Father, help your servant Paul

A distant reading 7^th-12^th century seals from the Dumbarton Oaks Byzantine Seal Online Catalogue

 

Paul Albert

12/6/21

Introduction

This paper discusses work in progress examining the Dumbarton Oaks Byzantine Seal Online Catalogue^[1]. The Catalogue describes seals used by people of the Byzantine empire with their written correspondence. For this project, a subset of 13,600 unique Byzantine seals from the Catalogue’s 15,100 seals^[2] was examined. Not only is this a large subset of seals in the catalogue, but it is also a significant subset of seals known to exist today (estimates of this number range from 80,000^[3] – 100,000^[4]).

Byzantine seals had the utilitarian functions of securing documents and authenticating the sender^[5], but they also had numerous emotive functions as well^[6]. In this regard, a common comparison is made between these seals and modern-day business cards.^[7] Certainly, a modern business card is expected to give utilitarian information about a person (company, title, contact information, etc.), but it is also expressly designed to carry along with it certain emotive signals as well. The weight of the card stock, the selection of font, the possible inclusion of photos, slogans, mission statements, etc. are all meant to convey an emotive signal about some important essence of the card bearer.^[8]

From this perspective, a longitudinal study of Byzantine seals can offer us a unique window into Byzantine culture and the individuals within it; a glimpse into general trends both in overall cultural conventions and individual aspirations. Taken as a whole, the seals allow a unparalled window into examining many facets of Byzantine culture and society.

With a few notable exceptions^[9], past examinations of Byzantine seals have usually focused on only a small set of seal exemplars for analysis. The goal of this paper is different. Rather than focus on a small set of seals, this paper seeks to look at patterns that exist among the whole through a “distant reading” approach.

By “distant reading,” I refer to the theoretical approach popularized by Franco Moretti^[10] to use computational analysis to describe regularities and differences found in textual corpuses. In this specific project, the corpus of text examined includes the text directly inscribed on the seal itself, as well as the text used by Dumbarton Oaks in the catalogue to describe the obverses (front) and reverses (back) of the seals.

“Capta” not “Data”

In discussing the nature of information analyzed in the Digital Humanities, Johanna Drucker popularized the idea of using the term “capta” rather than “data” to describe the type information this paper studies. Drucker points out that there is a phenomenological distinction between information that might be recorded by an instrument (such as size, weight, metallic composition, etc.) and types of information that are constructed by human interpretation. She states:

Capta is “taken” actively while data is assumed to be a “given” able to be recorded and observed. From this distinction, a world of differences arises. Humanistic inquiry acknowledges the situated, partial, and constitutive character of knowledge production, the recognition that knowledge is constructed, taken, not simply given as a natural representation of pre-existing fact.^[11]

While the present paper somewhat blithely presents figures in a way that might make lead the reader to think of the information being analyzed as being clear cut and binary, I hope the reader keeps in mind that information being analyzed has many levels of ambiguity and human interpretation.

For example, even something as seemingly clear cut as the century of the date of a seal should belong to depends highly on human interpretation.^[12] With the small exception of seals for emperors or other notable persons (where specific years are certain), the process of ascribing a date range to a seal is inherently a subjective exercise. All told, only about 5% of the 13,600 unique seals used in this project have enough information for specific years to be assigned to a seal. For the other 95% of the seals, where only a range of centuries could be estimated, the cataloguers at Dumbarton Oaks look at many seal features but most closely at the epigraphy of inscribed text (see Footnote ).

Further ambiguities in our information arise from the specific vagaries in the cataloguing process itself and the seals themselves. The Online Catalogue built by Dumbarton Oaks is the product of many hands across many years (as are the seals it describes). For example, there are various ways of marking a saint (“Saint,” “St.” or “St ”), various ways of marking the Virgin Mary (“Virgin,” “Mother of God,” “Theotokis,” etc.), various ways of marking Christ (“Lord,” “Christ,” “Savior,” etc.) that differ between the various seals and fields in the catalog. While these “fuzzy” differences are easily understood by the human reader, the differences need to be identified and reconciled for a computational analysis seeking patterns between the seals.

Further ambiguities in the information studied arise from the nature in which the information was coded and captured through “web scraping.” For example, different practices are used in the Catalogue in distinguishing Greek text from its English translation. While these distinctions seem perfectly clear to the viewer of the Online Catalogue, they do pose problems for a “web scraper” seeking to categorize the information scraped from a web page.

“So, what is the point?” you, the reader, might be asking. If this information is inherently challenged, why spend time analyzing it? Ontologically, I would argue, this Digital Humanities project does exactly what historians have always done, look at anecdotal pieces of information, seek out patterns and draw conclusions. One can even argue that a Digital Humanities project forces the scholar to recognize ambiguities in information that are routinely overlooked when information is not forced into becoming numbers.

Still, to avoid the reader unconsciously assuming a false exactitude, this paper will generally round percentages, use thick translucent lines in charts rather than thin opaque lines and stay away from the word “data” and use the term “information” to describe what is being analyzed.

While we need to begin with a firm recognition that we are dealing with “capta” and not “data” for this analysis, there are sufficient regularities in the information compiled and captured from the catalogue to allow for some general patterns to be analyzed. The fact that we are dealing with a relatively large number of items here gives some comfort that while all the nuances of each seal might not be perfectly captured in the binary nature of a computational approach, there is enough information that possesses enough regularities for some general conclusions to be reliably drawn.^[13]

To help with the reader further explore my analysis, I have built an online interactive dashboard where my findings can be explored. The dashboard allows the user to explore aggregate patterns (the whole) as well as explore the images and captured information from the individual seals that form these patterns (the parts). [A preliminary version can be found here, it is my hope that Dumbarton Oaks will soon modify its web server settings to allow for the images of seal obverse and reverse to be displayed interactively on the dashboard.]

Analytical Approach

For this project, I only look at a subset of online catalogue seals that have been ascribed to the 7^th through 12^th centuries, are noted to have been in Greek or Latin languages and presented usable text for the various fields used in my analysis. All told, this is a set of 13,600 unique seals.

To capture the ambiguous nature of the century a seal might belong to, I have decided to count seals for every century they have been attributed to. If a seal is listed as both 8^th and 9^th century, then this analysis will count that seal for both centuries. Roughly 45% of the seals are listed for two centuries (a few of these had specific dates such as such as BZS.1947.2.344 “Alexios I Komnenos (1081–1118),” but the vast majority had a much more general dating such as BZS.1947.2.344 “Heliobos monk (tenth/eleventh century).” By (double) counting these seals across each century in the multiple centuries attributed, we obtain a set of 17,346 unique records from our 13,600 unique seals

Figure . Number of seals by century attribution from Dumbarton Oaks online catalogue selected for project analysis. 13,600 unique seals, 17,346 unique records when counted by century attribution (for example, a seal ascribed to both the 10^th and 11^th century is counted in both centuries).

In general, three different information fields from the online catalogue were developed and analyzed for this project.

1. Inscribed Text Field – the text inscribed on the seal (contains the Greek/Latin text with English translation). It was often a practice in seals to use abbreviations in the actual words inscribed on a seal. One example from this field would be – “Christ, help your servant Stephen.”
2. Obverse Description Field – an English description of what the obverse (front) of the seal shows. One example from this field would be – “Bust of St. Michael holding scepter with trefoil and globe. Border of dots.”
3. Reverse Description Field – an English description of what the reverse (back) of the seal shows. One example from this field would be – “Inscription of seven lines within linear border. Border of dots.”

Key Findings

Overall, 70% of the seals studied go beyond giving utilitarian information about the sender to also include explicit written invocations seeking divine assistance to directly benefit the sender (for example, “Lord, help Theodore, notarios”). As an emotive signal, such an invocation could signal the piety (and trustworthiness) of the sender. From the sender’s individual standpoint, such an invocation might also have been intended as a “living prayer” to be echoed and reinforced every time the seal was read by its receiver.

As can be seen in , these invocations were made to different divine beings and the explicit terms used seeking of assistance varied.

Sample Inscribed Text Field entries considered to explicitly invoke divine assistance:

Lord, help your servant Leo.

Lord, help Andronikos, sebastokrator, born in the purple.

(Lord, help) Theophylaktos Synadenos.

Martyr, may you protect Manuel Komnenodoukas.

All holy One, may you protect the owner of this letter.

Forerunner, watch over John Trichas.

Jesus Christ, King of Kings. Mother of God, help John, despotes.

May you, Virgin, protect Eumathios Philokales, doux and protonobellisimos.

Mother of God, help your servant Constantine.

Theotokos, help Constantine Loupos, dishypatos.

Table . Examples of seals where inscribed text explicitly invoked divine assistance.

Seals of the 8^th – 10^th centuries most characteristically invoked divine assistance

Across the seal records examined, a clear trend of the percentage of seals for each century that invoke divine assistance in the Inscribed Text Field.^[14]

Figure . Percentage of seals invoking divine assistance shows remarkable change across time

Starting in the 7^th century, less than 50% of our seals expressly sought divine assistance in their text inscriptions. However, by the 9^th century, that percentage rose to over 90, decreasing to less than 50% by the 12^th century (see ).

The main phenomenon this paper seeks to explore is the startling conformity of 9^th century seals to explicitly invoke divine assistance for the seal owner. First, we examine other seal characteristics that might correlate with this practice. Next, we discuss some ideas about why this practice might have existed.

Seals expressly invoking divine assistance mostly used the term “Help” (βοήθει)

While the seals of the 9^th century, compared to other centuries, show striking conformities in invoking divine assistance, they mostly seek the same form of assistance as seals from other centuries. Specifically, the explicit term used for the type of assistance sought is generally the word “help” (βοήθει). The biggest change in terms used was in the 12^th century where a sizable portion of the seals (about 40%) invoking divine assistance adopt the term “Protect” (σκέποις). See Figure 3.

Figure . Percentage of terms used in invoking divine aid (Inscribed Text Field)

Who help is sought from changes over the centuries

While over 90% of our 9^th century seals expressly invoke divine assistance, those that do seek help are split in seeking help from the Virgin or Christ. In contrast, seals of the 7^th century almost entirely seek help from the Virgin.^[15] See .

Figure . Who was help sought from? Divine entity appealed to by invocations in Interpreted Text Field.

Starting in the 11^th century there is a marked increase in seeking divine help from other than Christ or the Virgin. Usually, if not Christ or the Virgin, the divine entity invoked is a Saint.^[16]

Seals in the 9th century, where assistance was most invoked, also tended to have the most inscribed words

Strongly correlated with the conformity of 9^th century seals to invoke divine assistance () is the number of interpreted words in the Inscribed Text Field. I use the word “interpreted” to indicate that many of seals contain abbreviations (or illegible/missing text) that are expanded into full words in the online catalog. As an example, for this particular seal, the obverse bears the inscription “Κε βθ τῷ” and the reverse the inscription “Ῥω..ν σπαθ β νοταρ τῆς σακέ.λ,.” The Online Catalogue interprets this inscription to be “Κύριε βοήθει τῷ σῷ δούλῳ Ῥωμανῷ πρωτοσπαθαρίῳ καὶ βασιλικῷ νοταρίῳ τῆς σακέλλης” (Lord, help your servant Romanos, protospatharios and imperial notarios of the sakelle). In this example, nine inscribed “words/abbreviations” were expanded into 12 individual words by the cataloguers.

If (and I realize this is a big if) the seals are relatively uniform both in condition and how they use abbreviations across the centuries, seals of the 9^th century show a marked verbosity. On average, they contain around 10 interpreted words per seal versus about 4 words for the 7^th century and 6 words for the 12^th century. See .

Figure . Average number of interpreted words per seal (top) versus percentage of seals invoking assistance (bottom). For word count, interpreted words in Greek/Latin counted, not English translation. (Correlation, R2 = 0.90, p < 0.05)

Seals for centuries with most invocations show the least variety in words choice

While the seals of the 9^th century are more verbose than those of other centuries, the actual wordings they use are much more conventionally formulaic. To test for this, I developed a measure called the Lexical Variability Coefficient. For each century, I count the number of unique words used (e.g., “mother” would be counted as one unique word) and divided this by the total number of words used overall in the Interpreted Text Field (e.g., “mother” would be counted every time the word is used across all seals for the century).^[17] See .

Figure . Lexical Variability Coefficient (Number of Unique Words / Total Number of Words) (top) versus percentage of seals invoking assistance (bottom). In the 9th century, for every 10 additional words inscribed on seals, less than 1 new word was introduced overall. In the 12th century, this was more than double. (Correlation, R2 = 0.74, p < 0.05)

Notably, while the seals of the 8^th-10^th centuries use the most words (, top), they say the fewest different things (, top)!

Seal size (diameter) not correlated with the number of words inscribed

So far in this analysis, we have mostly examined the Catalogue’s description of the text inscribed on our seals. Inspired by the strong correlation between average number of words used per seal and the percentage of seals invoking assistance (), it seems reasonable to ask if there might be other attributes listed in the Catalogue that, in turn, might directly impact the number interpreted words inscribed.

One obvious attribute to examine in this regard is seal size. Arguably, the larger a seal is, the more words that can be inscribed onto it. Indeed, seals of the 9^th century tend to be larger in diameter on average (26 mm) than those of the 7^th (25 mm) and the 12^th (23 mm) and have more words (). However, 8^th century seals, for example, aree on average the largest (26.5 mm) across the centuries but tend to have fewer interpreted words than those of the 9^th century. In general, for the seals examined, seal size has little correlation with seal verbosity (and consequently, little correlation with whether seals invoked divine assistance). See .

Figure . Average Seal Size, Seal Verbosity and Proportion invoking divine aid across centuries. (Weak correlation between average seal diameter and % seals invoking assistance across centuries, R2 = 0.05, p > 0.5)

Practice of showing religious figural image on obverse only moderately correlated with whether divine assistance is invoked

If the seal size has little correlation with the number of words inscribed, perhaps there are other features that might? For example, it might be argued that the practice of showing a figural image on a seal gave its owner less of a field for words and less of an opportunity to invoke divine assistance. ^[18]

Figure . Percentage of seals by century where the obverse portrayed a figural religious image (top) versus percentage of seals invoking assistance (bottom) (Correlation R2 = 0.79, p < 0.05)

All else being equal, there is strong support for the idea that showing figures on a seal’s obverse negatively correlated with inscribed invocations for assistance. From a practical standpoint, there would less room for inscribed text. Of course, it might also be argued that portraying a religious figure on a seal might be considered as an implicit invocation for assistance (making a textual invocation less important). See .

Devoting space on the seal’s reverse to inscribed text doesn’t generally lead to more invocations

In considering possible seal attributes that might correlate with the practice of inscribed invocations, we have looked at both the overall size of the seal and whether the obverse of the seal had a religious figural image. The next and last seal attribute examined for this paper is the seal reverse. For the majority of the seals studied (about 90%), the Catalogue Reverse Description Field suggests the reverse is mostly text. However, across the centuries, this practice only shows mixed correlation with the practice of inscriptions expressly invoking divine assistance. See .

Figure . Percentage of seals where reverse is listed as mostly being an inscription (top) versus percentage of seals invoking assistance (bottom) (Correlation R² = 0.64, p > 0.05)

From the 7^th century to the 9^th century, the proportion of seals containing mostly an inscription on the reverse and the proportion of seals explicitly invoking assistance seem to track quite nicely. However, this is not the case looking from the 9^th century to the 12^th century. It is not that seals of these centuries had less of an easy opportunity to invoke assistance in their text, it seems a deliberate choice.

Discussion

The central finding of this paper is that seals in the 9^th century are much more likely to expressly invoke divine assistance in their inscriptions. What could explain this trend?

Could it be that people in the 9th century were much more pious than those, say, in the 7th or 12th century?

As silly as this proposition might sound to a Byzantinist, it is a question that should not be dismissed out of hand. If anything, recent research into Byzantine culture and society has highlighted the fact that it is far more nuanced and heterogenous than had been postulated in the past.^[19]

A scholarly examination of this question cannot rely on sphragistic evidence alone. Some other points of information that might help inform this question would be to look at some other indicators of relative piety such as distant readings of other types of text (correspondence, grave markers, laws, legal decisions, etc.) and trends in the establishment of new churches or monasteries.

Could it be that there were historically unique factors in the 9th century that led to the relatively uniform practice of expressly invoking divine assistance on seals?

Certainly, the 8^th and 9^th centuries are notably marked by a religious polarization around icons. On one hand, the iconoclasm debate revolved around the idea of how to properly worship the divine (not on the question of piety itself). From this perspective, iconoclasm in itself might not have had much direct impact on the practice of inscribed invocations. On the other hand, perhaps this theological polarization might have encouraged the owners of seals to search for a commonality to be expressed in the seals they commissioned and sent. Perhaps the turbulence of the time uniquely encouraged people to expressly invoke divine assistance?

Could it just be a change in cultural conventions (i.e., fashion) in what senders inscribed on their seals?

While I argue that attributes on seals offer a personal expression of the individual, they are also a product of cultural conventions (much like modern day business cards). Perhaps, rather than indicating different levels of personal piety across centuries, the practice of inscribed invocations for assistance that peaked in the 9^th century merely marks changes in fashion^[20] or just some sort of random drift^[21] in cultural conventions?

Could this just be an artifact of the cataloguing process?

Finally, there is a possibility that an explicit invocation for divine assistance might, in itself, guide the cataloguer to ascribe an 8^th-10^th century epoch to the seal. Dating a seal is an inherently subjective hermeneutic process.

Conclusion

The value of any historical analysis is not so much its ability to reach definitive answers, but in its ability to help inspire new and better questions. The central phenomenon this analysis presents is the relatively striking conformity among seals from the 8^th – 10^th century to invoke divine assistance. The new (and, hopefully, “better”) question to explore is “Why?”

