Hearing Voices: Reapproaching Medieval Inquisition Records

Abstract

The records of medieval heresy inquisitions have been a subject of controversy ever since their rediscovery by historians. The detail they convey of specific social interactions has continued to inspire generations of scholars, while the coercive conditions of their production have placed strong caveats over their interpretation. This article offers a comprehensive review of the debate on the uses of inquisition records, encompassing scholarship across multiple languages and schools of thought. It also highlights some shortcomings in that debate, e.g., the overrepresentation of inquisitors’ choices; the claim that the use of torture led automatically to reproducing outlandish inquisitorial fears; and the idea that exceptional detail correlates with reliability. The article concludes with the proposal of the Dissident Networks Project (DISSINET) to use structured data within a new variety of quantitative history. This method, founded on the Computer-Assisted Semantic Text Modelling approach that DISSINET has pioneered, is well-suited to addressing the biases of inquisition documents and opening them to scrutiny, thus providing a significant complement to close reading.

In the summer of 1388, Giacomo Bech, a Piedmontese suspected of heresy, was interrogated by the Dominican Antonio di Settimo in Turin.1 The opening exchanges on 23 July were cagey. Against a background of the political opposition from the count of Savoy, the inquisitor appears determined (Merlo 1977, pp. 132–33): questions were loaded, and there was a clear expectation that Giacomo fitted a heretical mould that even crossed multiple inquisitorial categories. He was asked whether he knew the valdenses in various places, and whether he had been to their ‘synagogues’, even though the religious practices on which he was questioned included those associated with the Cathar heresy (e.g., the consolamentum) as well as the Waldensians. For his part, Giacomo confirmed the orthodoxy of his faith, denied any heretical practices or any knowing involvement with heretics, and explained his contacts with others condemned or accused with reference to his broader social life and economic activities. Had he known Isabella, the wife of Uberto Cappelli in Pessinetto? He did not think so, but could not be sure, since he was a man who had ‘spoken and conversed with many women in the said place of Pessinetto and elsewhere’ (Patria and Pazé 2016, p. 258). Had he been in Villastellone? Why yes, he ‘stayed for two months or thereabouts and was selling wine there with Enrico Pometto’, a business associate from Chieri (Patria and Pazé 2016, p. 260). By his later depositions between 21 and 28 August, however, things had changed. The text notes that he returned to confess ‘of his own free will, without torture and outside the place of torture’, an acknowledgement, in fact, that he had in the meantime been tortured. He gave far more, however, than just positive affirmation to the suspicions he was under. Rather, he described in detail his journey from taking ‘the habit of those who called themselves the apostles or [brothers] of the poor life in Pontolino (…) from Giovanni di Pornassio’ through to deep involvement in a nexus of ‘heresy’ that stretched far to the east (Patria and Pazé 2016, pp. 270–73). While involved in the latter, he learnt beliefs usually associated with the Cathar heresy (e.g., ‘that God did not create or make any visible thing’) (Patria and Pazé 2016, p. 271), and was allegedly even sent ‘to Slavonia to learn the aforesaid doctrine fully and completely from the masters living there in a place called Bosnia’, though he failed to reach his destination (Patria and Pazé 2016, p. 273).

Giacomo’s confession is one that is well known to heresy scholars (e.g., Lea 1888, pp. 255–56; Merlo 1977, pp. 30, 43, 53–54, 73; Utz Tremp 2008, pp. 228–35, 241–43). It reminds us of many of the reasons why the surviving records of inquisitorial trials continue to both fascinate and frustrate as evidence of medieval religious dissidence and its investigation. On the one hand, for all that Giacomo’s voice is portrayed in the third person, his testimony feels vivid, whether it be the descriptions of his itinerant lifestyle, his crafty allusion to his popularity with women, or the twists and turns, connections and experiences found within the dissident conversion narrative he offered. We can understand here why historians have sometimes seen such words as an inviting window into a world of lay society, culture and religion that so often seems inaccessible in other pre-modern sources (e.g., Le Roy Ladurie 1975; Ginzburg 1999, 2002). And yet, despite these appearances, it is also, as many others have emphasised (e.g., Grundmann 1965; Biller 2001b, pp. 345–67), a type of text that is wrapped up in the conceptions and outlooks of churchmen. While the confession is remarkable and not without idiosyncrasy, Antonio surely had his own expectations when he subjected Giacomo to torture: were the latter’s Bosnian connections real or, as seems far more likely, prompted by ecclesiastical commonplaces concerning the Eastern origins of dualist heresy? And if that was untrue, then what else? ‘Cathar’ beliefs and practices had been almost unheard of in inquisitorial records for decades, and yet suddenly re-emerged here, further stretching the reader’s credulity. At another level meanwhile, we have a tête-à-tête at trial, a duel between the interrogator and the interrogated that is gripping but shifting, with the truth of both Giacomo’s life and even the interrogation itself contested and somewhat obscured by the contest and its rendering as text.

If the debate around what such records really represent might seem interminable—and it indeed has a long history reaching as far back as medieval (e.g., Friedlander 1996, pp. 95, 189; see also Friedlander 2000, p. 87; Paolini and Orioli 1982, p. 49) and then Protestant (e.g., van Limborch 1692, p. 276) debates over inquisitors’ proceedings—, it is in part because all these positions and outlooks are essentially correct. The elements are entwined, and most historians would acknowledge that their task is to recognise them and work through them as best they can and as far as needed for their research aims. It is a challenging task. What is more, while rarely acknowledged, it is actually even more challenging for the vast bulk of inquisitorial records that are less colourful than the attention-grabbing trial of Giacomo Bech. Take, for instance, the record of a sentence that the Dominican Peter Seila handed down at an early inquisitorial trial (1241) in Montauban, Languedoc:

Laurence, wife of Guillaume Faidit, said that when her husband rendered himself a heretic, she followed him to Villemur, and saw those heretics there who received her husband. She also sought advice from the Waldensians about the illness of her son, heard the preaching of the Waldensians four times, gave them bread and wine, and believed that Waldensians were good men. She will go to Le Puy, Saint-Gilles, Santiago, San Salvador, and Saint-Denis.2

Here, there are perhaps the same broad issues: we have a suspect’s reported confession; there is doubtless an attempt by an inquisitor to classify and standardise their crime (the belief that ‘valdenses essent boni homines’ occurs very frequently in the same document); and beneath the verdict, there surely lay some sort of adversarial conversation between the two parties. But the workings of these processes are far fainter than in Bech’s trial: here it is not immediately obvious how the careful reader might get a handle on the exact nature of the potential influences that were integrated into the finished text.

