As I discussed in this post, Laura Mandell et al., quoting from Colin Ware, argues that the first things we see in data visualisations are “errors and artifacts”, caused by how the data is structured, by how it is collected, and even by historical events, such as a publishing house fire destroying documents (section 1). For example, I believe I fell into such a trap when I looked at the data for “11A1 God the Creator”. (Note: Please click on any visualisation to open a full-size version in a new tab.)
The above line graph shows the number of emblems published per year, and is coloured to show present-day nations, revealing what I hesitatingly described as a “French” period (1540 to 1610s), then a gap of approximately thirty to forty years, followed by a “German” period (1650s to 1780). I was particularly tantalised by the idea that the gap corresponded with the Thirty Years’ War, even though I still saw such an interpretation to be a stretch based on little evidence.[1] Not only does the ‘no emblem’ gap disappear in the data for all 1,517 emblems, but changing the type of visualisation shows that gaps between emblem publication were more common for 11A1 tagged emblems (Fig. 2).
The above visualisation, unlike the line graph, highlights the discontinuity between publications, and is thus a more accurate structuring of the data.
So, prior to examining and interpreting the data, it seems appropriate to discuss some of the potential sources of errors and artifacts in my data, while also looking at choices made to minimise such errors, as appropriate. As such, this post will look at issues with the creation of my dataset, with Iconclass notations, with Tableau, and with the bibliographical location metadata. Lastly, I will look at what the data itself cannot reveal, showing the need for further research on the topic of the Christian denominations of emblematists.
1,517 Emblems: A Subset of a Subset – Potential for Artifacts, Potential for Accuracy
As I noted at the end of Pt. 1, my data is a subset of the emblems available through Emblematica Online; its collection potentially being a subset of existent emblem books. Two observations must thus be made up front, regarding my choices, and regarding the prevalence of Germany. First, my selection of emblems was not random; I selected particular symbols, as indicated by certain Iconclass notations, and I did not even select all of the possible Christian religion notations. Therefore, there is the potential that my data is slanted toward my interests. However, since my choices were solely based on particular symbols, the choice should not directly affect the bibliographical data representative of each symbol. I did not remove any emblems within a notation’s dataset because of language, publication location, or even the co-occurring notations. So each requested notation’s dataset is fully representative of all the emblems that have been marked-up so far.
Second, my data seems to be slanted towards German emblems, which could be an artifact. Looking at the overall language, location and city data for my 1,517 emblems (next three visualisations below), one thing becomes immediately clear; Germany rules!
The above three diagrams show a dominating German presence. German is the language most frequently used in these emblems, followed by Latin and then French, the latter appearing in less than half of the number of emblems German does. Over one thousand of the emblems in the set were published in present day Germany, meaning that approximately two thirds of all the emblems in this dataset come from German cities. The third visualisation shows that the four top producing cities in this set are German ones: Nuremberg, Frankfurt, Munich and Augsburg. Paris, the first non-German city, comes in fifth place. Why is this, and is it a problem?
The prevalence of Germany could, hypothetically, be an artifact. Since my data is dependent on the emblems marked-up by the various academic institutions associated with Emblematica Online, it seems prudent to acknowledge the fact that they may have chosen which emblem books to mark-up first based on their interests. So I checked their language totals for Books and Emblems, as available from the “Filters” sidebar while browsing. According to the site, most of their books are in Latin (total = 619), whereas there are only 267 German books, less than half of the Latin texts.[2] Similarly, German comes in second in marked-up emblems; there are about 6,500 more in Latin. So it appears that my data was not shaped by any preference towards German emblem books by the scholars behind Emblematica Online.
Indeed, further research revealed that at least some of the patterns observable in my data are accurately indicative of the history of emblem books. Emblem books were often printed in multiple editions, and translated into other languages. Andrea Alciato’s “work alone ran into more than one hundred editions within a century (including vernacular translations” (Visser 141). As father of the genre, he has received more attention. Peter M. Daly traces the appearances of Alciato’s reprints from the 1530s to the 1620s; from Augsburg to Paris, Venice, Lyon, Frankfurt and Padua (n. pag.).[3] The visualisation below shows just such a transition in Alciato’s work within my data set.
So regarding the father of the genre, my smaller data set is fairly representative of the larger picture, which bodes well for its overall accuracy.