Methodology

To “scrape” (iterate through the various pages and capture information about items) the Dumbarton Oaks Byzantine Seals Online Catalogue, I used a tool called Octoparse. First, I programmed Octoparse to iterate through each page of the Catalogue (with no filters selected), select each seal listed to get an item view page and then pull information listed about each seal into selected fields. While this gave base information about each seal, it did not give information about how the seal had been tagged by Century, Location, Title, Office, or Language. To get this information, I then programmed Octoparse to iterate through each of the filter types (Century, Location, Title, Office, Language) and capture this accession numbers of seals that matched these filters. This led to six individual data files.

To clean the information, I used a tool called Tableau Prep which allowed me to join the six data files for each seal to each other to create a master file with records for each seal that contained both information contained in the item view page and the filters.

To analyze and visualize the data, I mostly used a tool called Tableau Desktop. In addition, in order to develop the Lexical Variability Coefficient, I used two Browserling.com tools here and here.

It is my plan to create a public GitHub repository to archive both the raw information captured and cleaning process for future scholars to both examine my analysis and build on it.

Bibliography

Bentley, R. Alexander, Matthew W. Hahn, and Stephen J. Shennan. “Random Drift and Culture Change.” Proceedings: Biological Sciences 271, no. 1547 (2004): 1443–50.

Cotsonis, John A. The Religious Figural Imagery of Byzantine Lead Seals I: Studies on the Image of Christ, the Virgin and Narrative Scenes. London: Routledge, 2020. https://doi.org/10.4324/9780429327193.

———. The Religious Figural Imagery of Byzantine Lead Seals II: Studies on Images of the Saints and on Personal Piety. London: Routledge, 2020. https://doi.org/10.4324/9780429327216.

Drpić, Ivan. “The Patron’s ‘I’: Art, Selfhood, and the Later Byzantine Dedicatory Epigram.” In Epigram, Art, and Devotion in Later Byzantium, 89:895–935. Cambridge University Press, 2014. http://www.jstor.org/stable/43577194.

Drucker, Johanna. “Humanities Approaches to Graphical Display.” Digital Humanities Quarterly 005, no. 1 (March 10, 2011).

Kazhdan, A. P., and Ann Wharton Epstein. Change in Byzantine Culture in the Eleventh and Twelfth Centuries. University of California Press, 1990.

Moretti, Franco. “Conjectures on World Literature.” New Left Review, no. 1 (February 1, 2000): 54–68.

Oikonomides, Nicolas. “The Usual Lead Seal.” Dumbarton Oaks Papers 37 (1983): 147–57. https://doi.org/10.2307/1291481.

“Politics and Government in Byzantium: The Rise and Fall of the Bureaucrats (New Directions in Byzantine Studies): Shea, Jonathan, Stathakopoulos, Dionysios.” Accessed November 29, 2021. https://www.amazon.com/Politics-Government-Byzantium-Bureaucrats-Directions/dp/0755601939.

Seibt, Werner. “The Use of Monograms on Byzantine Seals in the Early Middle-Ages (6th to 9th Centuries).” Parekbolai. An Electronic Journal for Byzantine Literature 6, no. 0 (June 19, 2016): 1–14. https://doi.org/10.26262/par.v6i0.5082.

1. Many thanks to Jonathan Shea of Dumbarton Oaks for his help and guidance in answering questions about the Catalogue itself and how to interpret the information it contains. All errors made in the analysis are solely the responsibility of the author. The Dumbarton Oaks Online Catalogue of Byzantine Seals can be viewed and searched here. ↑

2. On 11/8/21, when the data was scraped from the Online Catalogue, the catalogue contained 15,100 seals of the approximately 17,000 seals in Dumbarton Oaks entire collection. Additions to the Online Catalogue are made on almost a daily basis. In selecting the set of 13,600 seals examined from the available 15,100 records available online on this date, seals falling outside the 7^th – 12^th centuries, not coded as being in Greek or Latin and not having a searchable Inscribed Text Field entry were excluded. ↑

3. Cotsonis, The Religious Figural Imagery of Byzantine Lead Seals II, 4. ↑

4. Seibt, “The Use of Monograms on Byzantine Seals in the Early Middle-Ages (6th to 9th Centuries),” 2. ↑

5. For an overview of the role and importance of seals, see Oikonomides, “The Usual Lead Seal.” ↑

6. ↑

7. “Politics and Government in Byzantium: The Rise and Fall of the Bureaucrats (New Directions in Byzantine Studies): Shea, Jonathan, Stathakopoulos, Dionysios,” 10. ↑

8. For a fascinating discussion of the emotive and performative functions of Byzantine epigrams that directly considers how written text was used to reflect a Byzantine performative-self see Drpić, “The Patron’s ‘I.’” (especially p. 98). ↑

9. Particularly impressive is the work of John Cotsonis who was doing “Digital Humanities” before Digital Humanities was “cool.” For a compilation of many of his articles quantitatively examining Byzantine seals, see Cotsonis, The Religious Figural Imagery of Byzantine Lead Seals I; Cotsonis, The Religious Figural Imagery of Byzantine Lead Seals II. ↑

10. Moretti, “Conjectures on World Literature.” ↑

11. Drucker, “Humanities Approaches to Graphical Display.” ↑

12. For a discussion on the Dumbarton Oaks cataloging process see here. ↑

13. Unfortunately, I find that many people in the humanities are more likely to dismiss a quantitative finding that wrestles with shades of gray than they will dismiss a qualitative finding built on similarly nuanced information. ↑

14. Key words searched for in the Inscribed Text Field to code for an invocation were “help,” “aid,” “protect,” “watch over” or “save.” While in some instances, these invocations might be made a non-divine being, such as an emperor, the instances are so few as not be statistically significant. ↑

15. Key words used in the Inscribed Text Field to assign Christ include “Christ,” “Jesus,” “Savior” and “Lord” (there are a small set of seals where the term “lord” is used, but do not denote Christ). Key words used to assign the Virgin include “Mother of God,” “Virgin,” “Lady” and “Mother of the Divine Word.” The programmatic logic used for the analysis favors “Christ” over “Virgin.” For example, if the Inscribed Text Field contained the words “Mother of God and Lord” the logic would assign “Christ” over the “Virgin” regardless of the fact the Virgin was mentioned first. This logic should not skew general findings, only 375 seals (about 2% of the total number of seals) have Inscribed Text Fields that contain keywords for both “Christ” and the “Virgin. ↑

16. For a discussion regarding the depiction of saints on seals as a reflection of general cultural trends, see Cotsonis, The Religious Figural Imagery of Byzantine Lead Seals I, 73. ↑

17. and use slightly different approaches for counting total words in the Greek/Latin interpreted text that give slightly different results. My working hypothesis is that one of the methods used for calculating words would consider “. . .” (with the spaces between periods) as three individual words while the other approach would count it as one. While the differences between algorithms might slightly alter my findings, I consider the impact to be relatively inconsequential. Nevertheless, this phenomenon requires additional investigation in the future. ↑

18. The Obverse Description Field offers unique challenges to a “distant reading” exercise. Because descriptions could be quite complicated in this field (e.g., Mother of God holding the Christ Child with a cross above), this analysis seeks to capture the most prominent element of the Obverse Description Field by programmatically considering the relative position of key search words. In the example above, the Obverse would be considered as primarily featuring the Virgin rather than Christ or a Cross because “Mother of God” is mentioned before other key words in the field text. Seals considered to portray a religious figural image are those that portray Christ (which was relatively rare), the Virgin (very common) or a Saint (fairly common). ↑

19. See, for example, Kazhdan and Epstein, Change in Byzantine Culture in the Eleventh and Twelfth Centuries. ↑

20. For a discussion of changing fashions used in Byzantine seal invocative monograms, see Seibt, “The Use of Monograms on Byzantine Seals in the Early Middle-Ages (6th to 9th Centuries),” 8–10. ↑

21. Bentley, Hahn, and Shennan, “Random Drift and Culture Change.” ↑

Father, help your servant Paul

A distant reading 7^th-12^th century seals from the Dumbarton Oaks Byzantine Seal Online Catalogue

Paul Albert

12/6/21

Introduction

This paper discusses work in progress examining the Dumbarton Oaks Byzantine Seal Online Catalogue^[1]. The Catalogue describes seals used by people of the Byzantine empire with their written correspondence. For this project, a subset of 13,600 unique Byzantine seals from the Catalogue’s 15,100 seals^[2] was examined. Not only is this a large subset of seals in the catalogue, but it is also a significant subset of seals known to exist today (estimates of this number range from 80,000^[3] – 100,000^[4]).

Byzantine seals had the utilitarian functions of securing documents and authenticating the sender^[5], but they also had numerous emotive functions as well^[6]. In this regard, a common comparison is made between these seals and modern-day business cards.^[7] Certainly, a modern business card is expected to give utilitarian information about a person (company, title, contact information, etc.), but it is also expressly designed to carry along with it certain emotive signals as well. The weight of the card stock, the selection of font, the possible inclusion of photos, slogans, mission statements, etc. are all meant to convey an emotive signal about some important essence of the card bearer.^[8]

From this perspective, a longitudinal study of Byzantine seals can offer us a unique window into Byzantine culture and the individuals within it; a glimpse into general trends both in overall cultural conventions and individual aspirations. Taken as a whole, the seals allow a unparalled window into examining many facets of Byzantine culture and society.

With a few notable exceptions^[9], past examinations of Byzantine seals have usually focused on only a small set of seal exemplars for analysis. The goal of this paper is different. Rather than focus on a small set of seals, this paper seeks to look at patterns that exist among the whole through a “distant reading” approach.

By “distant reading,” I refer to the theoretical approach popularized by Franco Moretti^[10] to use computational analysis to describe regularities and differences found in textual corpuses. In this specific project, the corpus of text examined includes the text directly inscribed on the seal itself, as well as the text used by Dumbarton Oaks in the catalogue to describe the obverses (front) and reverses (back) of the seals.

“Capta” not “Data”

In discussing the nature of information analyzed in the Digital Humanities, Johanna Drucker popularized the idea of using the term “capta” rather than “data” to describe the type information this paper studies. Drucker points out that there is a phenomenological distinction between information that might be recorded by an instrument (such as size, weight, metallic composition, etc.) and types of information that are constructed by human interpretation. She states:

Capta is “taken” actively while data is assumed to be a “given” able to be recorded and observed. From this distinction, a world of differences arises. Humanistic inquiry acknowledges the situated, partial, and constitutive character of knowledge production, the recognition that knowledge is constructed, taken, not simply given as a natural representation of pre-existing fact.^[11]

While the present paper somewhat blithely presents figures in a way that might make lead the reader to think of the information being analyzed as being clear cut and binary, I hope the reader keeps in mind that information being analyzed has many levels of ambiguity and human interpretation.

For example, even something as seemingly clear cut as the century of the date of a seal should belong to depends highly on human interpretation.^[12] With the small exception of seals for emperors or other notable persons (where specific years are certain), the process of ascribing a date range to a seal is inherently a subjective exercise. All told, only about 5% of the 13,600 unique seals used in this project have enough information for specific years to be assigned to a seal. For the other 95% of the seals, where only a range of centuries could be estimated, the cataloguers at Dumbarton Oaks look at many seal features but most closely at the epigraphy of inscribed text (see Footnote ).

Further ambiguities in our information arise from the specific vagaries in the cataloguing process itself and the seals themselves. The Online Catalogue built by Dumbarton Oaks is the product of many hands across many years (as are the seals it describes). For example, there are various ways of marking a saint (“Saint,” “St.” or “St ”), various ways of marking the Virgin Mary (“Virgin,” “Mother of God,” “Theotokis,” etc.), various ways of marking Christ (“Lord,” “Christ,” “Savior,” etc.) that differ between the various seals and fields in the catalog. While these “fuzzy” differences are easily understood by the human reader, the differences need to be identified and reconciled for a computational analysis seeking patterns between the seals.

Further ambiguities in the information studied arise from the nature in which the information was coded and captured through “web scraping.” For example, different practices are used in the Catalogue in distinguishing Greek text from its English translation. While these distinctions seem perfectly clear to the viewer of the Online Catalogue, they do pose problems for a “web scraper” seeking to categorize the information scraped from a web page.

“So, what is the point?” you, the reader, might be asking. If this information is inherently challenged, why spend time analyzing it? Ontologically, I would argue, this Digital Humanities project does exactly what historians have always done, look at anecdotal pieces of information, seek out patterns and draw conclusions. One can even argue that a Digital Humanities project forces the scholar to recognize ambiguities in information that are routinely overlooked when information is not forced into becoming numbers.

Still, to avoid the reader unconsciously assuming a false exactitude, this paper will generally round percentages, use thick translucent lines in charts rather than thin opaque lines and stay away from the word “data” and use the term “information” to describe what is being analyzed.

While we need to begin with a firm recognition that we are dealing with “capta” and not “data” for this analysis, there are sufficient regularities in the information compiled and captured from the catalogue to allow for some general patterns to be analyzed. The fact that we are dealing with a relatively large number of items here gives some comfort that while all the nuances of each seal might not be perfectly captured in the binary nature of a computational approach, there is enough information that possesses enough regularities for some general conclusions to be reliably drawn.^[13]

To help with the reader further explore my analysis, I have built an online interactive dashboard where my findings can be explored. The dashboard allows the user to explore aggregate patterns (the whole) as well as explore the images and captured information from the individual seals that form these patterns (the parts). [A preliminary version can be found here, it is my hope that Dumbarton Oaks will soon modify its web server settings to allow for the images of seal obverse and reverse to be displayed interactively on the dashboard.]

Analytical Approach

For this project, I only look at a subset of online catalogue seals that have been ascribed to the 7^th through 12^th centuries, are noted to have been in Greek or Latin languages and presented usable text for the various fields used in my analysis. All told, this is a set of 13,600 unique seals.

To capture the ambiguous nature of the century a seal might belong to, I have decided to count seals for every century they have been attributed to. If a seal is listed as both 8^th and 9^th century, then this analysis will count that seal for both centuries. Roughly 45% of the seals are listed for two centuries (a few of these had specific dates such as such as BZS.1947.2.344 “Alexios I Komnenos (1081–1118),” but the vast majority had a much more general dating such as BZS.1947.2.344 “Heliobos monk (tenth/eleventh century).” By (double) counting these seals across each century in the multiple centuries attributed, we obtain a set of 17,346 unique records from our 13,600 unique seals

Figure . Number of seals by century attribution from Dumbarton Oaks online catalogue selected for project analysis. 13,600 unique seals, 17,346 unique records when counted by century attribution (for example, a seal ascribed to both the 10^th and 11^th century is counted in both centuries).

In general, three different information fields from the online catalogue were developed and analyzed for this project.

1. Inscribed Text Field – the text inscribed on the seal (contains the Greek/Latin text with English translation). It was often a practice in seals to use abbreviations in the actual words inscribed on a seal. One example from this field would be – “Christ, help your servant Stephen.”
2. Obverse Description Field – an English description of what the obverse (front) of the seal shows. One example from this field would be – “Bust of St. Michael holding scepter with trefoil and globe. Border of dots.”
3. Reverse Description Field – an English description of what the reverse (back) of the seal shows. One example from this field would be – “Inscription of seven lines within linear border. Border of dots.”

Key Findings

Overall, 70% of the seals studied go beyond giving utilitarian information about the sender to also include explicit written invocations seeking divine assistance to directly benefit the sender (for example, “Lord, help Theodore, notarios”). As an emotive signal, such an invocation could signal the piety (and trustworthiness) of the sender. From the sender’s individual standpoint, such an invocation might also have been intended as a “living prayer” to be echoed and reinforced every time the seal was read by its receiver.

As can be seen in , these invocations were made to different divine beings and the explicit terms used seeking of assistance varied.

Sample Inscribed Text Field entries considered to explicitly invoke divine assistance:

Lord, help your servant Leo.

Lord, help Andronikos, sebastokrator, born in the purple.

(Lord, help) Theophylaktos Synadenos.

Martyr, may you protect Manuel Komnenodoukas.

All holy One, may you protect the owner of this letter.

Forerunner, watch over John Trichas.

Jesus Christ, King of Kings. Mother of God, help John, despotes.

May you, Virgin, protect Eumathios Philokales, doux and protonobellisimos.

Mother of God, help your servant Constantine.

Theotokos, help Constantine Loupos, dishypatos.

Table . Examples of seals where inscribed text explicitly invoked divine assistance.

Seals of the 8^th – 10^th centuries most characteristically invoked divine assistance

Across the seal records examined, a clear trend of the percentage of seals for each century that invoke divine assistance in the Inscribed Text Field.^[14]

Figure . Percentage of seals invoking divine assistance shows remarkable change across time

Starting in the 7^th century, less than 50% of our seals expressly sought divine assistance in their text inscriptions. However, by the 9^th century, that percentage rose to over 90, decreasing to less than 50% by the 12^th century (see ).

The main phenomenon this paper seeks to explore is the startling conformity of 9^th century seals to explicitly invoke divine assistance for the seal owner. First, we examine other seal characteristics that might correlate with this practice. Next, we discuss some ideas about why this practice might have existed.

Seals expressly invoking divine assistance mostly used the term “Help” (βοήθει)

While the seals of the 9^th century, compared to other centuries, show striking conformities in invoking divine assistance, they mostly seek the same form of assistance as seals from other centuries. Specifically, the explicit term used for the type of assistance sought is generally the word “help” (βοήθει). The biggest change in terms used was in the 12^th century where a sizable portion of the seals (about 40%) invoking divine assistance adopt the term “Protect” (σκέποις). See Figure 3.