This paper therefore has two different objectives. Firstly, synthesising recent discussion as well as providing our own considerations, we seek to provide a careful delineation of the various ‘biases’ that influenced both the creation of inquisition records and their journey to the modern reader. While these biases came together in different ways to produce sources as apparently different in style as those we have just seen, they are categorisable in such a way that the challenge of interpretation can be better understood across all documents. Secondly, we address the issue of approaches to these sources. Here we acknowledge both classical source critical approaches and the emergence of discourse analysis techniques as a way of reframing the problem of bias, but we also seek to confront the continuing reliance on ‘exceptional’ passages, which pushes records like that concerning Laurence further to the margins. The approach we are pioneering within the Dissident Networks Project (DISSINET, https://dissinet.cz, accessed on 1 October 2022), by contrast, is founded precisely on the underlying commonality between the two examples we have seen here. Almost every sentence is focused intently on interactions and relationships, whether it concerns what allegedly went on among dissidents or what went on at trial: between people, places, and things both material and immaterial. While a simple observation, it in fact suggests a route into transforming these sources into structured data with surprisingly little loss of semantic nuance. Computer-Assisted Semantic Text Modelling (CASTEMO), an approach developed by the DISSINET team, aims to offer precisely this sort of transformation (Zbíral et al. 2022b). The collection of data via this method in turn allows for the use of computational techniques as a powerful complement to qualitative analysis: building on the semantic text modelling we propose, they have the potential to analyse the aforesaid biases, to move beyond them in new research directions, and to address the records in their bulk as well as in their discursive complexity.

1. Biased Knowledge

If the complex of influences that shaped inquisitorial trial records can frustrate, they are also part of the continued appeal of these documents as objects of study. Perhaps more than most medieval sources, these texts provide some unusually strong indications of the conditions under which they were produced: not to the extent that the quandaries we have mentioned can be easily resolved, but enough to bring them into enticing focus. It is thus little surprise that such biases themselves have become the object of systematic inquiry.

From a classic source critical perspective, represented for instance by Herbert Grundmann, Alexander Patschovsky, Grado G. Merlo, Andrea Del Col and in many ways also Caterina Bruschi, the biases of our sources are often conceived of in terms of filters that have affected the information as it has come down to us. Once the effect of these filters is understood, we can try to compensate interpretatively for them in different ways, depending on the research question (e.g., Grundmann 1965; Patschovsky 1991, p. 255; Merlo 1977, pp. 9–15; Del Col 1994; Bruschi 2009, pp. 14–26).

More recently, these perspectives on bias have been both critiqued and complemented by another that stresses the active production, rather than just selection and distortion, of knowledge by the discourses active at trial. This shift is connected with the wider ‘linguistic turn’ in historical disciplines, and more generally with postmodern reinterpretation of power as primarily productive rather than simply repressive (cf. Cárcel 2000, p. 103; Arnold 2003, p. 63). Thus, John Arnold draws into question the possibility of eluding inquisitorial discourse: ‘The language of inquisition not only permeates the registers, it also creates them. The project is therefore not so much one of ”reading against the grain” as reaching an accommodation with the language of power by bringing to light the particular interplay between the hegemonic and the subaltern’ (Arnold 2001, p. 121; see also Arnold 1998a, pp. 380–82). The focus is thus shifted somewhat away from simply analysing religious dissidence and/or inquisition through the sources, and towards dissecting the ‘productive’ power-imbalanced interactions that forged the texts themselves.

The ‘classic’ and ‘discursive’ approaches probably need to coexist, since each of them has different possibilities and limitations, and each allows us to address a different set of questions. The latter approach is especially useful for interpreting the rich narrative detail found in inquisition registers, and turning it into evocative extrapolations illuminating larger cultural patterns and processes. The former approach, on the other hand, should by no means be cast aside. Its techniques are in fact not incompatible with the idea of inquisition records as actively produced knowledge; rather, they can encourage us to define more precisely the different biases affecting the process of production, and thus set some clear criteria upon which to judge the knowledge conveyed by our sources.

However one perceives these biases, it is indeed necessary to take a step back and digest the full fruits of these approaches for understanding the range of influences that have shaped the corpus of trial records available to us and the rich variety they have created. Among these biases, types can be defined, each affecting the use of inquisitorial records in different ways and, depending on the document, with different severities. Firstly, there are coverage biases, introduced by two main factors: the selection of people who were summoned or came to confess, and the contingencies that have affected the textual transmission of inquisitorial records. A second type is interrogation biases, introduced by the interaction at trial: the choices, actions, and cultural constraints not only of the inquisitor but of all participants. Thirdly, and finally, there are recording biases, conditioned by the choices made by inquisitorial staff over what and how to record.

2. Coverage Biases

Starting with what precedes not only the written record but even the trial itself, the inquisition material is determined by the selection of deponents who were summoned or came to confess. This means that modern-day readers will never know about the innumerable idiosyncratic Menocchios, and whole dissident cultures will remain completely lost to us. Their heterodoxy might have not been considered so serious or might never have been recognised as heterodoxy in the first place (Forrest 2005, p. 13). As underlined by Georges Duby (1968, p. 399), Robert I. Moore (2007, pp. 66–67) and others, in medieval Europe, orthodoxy was enforced with quite different degrees of intensity and precision at different times and places.

Another major factor in the selection of deponents was the reach of inquisitions in both geographic and social terms. Contrary to the popular image, inquisitors had quite limited resources at their disposal, often struggling with lack of time, finances, personnel, and support from secular or ecclesiastical authorities, or even with open opposition (Dossat 1959; Lansing 1998, pp. 149, 151–53; Thompson 2004, vol. 1, pp. 701–30; Given 1997, pp. 111–40; Moore 2019, pp. 113–14, 118–19, 239–43; Hill 2019, pp. 60–61, 66, 85–88). The limited geographical coverage of inquisitorial records (Roche 2005, pp. 209, 254–55, 370–71; Brenon 2003, p. 207) was doubtless affected by low effective reach of tribunals (Mentzer 1984, p. 148), which struggled to penetrate more remote areas, as well as those governed by more reticent lay authorities. In some areas, inquisitors were nominated but seem to have been comparatively less active due to contingencies other than the lack of institutional rooting and support.3 Socially, inquisitors needed to take into account the realities of power. Processes that targeted the highest ranks of society, especially the high nobility, were rare. The murder of the first papal inquisitor ever appointed, Conrad of Marburg, after he accused Count Henry III of Sayn and other prominent nobles of heresy, provides a telling example of the risks such lack of tact involved (Sullivan 2011, pp. 91–92). On the other hand, some inquisitors appear to have set their sights on specific classes of a more middling type. For instance, Jean-Paul Rehr argues that the medieval inquisitorial campaign with the greatest known coverage—the inquisition of the inhabitants of Lauragais in 1245–1246, with the extant part of its records amounting to more than 5500 deponents—betrays a disproportionate interest in inculpating local consular families, at least as concerns the two settlements he has already explored (Rehr 2019, pp. 29–31). John Hine Mundy has shown that urban patricians of Toulouse, their rights already under threat from multiple sides, also felt the impact of the inquisition particularly harshly in the mid-thirteenth century (Mundy 1985, pp. 27–33). Other more prosaic social factors could also have an effect. Bernard Gui seems to have shown some leniency towards young people: he noted down their age, presumably as an alleviating circumstance, and avoided the public shaming of a child at the sermo generalis (Hill 2019, p. 155). The aforementioned Peter Seila also appears to have given crimes committed in youth special treatment (Feuchter 2007, p. 333). It is thus plausible that some younger people were even able to avoid summons, and are underrepresented in the records.

