How do trial records tell stories of heresy? A computational analysis of narrative in Peter Seila's 1241-42 inquisition register

In an article recently published in Historical Research (Oxford University Press), Robert L. J. Shaw, Tomáš Hampejs, and David Zbíral study the 649 summaries of guilt in the mid-thirteenth century sentence register of Peter Seila, one of the first inquisitors appointed in Languedoc. Following the Computer-Assisted Semantic Text Modelling (CASTEMO) approach, we encoded every clause of these summaries as a structured data statement, preserving their original order. This allowed us to analyse both the types of crimes recorded and the sequence in which they were narrated. Our key findings include:

• A narrow inquisitorial vocabulary dominates, but its application was flexible. Just 20 lexico-semantic crime types account for 84% of all 3,789 crimes recorded. This formulaic quality doubtless reflects inquisitorial and notarial priorities. Yet variations in how crimes were framed also reveal openness to deponent explanations. The representation of narrative components retained flexibility rather than rigid typecasting: for instance, while there are clear differences between crimes associated with different dissident sect labels, their profile varies across regions.
• More efficient summarization occludes the more explorative qualities of underlying interrogations. Nearly a third of summaries contain non-adjacent repetitions of crime types: traces of underlying depositions that must have explored individual occasions of contact biographically rather than simply checking off crime types. Yet over the nine months of sentencing (Ascension week 1241–Lent 1242), summaries became progressively shorter and more organized by aggregated crime types, suggesting a growing notarial confidence in constructing composite narratives.
• Crime sequences follow a similar order to a near-contemporary question list, yet with revealing variations. Permutation testing confirms that the ordering of crimes in Peter's summaries is typically close to that found in the question list of the earliest inquisitorial manual: the Narbonne Order of Procedure, written in or shortly after 1244. Yet sequence variations that suggest flexibility in underlying interrogations also abound. It is also the later, more aggregated summaries that most overtly resemble the Narbonne Order model, suggesting that the increasingly proficient notarial refinement of narrative in the register influenced the question list’s compilers.
• Administrative processes shape the "construction" of heresy. The growing refinement of dissident narratives over time is more consistent with prosaic administrative needs (for instance, to better organize crimes to assist sentencing) than a simple imposition of inquisitorial expectations of heresy. Nevertheless, their increasingly efficient presentation may well have laid the groundwork for more perfunctory or preconceived investigations by others.