Figure . Percentage of terms used in invoking divine aid (Inscribed Text Field)

Who help is sought from changes over the centuries

While over 90% of our 9^th century seals expressly invoke divine assistance, those that do seek help are split in seeking help from the Virgin or Christ. In contrast, seals of the 7^th century almost entirely seek help from the Virgin.^[15] See .

Figure . Who was help sought from? Divine entity appealed to by invocations in Interpreted Text Field.

Starting in the 11^th century there is a marked increase in seeking divine help from other than Christ or the Virgin. Usually, if not Christ or the Virgin, the divine entity invoked is a Saint.^[16]

Seals in the 9th century, where assistance was most invoked, also tended to have the most inscribed words

Strongly correlated with the conformity of 9^th century seals to invoke divine assistance () is the number of interpreted words in the Inscribed Text Field. I use the word “interpreted” to indicate that many of seals contain abbreviations (or illegible/missing text) that are expanded into full words in the online catalog. As an example, for this particular seal, the obverse bears the inscription “Κε βθ τῷ” and the reverse the inscription “Ῥω..ν σπαθ β νοταρ τῆς σακέ.λ,.” The Online Catalogue interprets this inscription to be “Κύριε βοήθει τῷ σῷ δούλῳ Ῥωμανῷ πρωτοσπαθαρίῳ καὶ βασιλικῷ νοταρίῳ τῆς σακέλλης” (Lord, help your servant Romanos, protospatharios and imperial notarios of the sakelle). In this example, nine inscribed “words/abbreviations” were expanded into 12 individual words by the cataloguers.

If (and I realize this is a big if) the seals are relatively uniform both in condition and how they use abbreviations across the centuries, seals of the 9^th century show a marked verbosity. On average, they contain around 10 interpreted words per seal versus about 4 words for the 7^th century and 6 words for the 12^th century. See .

Figure . Average number of interpreted words per seal (top) versus percentage of seals invoking assistance (bottom). For word count, interpreted words in Greek/Latin counted, not English translation. (Correlation, R2 = 0.90, p < 0.05)

Seals for centuries with most invocations show the least variety in words choice

While the seals of the 9^th century are more verbose than those of other centuries, the actual wordings they use are much more conventionally formulaic. To test for this, I developed a measure called the Lexical Variability Coefficient. For each century, I count the number of unique words used (e.g., “mother” would be counted as one unique word) and divided this by the total number of words used overall in the Interpreted Text Field (e.g., “mother” would be counted every time the word is used across all seals for the century).^[17] See .

Figure . Lexical Variability Coefficient (Number of Unique Words / Total Number of Words) (top) versus percentage of seals invoking assistance (bottom). In the 9th century, for every 10 additional words inscribed on seals, less than 1 new word was introduced overall. In the 12th century, this was more than double. (Correlation, R2 = 0.74, p < 0.05)

Notably, while the seals of the 8^th-10^th centuries use the most words (, top), they say the fewest different things (, top)!

Seal size (diameter) not correlated with the number of words inscribed

So far in this analysis, we have mostly examined the Catalogue’s description of the text inscribed on our seals. Inspired by the strong correlation between average number of words used per seal and the percentage of seals invoking assistance (), it seems reasonable to ask if there might be other attributes listed in the Catalogue that, in turn, might directly impact the number interpreted words inscribed.

One obvious attribute to examine in this regard is seal size. Arguably, the larger a seal is, the more words that can be inscribed onto it. Indeed, seals of the 9^th century tend to be larger in diameter on average (26 mm) than those of the 7^th (25 mm) and the 12^th (23 mm) and have more words (). However, 8^th century seals, for example, aree on average the largest (26.5 mm) across the centuries but tend to have fewer interpreted words than those of the 9^th century. In general, for the seals examined, seal size has little correlation with seal verbosity (and consequently, little correlation with whether seals invoked divine assistance). See .

Figure . Average Seal Size, Seal Verbosity and Proportion invoking divine aid across centuries. (Weak correlation between average seal diameter and % seals invoking assistance across centuries, R2 = 0.05, p > 0.5)

Practice of showing religious figural image on obverse only moderately correlated with whether divine assistance is invoked

If the seal size has little correlation with the number of words inscribed, perhaps there are other features that might? For example, it might be argued that the practice of showing a figural image on a seal gave its owner less of a field for words and less of an opportunity to invoke divine assistance. ^[18]

Figure . Percentage of seals by century where the obverse portrayed a figural religious image (top) versus percentage of seals invoking assistance (bottom) (Correlation R2 = 0.79, p < 0.05)

All else being equal, there is strong support for the idea that showing figures on a seal’s obverse negatively correlated with inscribed invocations for assistance. From a practical standpoint, there would less room for inscribed text. Of course, it might also be argued that portraying a religious figure on a seal might be considered as an implicit invocation for assistance (making a textual invocation less important). See .

Devoting space on the seal’s reverse to inscribed text doesn’t generally lead to more invocations

In considering possible seal attributes that might correlate with the practice of inscribed invocations, we have looked at both the overall size of the seal and whether the obverse of the seal had a religious figural image. The next and last seal attribute examined for this paper is the seal reverse. For the majority of the seals studied (about 90%), the Catalogue Reverse Description Field suggests the reverse is mostly text. However, across the centuries, this practice only shows mixed correlation with the practice of inscriptions expressly invoking divine assistance. See .

Figure . Percentage of seals where reverse is listed as mostly being an inscription (top) versus percentage of seals invoking assistance (bottom) (Correlation R² = 0.64, p > 0.05)

From the 7^th century to the 9^th century, the proportion of seals containing mostly an inscription on the reverse and the proportion of seals explicitly invoking assistance seem to track quite nicely. However, this is not the case looking from the 9^th century to the 12^th century. It is not that seals of these centuries had less of an easy opportunity to invoke assistance in their text, it seems a deliberate choice.

Discussion

The central finding of this paper is that seals in the 9^th century are much more likely to expressly invoke divine assistance in their inscriptions. What could explain this trend?

Could it be that people in the 9th century were much more pious than those, say, in the 7th or 12th century?

As silly as this proposition might sound to a Byzantinist, it is a question that should not be dismissed out of hand. If anything, recent research into Byzantine culture and society has highlighted the fact that it is far more nuanced and heterogenous than had been postulated in the past.^[19]

A scholarly examination of this question cannot rely on sphragistic evidence alone. Some other points of information that might help inform this question would be to look at some other indicators of relative piety such as distant readings of other types of text (correspondence, grave markers, laws, legal decisions, etc.) and trends in the establishment of new churches or monasteries.

Could it be that there were historically unique factors in the 9th century that led to the relatively uniform practice of expressly invoking divine assistance on seals?

Certainly, the 8^th and 9^th centuries are notably marked by a religious polarization around icons. On one hand, the iconoclasm debate revolved around the idea of how to properly worship the divine (not on the question of piety itself). From this perspective, iconoclasm in itself might not have had much direct impact on the practice of inscribed invocations. On the other hand, perhaps this theological polarization might have encouraged the owners of seals to search for a commonality to be expressed in the seals they commissioned and sent. Perhaps the turbulence of the time uniquely encouraged people to expressly invoke divine assistance?

Could it just be a change in cultural conventions (i.e., fashion) in what senders inscribed on their seals?

While I argue that attributes on seals offer a personal expression of the individual, they are also a product of cultural conventions (much like modern day business cards). Perhaps, rather than indicating different levels of personal piety across centuries, the practice of inscribed invocations for assistance that peaked in the 9^th century merely marks changes in fashion^[20] or just some sort of random drift^[21] in cultural conventions?

Could this just be an artifact of the cataloguing process?

Finally, there is a possibility that an explicit invocation for divine assistance might, in itself, guide the cataloguer to ascribe an 8^th-10^th century epoch to the seal. Dating a seal is an inherently subjective hermeneutic process.

Conclusion

The value of any historical analysis is not so much its ability to reach definitive answers, but in its ability to help inspire new and better questions. The central phenomenon this analysis presents is the relatively striking conformity among seals from the 8^th – 10^th century to invoke divine assistance. The new (and, hopefully, “better”) question to explore is “Why?”

Methodology

To “scrape” (iterate through the various pages and capture information about items) the Dumbarton Oaks Byzantine Seals Online Catalogue, I used a tool called Octoparse. First, I programmed Octoparse to iterate through each page of the Catalogue (with no filters selected), select each seal listed to get an item view page and then pull information listed about each seal into selected fields. While this gave base information about each seal, it did not give information about how the seal had been tagged by Century, Location, Title, Office, or Language. To get this information, I then programmed Octoparse to iterate through each of the filter types (Century, Location, Title, Office, Language) and capture this accession numbers of seals that matched these filters. This led to six individual data files.

To clean the information, I used a tool called Tableau Prep which allowed me to join the six data files for each seal to each other to create a master file with records for each seal that contained both information contained in the item view page and the filters.

To analyze and visualize the data, I mostly used a tool called Tableau Desktop. In addition, in order to develop the Lexical Variability Coefficient, I used two Browserling.com tools here and here.

It is my plan to create a public GitHub repository to archive both the raw information captured and cleaning process for future scholars to both examine my analysis and build on it.

Bibliography

Bentley, R. Alexander, Matthew W. Hahn, and Stephen J. Shennan. “Random Drift and Culture Change.” Proceedings: Biological Sciences 271, no. 1547 (2004): 1443–50.

Cotsonis, John A. The Religious Figural Imagery of Byzantine Lead Seals I: Studies on the Image of Christ, the Virgin and Narrative Scenes. London: Routledge, 2020. https://doi.org/10.4324/9780429327193.

———. The Religious Figural Imagery of Byzantine Lead Seals II: Studies on Images of the Saints and on Personal Piety. London: Routledge, 2020. https://doi.org/10.4324/9780429327216.

Drpić, Ivan. “The Patron’s ‘I’: Art, Selfhood, and the Later Byzantine Dedicatory Epigram.” In Epigram, Art, and Devotion in Later Byzantium, 89:895–935. Cambridge University Press, 2014. http://www.jstor.org/stable/43577194.

Drucker, Johanna. “Humanities Approaches to Graphical Display.” Digital Humanities Quarterly 005, no. 1 (March 10, 2011).

Kazhdan, A. P., and Ann Wharton Epstein. Change in Byzantine Culture in the Eleventh and Twelfth Centuries. University of California Press, 1990.

Moretti, Franco. “Conjectures on World Literature.” New Left Review, no. 1 (February 1, 2000): 54–68.

Oikonomides, Nicolas. “The Usual Lead Seal.” Dumbarton Oaks Papers 37 (1983): 147–57. https://doi.org/10.2307/1291481.

“Politics and Government in Byzantium: The Rise and Fall of the Bureaucrats (New Directions in Byzantine Studies): Shea, Jonathan, Stathakopoulos, Dionysios.” Accessed November 29, 2021. https://www.amazon.com/Politics-Government-Byzantium-Bureaucrats-Directions/dp/0755601939.

Seibt, Werner. “The Use of Monograms on Byzantine Seals in the Early Middle-Ages (6th to 9th Centuries).” Parekbolai. An Electronic Journal for Byzantine Literature 6, no. 0 (June 19, 2016): 1–14. https://doi.org/10.26262/par.v6i0.5082.

1. Many thanks to Jonathan Shea of Dumbarton Oaks for his help and guidance in answering questions about the Catalogue itself and how to interpret the information it contains. All errors made in the analysis are solely the responsibility of the author. The Dumbarton Oaks Online Catalogue of Byzantine Seals can be viewed and searched here. ↑

2. On 11/8/21, when the data was scraped from the Online Catalogue, the catalogue contained 15,100 seals of the approximately 17,000 seals in Dumbarton Oaks entire collection. Additions to the Online Catalogue are made on almost a daily basis. In selecting the set of 13,600 seals examined from the available 15,100 records available online on this date, seals falling outside the 7^th – 12^th centuries, not coded as being in Greek or Latin and not having a searchable Inscribed Text Field entry were excluded. ↑

3. Cotsonis, The Religious Figural Imagery of Byzantine Lead Seals II, 4. ↑

4. Seibt, “The Use of Monograms on Byzantine Seals in the Early Middle-Ages (6th to 9th Centuries),” 2. ↑

5. For an overview of the role and importance of seals, see Oikonomides, “The Usual Lead Seal.” ↑

6. ↑

7. “Politics and Government in Byzantium: The Rise and Fall of the Bureaucrats (New Directions in Byzantine Studies): Shea, Jonathan, Stathakopoulos, Dionysios,” 10. ↑

8. For a fascinating discussion of the emotive and performative functions of Byzantine epigrams that directly considers how written text was used to reflect a Byzantine performative-self see Drpić, “The Patron’s ‘I.’” (especially p. 98). ↑

9. Particularly impressive is the work of John Cotsonis who was doing “Digital Humanities” before Digital Humanities was “cool.” For a compilation of many of his articles quantitatively examining Byzantine seals, see Cotsonis, The Religious Figural Imagery of Byzantine Lead Seals I; Cotsonis, The Religious Figural Imagery of Byzantine Lead Seals II. ↑

10. Moretti, “Conjectures on World Literature.” ↑

11. Drucker, “Humanities Approaches to Graphical Display.” ↑

12. For a discussion on the Dumbarton Oaks cataloging process see here. ↑

13. Unfortunately, I find that many people in the humanities are more likely to dismiss a quantitative finding that wrestles with shades of gray than they will dismiss a qualitative finding built on similarly nuanced information. ↑

14. Key words searched for in the Inscribed Text Field to code for an invocation were “help,” “aid,” “protect,” “watch over” or “save.” While in some instances, these invocations might be made a non-divine being, such as an emperor, the instances are so few as not be statistically significant. ↑

15. Key words used in the Inscribed Text Field to assign Christ include “Christ,” “Jesus,” “Savior” and “Lord” (there are a small set of seals where the term “lord” is used, but do not denote Christ). Key words used to assign the Virgin include “Mother of God,” “Virgin,” “Lady” and “Mother of the Divine Word.” The programmatic logic used for the analysis favors “Christ” over “Virgin.” For example, if the Inscribed Text Field contained the words “Mother of God and Lord” the logic would assign “Christ” over the “Virgin” regardless of the fact the Virgin was mentioned first. This logic should not skew general findings, only 375 seals (about 2% of the total number of seals) have Inscribed Text Fields that contain keywords for both “Christ” and the “Virgin. ↑

16. For a discussion regarding the depiction of saints on seals as a reflection of general cultural trends, see Cotsonis, The Religious Figural Imagery of Byzantine Lead Seals I, 73. ↑

17. and use slightly different approaches for counting total words in the Greek/Latin interpreted text that give slightly different results. My working hypothesis is that one of the methods used for calculating words would consider “. . .” (with the spaces between periods) as three individual words while the other approach would count it as one. While the differences between algorithms might slightly alter my findings, I consider the impact to be relatively inconsequential. Nevertheless, this phenomenon requires additional investigation in the future. ↑

18. The Obverse Description Field offers unique challenges to a “distant reading” exercise. Because descriptions could be quite complicated in this field (e.g., Mother of God holding the Christ Child with a cross above), this analysis seeks to capture the most prominent element of the Obverse Description Field by programmatically considering the relative position of key search words. In the example above, the Obverse would be considered as primarily featuring the Virgin rather than Christ or a Cross because “Mother of God” is mentioned before other key words in the field text. Seals considered to portray a religious figural image are those that portray Christ (which was relatively rare), the Virgin (very common) or a Saint (fairly common). ↑

19. See, for example, Kazhdan and Epstein, Change in Byzantine Culture in the Eleventh and Twelfth Centuries. ↑

20. For a discussion of changing fashions used in Byzantine seal invocative monograms, see Seibt, “The Use of Monograms on Byzantine Seals in the Early Middle-Ages (6th to 9th Centuries),” 8–10. ↑

21. Bentley, Hahn, and Shennan, “Random Drift and Culture Change.” ↑

Paul Albert

12/5/18

When bankers get together, they discuss art.

When artists get together, they discuss money.

Oscar Wilde

Setting the Stage

While the history of art is replete with stories of talented artists dying unknown and penniless, it is worthwhile to ask, “how did other roughly equivalently talented artists gain fame and fortune in their lifetimes?”. For, I believe, that if one were to ask any artist which they would prefer, living poor and unknown or living rich and famous, they all would state a preference for the latter.^[1] Like as with you and me, artists can only aspire for posterity, but, day to day, actually measure their own worth and put bread on their table by how their efforts are recognized in their lifetime.

Figure 1. The Artist in Despair, Adam Elsheimer, Rome, circa 1600

In 1556, Giorgio Vasari visited the house of the sculptor Leone Leoni in Milan.^[2] Ten years later, Vasari wrote about Leoni’s opulent lifestyle and noted that “were Michelangelo to return to life… he would say that the art that caused him to be considered exceptional had become another thing” because artists now lived as princes.^[3]

Figure 2. Casa degli Omenoni, Milan, Leone Leoni, 1562

In contrast, Michelangelo lived a much less opulent lifestyle than Leoni and was known constantly complain about being poor. ^[4] Despite living in relatively conditions, when Michelangelo died in 1564, he left more liquid wealth behind than it cost the Medici family to purchase the Pitti Palace 15 years earlier.^[5]

Figure 3. Pitti Palace, Florence, 1548

To start our historiographical investigation into what makes success in the artist’s life, I’d like to begin with writings of Georgio Vasari, which form a foundational cornerstones of modern western art history. In his Lives of the Artists (Vite), first published in 1550 and revised in 1568, Vasari devotes the bulk of the book discussing the lives of 250 artists and their work. Ernst Gombrich calls the Vite “perhaps the most famous and, even today, the most read work of the older literature of art” and notes “even today art history remains under his spell”^[6]

In discussing the Vite, it is important to note that it does not seek to be mostly a work of history or a work mostly about a theory of art.^[7] The Vite contains a few sections in which Vasari introduces overarching themes, but the bulk of the Vite are stories about individual artists and their works. It is a book that is meant to be pedagogical^[8] about how an artist should live as well as theoretical in how art has developed. In his pedagogic stories, Vasari might make an aside to us, the reader, or he might have one of his characters make a point he wishes us to know.