A final coverage bias results from the haphazard preservation of inquisition records.4 Not that we know of any specific ‘archival politics’ that would systematically distort the coverage of certain regions, periods, religious cultures, or social groups. For instance, even seventeenth-century Protestant scholars—who, driven by a clear confessional agenda,5 played an important role in preserving material on medieval heresy including inquisition records6—do not seem to have omitted inquisitorial material on the Cathars in favour of that on the Waldensians, albeit that the latter played a far greater role in their foundation narratives. But even in the absence of any credibly demonstrated conservation biases, the hazards of history alone have reduced inquisitorial records to but a fraction of their original volume. After medieval losses, sometimes due to destruction by those implicated by the records (Given 1989; Kras 2020, pp. 327–79), the sixteenth century saw further dwindling of medieval inquisitorial material. There is evidence from various regions for the use of inquisition records for bookbinding.7 The dispersion and loss of inquisitorial documents has continued ever since (Albaret 2001, pp. 23, 26), with an important role played, in France and Italy, by the revolutionary upheavals at the end of the eighteenth century (e.g., Benedetti 2006a, pp. 18, 26; Tavuzzi 2007, p. 33; Parmeggiani 2018, p. 416). Against this context, the mission led by Jean de Doat on behalf of Jean-Baptiste Colbert between 1665 and 1670, which produced seventeen large volumes copied from inquisitorial archives as part of larger 248 volume collection of Languedocian archival material (Kolmer 1979; Biller et al. 2011, pp. 20–26),8 has doubtless resulted in the overrepresentation of the region in historical study.

In other regions such as Germany, Austria, Bohemia, and Poland, we fail to pinpoint specific waves and causes of documentary losses, but they were considerable. For instance, we have but fragments from the extensive inquisition led by Gallus of Neuhaus in Southern Bohemia in 1335–1353/5. Patschovsky (1979, pp. 18–24), in his extrapolation from the known original volume of the register and from the fact that the extant fragments mention over 300 persons, estimates that the total document would have mentioned around 4400 persons, which outmatches most medieval inquisitions we know of. Biller (2001b, p. 311) points out how much our image of medieval heresy would probably have to change and how much other regions might have overshadowed Languedoc if records which once existed were better preserved.

Despite such losses, however, it is important to stress that the extant registers normally do not contain random collections of isolated testimonies but rather one or several sets of interconnected records: for instance on various dissident activities in one settlement or on a reasonably compact social network that the inquisitors uncovered (cf. Esch 2003, p. 250). Therefore, while the white spaces can never be truly compensated for, some kinds of research will be affected far less than others.

3. Interrogation Biases

Another important set of biases results from interaction at trial: the choices of its participants, their expectations of one another, and the dynamics and circumstances of interaction. Even the briefest and driest records convey at least a faint echo of this interaction, albeit rearranged in line with the needs of legal procedure (Kuehn 1989, p. 516), and necessarily partial.

When considering potentially distorting factors within the interrogation, the effect of torture and its threat often comes as the first association. Torture was officially introduced as an option in heresy trials by Innocent IV in the bull Ad extirpanda in 1252.9 While it is generally acknowledged that medieval inquisitors of heresy employed it comparatively rarely,10 many historians have seen this date as a turning point. For instance, Robert I. Moore (1994, p. 19) claimed that from the introduction of torture, ‘the statements of heretics are a rich source of information on the fears of the inquisitors, but not on the beliefs of the heretics’ (see similarly Patschovsky 1991, p. 252). In a more recent book, he acknowledges that medieval inquisitors used torture rarely, but when they did, ‘its use meant that the expectations of a prosecutor who sincerely believed that he confronted a terrible and urgent danger would always be confirmed’ (Moore 2012, p. 301). According to Mario Sbriccoli (1991, p. 23), ‘[w]ith the use of force, the defendants would be summoned to become their own accusers, and identify themselves with the objectives of the inquisitor/judge’. Suspects thus became mere objects of trial, and ‘practically voiceless’ (see similarly Solvi 1998, p. 392).

Torture did make people speak, but the comfortable claim that it largely made subjects reproduce the ‘fears of the inquisitors’ deserves far greater debate. We do not doubt that this could and did happen, as already seen with Giacomo Bech. Further examples can be found in trials against the knights Templar and witches (if we extend the notion of ‘inquisitors’ beyond papal investigators of heresy). But in many instances, torture was followed by completely mundane accounts lacking in extravagance that are very far from what might be considered the worst inquisitorial fears, such as the idea of organised heretical sects and conspiracies (e.g., Paolini and Orioli 1982, vol. 1, pp. 43–47, cf. vol. 2, pp. 600–1). In some cases, might it have even produced factual information that might otherwise never have been admitted? The question has in fact proved unanswerable, even in the modern context: empirical study is impossible, and theoretical approaches remain inconclusive.11 In and of itself, the use of torture ultimately thus provides no reliable steer on the quality of information elicited.

It is also impossible to judge how the quality of information produced under torture actually compares to that offered in response to other coercion techniques. A more widely used method was incarceration, which doubtless also provided a major incentive to offer something to the inquisitors, who noted its effectiveness.12 Beneath this, there was a whole gamut of other ways in which psychological pressure could be exerted (e.g., Sbriccoli 1991, p. 18; Fois 2015, p. 83), such as the threat of financial fines and confiscations. Confession could also become a matter of dealmaking: deponents were sometimes given—or indeed they extracted in advance—the promise of not receiving harsh punishment in return for providing information (Biller et al. 2011, pp. 81–82).

