/past/.