What Vasari engages in the Vite is known as “common-sense talk,”^[9] where aphorisms are put forward that can often disagree with each other, much as proverbs from the bible. As example, using aphorisms not found in the Vite, an aphorism such as “absence may make the heart grow fonder” might be made at one point, but the statement “out of sight, out of mind” might be made at a different point. In addition, Vasari, like his fellow humanist colleagues in the early modern age, also heavily borrowed terms and rhetorical conventions from the ancients. The result is that sometimes we are not able to tell if what Vasari is telling us is more formula than fact. In speaking about a humanist writer in Vasari’s time, Michael Baxandall notes “It is true that he would hardly have said these things if he had thought them obviously untrue, but they are something less than propositions springing direct from experience.”^[10]

Given that the Vite was never intended to be a historical work as we know one and sometimes rhetorical formula might overshadow facts as known to Vasari, any attempt to gain fine-grain insight into Vasari’s thought needs to be tempered with a high degree of caution. With these caution klaxons ringing in our ears, let us now ask “what are the ingredients that lead to riches in artists’ lives in the Vasari’s Lives of the Artists?”

To begin, Vasari wants us to think that while many artists are poor,^[11] it makes for a poor artist if she above all else seeks to become rich. Vasari writes:

And to tell the truth, those artists rarely succeed in being excellent whose final and principal aim is gain and profit not glory and honor, even if they have admirable talent. (Life of Rustici)^[12]

Vasari, like the artist/historian Leon Battista Alberti a century before,^[13] argues that a lust for money inhibits the artist to express herself with freedom and therefore inevitably results in a stunted product. He proposes instead that the artist should begin by seeking both glory and honor and that riches will likely follow.^[14] Fittingly, this very proposition is echoed in the conception and gestation of the Vite itself. In 1647, Paoli Giovi, a friend of Vasari, wrote to Vasari to encourage Vasari’s efforts on the Vite. Giovi suggests that the work will bring Vasari fame and riches and notes that “gain comes from praise, but praise does not come from gain.”^[15]

While Vasari is clear that money should not be the guiding ambition for the artist, he also wants us to believe that a lack of money and just rewards is injurious to art in general and the artist in particular.

It may be believed and therefore affirmed that, if just remuneration existed in our century, even greater and better works than the ancients ever executed would, without a doubt, he created. But being forced to struggle more with Hunger than with Fame, impoverished geniuses are buried and unable to earn a reputation (which is a shame and a disgrace for those who might be able to help them but take no care to do so). (Preface to Part Three)^[16]

Beyond wealth, then, what would have constituted success during the life of an artist for Vasari? I believe he would have said technical/aesthetic excellence, ingenuity, virtue, honor and fame. But for Vasari, there is no absolute formula that will guarantee success. Rather than suggest an absolute path for success, Vasari seeks to understand what environmental conditions that might lead to success. One of those critical conditions is the artistic competition. That the artist be aware of and strive to surpass her past and present peers’ technical achievements.

Donald Preziosi, in highlighting this dynamic, notes that Vasari’s Vite

…constituted a systematic attempt to account for the apparent contradictions in the relativity of artistic reputation—the fact that artists could be considered justly great at a particular time and place even though their accomplishments might be seen by later generations, and with equal justification, as less great or as artistically incomplete.^[17]

Preziosi points to these words from Vasari from the Preface to Part One of the Vite to support this point.

As the men of the age were not accustomed to see any excellence or greater perfection than the things thus produced, they greatly admired them, and considered them to be the type of perfection, barbarous as they were.

While Vasari states that artistic achievement can be measured by a scale set by the efforts of past and present achievements,^[18] Vasari does not suggest that all is relative when it comes to what make great art. Indeed, an artist’s success is also the result of divine providence. By example, we can look at the artist Vasari considers to be the most successful and perfect artist of all time – Michelangelo.

In describing the birth of Michelangelo, the exemplar of the great artist, Vasari in his 1568 Vite writes that

There was born a son, … to Lodovico di Leonardo Buonarroti Simoni…near the Sasso della Vernia, where S. Francis received the Stigmata… which son he gave the name Michelagnolo … inspired by some influence from above… he wished to suggest that he was something celestial and divine…^[19]

Figure 4. Creation of Adam, Sistine Chapel, Michelangelo 1508-1512

In this passage, Vasari tells us that Michelangelo’s father was divinely inspired by God to choose the name Michelangelo for his son. A name that literally means “messenger who resembles God.”^[20]

Vasari’s writings not only point to an extra-worldly nature for Michelangelo at birth^[21], his writings also point to a heavenly nature in death. As with the legends of saints^[22] that Vasari would have had heard countless times, Vasari records that after Michelangelo’s death, and his body laid in a coffin for twenty-two days, “we found it so perfect in every part, and so free from any noisome odor…as if he had passed away but a few hours before.”^[23]

Nor was Michelangelo the only artist Vasari attributed divine providence to, Cimabue, Giotto and Leonardo, among others, enjoy that distinction as well. While Vasari certainly looked to the ancients as an exemplar for the Vite, it would be hard to suggest that his use of the term “divine” was merely formulaic. Indeed, Ernst Kris and Otto Kurz state that Vasari so often correlates the hand of the artist with the hand of God and proposed that the Vite is as much a theological work as a historical work.^[24]

In Vasari’s theology, divine providence might play a part in an artist’s success, but it does not determine the artist’s success. Rather than relying on some form of predeterminism, Vasari’s grants artists a freedom of agency to use their talents to gain fame and its consequent riches or to act without virtue and never attain fame. To Vasari, while talent might be divinely planted, the fruits gained from this talent depend on a soil enriched by the virtuous efforts of the artist.^[25]

However, there is more than virtue and hard work that leads to success. Lurking in the background of the Vite’s theology is also the idea that fortune, chance, plays a role in artistic success. Machiavelli’s Prince, preceding the Vite by about 35 years, explicitly writes that one half of person’s achievements can be attributed to fortune and one half can be attributed to virtú.^[26] While Vasari does not offer such a definitive formula for the role of chance, fortune undoubtedly plays an important role in the Vasari’s core conceptions of artistic success.

Consider the frontispiece of the 1550 Lives shown in Figure X:

Figure 5. Title page of first version of Vasari Lives, 1550

On this prefatory page, we see many of the themes contained in the Vite. The Vite’s civic chauvinism is displayed in portraying the city of Florence, where god saw fit to provide the soil for the Renaissance, and the Medici crest reflecting whom the book is dedicated to is displayed on top. On the right we see Apollo looking to the left at Fame, who is holding a torch. On the plinth upon which Fame is supported is a ship with full sails sitting on top of a turtle. This device is not only a reference to Cosimo I^[27] the duke to whom the Vite is dedicated to, it is also can serve as a reminder of good fortune, which is also frequently portrayed as a boat with full sails.^[28]

In general, rather than positing fortune as a force that leads to fame, Vasari and his contemporaries tended to portray fortune more as a condition that needed to be conquered in order to achieve fame.^[29],[30] This idea that the vicissitudes of fortune are a force that needs to conquered is beautifully illustrated by Vasari in the ceiling fresco that he completed by 1548 for his own house in Arezzo.

Figure 6. Virtue struggling with Fortune and Envy, Giorgio Vasari, Casa da Vasari, Arezzo, 1548

Vasari wrote that in the ceiling to this room dedicated to the theme of Fortune:

I did … a large painting in the middle, that contains life-size figures of Virtue with Envy, under her feet, and Fortune, gripped by the hair, while she beats both. A circumstance that gave great pleasure then is that in going round the room Fortune at one place seems above Envy and Virtue, and at another Virtue is above Envy and Fortune, just as it is often the case in reality.^[31]

This opportunity to see Virtue triumphing over Fortune and Fortune triumphing over Virtue in equal measure as one walks around the room visually echoes Machiavelli’s earlier written suggestion that Fortune and Virtue have equal contributions to a person’s achievements.^[32] Significantly, by having Virtue hold onto a lock of Fortune’s hair, Vasari also suggests that Fortune can sometimes bring a favorable wind, and when that happens, one needed to grasp and use the force of good fortune.

In the case of the artist Pinturicchio, Vasari notes

Just as there are many aided by Fortune who have not been endowed with much talent, so, on the contrary, there are countless talented men persecuted by adverse and hostile Fortune. Thus, it is common knowledge that Fortune adopts as her children those who depend upon her without the help of any talent, for she likes to raise up some with her own favour who would never have been recognized through their own worth. This is what happened with Pinturicchio from Perugia, who, even though he executed many works and was assisted by various people, none the less enjoyed a much greater reputation than his works deserved.^[33]

The Standing Ovation Model

To further investigate Vasari thoughts on success in artist’s lives, I’d like us to think about a conceptual model that social theorists call the Standing Ovation Problem.^[34] In this model, we are in a theater where a performance has just ended and asked if the performance will be given a standing ovation. For the sake of our analysis let us equate a standing ovation with fame and riches in an artist’s life.

I suggest that Vasari argues that it is the quality of the performance itself that determines whether a standing ovation will be awarded. That each actor’s performance is based on the artist’s competition with the other artists on stage that night as well as with other artists whose performances from the past they have seen. Indeed, as previously noted, Vasari sees this form of referential competition as a vital and indispensable force that furthers the history of art and nurtures artistic progress.^[35],[36]

Based on his historic mindset and essentialist aesthetic^[37], I doubt that Vasari would fully embrace the idea that the audience for the performance also acts in such a self-referential way. As already noted, Vasari would be willing to say the audience measures quality based on the performances they have seen. But beyond this principal, it is key to note that while performers compete with each other in Vasari’s writings, Vasari does not offer audience members the ability to cooperate with each other in deciding whether to give a performance a standing ovation. To Vasari’s the artists are given more freedom and agency to interact in more complex ways than is granted the audience.

Vasari’s almost sole focus on the performer in his seminal writings resulted in a focus that dramatically tilted the biases of art historians for centuries afterward to look more to the performer than the audience in explaining artistic success. Catherine Soussloff observes:

Abby Warburg recognized the reliance of art historians on this primary literature that he criticized. He said that “our ordinary, enthusiastic, biographically orientated history of art” comes from Renaissance biographies by Vasari and others, resulting in a reliance on individual genius and “hero-worship” that he saw as particularly “peculiar to the propertied classes, the collector and his circle.^[38]

I suggest that that the degree of focus on the artist as myth has introduced a bias into art history, a bias that has prevented us from a more full and nuanced understanding of what leads to artist success. Pierre Bourdieu notes:

The ’charisma’ ideology … is undoubtedly the main obstacle to a rigorous science of the production of the value of cultural goods. It is this ideology which directs attention to the apparent producer, the painter, writer or composer, in short, the ’author’, suppressing the question of what authorizes the author… The question can be asked in its most concrete form: who is the true producer of the value of the work – the painter or the dealer? ^[39]

In short, what makes reputations is not . . . this or that ‘influential’ person, this or that institution, review, magazine, academy, coterie, dealer or publisher; it is not even the whole set of what are sometimes called ‘personalities of the world of arts and letters’; it is the field of production, understood as the system of objective relations between these agents or institutions and as the site of the struggles for the monopoly of the power to consecrate, in which the value of works of art and belief in that value are continuously generated.^[40]

Bourdieu goes on to brilliantly summarize this messy and self-referential process as “social alchemy.” Elizabeth Honig describes a very similar dynamic to Bourdieu’s operating in early modern Europe.

This second site of value-creation, which I will call the honor system … comes to encompass forms of discourse that evolvè in the Renaissance to validate the status of the artist, but it also incorporates complex modes of social behavior that are in themselves means of negotiating status. The objects that circulate in this system cannot and indeed must not be translatable into a purely monetary value, for they are signifiers of power, honor, and obligation to those who exchange them.^[41]

Against the proposition that it is hero-artists work itself that primarily accounts for artistic success, contemporary social theorists suggest that it is instead, other people influencing each other. The sociologist Duncan Watts notes:

…social epidemics require just the right conditions to be satisfied by the network of influence. And as it turned out, the most important condition had nothing to do with a few highly influential individuals at all. Rather, it depended on the existence of a critical mass of easily influenced people who influence other easy-to-influence people. When this critical mass existed, even an average individual was capable of triggering a large cascade—just as any spark will suffice to trigger a large forest fire when the conditions are primed for it.^[42]

Challenges to Narrating Success

To further understand how the narrative approach and implied causalities of Vasari’s Vite might impose a bias that posits artistic success as mainly a property of the performance, it is useful to contrast the findings of other scholars examining the phenomenon of success as a social outcome.

Duncan Watts, in a series of experiments, created a number of online “worlds” that he populated with different audiences, with over 14,000 subjects in total.^[43] He asked each of the audiences to listen to and rate songs and then displayed an updated average rating for each song. What he found at the end of the experiment was that the best rated songs in each of the worlds greatly diverged from each other (as well as from a control world where the audience was asked to rate songs but did not see what each other’s ratings were). For example, a song that was ranked number one in one world, was ranked 40^th in another world.^[44] Watt’s conclusion is that allowing his audience to see each other’s ratings introduces a measure of randomness based on the order in which each member rates each song.^[45] That if the experiment were run over again with the exact same audience and the exact same songs for the exact same worlds, the outcomes in each world would vary wildly if the order in which each audience member rated the song were changed. Artistic success is dependent on the path of how the artist perceptions are shared by the audience.^[46]

Figure 7. When subjects saw other subjects ratings of songs, they were much more likely to listen to the better rated songs.

Watt’s findings, that shared audience opinion is not just shaped by the artist’s performance, but also by the order in which opinions are shared offers a valuable methodological insight for art historians. It calls for the historian to not just look at patron/artist itself, but also how patrons related to other patrons and society at large.

In other work that this author has done, the self-referentiality of the audience in determining artistic success was further highlighted. Using a Getty Institute database of prices paid during the 17^th century in Rome by patrons directly to artists, I found that artist career earnings were much more impacted by the unique number of patrons the artist captured than the sheer number of commissions gained or even the number of years the artist worked.^[47]

This finding suggests that that “Artists carry with them in their prestige package a history of past relationships…. Present-day status is based on a position within a web of ties and also has embedded within it the history of past positions.”^[48] At least in 17^th century Rome, career rewards to an artist’s monetary value lay in their ability to bridge and connect patron networks to each other.

In a recently published study by Samuel Frailberger, et.al, the career trajectories of about one-half million artists across twenty-three thousand institutions over 26 years were examined.^[49] The findings dramatically highlight the idea of path dependence, that where an artist starts their career has a dramatic impact on career success. For example, artists that began exhibiting in low prestige galleries showed a much greater likelihood of dropping out of the market. Artist that began their careers with high prestige galleries are much more likely to maintain showing at high prestige galleries.

Figure . Artists beginning with more prestigious showings less likely to drop out of art market

Figure 9. Artist beginning with more prestigious showings much more likely to end with prestigious showings.

Self-Referentiality as Art

In 1970, the artist Hans Haacke created a pioneering art exhibit that exactly highlights the nature of audience self-referentiality in art.^[50] Visitors were encouraged to interact with a computer keyboard, very much a novelty at the time, and enter data about themselves such as age, gender and opinions on matters such as whether marijuana use should be punished. Their answers were instantly tabulated and shown on a large screen on the wall of the installation. According to Haacke, “the visitors, in effect, were producing a collective self-portrait in a participatory and self-reflective process.”^[51]

This paper argues that exactly this dynamic that underlies artistic success. It is not just the work of the artist that drives success, success is, instead, the product of its reception. As Albert-László Barabási summarizes:

Your success isn’t about you and your performance.

It’s about us and how we perceive your performance.

Or, to put it simply, your success is not about you, it’s about us.^[52]

Bibliography

Albert, Paul. “Art Market Dynamics in Seicento Rome.” Accessed December 3, 2018. http://visualhumanist.com/academic/81/.

Alberti, Leon Battista, and Cecil Grayson. On Painting and On Sculpture: The Latin Texts of De Pictura and De Statua. Phaidon, 1972.

Barabási, Albert-László. The Formula: The Universal Laws of Success. Little, Brown and Company, 2018.

Barolsky, Paul. “The Theology of Vasari.” Notes in the History of Art 19, no. 3 (2000): 1–6.

———. “What Are We Reading When We Read Vasari?” Notes in the History of Art 22, no. 1 (2002): 33–35.

Billig, Michael. Arguing and Thinking: A Rhetorical Approach to Social Psychology. Cambridge University Press, 1996.

Bonus, Holger, and Dieter Ronte. “Credibility and Economic Value in the Visual Arts.” Journal of Cultural Economics 21, no. 2 (June 1, 1997): 103–18. https://doi.org/10.1023/A:1007338319088.

Bourdieu, Pierre. The Field of Cultural Production: Essays on Art and Literature. Columbia University Press, 1993.

Bull, George. “Popes, Princes and Peasants: The Diversity of Patronage in Vasari’s ‘Lives.’” Journal of the Royal Society of Arts 135, no. 5370 (1987): 419–29.

Cheney, Liana De Girolami. Giorgio Vasari’s Teachers: Sacred & Profane Art. Peter Lang, 2007.

———. The Homes of Giorgio Vasari. Peter Lang, 2006.

Cioffari, Vincenzo. “The Function of Fortune in Dante, Boccaccio and Machiavelli.” Italica 24, no. 1 (1947): 1–13. https://doi.org/10.2307/476522.

Cuneo, Pia F. Artful Armies, Beautiful Battles: Art and Warfare in the Early Modern Europe. Brill, 2002.