The biggest skewing influence on the content of inquisitorial records, however, was the very situation of interrogation itself (Lerner 2007, pp. 4–5). The inquisitor exercised power over the deponents and elicited their narrative primarily by the choices he made in asking questions. These questions set the overall discourse. A palpable textual imprint of this bias remains where questions are explicitly recorded, but regardless of their preservation, they were always present: there was little spontaneity in the deponents’ narration (Del Col 2000, p. 67; Arnold 2003, pp. 70–71). The inquisitors had more than their individual experience and understanding of heresy and its social embeddedness to rely on when asking questions: a regular part of inquisitors’ handbooks were interrogatoria, lists of suggested questions, often adapted to particular heresies (E.g., Selge 1967, pp. 71–72; Anselm of Alessandria 1950, pp. 319–20; Gui 1926–1927, vol. 1, pp. 26–32, 76–82, 102–4, 156–74; cf. Grundmann 1965; Biller et al. 2011, pp. 100–2; Bueno 2015, pp. 68–69; Hill 2019, pp. 126–32). Recognising a topic covered by a known interrogatorium suggests the inquisitor’s question even if the latter is not explicitly recorded. This has led scholars to pay special attention to ‘exceptional’ topics and expressions which leave the pre-established schemes, but even there we must remain careful: the appearance of a less common topic not covered by inquisitorial manuals does not automatically mean that it was raised unprompted (Bruschi and Biller 2003a, p. 16; Arnold 2003, pp. 69–71; Biller 2003, p. 133). The importance of questions for the interpretation of inquisitorial records is thus capital and widely acknowledged (e.g., Grundmann 1965, p. 522; Grundmann 2019, pp. 126–79; Biller 2001b, p. 314; Biller 2006; Pegg 2001, pp. 45–51; Arnold 2004, pp. 233–42; Paul 2013, p. 288). Such questions were the main structuring principle of the conversation at trial, and the frameworks the interrogatoria created sometimes suggest quite a rigid approach to questioning (e.g., Kieckhefer 1979, pp. 29–30, 63). In any case, they determined the focus of the deposition, and shaped the discourse of the deponents’ answers, due to their often leading quality (Biget 1991; Bruschi 2009, pp. 22–23). Answers could even be extorted, for instance through the confrontation of deponents with other evidence—previous depositions by others, or their own.13 Some inquisitors’ manuals even recommend using bait questions, in which the investigator pretended to have evidence which in fact he did not (Hill 2019, p. 115). This practice is still in use in some legal systems14 despite indications in modern research that bait questions may seriously mislead not only the record but even the deponents’ perception and memory of events (Luke et al. 2017).

In spite of the fundamental importance of the inquisitors’ questions, we should not underestimate the space for expression that the deponents were given. Even where the influence of interrogatoria was present, the keenness to gather new evidence and snowball further suspects led, at times, to some relatively open questioning, and even less interrupted narration, such as in the well-known register of Jacques Fournier. Furthermore, however pronounced the imbalance of power may have been, deponents were far from passive and voiceless objects of inquisitorial power. The choices they made produce an absolutely crucial, and seriously underestimated, bias. They could choose to deny a charge, and their denial was often accepted (e.g., Biller 2004a, vol. 1, p. 455). They ultimately avowed what they felt they had to (given the pressure of the situation and the evidence already available to the inquisitors) and/or what they wanted to (given their own aims in the trial and their understanding of the inquisitors’ knowledge and expectations) (Bruschi 2009, pp. 23–24). While failing to reveal information to inquisitors under oath was a punishable offence, concealing or minimising the involvement of family members and close connections remained a regular feature (e.g., Biller 2001b, p. 313; Vilandrau 2001). Sometimes the attempts at concealing or misrepresenting facts were even coordinated, both in the sense of coalitions and of threatening witnesses to hide or confess some information.15 The coordination of testimonies must have been extremely widespread, even if it can only be directly attested where the inquisitors succeeded in uncovering it (e.g., Benedetti 1999, pp. 168, 192–24; see also Given 1997, pp. 96–97, 155–63; Biller 2001b, p. 313; Bruschi 2009, pp. 15–16). On the other hand, we also know about cases of keen collaboration, where implicating one’s contacts either served as a way to harm them or to extricate oneself (e.g., Théry 2003, p. 479; Bruschi 2009, p. 17).

Besides the choice of information by both the inquisitors and the deponents, we need to acknowledge the various cultural biases that conditioned those decisions. One aspect of the problem is the preconceptions on both sides. The inquisitors were obviously influenced by specific conceptions of their task and office (Caldwell Ames 2004, 2009), of different heresies, of laypeople’s culture and religiosity, and of the motivation and psychology of the people they were confronting. They also had their own technical language, within which deponent actions were inevitably framed. While inquisitorial handbooks, a broad genre of texts that might contain discussions of heresies as well as procedural counsel and formularies, could provide knowledge born of experience, they nevertheless served to solidify the inquisitors’ assumptions and the categories derived from them (Sackville 2011, pp. 135–53). Similarly, the deponents had their own cultural background and their theories about the inquisitorial understanding of illicit behaviour and belief, the focus and aim of investigations (Bruschi 2009, p. 17), and the kind of information their interlocutors would like to hear. Ginzburg understood these interaction dynamics as a problem of contact and translation between two distinct cultures, the learned culture of the inquisitors and the popular culture of the deponents (Ginzburg 1992, p. 162).16 For Arnold (2001, p. 12), it is less a bilateral clash of cultures, and more a polyphonic exchange of cultural resources or codes. Sometimes, the records give us a glimpse of interesting misunderstandings between the inquisitor and the deponent, when their languages simply failed to meet.17 At the same time, however, some of the inquisitors’ questions were aimed precisely at understanding better what the deponents meant, and there is evidence that some were willing to extend and correct their views on the basis of the information they gathered (e.g., Anselm of Alessandria 1950, p. 319; Biller 2003, pp. 135–37; Biller 2001a, p. 279; Biller 2004b, p. 264).18

In addition to cultural biases, there were various cognitive factors that helped to shape the interaction. For instance, Biller points to a gender bias in the deponents’ memory: it seems that men tended to remember better the involvement of other men, and conversely, women tended more often to give testimony about women (Biller 2001b, p. 313). Autobiographical memory thus itself represents a specific bias which deserves our attention. Arguably, formulations such as ‘s/he said s/he did not remember’ might attest to techniques of evasion (cf. Bueno 2010, pp. 382–91; Bueno 2015, pp. 76–78), but we should not be too quick in making this a general explanation given the obvious difficulties and even false recollections we all face in describing anything but the most recent events.