My data may also speak to the relationship between language and publication location. Cole, Han and Vannoy argue that emblem books “are more often written in the vernacular of the region where the book was published” (111); however, Saunders argues that emblem book publication was sometimes “international”; showing that French emblem books were first published in France, then in other European regions, such as the present-day nations of the Netherlands and Belgium (422). The following matrix visualisation shows the numbers of emblems in my data set, the x-axis showing present day nations, the y-axis showing languages.
Certainly my dataset shows that most books tended to be published in the local vernacular – eight hundred German emblems were produced in present-day Germany, and France produced 240 emblems in French – thus agreeing with Cole, Han and Vannoy. Still, neither Germany nor France produced all of the emblems in their vernaculars; for example, Germany produced 110 of the French emblems in this set, which corresponds with Saunders’ observations. As such, a little research helped show that even though my dataset seems to align with German books, that apparent ‘favouritism’ is not an artifact of the data. Since enough of the metadata seems to agree with emblem scholars, I am more confident with the bibliographical data’s overall accuracy.
Iconclass, Tableau and Location Metadata – Little Headaches
Headache #1: Errors and artifacts in the data popped up from some other sources, and should be acknowledged, even though they were not always fully addressed. I have written previously on the issues with using Iconclass as a source of supposedly objective numerical data.[4] Essentially, how an emblem gets marked-up depends on who is doing the marking-up, which brings in the digital humanities concerns with thoroughness and consistency. In Part 4, I will look at a situation where five almost identical emblems (likely adapted from each other) are marked in two quite different, and yet wholly fitting, ways. The result is two sets of Iconclass notations for essentially the same image, with only one co-occurring notation. Thus the choices of those doing the marking-up certainly affect the data in ways I cannot completely predict and control for. Still, the notations are still the best option for numerically analysing emblem images.
Headache #2: Tableau’s interaction with the data tables created a situation where emblems were being counted more than once because of the way different variables were sorted in the data; though I was able to minimise the potential of errors, through suggestions from John Brosz and some of my own ideas, I was unable to remove the problem completely. The problem occurred for a few reasons. First, duplication happened because some emblems are in more than one language. If an emblem had three languages, it would look like it added one emblem for each language. Such a result is desirable for the graphs dealing with languages, but not otherwise; so for those latter graphs, I essentially use a data set without language data, and that fixes unwanted duplication.
Second, some emblems appear in multiple “Requested Notation” datasets. One of the reasons was because Iconclass is a hierarchy. I requested that the the data for a notation would include the data for all its subcategories; however, I also sometimes asked for a separate dataset for a subcategory itself. That created the problem. For example, I asked for 11C1, and a separate set for the subcategories 11C11, 11C12 etc. That means the dataset for 11C1, which contains the full dataset of 11C11, effectively duplicated all of 11C11’s emblems. This issue was fixed by filtering out all subcategories in visualisations for the whole set of emblems, leaving only those requested notations highest up in the hierarchy.
However, some emblems were repeated in different requested notations that were not above or below each other in the hierarchy. For example, hypothetically, an emblem could be tagged with 11A1 and 11C16. There was no easy way to remove those duplicates, meaning that visualisations counting emblems for all the notations show 1,777 emblems instead of 1,517. However, the problem is removed completely when one focuses on individual requested notations; that is why I based my work in Part 3 on that view of the data, relying less on the visualisations that count emblems for all the notations.[5]
Headache 3: Perhaps the easiest data artifact to deal with was the location metadata. The publication location data given to me, came from transcriptions from the books themselves. That means a single place name was often rendered differently because of different languages, different spellings within the same language,[6] and transcription errors. For example, present day Antwerp, Belgium was listed as “A Anvers”, “Antuerpiae”, “Antverpiae” “Antwerpen”, and “t’Antwerpen”. To deal with that issue, I “normalised” the place names to their present day anglicised equivalents,[7] since I wanted an accurate count of emblems from each city. That includes adding the anachronism of the present day ‘nations’ these cities and towns belong to, simply because it helped me to conceptualise the locations within regions. Standardising the publication locations removed what would have been an error in the data, consolidating multiple labels for the same place into one; quite akin to combining data for long ‘s’ and short ‘s’ mentioned by Mandell et al. (section 1). Any error in translation is, of course, completely mine.[8]
What the Data Cannot Say: Location + Language ≠ Religion
“[Quantitative research] provides data, not interpretation.”