Deborah Cibelli. “Images of Fame and Changing Fortune in the First and Second Editions of Vasari’s ‘Vite.’” Explorations in Renaissance Culture 25, no. 1 (1999). https://doi.org/10.1163/23526963-90000207.

Delancey, Julia. Review of The Wealth of Michelangelo, by Rab Hatfield. Renaissance Quarterly 57, no. 1 (2004): 207–9.

Elizabeth Honig. “The Artist: Homo Economicus/ Femina Economica.” In A History of the Western Art Market: A Sourcebook of Writings on Artists, Dealers, and Markets, edited by Titia Hulst. Univ of California Press, 2017.

Fraiberger, Samuel P., Roberta Sinatra, Magnus Resch, Christoph Riedl, and Albert-László Barabási. “Quantifying Reputation and Success in Art.” Science, November 8, 2018, eaau7224. https://doi.org/10.1126/science.aau7224.

Giuffre, Katherine. “Sandpiles of Opportunity: Success in the Art World.” Social Forces 77, no. 3 (1999): 815–32. https://doi.org/10.2307/3005962.

Goldstein, Carl. “Rhetoric and Art History in the Italian Renaissance and Baroque.” The Art Bulletin 73, no. 4 (1991): 641–52. https://doi.org/10.2307/3045834.

Gombrich, E. H., and Max Marmor. “The Literature of Art.” Art Documentation: Journal of the Art Libraries Society of North America 11, no. 1 (1992): 3–8.

Kelley Helmstutter Di Dio. “Leone Leoni’s Signposting in Sixteenth-Century Milan.” In The Patron’s Payoff: Conspicuous Commissions in Italian Renaissance Art, edited by Jonathan K. Nelson and Richard J. Zeckhauser. Princeton University Press, 2014.

Kris, Ernst, and Otto Kurz. Legend, Myth, and Magic in the Image of the Artist: A Historical Experiment. Yale University Press, 1981.

Maginnis, Hayden B. J. “Matters of Money: Vasari on Early Italian Artists.” Notes in the History of Art 14, no. 1 (1994): 6–9.

Mezzatesta, Michael P. “The Façade of Leone Leoni’s House in Milan, the Casa Degli Omenoni: The Artist and the Public.” Journal of the Society of Architectural Historians 44, no. 3 (1985): 233–49. https://doi.org/10.2307/990074.

Michael. Baxandall. Painting and Experience in Fifteenth Century Italy: A Primer in the Social History of Pictorial Style. Oxford: Clarendon Press, 1972.

“Michelangelo (given Name).” Wikipedia, October 23, 2017. https://en.wikipedia.org/w/index.php?title=Michelangelo_(given_name)&oldid=806590218.

Page, Scott. “The Standing Ovation Problem – Miller – 2004 -.” Complexity, 2004. https://onlinelibrary-wiley-com.mutex.gmu.edu/doi/abs/10.1002/cplx.20033.

Pon, Lisa. “Michelangelo’s Lives: Sixteenth-Century Books by Vasari, Condivi, and Others.” The Sixteenth Century Journal 27, no. 4 (1996): 1015–37. https://doi.org/10.2307/2543906.

Preziosi, Donald. The Art of Art History: A Critical Anthology. Oxford University Press, 2009.

Rubin, Patricia Lee, and Maurice Rubin. Giorgio Vasari: Art and History. Yale University Press, 1995.

Salganik, Matthew J., and Duncan J. Watts. “Web-Based Experiments for the Study of Collective Social Dynamics in Cultural Markets.” Topics in Cognitive Science 1, no. 3 (July 2009): 439–68. https://doi.org/10.1111/j.1756-8765.2009.01030.x.

Shanken, Edward A. “Investigatory Art: Real-Time Systems and Network Culture.” Accessed December 4, 2018. https://necsus-ejms.org/investigatory-art-real-time-systems-and-network-culture/.

Soussloff, Catherine M. The Absolute Artist: The Historiography of a Concept. U of Minnesota Press, 1997.

Spear, Richard E., Philip Sohm, Christopher R. Marshall, Raffaella Morselli, Elena Fumagalli, and Renata Ago. Painting for Profit: The Economic Lives of Seventeenth-Century Italian Painters. New Haven, Conn: Yale University Press, 2010.

Tate. “Hans Haacke; Lessons Learned, Landmark Exhibitions Issue, Tate Papers No.12 Autumn 2009.” Tate. Accessed December 4, 2018. https://www.tate.org.uk/research/publications/tate-papers/12/lessons-learned.

Vasari, Giorgio. The Lives of the Artists. OUP Oxford, 1998.

Vasari, Giorgio, and Jean Paul Richter. Lives Of The Most Eminent Painters, Sculptors, And Architects, Volume 4. Arkose Press, 2015.

Verstegen, Ian. “Vasari’s Progressive (but Non-Historicist) Renaissance.” Journal of Art Historiography; Glasgow, 2011. http://search.proquest.com/docview/1017601236/citation/35DC9B0A4E184CEEPQ/5.

Watts, Duncan J. Everything Is Obvious: Once You Know the Answer. Crown Publishing Group, 2011.

———. “Is Justin Timberlake a Product of Cumulative Advantage?” The New York Times, April 15, 2007, sec. Magazine. https://www.nytimes.com/2007/04/15/magazine/15wwlnidealab.t.html.

1. Pierre Bourdieu, The Field of Cultural Production: Essays on Art and Literature (Columbia University Press, 1993), 79. ↑

2. Michael P. Mezzatesta, “The Façade of Leone Leoni’s House in Milan, the Casa Degli Omenoni: The Artist and the Public,” Journal of the Society of Architectural Historians 44, no. 3 (1985): 233, https://doi.org/10.2307/990074. ↑

3. Kelley Helmstutter Di Dio, “Leone Leoni’s Signposting in Sixteenth-Century Milan,” in The Patron’s Payoff: Conspicuous Commissions in Italian Renaissance Art, ed. Jonathan K. Nelson and Richard J. Zeckhauser (Princeton University Press, 2014), 149. ↑

4. Giorgio Vasari, The Lives of the Artists (OUP Oxford, 1998), 475. ↑

5. Julia Delancey, review of The Wealth of Michelangelo, by Rab Hatfield, Renaissance Quarterly 57, no. 1 (2004): 208. ↑

6. E. H. Gombrich and Max Marmor, “The Literature of Art,” Art Documentation: Journal of the Art Libraries Society of North America 11, no. 1 (1992): 5. ↑

7. Paul Barolsky, “What Are We Reading When We Read Vasari?,” Notes in the History of Art 22, no. 1 (2002): 33–35. ↑

8. Gombrich and Marmor, “The Literature of Art,” 5. ↑

9. Michael Billig, Arguing and Thinking: A Rhetorical Approach to Social Psychology (Cambridge University Press, 1996), 236. ↑

10. Carl Goldstein, “Rhetoric and Art History in the Italian Renaissance and Baroque,” The Art Bulletin 73, no. 4 (1991): 643, https://doi.org/10.2307/3045834. ↑

11. Richard E. Spear et al., Painting for Profit: The Economic Lives of Seventeenth-Century Italian Painters (New Haven, Conn: Yale University Press, 2010), 4. ↑

12. Patricia Lee Rubin and Maurice Rubin, Giorgio Vasari: Art and History (Yale University Press, 1995), 58. ↑

13. Leon Battista Alberti and Cecil Grayson, On Painting and On Sculpture: The Latin Texts of De Pictura and De Statua (Phaidon, 1972), 67. ↑

14. Hayden B. J. Maginnis, “Matters of Money: Vasari on Early Italian Artists,” Notes in the History of Art 14, no. 1 (1994): 7. ↑

15. Rubin and Rubin, Giorgio Vasari, 407. ↑

16. Giorgio Vasari and Jean Paul Richter, Lives Of The Most Eminent Painters, Sculptors, And Architects, Volume 4 (Arkose Press, 2015), 283. ↑

17. Donald Preziosi, The Art of Art History: A Critical Anthology (Oxford University Press, 2009), 15. ↑

18. Ian Verstegen, “Vasari’s Progressive (but Non-Historicist) Renaissance*,” Journal of Art Historiography; Glasgow, 2011, 4, http://search.proquest.com/docview/1017601236/citation/35DC9B0A4E184CEEPQ/5. ↑

19. Vasari and Richter, Lives Of The Most Eminent Painters, Sculptors, And Architects, Volume 4, 341. ↑

20. “Michelangelo (given Name),” Wikipedia, October 23, 2017, https://en.wikipedia.org/w/index.php?title=Michelangelo_(given_name)&oldid=806590218. ↑

21. Ernst Kris and Otto Kurz, Legend, Myth, and Magic in the Image of the Artist: A Historical Experiment (Yale University Press, 1981), 59. ↑

22. Lisa Pon, “Michelangelo’s Lives: Sixteenth-Century Books by Vasari, Condivi, and Others,” The Sixteenth Century Journal 27, no. 4 (1996): 1022, https://doi.org/10.2307/2543906. ↑

23. Catherine M. Soussloff, The Absolute Artist: The Historiography of a Concept (U of Minnesota Press, 1997), 35–36. ↑

24. Paul Barolsky, “The Theology of Vasari,” Notes in the History of Art 19, no. 3 (2000): 1. ↑

25. Maginnis, “MATTERS OF MONEY,” 7–8. ↑

26. Vincenzo Cioffari, “The Function of Fortune in Dante, Boccaccio and Machiavelli,” Italica 24, no. 1 (1947): 10, https://doi.org/10.2307/476522. ↑

27. Liana De Girolami Cheney and Liana Cheney, Giorgio Vasari’s Teachers: Sacred & Profane Art (Peter Lang, 2007), 203. ↑

28. Deborah Cibelli, “Images of Fame and Changing Fortune in the First and Second Editions of Vasari’s ‘Vite,’” Explorations in Renaissance Culture 25, no. 1 (1999): 116, https://doi.org/10.1163/23526963-90000207. ↑

29. Deborah Cibelli, 117. ↑

30. Cioffari, “The Function of Fortune in Dante, Boccaccio and Machiavelli,” 4. ↑

31. Liana Cheney and Liana De Girolami Cheney, The Homes of Giorgio Vasari (Peter Lang, 2006), 137. ↑

32. Pia F. Cuneo, Artful Armies, Beautiful Battles: Art and Warfare in the Early Modern Europe (Brill, 2002), 232. ↑

33. Vasari, The Lives of the Artists, 250. ↑

34. “The Standing Ovation Problem – Miller – 2004 – Complexity – Wiley Online Library,” accessed December 3, 2018, https://onlinelibrary-wiley-com.mutex.gmu.edu/doi/abs/10.1002/cplx.20033. ↑

35. George Bull, “Popes, Princes and Peasants: The Diversity of Patronage in Vasari’s ‘Lives,’” Journal of the Royal Society of Arts 135, no. 5370 (1987): 421. ↑

36. Michael. Baxandall, Painting and Experience in Fifteenth Century Italy: A Primer in the Social History of Pictorial Style. (Oxford: Clarendon Press, 1972), 26. ↑

37. Verstegen, “Vasari’s Progressive (but Non-Historicist) Renaissance*,” 7. ↑

38. Soussloff, The Absolute Artist, 87. ↑

39. Bourdieu, The Field of Cultural Production, 76. ↑

40. Bourdieu, 78. ↑

41. Elizabeth Honig, “The Artist: Homo Economicus/ Femina Economica,” in A History of the Western Art Market: A Sourcebook of Writings on Artists, Dealers, and Markets, ed. Titia Hulst (Univ of California Press, 2017), 83. ↑

42. Duncan J. Watts, Everything Is Obvious: Once You Know the Answer (Crown Publishing Group, 2011), 96. ↑

43. Matthew J. Salganik and Duncan J. Watts, “Web-Based Experiments for the Study of Collective Social Dynamics in Cultural Markets,” Topics in Cognitive Science 1, no. 3 (July 2009): 439–68, https://doi.org/10.1111/j.1756-8765.2009.01030.x. ↑

44. Duncan J. Watts, “Is Justin Timberlake a Product of Cumulative Advantage?,” The New York Times, April 15, 2007, sec. Magazine, https://www.nytimes.com/2007/04/15/magazine/15wwlnidealab.t.html. ↑

45. Watts, Everything Is Obvious, 79. ↑

46. Holger Bonus and Dieter Ronte, “Credibility and Economic Value in the Visual Arts,” Journal of Cultural Economics 21, no. 2 (June 1, 1997): 210, https://doi.org/10.1023/A:1007338319088. ↑

47. Paul Albert, “Art Market Dynamics in Seicento Rome,” accessed December 3, 2018, http://visualhumanist.com/academic/81/. ↑

48. Katherine Giuffre, “Sandpiles of Opportunity: Success in the Art World,” Social Forces 77, no. 3 (1999): 818, https://doi.org/10.2307/3005962. ↑

49. Samuel P. Fraiberger et al., “Quantifying Reputation and Success in Art,” Science, November 8, 2018, eaau7224, https://doi.org/10.1126/science.aau7224. ↑

50. “Investigatory Art: Real-Time Systems and Network Culture,” accessed December 4, 2018, https://necsus-ejms.org/investigatory-art-real-time-systems-and-network-culture/. ↑

51. Tate, “Hans Haacke; Lessons Learned, Landmark Exhibitions Issue, Tate Papers No.12 Autumn 2009,” Tate, accessed December 4, 2018, https://www.tate.org.uk/research/publications/tate-papers/12/lessons-learned. ↑

52. Albert-László Barabási, The Formula: The Universal Laws of Success (Little, Brown and Company, 2018), 24. ↑

Paul Albert

Fahad Aloraini

5/14/2018

Abstract:

This paper addresses a topic seldom discussed in the field of Agent-Based Modeling, the idea of aesthetic judgement. We argue that in selecting underlying fitness functions for computational models, the concept of fittingness (harmony/congruency) is too often overlooked in favor of just looking at fitness (functional utility). In considering aesthetic judgement, empirical grounding for a computational model can be found in the fields of Experimental and Computational Psychology that can help us build a fittingness function for aesthetic judgements of art. After discussing this grounding, we develop a model to better understand aesthetic judgements.

Introduction:

One of the joys of being a human being is our ability to be challenged and moved by art.

Figure 1. Neanderthal cave wall “painting.” Spain. 40,800 years old.

Many thinkers have spent time asking, “what is beauty?” This is a philosophical question that, by its nature, does not lend itself to empirical study alone.

Rather than ask this daunting question, we would like to ask the question, instead, “what do people find beautiful?” This is a question that is empirical in nature and a question that we believe is approachable through Agent-Based Modeling, though we have yet to find something we would call a complete solution.

In addressing this question, many have looked to either psychological or historical factors. For example, some evolutionary psychologists have hypothesized that people find pictures of landscapes that are better suited to hunting/gathering more beautiful than pictures of landscapes that are not (Falk & Balling, 2010). Some researchers have found evidence to suggest that markers for female beauty in various periods is somewhat correlated with markers for female fertility (Cunningham, 1986). Others have found when facial features are averaged out into a composite portrait, the resulting image is found more attractive than what is actually found in the median of faces (Langlois & Roggman, 1990).

Thinkers of a more philosophical bent have suggested that beauty is found in an optimal combination of expectations and surprises and look to information theory concepts such as order and entropy to try to explain beauty (Reber, Schwarz, & Winkielman, 2004; Rigau, Feixas, & Sbert, 2008).

Many contemporary art historians, while not necessarily discounting these ideas, tend to look at a greater social/historical context to explain why a particular work of art is found beautiful. To some extent, these thinkers focus on the idea of beauty as a socially constructed phenomenon.

The goal of this project is to develop an Agent-Based Modeling to test the idea that judgement of beauty is something that, at least in part, has a social and a psychological foundation.

To begin this journey, we’d like to examine a basic primitive of Agent-Based Modeling (ABM), the “Fitness Function.”

In his 2012 book, Entangled: An Archeology of the Relationships between Humans and Things, Ian Hodder makes the observation that the term “fit” has two different meanings (Hodder, 2012). On the one hand, it can describe what we believe is a traditional ABM approach, the fitness of a thing as measured by its utility and effectiveness. On the other hand, the term “fit” can also describe how well something fits in with everything else; it’s congruency and harmony with its overall environment; it’s “fittingness.”

Consider, for example, the scientific theory of Darwinism. Many have suggested that the test for a new scientific theory is based on a specific fitness function. If the theory can explain a wider scope of phenomenon more simply than prevailing theories, the new theory deserves to be adopted. Darwinism did explain a wider scope of phenomena more simply that previous theories. In this sense, it fulfilled a utility fitness function and was adopted (Hodder, 2016, p. 90).

However, Hodder also asks if Darwinism would have been adopted if it did not also display a fittingness with the overall intellectual milieu. In particular, the intellectual infrastructure being asked to incorporate Darwinism had already incorporated new harmonious theories in many other fields such as physics and geology. Like Darwinism, these new theories offered an empirical deterministic underpinning of natural phenomenon that had previously only been thought of mostly in theological terms.

Hodder argues that Darwinism certainly offered a more “fit” theory in its explanatory power but would never have been adopted if did not also offer a “fittingness” with the other theories of the time.

As another example, while a hydraulic drill can do a much better job than a hand axe in shaping stone, the hydraulic drill could never have been adopted in Neolithic times. It provides a better fit to the specific task, but not a better fittingness to the environment.

It is interesting to note that Hodder describes “fittingness” as a phenomenon he calls “nested” (Hodder, 2012, p. 114). Hodder suggests that there are multiple levels of fittingness that operate independently and reflexively in exactly the same way that Barbasi describes a scale free network as operating (Barabasi & Bonabeau, 2003).