The issue of memory draws us to the narrative structure of deponent recollections. They are, fundamentally, co-authored semi-structured biographical narratives. As David and Trainum put it in a modern context: ‘A confession is a certain type of storytelling. (...) Law enforcement is not functioning as a passive audience during the storytelling. (...) Although a confession might be ascribed to a suspect (e.g., the confession of John Doe), they are the product of a conversational collaboration between the suspect and the police’ (David and Trainum 2020, pp. 113–14). And since confessions are narratives, ‘applying a narrative analysis to their production becomes an important element in understanding how confessions are elicited, constructed, and told’ (David and Trainum 2020, p. 116). Narrative analysis of premodern trial evidence, in spite of the leading contributions by Natalie Zemon Davis and others (Davis 1983; Arnold 2019, p. 7; Bruschi 2009, pp. 21–24),19 still remains uncommon in practical research. Setting aside for a moment the key role of the notary who compiled the final version—a process discussed in the next section—the oral exchange already had a strong narrative structure, one that emerged from both the deponent and the inquisitor, the latter guiding the former with his questions. Within such exchanges, those describing their past tended to succumb to thematic patterns that played off the questioning and the circumstances they now found themselves in; the interpretation of one’s past from the perspective of one’s present situation (László 2008, p. 116; Esch 2003, p. 251) is a common feature of storytelling but particularly pertinent to the context of trials. Another widespread tendency of the way people narrate their past is disproportion in detail: some matters are well-covered while others remain underrepresented.20 Such influences seem alive in the rather improvised, sometimes even freewheeling nature of Giacomo Bech’s testimony, even before his torture. Self-presentation, a universal feature of human interaction (e.g., Goffman 1959; Baumeister 1982) and one especially relevant in a trial context, thus had different components ranging from premeditated strategies—the choices discussed above—to less conscious patterns that might arise situationally but had also been shaped by previous social interaction. Most deponents were standing trial for the first time in their lives, but they of course had engrained ways of presenting and justifying themselves that they would fall back on under scrutiny.

All this serves to underscore the fact that a serious inquiry into the interrogation biases needs to cover a much broader field than ‘inquisitorial distortions’. It needs to grasp the human dynamics of interrogation, in which both sides had their individual as well as cultural predispositions, influenced one another in interaction, and left their trace in the resulting narratives.

4. Recording Biases

Extant trial documents cover different aspects and phases of a process of text production, transcription, summarisation, and archivation. At each stage of this chain of textual production, there were choices at play that could have led to a particular selection and editing of the original information. At the same time, these biases were constrained by the need to create an authoritative document, something which favoured the preservation of well-grounded information on the actions of suspects.

Debates over the reliability of inquisitorial records have far too often focused on the degree to which we hear the inquisitors’ or the defendants’ voices. In fact, in the first instance, we hear the voice of the notaries. Ultimately, it was they who produced the trial records. They guaranteed the legal force of the process through the act of documentation, but they also provided their expertise in turning the oral exchange at trial, with all its back and forth, repetitions, and inconsistencies, into a reasonably coherent narrative: they were a third party in the process of narrativisation (Chiffoleau 1990, p. 306). Inquisitors were usually not trained in this non-trivial task, and relied heavily on notaries for it (e.g., Moore 2019, p. 91). Notaries were even sometimes commissioned by the inquisitors to conduct interrogations on their own (e.g., Bruschi 2009, p. 19 n. 18; Moore 2019, pp. 105, 111; Tavuzzi 2007, pp. 27–28).

Notaries were not mere employees of the inquisitorial office (Moore 2019, pp. 114–18). They were trusted public officials (Langeli 2004, pp. 55–71), acting under oath of recording faithfully the proceedings at hand, and for the breach of this oath they could be liable to penalties (Moore 2019, p. 91). There is even interesting evidence that in thirteenth-century northern and central Italy the inquisitors did not have full control of the ‘inquisitorial’ records, since the latter were not only written but even kept by notaries (Moore 2019, p. 92); if an inquisitor wanted to consult the official version of the records or make a transcription, he needed to ask the notary. The role of notaries, and their expertise in recording oral exchange at various occasions, is still under-researched and underestimated in the study of inquisitorial records (Benedetti 2006a, pp. 25–26; Del Col 2010, pp. 1118–19; more generally, Moore 2019, pp. 99, 119; Fois 2013).

The textual process that shaped inquisitorial records was a complex one. During the interrogation, the notary took cursory notes of what he thought worth recording (Forrest 2005, p. 70) based on instructions from the inquisitor, previous collaboration with him, but inevitably also his own judgement. There are some rare medieval examples where such notes taken directly at the interrogation, sometimes called a protocollum (Dossat 1959, p. 15),21 were considered as the final record of the interrogation, which was then archived (Kras 2020, p. 314). With this approach, which only became more common in the early modern era (e.g., Del Col 1998, p. CLXXXI), the record might be considered quite close to the actual oral delivery, but even such ‘live’ minute-taking involved quick decisions on what to record and what phrasing to choose (cf. Pegg 2001, p. 57). The choice concerning how much to record could have been conditioned by the workload of the tribunal (Bruschi 2009, pp. 16–17), but also by the goals of a specific investigation (Merlo 1979, p. 61). Besides the selection of information, notaries were, in most cases, the ones who translated the oral exchange between the inquisitor and the deponent from the respective vernacular to Latin—an often underlined process of transformation (e.g., Grundmann 1965, pp. 522–23; Lerner 2007, p. 5; Davis 1979, pp. 68–69; Sherwood 2012, p. 57) which has misled some scholars to ascribe special authenticity to the occasional vernacular passages found in the texts.22

If recording at trial was the first opportunity for selection, translation and framing, still others presented themselves in the subsequent editing phase. Most surviving records are in fact not the original minutes but reorganised clean copies (e.g., Forrest 2005, pp. 4–5) produced after the event. As far as we know, there is no medieval example where the notes taken at the interrogation (protocollum) and the clean copy of the deposition (depositio, confessio, dictum) are both extant, which would better enable us to understand this crucial part of the textual process.²³

Some of the extant codices are not even these original clean copies, but rather their transcriptions commissioned at a later date, mostly to safeguard their preservation, as conciliar legislation makes apparent (cf. Sherwood 2012, p. 68). Assembling records into such larger volumes constitutes another occasion for changes: casual omissions and losses, but potentially also some conscious choices. Several criteria were at play here. In some codices, the documents continue to follow, broadly speaking, a chronological order (e.g., Paolini and Orioli 1982); in others, they were assembled by case disregarding genre (e.g., Tanner 1997); still others are collections of documents of the same genre. The latter of these three approaches is the reason why we often either have a book of depositions without systematically knowing the sentences, or a book of sentences with summaries of transgressions but without more extensive deposition material.