-Franco Moretti (9)
Though images and symbols for God can certainly be called religious, I am hesitant to make any claim regarding the religious affiliation of the emblematists. My dataset could be interpreted as showing the emblems to be Protestant, but such a reading is unfounded. The data shows a large proportion of German emblems in this set, both in terms of language and in terms of production location, as Figures 3, 4 and 5 above indicate (linked here on the numbers). If one assumes Germany was a solely Protestant region during the heyday of emblem books, given the strong German links with Luther and the Reformation (Oxtoby and Amore 209-13), then it looks like the majority of these emblems were Protestant ones. However, there is no metadata accounting for religion directly. Additionally, there is the problem that emblems include what could be considered religious imagery. Given the destruction of religious iconography by Protestants, specifically Calvinists, during the 16th century (Stronks, “Literature” 227), and the tendency of Protestantism to avoid images (“Literature” 234), one may even question if Protestants would engage in emblematic visualisation of the persons of the Trinity. Els Stronks actually argues that it was Roman Catholic influence through emblem books that helped re-introduce religious imagery to Protestantism (“Literature” 228, 238; “Dutch Religious” 158). The German presence in the data, therefore, says nothing about the religious ideas expressed in the emblems.
Further research reveals that the religious aspect of emblems is quite complex. Some works can certainly be labeled as Protestant (Visser 142); or, as in the work of János Zsámboky, aka Johannes Sambucus, be shown to have Lutheran attitudes (Visser 157). Others are the work of Catholics (Visser 142; Knapp and Tüskés 38). Indeed, the Jesuits influenced the genre across Europe, authoring and/or producing many emblem books (Knapp and Tüskés 39; Saunders 422, 426), and emblem books were used as educative texts within Jesuit Colleges, both for the study of literature and poetics and for religious studies (Knapp and Tüskés 48). Then there are the books that exhibit Christian humanism from Italy (Visser 141; Knapp and Tüskés 90), with “an emphasis on a personal, intellectual, spirituality” and “a critical attitude towards” religious institutions (Visser 143, 148).
While some books can be thought of a belonging to one Christian denomination or movement, boundaries are also crossed. Sometimes emblematists “borrowed” emblems, to use Saunders’ term (420), from other works; taking a picture and changing the epigram and/or motto to the extent that an edition in a new language would be less a translation[9] and more an adaptation (Saunders 421; Knapp and Tüskés 89, 94, 99-100). Other times the image would be altered to change the import or effect of the emblem,[10] and sometimes new books would be printed combining emblems from multiple sources (Knapp and Tüskés 94). This practice means that religious ideas could migrate. Els Stronks discussed how Catholic emblem books influenced Protestant emblematic works within the Dutch Republic (“Dutch Religious” 143), sometimes printed in almost unchanged form (“Dutch Religious” 143, 156), sometimes with pictura, motto and epigram drastically adapted to better fit the different religious world-view (“Dutch Religious” 146). It is also entirely possible that a portion of emblem books were written to be accessible to readers of multiple Chritianities (see Visser 161). The humanist emblem books of Alciato[11] were found in both Catholic and Calvinist libraries, and read by priests and ministers (Visser 163). While it could be desirable to use the Iconclass notations with the language and location metadata to theorise the religious content of the emblems, further close reading of emblems and research into the individual creators’ religious outlook would be needed to make such arguments strong.
Endnotes
[1] Thanks goes to Dr. Michael Ullyot who tempered my enthusiasm even further, by mentioning there were very few data points for this notation – only thirty-six emblems – thus placing further doubt on the theory.
[2] Many books in my raw data were multilingual; listing two, three or more languages. Alison Saunders discusses these “polyglot productions”, noting examples from Antwerp in the early 1600s (423). So, when Emblematica Online says a book is German or Latin, that does not mean the language is exclusive; their metadata tends to list all the languages.
[3] Alison Saunders gives an almost identical list, though without dates (415).
[4] See my posts entitled: “A Problem of Standardisation” and “The Challenge of Hoover and McGann”.
[5] As you will see in Part 3, I looked at visualisations for each of the sixty-one categories separately, showing the number of emblems published in each location and each year, and in what language. Then I manually inputted relevant information, in order to examine the patterns typical of the God-symbols as a group, without the error of duplicated emblem count.
