Gratian, an otherwise unknown teacher, finished compiling and writing the Concordia Discordantium Canonum ("Agreement of Disagreeing Canons") or Decretum around 1140 in Bologna as a textbook for university students studying canon law. In 1996, Anders Winroth discovered that there were two versions or recensions of the Decretum, and that four surviving 12th-century manuscripts preserve the text of the first recension. Winroth argued that different authors (Gratian 1 and Gratian 2) compiled and wrote the two recensions, and that it was the first recension that the author finished and began to circulate around 1140, not the second.

Gratian compiled the Decretum out of canons taken from decrees of church councils, from papal letters, from Roman law, and from patristic authors like Augustine, Ambrose, Jerome, and Gregory. Gratian collected the texts for the canons that he used from formal sources (previous canonical collections) as opposed to material (original) sources. There are 1,860 canons in the first recension and 3,945 in the second. On his own authority as a jurist, he commented on the canons in the form of sayings or dicta. Gratian organized the canons and dicta in the first part of the Decretum into 101 distinctions and those in the second part into 36 cases. He introduced each of the cases in the second part with a hypothetical case statement that sets out the facts of the case and divides it into questions. In addition to the cases, the second part of the Decretum includes a treatise on penance (de Penitentia). The third part, a treatise on the sacraments (de Consecratione), was added to the Decretum at a late stage in the composition of the second recension. Other teachers of canon law added approximately 150 canons (paleae) to the second recension after it began to circulate.

Winroth’s identification of the Florence (Fd), Barcelona (Bc), Paris (P), and Admont (Aa) manuscripts as a first recension of Gratian’s Decretum has been generally accepted, but some of the conclusions he drew from the discovery remain controversial. Winroth argues that Gratian 1 and Gratian 2 were probably not the same person, because Gratian 2 makes much fuller use of the Codex and Digest of Justinian than does Gratian 1, whose Roman law sources were largely limited to excerpts from barbarian legal codes that he found in previous canonical collections. Pennington argues that Gratian made earlier and more sophisticated use of Roman law concepts and terminology than Winroth gives Gratian credit for, but his argument depends in part on evidence from the St. Gall (Sg) manuscript of the Decretum, which Winroth rejects as an abbreviation.

Moreover, Winroth rejects the hypothesis that a dictum that apparently refers to the Second Lateran Council of 1139 is an interpolation. He is consequently committed to a late date for the first recension (around 1140). Winroth originally argued for a late date (mid-1150s) for the second recension as well, but as evidence emerged that the second recension was in general circulation by 1150, he shifted to arguing for a period of just a few years between the first and second recensions, while continuing to insist that 1140 is the earliest possible date for the first recension, and that the first and second recensions are the work of two different individuals. Recently, Winroth reexamined the 12th-century tradition that Gratian had been a bishop, and argues that a "Gratian of Chiusi, bishop" mentioned in a Siena necrology, who may have died in 1144 or 1145, is the Master Gratian responsible for the first recension of the Decretum. Winroth considers this further evidence that Gratian 1 and Gratian 2 were not the same person.

The controversy over the identity of Gratian is an aspect of a broader debate over what the discovery of the recension preserved in Fd, Bc, P, and Aa adds to our understanding of the evolution of the Decretum. For the last fifteen years, that debate has revolved primarily around the question of whether Sg represents a recension of the Decretum earlier than the first recension. Winroth maintains that the first recension was a finished product that Gratian 1 released into circulation, and that Sg is an abbreviation of the first recension. Pennington, on the other hand, argues that both Sg and the first recension are snapshots taken at particular moments in a much longer process of continuous textual evolution (going back as far as the 1120s), and that Sg represents an earlier stage in the development of the Decretum than the first recension. The debate over Sg has been fought to a stalemate, and new evidence is needed to move the discussion forward.

The purpose of the project is to reconsider the debates over the identity of Gratian and the evolution of the Decretum between the first and second recensions using statistical analysis of word frequencies in the text as evidence. The first question concerns the identity of Gratian: whether or not the Decretum is the work of a single master who compiled the canons and composed the case statements and dicta organizing the presentation of and commenting on those canons. The result will be an assessment of the likelihood that the case statement and dicta were written by one, two, or more authors. Many topics that Gratian covered in the first recension were expanded upon in the second recension by the incorporation of additional canons and dicta. At least one entirely new topic – the status of Jews in canon law – was added in the second recension. The second question concerns the evolution of the text of the Decretum between the first and second recensions: what other entirely new topics were added to the Decretum in the second recension? The result will be the identification of additional entirely new topics, if any.

Ideally, electronic texts of good critical editions of the first and second recensions of the Decretum, following consistent orthographic conventions, would be available for analysis. In practice, what is available is the electronic text of Friedberg's problematic 1879 edition of the Decretum that the MGH used to produce its Gratian concordance. Friedberg's text can be transformed into a proxy for the second recension by removing de Consecratione and the paleae. The resulting second recension text can then be transformed into a proxy for the first recension by applying the differences found in the appendix of Winroth's The Making of Gratian's Decretum.

For authorship attribution, we are interested in analyzing common function words like conjunctions and prepositions in the case statements and the dicta – those parts of the text of the Decretum attributed since the 12th century to Gratian himself. Three samples are used: the case statements, which do not differ substantially between the first and second recensions; the first recension dicta; and a sample including dicta found only in the second recension, plus words that appear only in the second-recension forms of those dicta that differ between the first and second recension. The statistical frequency of the function words in each of the samples will then be analyzed using principal component analysis to determine whether it is possible to distinguish the author(s) of the case statements and the first and second-recension dicta from one another. Principal component analysis is concerned only with words that are indeclinable forms in Latin, so there is no need to lemmatize the samples.

To identify entirely new topics added to the second recension of the Decretum, we are interested in analyzing content words in both the dicta and the canons. Two samples are used: the first-recension dicta and canons; and a sample including dicta and canons found only in the second recension, plus words that appear only in the second-recension forms of those dicta and canons that differ between the first and second recensions. A list of common stop words found in Wortkonkordanz zum Decretum Gratiani will first be removed from the samples, which will then be lemmatized using the Schinke Latin word stemming algorithm. (The lemmatization algorithm uses an exclusion list of words ending in –que for which the ending is interpreted as part of the word rather than as an enclitic. Preliminary results indicate that the list needs further refinement.) Finally, the statistical frequency of unique or most distinctive words will be used to determine whether it is possible to identify topics which appear only in the second recension. If new topics are identified, they will be verifiable by close reading. Traditional scholarly methods can then be used to more fully explore their implications for our understanding of the evolution of the text of the Decretum between the first and second recensions.

Python and the Natural Language Toolkit (NLTK) will be used for corpus preparation. R and the Stylometry with R package (stylo) will be used for Principal Component Analysis and for analysis of Most Distinctive Words.