Sometimes we are even only left with new redactions, which significantly reframe the material, rather than with transcriptions (e.g., Välimäki 2019, p. 105; Del Col 2000, p. 66). A telling, if rare, example of such reframing and reorganisation is the accusatory libellus concerning Armanno ‘Pungilupo’. Armanno died in 1269 in an air of sanctity: his body was placed in the cathedral of Ferrara, and he came to be venerated as a saint with the bishop’s support. The Dominicans, however, had already gathered evidence of Armanno’s heterodoxy, and ultimately succeeded in condemning him post-mortem as a relapsed heretic. The inquisitorial libellus was designed precisely to prove Armanno’s guilt. Instead of simply transcribing earlier records, the Dominican editors abbreviated them, selected the most fragrant proofs of guilt, and organised them under rubrics in order to hammer home the main points of inculpation, while still paying attention to the names of witnesses and some of the circumstances (Zanella 1986, pp. 47–70).24 A recently discovered fragment of the original depositions (Bascapè 2002) allows for revealing comparisons with the condensed and reorganised versions in the libellus.

Even this brief overview of the complex textual process in which inquisitorial records were produced makes clear that each of its stages must have involved both conscious decisions on what and how to record or transcribe and deeper cultural conditioning of textual practices. In fact, not even every detail related to heresy was deemed worthy of recording. Famously, Bernard Gui instructs his fellow inquisitors in his manual:

It is worth noting and stressing in these matters that while as many questions, and sometimes different ones, should be asked as demanded by the variety of people and deeds in order to elicit and extort the truth more fully, it is not expedient that the full questioning be written down: rather, only those parts that touch the substance and nature of facts with greater verisimilitude and seem to better express the truth. For if one deposition is found to have a great abundance of questions, another containing fewer might seem diminished. Moreover, with too many questions written down in the process, agreement between witnesses’ depositions could hardly be achieved, [a danger] which must be considered and avoided.

(Gui 1926–1927, vol. 1, p. 32; cf. Bruschi 2009; Gui 2006, pp. 45–46)

Thus, there was a judicious selection of material, already pre-selected and moulded by questions, expectations, knowledge, and the processes of interaction at trial. Just as during the interrogation, there were more conscious and less conscious elements of selection, but in the process of recording, they were entirely enacted by inquisitorial staff. It has been much debated how anti-heretical texts, including trial documents, served to ‘construct’ heresy. But we need to specify this further. When looking at medieval heresy trial documents, we rarely find the most extreme, highly articulated ‘fears of the inquisitors’ such as organised counter-Churches and conspiracies. This is because what is being constructed in trial documents is primarily guilt of individuals; secondarily it is particular dissident rituals and beliefs; and only in the third place and often quite indirectly, it is heresy per se—heretical religion defined by these rituals, beliefs, and organised community. Outright counterfeiting of evidence could happen in medieval heresy trials,25 but only where the need to bend the process in a specific direction was so strong that it prevailed over other mechanisms at play. Much more frequently, recording and the textual transmission of trial evidence was yet another process of selection and reframing which, while far from unbiased, had a concern for factuality as well as a certain logic that today’s historians can hope to analyse and understand.

5. Beyond Bias: The Problem of Exceptionality

Acknowledging the different biases conditioned by coverage, interrogation, and recording is just the first step in engaging with them in everyday research practice. As mentioned earlier, researchers have approached these issues from multiple directions, not only for their own sake but to proceed beyond them: classical source criticism and discourse analysis approaches have made important contributions, and continue to inform and challenge each other. In one key respect, however, these techniques and their many variations and combinations have all led to an almost identical result, one we have already alluded to in the introduction. Greater awareness of bias has not led to greater coverage of the sources, but rather an almost overwhelming focus on exceptional texts and passages, deemed to convey more valuable testimony. With ‘classical’ approaches, the aim has not been to throw out evidence but rather to closely define its usefulness in line with particular questions. Robert Lerner’s reframing of the organised, highly radical images of the ‘Heresy of the Free Spirit’ as far better evidence of inquisitorial projection than of the more varied real religious outlooks of its suspected members remains a defining example of the power of classical source criticism (Lerner 2007). Nevertheless, for Lerner and his key influence Herbert Grundmann there is a sense that valuable information transpires most often where the records leave schemes established by the interrogatorium and inquisitorial knowledge (Grundmann 1965, although with a caveat at p. 538; Lerner 2007, p. 6). Alexander Patschovsky strikes a similar note in commenting on German evidence: he describes those passages where the questions and answers seem to break the mould as ‘rare sugars’ for historians to enjoy, amid the dry bread of arid, formulaic material (Patschovsky 1991, pp. 266–67).

The search for ‘rare sugars’ has not been limited to the material that these sources contain directly concerning medieval religious dissidence or its repression. The approaches of Emmanuel Le Roy Ladurie and Carlo Ginzburg, whose studies have adopted in part an ethnographic, micro-historical approach focused on medieval social life and popular culture, make sense against the background of increasingly critical readings of inquisition records (Le Roy Ladurie 1975; Ginzburg 1999, 2002). If the core religious subject matter of the source is so contestable given its origin, there is a certain logic in focusing on the incidental details, which inquisitors had seemingly less reason or ability to reframe. In his essay ‘The Inquisitor as Anthropologist’, Ginzburg indeed shows a particularly refined awareness concerning the challenge of interpreting sources that are fundamentally ‘dialogic or polyphonic’, both at the individual level—the face-off of inquisitor and deponent—and at the level of broader dialogue between their cultures. In his view, it is at moments of cultural misunderstanding that something more authentic emerges in the record of interrogation (Ginzburg 1992, esp. 159–61, 64).26

Ginzburg’s essay was in fact written in response to Renato Rosaldo’s critique of ethnographic approaches to inquisitorial records, a challenge which was later further reinforced by Kathleen Biddick. Drawing from postcolonial critical theory, the latter two authors question the very epistemic authority claimed by historians and offer analyses of the discourse not so much of inquisitorial records as of scholarship using them. In the view of Rosaldo, the heavy reliance on incidental, ethnographically appealing material in Le Roy Ladurie’s Montaillou ignores the power relations that lay behind the depositions which Jacques Fournier recorded, and falsely portrays the inquisitor himself as an ethnographer (Rosaldo 1986, pp. 77–97). For Biddick, the search for such ‘exotic pleasures’ likewise effaces inquisitorial power, and in doing so even repeats the violence of inquisitors against the subaltern to the aid of the historian’s standing (Biddick 1998, pp. 105–34).