There are, indeed, well known examples of the concept of fittingness being used to drive a computational model. For example, there are several cultural diffusion models where cultures that are more similar to each other are more likely to interact with each other and modify each other. (Axelrod, 1997; Epstein & Axtell, 1996, p. 74).

Nonetheless, there are many phenomena where fittingness could play a significant role but appears not to have been considered in many of the ABM efforts we’ve seen. For example, there are many companies who have seen their stock market value soar merely by issuing a press release discussing already disclosed initiatives along “hot technologies” such as java, the cloud, cyber-currrency, etc. Nothing about the companies’ fundamentals have changed, yet, by highlighting the company’s inclusion in a field of high flying stocks, the market drives the stock up. As another example, in organizational theory, the culture of an organization has often been called out as a driver of organizational behavior. We suggest that in looking to organizational culture as a salient factor influencing outcomes, we are in effect considering the idea of “fittingness.”

Several scholars have suggested that the traditional economic ideal of a rational actor needs to be rethought (Beinhocker, 2006). We hypothesize that, when we talk about human behavior, there are many instances where a thing’s fitness in terms of rational functional utility is somewhat hard to immediately discern in all the noise of everyday life. When this is the case, we believe that people turn to heuristic biases that are based much more on fittingness than fitness. Indeed, in the course of everyday human life, perhaps more decisions are made based on a thing’s fittingness than a thing’s fitness.

The more we’ve examined this idea of fittingness, the more convinced that we are that it deserves further attention in ABM. In some cases, rather than seeking fitness function to drive models, the field ABM might generally profit from seeking fittingness functions instead. In other cases, ABM might profit from more consciously informing a fitness function with a fittingness function as well.

As a philosophical aside, perhaps in the rush of prevailing scientific paradigms to embrace empiricism and its accompanying focus on functional utility, the pendulum has swung too far away from questioning a thing’s fittingness. In a very real sense, much of recent social science directions have been a movement to look also at the fittingness heuristics deployed in everyday life.

Empirical Experimental Psychology Grounding for Effort

The empirical grounding we look to for our effort is known as the Mere Exposure Effect (Zajonc, 1968). This is one of the most consistently documented findings in experimental psychology where it has been found that people’s liking for something increases with their exposure to it (Bornstein, 1989).

In other words, with some salient exceptions, the greater a thing’s fittingness with people’s memory, the more that thing is liked.

While we would never argue that liking a painting more is by definition, judging that painting more beautiful, we would argue that there is a significant correlation.

As an inspiration for our project, we look to the work of James Cutting from Cornell (Cutting, 2003, 2007, 2017). James Cutting sought to better understand how artistic cannons are maintained and chose to look at Impressionistic Art. He started with a sample set of pictures and then combed the 6,000,000 books in the Cornell library to determine the frequency with which each picture was replicated in print (Cutting, 2003). He believed that the more reproduced a work was in print, the greater was the likely mere exposure to that picture in the general culture.

It is important to note that Cutting’s efforts likely were not just capturing the prevalence of specific entire images in his subjects’ collective memory base. Because many of the pictures studied were iconic, those pictures could have been copied, in whole or part, by other artists and replicated into our culture. Further, the pictures themselves were products of artists who in turn were influenced by the composition and styles of other artworks that might still echo in our culture. Some of the echoes might remain fairly recognizable, but then again, might collide with other echoes or cultural structures and morph into a whisper of what they once were or become a shout of something significantly different.

The interplay of artistic creations is just one example of a class of phenomena where everything is influenced by everything else. It is a big beautiful reflexive mess.

For his first test, Cutting asked students to compare a set of two pictures and chose which one they like best. What he found, is that, consistent with other Mere Exposure Effect studies, the most published picture tended to be the most favorite.

Below are two of the pictures that were used in the test.

Figure 2. Sample pair of paintings used by Cutting

These two pictures are both by Paul Cezzane. In Cutting’s research, the picture on the right was found to be reproduced in literature 165 times while the picture on the left was reproduced only 30 times. When shown side by side, the test subjects preferred the picture on the right by a 65% margin.

Overall, consistent with other tests examining the Mere Exposure Effect, Cutting found that 59% of the variation in aesthetic judgement could be explained by looking at the logarithm of the frequency of base exposure to the picture (Cutting, 2003).

However, just using this test alone presents a problem. Certainly, the subjects judged the most published artwork as the most beautiful. But, that finding alone does not sufficiently prove exposure is what caused these judgements. Perhaps the reason these artworks were the most published to begin with was because they were somehow the most beautiful (cum hoc ergo propter hoc)?

To test for just this possibility, Cutting proceeded to do another test with a different cohort of subjects. In this test, Cutting somewhat randomly exposed a classroom of 116 subjects to the same set of images. However, over the course of a semester, Cutting exposed his subjects to the least published images four times more than the most published images. At the end of the semester, Cutting found his subjects now preferred the more published images relatively less often, with a correlation of only 48% to the logarithm of base exposure.

For 41 of the 51 pairs, in fact, the preference for the most published image went down. In the case of the Cezanne pictures above, the aesthetic preference for the most published image (the one on the right) went from 65% to 63%. In one case, the preference for the most published image went from down by 41% after the lesser published image was shown more often to the subjects.

In other words, by exposing his subjects more often to the less published images, Cutting was able to reverse the preferences in a statistically significant way.

Both of these test findings support the hypotheses that the “fittingness” of a painting can be tested by looking to understand the frequency of that painting in the memory of subjects and that the paintings more frequent in their memory would be judged to be more beautiful.

Computational Psychology Grounding for Effort

Given that we are looking at agent memory for our model, we decided to look to the Adaptive Control of Thought – Rational (ACT-R) algorithm used in cognitive computational psychology (Anderson & Milson, 1989). Specifically, we adopted a variant of a model optimized by Alexander Petrov for computational simulations (Petrov, 2006, Equation 2).

The ACT-R algorithm takes the logarithm of the frequency of exposures and modulates this using several other variables to derive an “arousal” value. We use this arousal value to have agents choose between two paintings.

Equation 1. Petrov revised ACT-R algorithm for computational simulation

In this equation, n is the frequency of exposure, t is the time since the i-th activation and d is the decay rate of activation. For the purposes of this project, we note that the generally accepted value in the literature for d is 0.5 (Petrov, 2006), but we will test 0.4 and 0.6 as well.

Pseudo Code for Model

1. “Replicate” Cutting’s experiment (null hypothesis):
   1. Create a virtual classroom of agents (i.e., subjects)
   2. Create an inventory of paintings with a base exposure frequency for each painting (in this case, the exact names and base publication frequencies for the actual Cutting test are used to establish the base inventory)
   3. At Tick 1, find the correlation (the R value) between paintings’ base exposure frequency and the ACT-R activation value for each painting
   4. Fill each agent’s memory with the inventory of paintings and their base exposure frequencies
   5. Randomly show the agents paintings from the inventory (at an initial ratio of 1:4 times for the most published/least published painting). Allow this ratio variable to be changed through a slider (Ratio of More:Less)
   6. Have each agent update their memory and ACT-R value as paintings are viewed additional times
   7. Tick / Repeat from Step E until run is set to end.
   8. Test for the correlation (the R value) between the base exposure frequency and choice over all painting pairs. As agents are more exposed to the less published images, this correlation should go down.
   9. Make the agents heterogeneous. Give agents different frequency distributions and ACT-R values for paintings.
   10. Allow for more granular representations of paintings (e.g., characteristics and traits) (Axelrod, 1997)
2. Introduce a “tastemaker” that impacts the frequency of paintings in agent’s memories
   1. Choose an agent to be tastemaker. At setup ask the tastemaker to choose a random painting as the tastemaker’s favorite.
   2. At each tick, have the tastemaker “promote” their favorite painting to the other agents and then have that painting replicated X times in agent memory (as set by a slider, “Tastemaker Impact”)
   3. Test for the correlation (the R value) between the base exposure frequency and choice over all painting pairs
   4. Test what happens when the tastemaker only promotes various paintings among the most published works in the inventory (top quartile)
   5. Test what happens when the tastemaker only promotes various paintings among the least published works in the inventory (bottom quartile)
   6. Test what happens when the tastemaker only sporadically promotes specific artwork (this should be about the same as reducing the Tastemaker Impact slider)
3. Spatialize the model
   1. Make agents heterogeneous
   2. At intervals, decide what is the favorite painting of a neighborhood and then add an extra N copies of that painting to the neighborhood’s memories

Discussion of Findings

In its present state, the model does not predict the actual values observed in Cutting’s subject for his test where he showed his subjects less published paintings four times more than published paintings. However, there is a rough congruence in the distribution of selections.

Figure 3. Comparison of preference distributions between Cutting’s findings and model’s findings

We hypothesize that, over many runs, when agents are more exposed to paintings that had a lower base level of publication, the correlation between agents’ preferences for the paintings that were less published initially would increase. This was not, in fact, what we found.

Figure 4. The longer the model run, the higher the correlation of preferences with the most published paintings. Each dot is the result of a single run, bars indicate the mean of the runs for the independent variable.

Next, we tested for what would happen if we varied the relative ratios of exposure between lesser published and more published paintings. We hypothesized that the greater the ratio of exposure to less published paintings, the less often agents would choose the more published paintings. To our chagrin, this was not the model’s result.

Figure 5. Model results not significantly impacted by varying ratios of exposures. Each dot is the result of a single run, bars indicate the mean of the runs for the independent variable.

Next, we chose to investigate the impact of changing the decay rate of the ACT-R algorithm used by our agents. We were not sure what the effect would be. We found that varying the decay rate had little impact on the correlation of preferences.

Figure 6. Model results not significantly impacted by changing ACT-R decay rate value. Each dot is the result of a single run, bars indicate the mean of the runs for the independent variable.

Finally, we introduced the idea of a Tastemaker. The Tastemaker is an agent that randomly selects and promotes a specific painting to all other agents who then update their memory to add an additional X copies of the painting (as set by a slider). Our overall findings indicate that, for future versions of the model, we would want to have the tastemaker promote more than one specific painting and test for what would happen if the tastemaker choses less published or more published paintings to promote.

We would hypothesize, all else being equal, that over time the influence of the tastemaker would decrease the correlation between agent preferences. As shown below, that is not what we found.

Figure 7. Adding a Tastemaker results in overall higher correlations contrary to initial hypotheses. Each dot is the result of a single run, bars indicate the mean of the runs for the independent variable.

To further test the model, we examined what would happen if we varied the number of times a promoted painting was put into the memory of all the agents. This variable is termed “Tastemaker Impact.” We caution that we have work to do on this part of the model in that the tastemaker only promotes a single, randomly drawn, painting. Nonetheless, we did discover a strong impact between this variable and the correlation of agent preferences as shown below.

Figure 8. Increasing Tastemaker Impact results in overall higher correlations. Each dot is the result of a single run, bars indicate the mean of the runs for the independent variable.

Finally, we examined how both varying Tastemaker Impact and time impacted the correlation of agent preferences. For the most part, the relationship of these variables remains fairly constant with our previous findings.

Figure 8. For the most part, the correlation of agent preferences and Tastemaker Impact holds fairly constant over the time of the run. Each line represents the correlation over time for different values of Tastemaker Impact.

If you, like us, believe that the ultimate measure of an analytical effort is not the answers gained, but instead the discovery of new questions to be asked, we believe that this effort has been a resounding success. We are ready to declare victory on that front. The results that we obtained from the present version of the model diverge so much from what we would expect, we believe that even more verification of the model needs to be done and points the way for future improvements.

Limitations of Findings

Our approach has a number of constraints and limitations but has helped us to ask better questions.

1. To the extent that beauty can be found independently of something being more liked, testing for the influence of the Mere Exposure Effect has intrinsic boundaries.
2. At some point, experimental psychology has found that liking can actually decrease after some point in time with repeated exposures (Montoya, Horton, Vevea, Citkowicz, & Lauber, 2017). For simplicity, we choose not to address that in our model. I personally find it humorous that one study found, that with bad art, repeated exposures never increased liking and actually increased disliking (Meskin, Phelan, Moore, & Kieran, 2013). While these dynamics are very amenable to computational modeling, for simplicity sake, we chose not to fully try to model this dynamic.
3. If fittingness indeed drives aesthetic judgement, it should be recognized that fittingness is multi-dimensional and driven by much more than the instant feeling one gets from looking at an image (Nanay, 2017). As noted before, the human mind likes to find patterns and the patterns found by looking at art extend far beyond the image itself. Every artwork is framed by an observer-dependent context that is informed by that observer’s memory and cognitive processes.
4. For all of our efforts to posit that there is deterministic mechanism that can influence aesthetic judgements, we are willing to admit, and indeed hopeful, that beauty itself might go beyond deterministic factors.
5. Our approach considers artwork traits as “categorical.” What this means is that, for our approach, a trait with a value of 11 is as different from a trait with a value of 99 as it is with a trait of a value of 12. What if instead, things are actually more graduated, and instead of categorical values, our trait values represented “intervals?” For example, what if a trait represented a color? Agents who liked value 11 would likely be more prone to like value 12 than value 99. After a fairly lengthy examination of this dynamic, which included reaching out to math professors on campus, we were unable to come up with a satisfactory algorithm to address that dynamic. For more on this limitation, see Appendix A.

As an observation, not as a criticism, we note, for example, that in Axelrod’s iconic cultural model (Axelrod, 1997), similarity between cultures is calculated by categorical similarity. The similarity of cultures is determined by how many trait values are exactly identical. Perhaps Axelrod would have gotten different results if he measured the similarity of cultures by treating trait values as interval values instead?

1. As a continuation of the last issue, it is questionable that a painting can be reduced to a set of interval values. The authors of this paper are divided on this question. To Paul, the reflexivity inherent in art judgement between pieces of art and the multi-dimensional nature of the judgements make this effort doomed to failure. However, Paul believes that we might be able to come close enough to ask valuable questions by coming up with random values to use in representing characteristics and traits of a painting. Fahad thinks that we can actually begin with an empirical set of paintings and quantify them sufficiently for a study. Fahad reminds us that a jpeg, for example, is an example of just such a process as reducing an image quantitatively.

Conclusion

Beauty might indeed lie in the eye of the beholder. We argue that a painting’s perceived beauty is very often a function of the fittingness of that painting with the beholders memory. Using the analytical tools offered by ABM and based on grounding in Experimental and Computational psychology, this paper lays the groundwork for such an exploration.

We suggest that a fittingness function be looked at much more often in the field of ABM and hope that our efforts have helped, at least dimly, light such a direction.

Bibliography

Anderson, J. R., & Milson, R. (1989). Human memory: An adaptive perspective. Psychological Review, 96(4), 703–719. https://doi.org/10.1037/0033-295X.96.4.703

Axelrod, R. (1997). The Dissemination of Culture: A Model with Local Convergence and Global Polarization. Journal of Conflict Resolution, 41(2), 203–226. https://doi.org/10.1177/0022002797041002001

Barabasi, A.-L., & Bonabeau, E. (2003). Scale-Free Networks. Scientific American, 288(5), 60–69.

Beinhocker, E. D. (2006). The origin of wealth: evolution, complexity, and the radical remaking of economics. Boston, Mass: Harvard Business School Press.

Bornstein, R. F. (1989). Exposure and affect: Overview and meta-analysis of research, 1968–1987. Psychological Bulletin, 106(2), 265–289. https://doi.org/10.1037/0033-2909.106.2.265

Cunningham, M. R. (1986). Measuring the physical in physical attractiveness: Quasi-experiments on the sociobiology of female facial beauty. Journal of Personality and Social Psychology, 50(5), 925.

Cutting, J. E. (2003). Gustave Caillebotte, French Impressionism, and mere exposure. Psychonomic Bulletin & Review, 10(2), 319–343. https://doi.org/10.3758/BF03196493

Cutting, J. E. (2007). Mere exposure, reproduction, and the impressionist canon. Partisan Canons (Durham, NC: Duke University Press) Pp, 79–93.

Cutting, J. E. (2017). Mere Exposure and Aesthetic Realism: A Response to Bence Nanay. Leonardo, 50(1), 64–66. https://doi.org/10.1162/LEON_a_01081

Epstein, J. M., & Axtell, R. L. (1996). Growing Artificial Societies: Social Science From the Bottom Up (First edition). Washington, D.C: Brookings Institution Press & MIT Press.

Falk, J. H., & Balling, J. D. (2010). Evolutionary Influence on Human Landscape Preference. Environment and Behavior, 42(4), 479–493. https://doi.org/10.1177/0013916509341244

Hodder, I. (2012). Entangled: an archaeology of the relationships between humans and things. Malden, MA: Wiley-Blackwell.

Hodder, I. (2016). Studies in Human-Thing Entanglement. Retrieved from http://www.ian-hodder.com/books/studies-human-thing-entanglement

Langlois, J. H., & Roggman, L. A. (1990). Attractive Faces Are Only Average. Psychological Science, 1(2), 115–121.

Meskin, A., Phelan, M., Moore, M., & Kieran, M. (2013). Mere Exposure to Bad Art. The British Journal of Aesthetics, 53(2), 139–164. https://doi.org/10.1093/aesthj/ays060

Montoya, R. M., Horton, R. S., Vevea, J. L., Citkowicz, M., & Lauber, E. A. (2017). A re-examination of the mere exposure effect: The influence of repeated exposure on recognition, familiarity, and liking. Psychological Bulletin, 143(5), 459–498. https://doi.org/10.1037/bul0000085

Nanay, B. (2017). Perceptual Learning, the Mere Exposure Effect and Aesthetic Antirealism. Leonardo, 50(1), 58–63. https://doi.org/10.1162/LEON_a_01082

Petrov, A. A. (2006). Computationally efficient approximation of the base-level learning equation in ACT-R. In Proceedings of the seventh international conference on cognitive modeling (pp. 391–392). Citeseer.