[6] The interchangeable U and V in Latin being one of the easiest to deal with.
[7] For the Latin place names I used the Association of College and Research Libraries’ “Latin Place Names” translation database. The other languages were easier to translate in general, often popping up quite easily in Google and confirmed from multiple sources, including municipal websites.
[8] I should note that the direct transcriptions are retained in the raw data; available for later double checking, in case a translation is mistaken.
[9] Still, some translations did stay close to their sources (Knapp and Tüskés 105).
[10] See Stronks, “Dutch Religious” 152 for an example that changes reader interaction with an emblem.
[11] It is even difficult to place Alciato, the father of the genre, into either the Protestant or Catholic camp, because of his humanism. He was honored for his work in law by the Vatican (Visser 154). His letters show that he had some reformist attitudes (Visser 154), at other times he vigorously defended the Catholic hierarchy and the primacy of the Pope (Visser 155-56).
Works Cited
Cole, Timothy W., Myung-Ja K. Han, and Jordan A. Vannoy. “Descriptive Metadata, Iconclass, and Digitized Emblem Literature.” Proceedings of the 12th ACM/IEEE-CS joint conference on Digital Libraries. New York: Association for Computing Machinery, 2012. 111-20. ACM Digital Library. Web. 28 Jun. 2015.
Daly, Peter M. “Bibliography of Studies of Andread Alciatus and His Emblems.” Andreas Alciatus. 2 vols. Toronto: U of Toronto P, 1985. N. pag. ProQuest ebrary. Web. 22 Apr. 2016.
Emblematica Online. University of Illinois Board of Trustees, 2015. Web. 25 Apr. 2016.
Knapp, Éva, and Gábor Tüskés. Emblematics in Hungary: A Study of the History of Symbolic Representation in Renaissance and Baroque Literature. Trans. András Török and Zsuzsa Boronkay. Tübingen, Ger.: Max Niemeyer Verlag, 2003. ProQuest ebrary. Web. 20 Apr. 2016.
Mandell, Laura, et al. “How to Read a Literary Visualisation: Network Effects in the Lake School of Romantic Poetry.” Digital Studies / Le Champ Numérique 3.2 (2012): n. pag. Web. 11 Feb. 2015.
Maxwell, Robert L. “RMBS/BSC Latin Place Names File.” RBMS: Rare Books and Manuscripts Section. Association of College and Research Libraries, 2012. Web. 25 Mar. 2016.
Moretti, Franco. Graphs, Maps, Trees: Abstract Models for a Literary History. London: Verso, 2007. Print.
Oxtoby, Willard G. and Roy C. Amore. “Christian Traditions.” World Religions: Western Traditions. 3rd ed. Eds. Willard G. Oxtoby and Amir Hussain. Don Mills, Ont.: Oxford UP, 2011. 164-265. Print.
Saunders, Alison. “French Emblem Books or European Emblem Books: Transnational Publishing in the Sixteenth and Seventeenth Centuries.” Bibliothèque d’Humanisme et Renaissance 61.2 (1999): 415–27. JSTOR. Web. 20 Apr. 2016.
Stronks, Els. “Dutch Religious Love EmblemsL Reflections of Faith and Toleration in the Later 17th Century.” Literature & Theology 23.2 (2009): 142-64. Oxford Journals. Web. 20 Apr. 2016.
—. “Literature and the Shaping of Religious Identities: The Case of the Protestant Religious Emblem in the Dutch Republic.” History of Religions 49.3 (2010): 219-53. JSTOR. Web. 20 Apr. 2016.
Visser, Arnoud. “Escaping The Reformation in The Republic of Letters: Confessional Silence in Latin Emblem Books.” Church History and Religious Culture 88.2 (2008): 139-67. ATLA Religion Database with ATLASerials. Web. 20 Apr. 2016.
Finding God in the Metadata: Distant Reading and Emblem Studies by Aaron Ellsworth is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
Aaron,
wonderful to see that even at this early stage tagged emblems can trigger this kind of research. As the person responsible for the Iconclass tagging of the German emblem books at Illinois and the Love emblems at Utrecht (and a portion of the Alciatus and French emblems at Glasgow), I am very happy to see this type of analysis emerge. You are quite right though, about the various biases in the dataset. These are due in part to the selection of books which depended on the focus of the various projects. Should you want to work with a larger dataset, or questions of content and choice answered, do let me know.