Such challenges have proved inspirational to discourse analysis approaches, which emphasise the ‘productive’ quality of the imbalanced power relations inherent at every stage of these documents’ creation. In John Arnold’s work, for instance, the construction of the ‘confessing subject’ by the language of trial interaction has taken centre stage (Arnold 2003, p. 64; Arnold 2001). Following Ginzburg’s Bakhtinian inspirations, Arnold also speaks about the ‘heteroglossia’ of inquisitorial records: not only are they composed of multiple voices, but those voices are intractably entangled even within (as well as between) the actors involved. There is, overall, a holistic quality to his outlook, attempting to take in the full complexity of interweaving influences within the trial discourse and even upon the historian, without assuming that any one element can simply be stripped away. It is thus unsurprising that Arnold (1998a, pp. 382–83) is quite critical of the emphasis on ‘exceptional’ passages that he found especially among the earlier generation of ethnographically inclined authors like Ginzburg. Nevertheless, in practice, the effect of his own approach is arguably not so dissimilar. Arnold’s analysis of the construction of deponent subjectivity, while pathbreaking, focuses in great depth on a relatively small selection of very rich depositions from Jacques Fournier’s register. He does not give up hope of accessing the subaltern voices of the accused in some way (Arnold 1998a, pp. 383–84), but this requires even more selectivity concerning what breaks free of more typical inquisitorial expression: ‘[E]very telling-of-the-self, although it must ‘succeed’ in the sight of inquisitorial authority, is also the opportunity—indeed, demand—for another moment of ‘self-making.’ (...) [T]he records (...) always produce language that exceeds the inquisitorial categories’ (Arnold 2001, p. 114). Elsewhere, his study of Lollard trial records is founded on the same search for ‘“excess” of speech that eludes easy categorisation, and which arguably troubled the interrogators in its unfixity, and thus prompted its record in writing’ (Arnold 2019, p. 8).

The focus on elements that ‘exceed’ expectations thus remains a critical feature on all sides of the debate concerning inquisition records. Bruschi, whose work responds to Arnold and defends aspects of the classical and Ginzburgian approaches, strikes an almost identical chord when she speaks of the importance of identifying ‘surpluses’ of language in disentangling inquisitorial and deponent voices. Our advancing knowledge of the complexity of the biases (or productive influences) has arguably narrowed the field even further, and rendered selection of material an amorphous challenge: Bruschi falls back on the historian’s divinatio to draw out the surpluses (Bruschi 2009, p. 29), while Arnold’s process calls for constant self-analysis of their own aims wherever they attempt to engage with the subaltern. The stakes are raised further—and in a more questionable manner—when ‘excesses’ or ‘surpluses’ are not only seen as the most interesting material, but implicitly the most ‘authentic’ material (e.g., Bruschi 2009, pp. 14–15) a tendency that risks obscuring the reflexive nuance that is now commonly espoused. Most critically, however, the end result leaves much of the material outside the scope of analysis.

6. From Exceptional Anecdote to Serial Complexity

As Andrea Del Col (2000, p. 62) argues, ‘[t]he criteria on whose grounds we should conduct such historical work must not be valid for only a few processes, those considered “dialogical”, but for all of them’. We also concur with Thomas Kuehn (1989, p. 519), who has cautioned that ’[t]he plurivocality of legal language cannot be grasped fully by reading a handful of documents relating to a single case’. How, then, can we build upon the fruits of careful source criticism, micro-historical approaches, and discourse analysis in a way that opens up a fuller range of inquisitorial evidence to systematic analysis, and allows us to treat even-handedly documents as divergent in style as those concerning Giacomo Bech and Laurence?

As observed in the introduction, a good point of departure is provided by the focus of inquisition documents on relational information, driven by the inquisitors’ need to assess guilt, to investigate exhaustively, and to imbue the resulting record with authority. Heresy trial records are extremely rich in details of who is related to whom, who did what with whom, where and when, and the content and circumstances of the interaction.27 We can track the various types and degrees of relations between people, as well as relate people to places (e.g., place of residence, presence in somebody’s house), objects (e.g., transfer of objects between people) and events (e.g., gatherings), and ideas and attitudes to both people and their interactions (e.g., beliefs communicated on some occasion). Similarly, the procedure of the trial itself and the process of recording is evidenced primarily through often copious quantities of relational information, which were seen to guarantee the validity of documents: who interrogated the deponent, who assisted the inquisitor, who witnessed the trial and who produced the textual record, when and where all this occurred, and the relationships between items of information produced through interaction (i.e., between questions and answers, between deponent admissions, between the confessions of different people, between proceedings at different places and times). Overall, these patterns may be thematically subdivided into three categories: social, spatial, and discursive.

It is a normal part of the historians’ craft to read through the text closely and note down those relations that are of interest for the research at hand in order to identify such patterns. Some of those patterns are indeed accessible to individual memory, classical note-taking, and informed abductive reasoning by an experienced and sensitive reader, in the manner famously promoted by Ginzburg (Ginzburg 1980). However, when taken in their totality, these relations easily amount to tens of thousands in a single inquisition register, clearly rendering many important patterns completely beyond the reach of an individual mind. Furthermore, even for patterns that might be in reach, we usually only notice those which we already expect, either because they correspond with our interests and theoretical outlooks or resound in an ‘exceptional’ passage that has grasped our attention. There are doubtless other less expected and less ‘searched for’ patterns, patterns that occur in more formulaic material as well as in more engaging content, that we will completely miss simply because we fail to systematically record the information which might make them apparent.

The call for a more systematic approach to the relational patterns found in inquisition records is not entirely new. From a socio-spatial perspective focused on the dissidents themselves, it was promoted at the Annales-influenced conference on ‘Hérésies et sociétés dans l’Europe pré-industrielle’, held in Royaumont in 1962, where Georges Duby called for a new and comprehensive ‘cartography of heresy (…) to pinpoint the receptive places, (…) the paths it took’, and to ‘pave the way for attempts at a social interpretation’ (Duby 1968, p. 402). Two issues, however, prevented the sort of comprehensive relational mapping envisioned by Duby from finding much immediate response. One was technical: neither the data infrastructure nor analytical methods required to build a big picture from micro-level relations were well-developed at the time of Royaumont. The growing interest in inquisition records thus rarely amounted to a systematic quantitative inquiry into patterns that these sources are so well placed to elucidate.28 Another issue is the changing fashions of research. As described, the interest in the richest (i.e., most exceptional) narratives offered by inquisition records came increasingly to the fore with the anthropologically inspired research of the 1970s and 1980s; moreover, the criticisms of such approaches inspired by the cultural turn only enhanced the focus on such material. The deepened awareness of the challenges of source interpretation that these developments have engendered has made it more difficult overall to imagine how one could viably employ systematic, quantitative approaches across inquisitorial corpora without the charge of positivism.