Reber, R., Schwarz, N., & Winkielman, P. (2004). Processing Fluency and Aesthetic Pleasure: Is Beauty in the Perceiver’s Processing Experience? Personality and Social Psychology Review, 8(4), 364–382. https://doi.org/10.1207/s15327957pspr0804_3

Rigau, J., Feixas, M., & Sbert, M. (2008). Informational Aesthetics Measures. IEEE Computer Graphics and Applications, 28(2), 24–34. https://doi.org/10.1109/MCG.2008.34

Zajonc, R. B. (1968). Attitudinal effects of mere exposure. Journal of Personality and Social Psychology, 9(2, Pt.2), 1–27. https://doi.org/10.1037/h0025848

Appendix A: Challenges in representing artwork for modeling purposes

Suppose we have a given artwork “Y” that is represented as “01 11” and a given agent’s memory as “08 11” and “02 11.” Each artwork has two characteristics (number positions) and each characteristic is given has a trait value.

We could hypothesize that an artwork’s quantification reflects a categorical value; that 08 11 and 02 11 are both equally different from 01 11 because they both have one characteristic that is different from the referent 01 11. This would be equivalent to taking the Hamming Distance from the referent to each artwork in memory.

In theory, this approach would most closely replicate Cutting’s tests.

However, it might not be that simple. Each characteristic’s trait value might not be a unique value but rather a value that lies on a continuum. Suppose the first position represents dominant color and 01 represents the color Blue.

Repeatedly exposing agents not to the exact color Blue (01), but similar values (say 00 and 02), could increase the agent’s appreciation of “blueness.” To consider this effect, we would want to treat each trait value as an interval value versus treating each trait value as a unique categorical value.

If we wanted to treat a painting’s trait value as a categorical value, one would be tempted to look in each agent’s memory and find the modes of each characteristic.

For example, suppose Agent 5 had the following values for Characteristic 1 with the following frequencies in their memory (size 100):

Characteristic 1 Value	Frequency
11	40
02	32
01	28

Now, let’s suppose we have a new referent artwork to evaluate whose Characteristic 1 is 03.

The question to ask is do we evaluate “03” as a difference in category value, in which case “03” is equally different from all the values in memory, or a difference in interval value, in which case “03” is less different from the values “02” and “01” than from the value “11?”

Suppose, for now, we want to test our referent number for differences in interval value.

The mean of Agent 5’s memory, as shown above, is 5.28. If we wanted to consider how distant “03” is from all the values in Agent 5’s memory, it is tempting to ask how “03” compares to that value.

However, if instead, we created a row in a spreadsheet for each memory value, and then tested X values as a new referent X(0,20) and asked how distant X was from each memory in memory (absolute(reference number – memory number) and then summed these values to find Total Distance, we would find that it is not when the referent number approaches the mean of the values that Total Distance is minimized.

Figure 9. Sample value with the lowest total distance from sample distribution not arithmetic mean and not the geometric mean.

–If page is not displayed in full width, please click on title link above —

---

Paul Albert
Art History 599-1: Creating Value: Making and Selling Art in Early Modern Europe
George Mason University
May 12, 2017

---

Section I: Overview

Challenge:

Art History scholarship has reached many rich and nuanced judgements about its subject by looking through a lens based on qualitative analysis. This is a very powerful lens whose interpretive power is well proven.

However, with a few notable exceptions, Art History scholarship has not studied its subject through the lens of quantitative analysis. This is a different type of lens that can help further our understanding of the subject.

Technology has provided art historians new and exciting ways to access materials to study. Equally important, technology has also provided art historians new and exciting ways to interpret data.

In speaking about the growing field of the Digital Humanities, Johanna Drucker observed that the goal for the Digital Humanities is not just making materials available electronically but also “the study of ways of thinking differently about how we know what we know and how the interpretive task of the humanist is redefined in these changed conditions.”^[1]

The goal of this paper is to explore, for one specific topic, how both qualitative and quantitative analyses might act as foils to one another. My hope is to show how such a bi-modal approach can lead to a richer understanding of the topic and highlight promising avenues for new research.

Topic to Address:

The specific topic this paper seeks to examine is the dynamics of the Art Market for painters in Seicento Rome. As opposed to looking at a specific painter, a specific painting or a specific patron, I seek to see if we can gain a more generalized understanding of key factors driving overall earnings and overall sales for artist and patron activity reported in a specific dataset.

Dataset Foundation for Analysis:

Professor Richard Spear, as part of his seminal research on the Seicento Rome market for paintings, compiled a dataset of prices paid to painters for about 1,000 individual commissions.^[2],[3]

The dataset created by Professor Spear contains information on the painter commissioned, the patron, the price, characteristics of the works such as subject matter, title and type of painting and information about the works destination for display. To advance scholarly knowledge, the Getty Institute has provided the Spear Payments to Artist dataset online to be queried and used for further research. ^[4]

The Spear Payments to Artists (P2A) Dataset was compiled from many reference sources (see Figure 7 in this paper) that listed commissions and then adjusted prices paid to reflect prices in Roman Silver Scudi (a currency that showed little inflation over the period studied). The dataset was pruned to include only sales in the primary market, by a painter directly to a patron, and only sales by artists who resided in Rome. To capture only the dynamics of the Rome art market, works by artists who normally resided in Rome but carried out the work in a different city were excluded and works by “foreign” artists done while in Rome were included.

Due to its very nature, the P2A Dataset offers an incomplete snapshot of a specific submarket, a market for high-end paintings. Considerable qualification needs to be given to be given as to whether findings based on it are truly representative of the overall market. In Section IV, I will examine this question further.

However, as an intellectual exercise, this paper proposes to posit the dynamics found in the P2A Dataset as true in order to question what the implications of the findings would be if they were truly representative of the topic.

At worst, this analysis seeks to develop “straw man hypotheses” to test current understanding of the topic. At best, this analysis hopes to help develop a more nuanced understanding of the topic and stimulate new lines of inquiry and discussion.

Figure 1. Prices to Artists Dataset Overview – Interactive Dashboard
(highlight elements for additional information)

Figure 1, above, is an interactive dashboard profiling the Prices to Artists Dataset.  Further information about each record can be displayed by highlighting the circles which each represent a separate commission.  Data can be filtered by using the controls on the right.

Summary of Findings:

---

Key Finding 1: Total painter earnings are best correlated to the size of their patron network (the distinct count of patrons).

Key Finding 2: Total patron spending is best correlated to the overall “aura” of the painters they collected and the breadth of painters they collected.

---

These findings illuminate the social underpinnings behind the specific market in Seicento Rome covered by the dataset. These key findings will be examined in Section III of this paper. Before addressing them, let us first look at the historical context of the market dynamics for painters and patrons in Seicento Rome.

Section II: Historical Context

What makes Seicento Rome unique on the Demand Side?

Rome, compared to other cities of the same period, had a markedly different economic engine underlying its economy, the church. Unlike other cities where power concentrated based that city’s unique economic production (such as cloth manufacturing, cloth dying or banking), the amount of economic power in Rome and its distribution was largely driven by the fortunes of the church. ^[5]

Second, unlike other cities where economic power might be concentrated in one or a few noble families for a relatively lengthy period of time, wealth distribution and its accompanying ability and need to display status in Rome would change quickly and dramatically with the changing of the non-hereditary papal reign.^[6]

All told, coinciding with the period that the bulk of the dataset covers, from 1585 to 1700, Rome saw 17 different popes for an average reign of about seven years. During this period, three popes had reigns that lasted less than a year. Pope Leo XI, a Medici pope, only managed to live 29 days into his reign in 1605.^[7]

Figure 2. Patron Spending by Papal Reign – Interactive Dashboard
(highlight elements for additional information)

Patron spending, as shown in the dataset, is broken out by papal reign. What is most striking about this dashboard is how the spending by families changed based on the head that the papal tiara sat on.^[8]

The fleeting nature of papal power encouraged a “carpe diem” attitude to art patronage.^[9] This disposition is clearly seen in this dashboard. The impact of the expectation for the quick changes of papal power is noted by Francis Haskell quoting from a 1620 letter to Lord Arundel:

It [is] a strange and unnaturall thing that in that place [Rome], contrary to all others, the long life of the Prince is sayd to be the ruyne of the people; whose wealth consists in speedy revolutions, and the oft new preparations of new hopes in those that aspire to rise by new families, who… are loath to blast their future addresses by spending to court those that are despaired of.^[10]

There are a few other factors affecting the demand for painting that bear briefly mentioning about Seicento Rome. The rebuilding of St. Peters, started by Pope Julius II in 1505, was still proceeding during the Seicento. In addition, during this period, the counter-reformation was in well underway. Not only were many new churches being built to house new orders, there was an overall increased demand for new types of art to support the Council of Trent’s dictums.^[11]

Finally, compared with other periods, the bar to display patron magnificence was significantly higher than in the past. Painter’s brushes were needed not only for new chapels, but also for much larger and much more ornate palazzos.

Having briefly discussed the distinctive historical/economic factors on the demand side of the equation that characterized Rome in the Seicento, let us turn to distinguishing factors on the supply side of the equation.

What makes Seicento Rome unique on the supply side?

Certainly, Rome was very special for the educational opportunities it offered during the period. These opportunities include the many ancient roman sculptures available for study, the many public and private collections available to visit and the many opportunities to receive instruction and practice with fellow students.

Rome was also special, for the time, for the ease in which artists could come into the city and pursue the profession.^[12] Though it did have a formal Academy for distinguished painters, Rome did not effectively regulate painters from other Italian cities or foreigners from other countries coming into Rome to compete for the patron’s wallet.^[13]

Rome was indeed unique in the educational and work opportunities it offered the aspiring artitst. Indeed, for this period, Rome has been called the predominant artistic center in Europe. Rome was the city that painters in the Seicento were encouraged to go to if they wanted to get rich.^[14]

Section III: Discussion of Key Findings

Key Finding 1: Total painter earnings are best explained by the size of their patron network.

Figure 3. Artist Total Earnings Factors – Interactive Dashboard
(highlight lines and circles for additional information)

Figure 3, above, offers an interactive dashboard showing the correlation between painter Total Earnings and factors found relevant in the dataset. The dropdown box controls which factor can be used for ranking painters.

For this Figure, and many other dashboards shown in this analysis, the dashboard provides the ability to filter the data to reflect sub-markets in the dataset. For example, setting Patron Type to “Institutional” would filter the data so that only records where the Patron Type is “Institutional” are analyzed.

Values for Work Display Type (Private/Public/Unspecified) was specified for each record by Richard Spear and included in the original dataset. The field Patron Type was created by this paper’s author based on the values given in the Patron field (e.g., St. Peters would be listed as “Institutional”). As has been noted, exploring this type of distinction between patrons can offer additional insights for scholarship.^[15]

A visual analysis of Figure 3 shows that some factors can be seen to be more correlated with Total Earnings than other factors. While the eye can perceive trends between the different factors and Total Earnings, none of the factors shown offer a complete and unambiguous one-to-one correlation. To help supplement the visual correlation that the eye can find, we can statistically test for factor correlation using simple regression analysis techniques.

A regression analysis is a statistical test that examines how variations in one measure are affected by variations in another measure. Key results from the test are measures of how correlated a change in one variable is with the other variable (the R²) and how large of a change each change in one variable affects the other variable (the Slope).

Figure 4: Total Artist Earning Factors Correlation Analysis – Interactive Dashboard.
(highlight lines and circles for additional information)

Key quantitative measures found by the regression analyses can be displayed by highlighting the trend lines in Figure 4. To better aid a comparison of the factors, Table 1 lists the Correlation Strength (R²) and Slope for each factor.

Table 1. Factor Correlation to Total Artist Earnings

Factor	Correlation Strength (R2)	Slope
Patron Network Size	63%	1,376
Number of Commissions	56%	706
Patron Aura	54%	0.14
Years Worked	38%	305

Paton Network Size:

The analysis shows that variance in the Patron Network Size explained 63 % of the variance in Total Earnings among artists in the dataset. Further, for every new patron an artist sold to  career earnings are correlated to increase by 1,376 scudi.

As a correlation factor, Patron Network Size is more significant than any of the other factors examined. Indeed, while adding a new patron explained an overall career revenue increase of 1,376 scudi to a painter, the analysis shows that across the dataset, painters could only expect 706 scudi in career Total Earnings for capturing a new commission or 305 scudi for working an additional year.

The importance of the Patron Network Size as a predictor of Total Earnings highlights the social nature of the Seicento Rome market the dataset captures. The “social power” of the artist, as measured by the size of their network, explains a large amount of the artist’s economic success.

While the perceived skill of the artist was a necessary prerequisite for economic success, just as important to the artist’s wallet was the skill of the artist in building their network. It was not just what the artist was able to accomplish with their brush, it was also whom the artist knew (and the connections of the artist’s connections) that led to higher earnings.^[16]

In being incented to grow their Patron Network, artists were not only incented to think about the wishes of their patrons, but just as importantly, the audience of their patrons. While this objective might have been implicitly incented by the Seicento Rome, it has been explicitly embraced by today’s social media marketers. In contemporary social media marketing, there is a clear and stated understanding that success is not gained so much by speaking to one’s primary audience as by reaching and speaking to the audience of one’s audience. ^[17]

This implicit incentive, found in the dataset, would likely have motivated the painter’s brush not just to consider the wishes of their patrons, but also what the patron’s audience would find valuable. The painter was incented to go “viral.”

The fact that many artists of the period employed a practice of “gifting” their work is often discussed when considering how different works could be valued against one another in determining market dynamics. In many cases, the practice of gifting a painting would be made with the expectation for payment. However, in some cases, there appears to be other motivations and pure outright gifting.^[18]

In 1683, the painter Niccolò Cassana wanted to place a self-portrait in the Medici collection and sought the help of Matte del Teglio, a Florentine agent who agreed to write to Cosimo III. Teglio writes to Cosimo III “I want to quiet an ambitious painter here [in Venice] who cannot be calmed with reason…the painter… wants to send his own self-portrait because there is one in your collection by Liberi and Bombelli.” ^[19]

While Cassana might expressly be motivated in this to serve his pride, and might have indeed expected a gift in return for his painting in this case, the practice of outright gifting to build relationships was common in the period. Outright gifting of a painting was not only a way to potentially gain future sales from the recipient, but done sparingly and strategically, could have been a way to play to the audience of their patron. Being able to claim that one’s work was part of a prestigious collection would have helped build the artist’s brand. If displayed, the work would introduce the artist to new potential clients. This practice of outright gifting could be seen generally as a long game by the artist to build their social power and Total Earnings.

Primary sources of the Rome Seicento suggest a clear recognition of the artist’s need to build a brand and go “viral.” The painter Giovanni Battista Passari, when speaking about Lanfranco’s career and support from a patron in this endeavor, Passari noted at the time how necessary a “favorevole aperture” (favorable coming out) was to an artist and stated, “Truth be told, to launch one’s name it is critical to have the protection of a patron from the start.”^[20]

The wish for a “favorevole aperture” in building an anchor point for artist valuation is exactly echoed today by companies seeking to build an anchor point for their stock valuation by having a successful initial public offering.

To grow their Patron Network, artists were told it was necessary to be perceived as inhabiting the same social field as the patron^[22]. In the Giovanni Battista Armenini 1587 treatise on painting, De’ veri precetti della pittvra, Richard Goldwaithe notes that Armenini spends the bulk of the treatise focusing on how an artist should interact with his patron – “he should counduct himself in accordance with the norms of the client’s social status”^[23]

The market incentive to grow Patron Network Size would likely have been greater in Rome as compared to other cities in the Seicento. As compared to Rome, other cities had more effective guild regulations barring market entry and the concentration of economic power led often to the practice of court painters (which was much rarer in Rome).

Number of Commissions:

Interestingly, while the average price for all commissions across the dataset is 453 scudi, the regression analysis suggests that for every new commission gained, regression analysis of this factor finds that the artist could expect 706 scudi in career Total Earnings for each new commissions.

This dynamic suggests that, for the market reflected in the dataset, there was a multiplier effect to each commission as it related to career Total Earnings.  For every 1 scudo earned in a single commission, the artist would see 1.6 scudi in career earnings.

Patron Aura:

It was not just painters that sought to promote and increase a brand identity for themselves. The drive to claim distinction permeated the social fabric in Seicento Rome and was as prevalent as the air the artist breathed. Patrons also sought to establish and burnish their own brands.

While the P2A Dataset cannot capture all of the dynamics of the social game inherent in this drive for distinction, we can try to measure the Patron Aura quantitatively. For the purposes of this analysis, the Patron Aura factor is calculated as the measure of the patron’s total expenditures across the dataset. This measure is then summed for all the patrons that collected an artist and compared with the artist’s Total Earnings across the dataset.

Not surprisingly, among institutional patrons, the Vatican shows the highest Patron Aura. For private patrons, the three highest Patron Auras were the Borghese Family, the Barberini Family and the Aldobrandini Family.

In deciding the best descriptive term to use for the factor “Patron Aura,” I was influenced by the work of Walter Benjamin who speaks about the aura of an image being driven by the uniqueness of that image.^[24] In much the same way, each patron represents a unique contribution to the artist’s brand.

In examining the factors correlating with a painter’s Total Earnings, this analysis found that it is not just the size of the artist’s Patron Network that explains earnings; earnings are also highly explained by the patrons’ characteristics that the artist sold to. It is not just how many people an artist sold to, it is also whom the artist sold to.

Indeed, the analysis of the dataset shows that artist Total Earnings are more explained by the collective Patron Aura of the clients they sold to than the number of sales they made or the years that they worked. Variance in the collective Patron Aura of all the patrons an artist gained is able to explain 54% of the variance in artist Total Earnings in the dataset.

While this paper seeks to examine the relationship behind artist Total Earnings and Patron Aura, Frederico Etro used the P2A Dataset to examine, in part, the relationship between factors in determining the market valuation of specific paintings. One of Etro’s findings indicates that overall, a patron who spent relatively more on paintings would pay relatively less for a given painting, all else being equal.^[25]

This finding tantalizingly indicates that the market dynamics as shown in the dataset incented artists to provide a discount based on Patron Aura. Like the practice of gifting, this can be seen as a long game strategy to maximize Total Earnings.

Years Worked:

Years Worked is measured by looking at the years between the date of the first sale by an artist and the last sale by that artist across the dataset.

Given the specific boundaries of the dataset, work by artists done while not in Rome are excluded.  Career Years is a measure that is somewhat problematic. Notably, an artist who had a good name could come to Rome and see more earnings in a set number of years than an artist who was struggling to make their name for the same number of years and left Rome to work elsewhere.

Nonetheless, since this analysis compares, for many artists overall, Years Worked while in Rome with Total Earnings while in Rome, Years Worked is a factor worth considering and correlative. 38% of the variance in Total Earnings can be explained by the variance in Years Worked.

If we look at this factor, what is striking is how relatively less the number of years worked explains earnings than the other factors considered above for the dataset. Overall, more years of effort did not always equal more earnings.

Figure 5. Artist Years Worked and Earnings – Interactive Dashboard
(highlight elements for additional information)

Seicento Rome was host to a number of notable painters; however, it is particularly striking to note how unique Caravagio was compared to his peers. This is not only demonstrated in Figure 5 above, but also in Figure 4.

Wallet Share:

Modern business theory urges firms to focus on wallet share over market share. Selling more things to an existing network of clients is said to be more profitable than trying to build a bigger network of clients for one particular thing.

For the P2A Dataset, Wallet Share can be measured by looking at the percentage of the total Patron Purchases an artist is able to capture. Analysis of the data does show a 38% correlation between Total Earnings and Wallet Share, the same degree of correlation found for Years Worked.

However, because the dataset only contains information on the purchases of paintings, it, by definition, does not capture any other type of goods for sale by the painter other than paintings.

Artists in the dataset might have sought to capture patron Wallet Share by providing more than the painting to their patrons. A portion of the artists in the P2A Dataset were not only were painters, but also earned income from their patrons as architects, “decorators,” and “building subcontractors.”

Though the dataset only allows us to compare artist Total Earnings based on the sale of painting goods, it is important to note that many of the highest Total Earnings painters enjoyed considerable success in complementary endeavors. Indeed, it is extremely possible that achieving success as an architect and success as a painter might have played off one another as artists sought to grow the value of their network to maximize Total Earnings.

Key Finding 2: Total patron spending is best explained by the “aura” of the painters they collected and the breadth of the painters they collected.

Just as Finding 1 highlight the social factors underlying Seicento Rome art markets so to do social factors come to the foreground when looking at patron Total Spending.

Figure 6. Patron Spending Factors – Interactive Dashboard
(highlight lines and circles for additional information)

Figure 6, above, allows a visual inspection of the correlation between patron Total Spending and various factors found in the P2A Dataset. To help supplement the visual correlation that the eye can find in Figure 6, as we did above for Figure 4, we can statistically test for factor correlation using simple regression analysis techniques.

Figure 7: Total Patron Spending Factors Correlation Analysis – Interactive Dashboard.
(highlight lines and circles for additional information)

Table 2. Factor Correlation with Patron Spending

Factor	Correlation Strength (R2)	Slope
Artist aura	55%	0.11
years collecting	45%	689
Distinct artists	40%	526
Number of Purchases	32%	245

Artist Aura:

In determining the market value of any given work of art, the artist’s name has often been mentioned as a key factor driving value. Artist Aura, for this analysis, is the measure of the total earnings for an artist across the dataset. 61% of total Patron Spending variation can be explained by the variation in Artist Aura.

In other words, if we sum up all the artist Total Earnings for the artists that were collected by a patron and then compare that with the total spending for the patron, patrons who spent more tended to collect artists that earned more.

Certainly, there is a bit of a recursive dynamic in this correlation. Artists in the dataset who sold the most to patrons who spent the most would have tended to earn more than their peers. However, there are other alternative market scenarios that could have been possible. For example, if patrons were looking merely to maximize the the number of artists they collected or the number of works collected, this correlation between artist Total Earnings and patron Total Spending would not be as strong in the dataset.

As with the Patron Aura factor examined in Finding 1, the importance of the Artist Aura as an explanatory factor highlights the social nature underlying the dynamics of the Seicento Rome art market. In the case, variance in Artist Aura, as a measure for how valued an artist is by the patron’s peers explains 55% of the variances in patron overall Patron Spending.

Years of Collecting:

Patron Years of Collecting was calculated by looking at the time span between the year of the first purchase and the last purchase in the dataset. Across the dataset, 45% of the variation in overall Patron Spending is explained by variations in Years of Collecting.

Because the overall analysis looks at both Institutional Patrons (such as the Vatican with 112 Years of Collecting) and Private Patrons (such as the Fabrizio Valguarnera with 18 Years of Collecting), grouping these patrons together might be problematic. However, in fact, there remains a striking consistency when these sub-markets are compared. For Institutional Patrons, 42% of the variation in Total Spending is explained by variations in Years of Collecting, while for Private Patrons, 46% is explained.

Count of Distinct Artists:

The analysis of the dataset shows that 40% of the variation in patron Total Spending is explained by the Count of Distinct Artists.

Works of painting in the Seicento period were characterized by a play of the overall parts to the whole. Focus was not on a single figure in the painting, but by the interplay of all the figures together. As opposed to a single musical note, paintings of the Seicento strove to create a chord.

Picture 1. Pietro da Cortona, Palazzo Barberini, Rome, 1633.

In a meaningful way, it seems that this visual aesthetic embodied in singular works was echoed on a larger scale by patrons themselves as they sought distinction in the collection and display of multiple works and multiple artists.^[26] To the patron, the whole ensemble was likely to seen to be greater than the sum of the parts.^[27]

In a sense, to differentiate themselves in the market, Seicento painters might have sought to build for themselves by striking a distinguishable note. In contrast, to distinguish themselves socially, patrons seem to have sought to differentiate themselves by displaying a chord comprising a harmonious collection of these notes. This dynamic suggests patrons would not only seek to display a wide variety of painting styles and types, but, inasmuch as the artists’ names were salient characteristics of paintings, collect different artists.

Indeed, the value of any particular work purchased by any particular patron was undoubtably influenced by the other works/painters the patron had already collected.

Number of Purchases:

While the Number of Purchases does correlate with the Total Spending by a patron, explaining 32% of the variation between factors, it is notable that a stronger correlation was not found such as with other factors examined. The relative weakness of this correlation highlights the distinct nature of the Art Market as found in the dataset.

While one record might be for decorating an entire gallery or chapel and another purchase might be for a single easel painting, if the Art Market shown in the dataset acted like a pure commodity market (for example, grain), overall the number of purchases would likely have a much stronger correlation with total spending in the dataset.

Section IV: Dataset Sources and Representativeness

Dataset Sources

The Spear/Getty Prices to Artist Dataset presents data that was preserved in by history, in sources such as exhibition catalogs and books about painters and patrons. All told, for the 818 records in the dataset that were analyzed, data from over 149 different historical records were recorded for 129 artists.

Figure 8. Dataset Reference Sources – Interactive Dashboard

Representativeness of the Dataset

As remarkable as the effort to collect the data in the dataset was, a dataset built on reference sources such as books focused on particular artists and patrons or catalogs from museum exhibitions is inherently bounded. The sources from which the dataset is drawn means that the dataset only contains works and artists that history has deemed worth noting. This results in a dataset intrinsically skewed to reflect the high end of the market for paintings in Seicento Rome, a specific sub-market of a larger market.

Many of the findings in this analysis of the dataset would likely be different for other sub-markets in the period. For example, artists on the low end who sold in the open market would likely have seen their Total Earnings much more correlated with the Number of Sales or Years of Work than those contained in the dataset.

Much of Art History is built on the qualitative judgement of scholars piecing together isolated historical anecdotes and building on the work of one another. The dataset examined in this analysis is at worst, just an incomplete set of individual historical anecdotes. At best, however, by collecting a large number of historical anecdotes and allowing for a quantitative analysis lens to be used, the dataset provides a qualitatively different type of research resource whose sum offers more insight than its individual parts.

As a quick test for how complete the data set records are for the artists that it tracks, we can do a bit of back of the envelope estimation. In 1624, Guilio Mancini estimated that a good painter in Rome could earn between 3 to 6 scudi per day. ^[28] This would equate to between 750-1,500 scudi a year.

Looking at the Spear/Getty data set and taking the number of years between each artist’s first commission and last commission recorded, we find the 129 artists in the dataset show a combined work span of 1,633 years and average annual earnings of 226 scudi a year.

If Mancini’s range is accurate, this would suggest that the payments to artists recorded in the dataset would have accounted for 15% to 30% of the total earnings for the work years covered and the artists listed. While not the desired 100%, this suggests the dataset can be seen as fairly representative of the bigger picture.

In the final analysis, we must take as a given that the dataset offers just a fragmentary view of a particular sub-market.  Ultimately, the most important test of the dataset is how well it can be used to inform our understanding of “how we know what we know” ^[29] and point to new avenues of inquiry.

Section V: Conclusion

The findings of this analyses of the high-end market dynamics in Seicento Rome between overall painter earnings and overall patron spending illustrate the importance of social networks to the Art Market. Painters maximized their earnings the most by building a larger network of patrons. Patron who spent the most focused on artists that their peers found the most economically worthy.

Qualitative analysis should never be given a second place to quantitative analysis in the practice of Art History. However, by using the new tools for quantitative analysis available to the scholar, we can gain a richer understanding of the subject and possibly reach new understandings and discover new directions for research.

Appendix: Methodological Considerations

Dataset Modifications:

• Excluded records where the patron or the date of work was not specified. Since many correlations were based on these variables, I considered it the most conservative way to get the cleanest data possible. The reduced the total number of records available for analysis from 954 to 818.
• Specified the date of a work to be the average of when the work was begun and when the date of the work was ended.
• Converted all text notes on payments to an actual figure. In a few cases, personal judgement was exercised when payment was given as an approximate range to come up with a figure.
• Created fields to reflect Patron Type (Institutional/Private), Number of Works in Commission (One/Several), and calculated fields for Papal Reign
• Grouped Patron values to reflect patron families (e.g., consolidated all Barbarini patrons into one value, “Barbarini Family.”)

Figure 9. Dataset Modifications – Interactive Dashboard
All changes made to the original Spear/Getty Prices to Artists Dataset for key factors used in analysis

Bibliography:

Ago, Renata. “Splendor and Magnificence.” In Display of Art in the Roman Palace 1550-1750, 2014.

Bratu Hansen, Miriam. “Benjamin’s Aura.” Critical Inquiry 34, no. 2 (January 1, 2008): 336–75. doi:10.1086/529060.

Burke, Jill, and Michael Bury. Art and Identity in Early Modern Rome. Aldershot, England ; Burlington, VT: Ashgate, 2008.

“Catholic Encyclopedia,” n.d. http://www.newadvent.org/cathen/12272b.htm.

Derek Thompson author. Hit Makers: The Science of Popularity in an Age of Distraction, 2017.

Drucker, Johanna. SpecLab: Digital Aesthetics and Projects in Speculative Computing. University of Chicago Press, 2009.

Etro, Federico, Silvia Marchesi, and Laura Pagani. “The Labor Market in the Art Sector of Baroque Rome.” Economic Inquiry 53, no. 1 (January 2015): 365–87. doi:10.1111/%28ISSN%291465-7295/issues.

Gail Feigenbaum editor, and Francesco Freddolini editor. Display of Art in the Roman Palace, 1550-1750, 2014.

Getty Institute. “Payments to Artists Database,” n.d. http://www.getty.edu/research/tools/provenance/payments_to_artists/.

Goldwaithe, Richard A. “The Painting Industry in Early Modern Italy.” In Painting for Profit. Yale University Press, 2010.

Haskell, Francis. Patrons and Painters: A Study in the Relations Between Italian Art and Society in the Age of the Baroque. Yale University Press, 1980.

Kent, D. V. Cosimo De’ Medici and the Florentine Renaissance: The Patron’s Oeuvre. New Haven: Yale University Press, 2000.

Michael. Baxandall. Painting and Experience in Fifteenth Century Italy: A Primer in the Social History of Pictorial Style. Oxford: Clarendon Press, 1972.

Patrizia. Cavazzini. Painting as Business in Early Seventeenth-Century Rome. University Park, Pa.: Pennsylvania State University Press, 2008.

Spear, Richard E. “A Database of Prices Paid to Painters in Seventeenth-Century Rome.” Getty Research Journal, no. 2 (2010): 147–50.

———. “Scrambling for Scudi: Notes on Painters’ Earnings in Early Baroque Rome.” The Art Bulletin 85, no. 2 (June 1, 2003): 310–20. doi:10.1080/00043079.2003.10787074.

Spear, Richard E., Philip Sohm, Christopher R. Marshall, Raffaella Morselli, Elena Fumagalli, and Renata Ago. Painting for Profit: The Economic Lives of Seventeenth-Century Italian Painters. New Haven, Conn: Yale University Press, 2010.

Warwick, Genevieve. “Gift Exchange and Art Collecting: Padre Sebastiano Resta’s Drawing Albums.” The Art Bulletin 79, no. 4 (1997): 630–46. doi:10.2307/3046279.

1. 1. Johanna Drucker, SpecLab: Digital Aesthetics and Projects in Speculative Computing (University of Chicago Press, 2009), xii. ↑

1. 1. Richard E. Spear, “A Database of Prices Paid to Painters in Seventeenth-Century Rome,” Getty Research Journal, no. 2 (2010): 147–50. ↑

1. 1. Richard E. Spear, “Scrambling for Scudi: Notes on Painters’ Earnings in Early Baroque Rome,” The Art Bulletin 85, no. 2 (June 1, 2003): 310–20, doi:10.1080/00043079.2003.10787074. ↑

1. 1. Getty Institute, “Payments to Artists Database,” n.d., http://www.getty.edu/research/tools/provenance/payments_to_artists/. ↑

1. 1. Spear, “Scrambling for Scudi,” 310. ↑

1. 1. Richard E. Spear et al., Painting for Profit: The Economic Lives of Seventeenth-Century Italian Painters (New Haven, Conn: Yale University Press, 2010), 34. ↑

1. 1. “Catholic Encyclopedia,” n.d., http://www.newadvent.org/cathen/12272b.htm. ↑

1. 1. Jill Burke and Michael Bury, Art and Identity in Early Modern Rome (Aldershot, England ; Burlington, VT: Ashgate, 2008), 4. ↑

1. 1. Gail Feigenbaum editor and Francesco Freddolini editor, Display of Art in the Roman Palace, 1550-1750, 2014, 7. ↑

1. 1. Francis Haskell, Patrons and Painters: A Study in the Relations Between Italian Art and Society in the Age of the Baroque (Yale University Press, 1980), 3. ↑

1. 1. Spear et al., Painting for Profit, 34. ↑

1. 1. Patrizia. Cavazzini, Painting as Business in Early Seventeenth-Century Rome (University Park, Pa.: Pennsylvania State University Press, 2008), 19. ↑

1. 1. Ibid. ↑

1. 1. Spear et al., Painting for Profit, 8. ↑

1. 1. Michael. Baxandall, Painting and Experience in Fifteenth Century Italy: A Primer in the Social History of Pictorial Style. (Oxford: Clarendon Press, 1972), 5. ↑

1. 1. Patrizia. Cavazzini, Painting as Business in Early Seventeenth-Century Rome, 130. ↑

1. 1. Derek Thompson author, Hit Makers: The Science of Popularity in an Age of Distraction, 2017, 215. ↑

1. 1. Genevieve Warwick, “Gift Exchange and Art Collecting: Padre Sebastiano Resta’s Drawing Albums,” The Art Bulletin 79, no. 4 (1997): 630–46, doi:10.2307/3046279. ↑

1. 1. Spear et al., Painting for Profit, 19. ↑

1. 1. Ibid., 77. ↑

1. 1. Eric C. Chang, Yan Luo, and Jinjuan Ren, “Cross-Listing and Pricing Efficiency: The Informational and Anchoring Role Played by the Reference Price,” Journal of Banking & Finance 37, no. 11 (November 2013): 4449–64, doi:10.1016/j.jbankfin.2012.12.018. ↑

1. 1. Haskell, Patrons and Painters, 20. ↑

1. 1. Richard A. Goldwaithe, “The Painting Industry in Early Modern Italy,” in Painting for Profit (Yale University Press, 2010), 288. ↑

1. 1. Miriam Bratu Hansen, “Benjamin’s Aura,” Critical Inquiry 34, no. 2 (January 1, 2008): 340, doi:10.1086/529060. ↑

1. 1. Federico Etro, Silvia Marchesi, and Laura Pagani, “The Labor Market in the Art Sector of Baroque Rome,” Economic Inquiry 53, no. 1 (January 2015): 380, doi:10.1111/%28ISSN%291465-7295/issues. ↑

1. 1. Renata Ago, “Splendor and Magnificence,” in Display of Art in the Roman Palace 1550-1750, 2014, 67. ↑

1. 1. D. V. Kent, Cosimo De’ Medici and the Florentine Renaissance: The Patron’s Oeuvre (New Haven: Yale University Press, 2000), 367. ↑

1. 1. Spear et al., Painting for Profit, 22. ↑

1. Drucker, SpecLab, xii. ↑