Three interconnected strategies can now, we believe, help us overcome those limitations. Firstly, we propose a serial and computer-assisted approach to data collection capable of capturing every aspect of relational information within the text (i.e., both that which concerns dissident activities, and that which concerns trial process), a procedure which at first provides a comprehensive semantic model of the source itself. Secondly, data thus collected enables the computational analysis of social, spatial and discursive patterns, working from the micro-level to the macro. Thirdly, building on these foundations, we can undertake a new computer-assisted approach to source criticism itself. Within the Dissident Networks Project (DISSINET, https://dissinet.cz), we have set out to put precisely these three strategies into practice.

With regard to the first point—the semantic modelling of sources, comprehensive structured data are of course not necessary for all historical studies founded on inquisitorial records. However, if we are to capture patterns of relations between tens, hundreds or even thousands of persons, groups, places, objects, events and concepts, and test hypotheses about these patterns systematically and at scale, we will very early on confront the limitations of the ways and means through which historians usually gather information. We will then be forced, not by any wide-eyed optimism concerning the digital humanities but by immediate research needs of utmost practicality, to engage in a more systematic transformation of our sources into structured data.

How should the historian approach the task of collecting structured data? There are multiple options, ranging from simple tables through to databases based on a complex data model or ‘ontology’. In the case of inquisition records, it is of particular importance that the data model is well-adapted to information on social and spatial relations as well as discursive characteristics. More specifically, it should be adapted to both relational information on dissident activities and relational information concerning the trial and its recording—as well as the contextual relationship between the two. Take the following example: ‘But next day he [Petrus Pictavini] heard the aforesaid Raymundus Petri telling him that on that night the aforesaid heretics [Bonetus de Quesinis and his companion] had hereticated the said sick man [Raynaldetus of Soricino]’ (Biller et al. 2011, pp. 692–93). The aim should not be to simply record that ‘Bonetus de Quesinis and his companion, Cathars, consoled Raynaldetus of Soricino’, but also that this information is claimed to have been reported by Raymundus Petri to Petrus Pictavini who then deposed it in front of a specific inquisitor and in a specific setting. The available contextual details (place, time, etc.) of all such relations, whether outside or within the trial, should thus be preserved. To map the discursive patterns of the documents accurately, the nuances of the original language should also be preserved: it is thus necessary to capture that the wording is ‘heretici’ and ‘hereticaverunt’ rather than ‘Cathars’ and ‘consoled’. The documents themselves meanwhile should be associable with metadata such as the date and place of production, roles (notary, inquisitor, defendant, assessor), genre etc., not only providing still further context for the collected data but doing so in a form which can be instantly factored into the analysis.

Modelling the source in this level of detail might seem excessive for some more limited research questions, but collecting data consistently on all these relational aspects allows for information to be seen in its full trial and source context, as far as it is available to us. This procedure then not only captures what is reported of dissidence, but also the evidence of the biases we have described thus far. The historians collecting data thus need not make up their mind on the questions those biases raise at the very moment of data collection based merely on feel and first impressions, risking idiosyncrasies and inconsistencies in approach. Rather they can proceed assured that they are comprehensively capturing the information necessary to review these biases in a more systematic way, a matter we will return to shortly.

The ideal approach for capturing this level of detail transforms textual sentences into highly structured and interrelated statements. So as to remain as close to the original sources as possible, these statements can be based on the structure of natural languages with predicate(s), subject(s), and object(s) as the basic framework, each of which can be developed by modifiers: in the case of predicates, by adverbs and adverbial clauses stating, for example, the location of an action, its time (often in the sense of relative rather than absolute chronology), duration, frequency, its causes or circumstances (e.g., by relation to other statements), etc.; in the case of actants—subjects and objects—by adjectives (‘the Waldensians came to see his ill wife’) and appositions (‘Bernarda, daughter of Peter’). It is also crucial for later interpretation to record the modality of these statements. For example, a statement may hold a question on, a doubt about, or even a negation of the action denoted by the predicate, rather than the assertion that the action really happened. Through such semantic modelling, a researcher can record virtually anything natural language allows, including features such as multiple subjects, predicates, and objects (e.g., ‘Peter and William saw, housed, and shared meals with heretics and Waldensians’) and fuzzy data (‘6 or 7 years before that’, ‘at different places’, ‘several times’, ‘and another person she does not know the name of’). Since the basic unit of data collected is a statement rather than a ‘fact’, this approach allows historians to represent conflicting information, distinguish epistemic levels (is this what the source actually says, or an interpretation of the source, or something which is added independently from the source as editorial inference?), and at all times trace back the original context of any piece of information. This context might include, for example, the order in which the information was given, and the interplay of questions and answers in a deposition record.

These features are exactly what Computer-Assisted Semantic Text Modelling (CASTEMO) offers (Zbíral et al. 2022b). CASTEMO is a manually directed approach to capturing sources as structured data. Its workflow is not one of marking up digital full text editions—it does not, in fact, require such editions to have been prepared—but rather of effectively recreating the words of textual sources in the form of syntactic data statements capable of representing almost every semantic nuance. While pioneered within DISSINET, it is now open to use by all researchers, and not only those studying inquisition records, since we envisage that the method will prove pertinent to those working on other sources and topics but with similar needs and challenges. We have also developed an open-source software environment, InkVisitor (Zbíral et al. 2022a), to enable CASTEMO data collection to take place in the most user-friendly and efficient manner.

CASTEMO data collection remains comparatively time-intensive, especially in the form pursued within DISSINET, where we are often capturing sources in their entirety so that the same data can be used to answer the widest possible array of questions. The more selective coding of sources—certain sections, sentences on certain topics or containing certain features, etc., as appropriate to one’s research interests, however, can still offer many of the same benefits, providing an option for researchers whose time and resources are more limited. However far one takes it, the pay-off is a structured dataset that allows for the thorough exploration and analysis of emergent patterns in a ‘scalable’ manner, extending from the micro- to the macro-level. As already mentioned, three types of patterns appear of special interest for research into medieval dissent and inquisition: social, spatial, and discursive. These can be explored through a combination of computational methodologies. While there is only room to discuss just a sample of these approaches here, there are, we feel, several that seem particularly well-adapted to these patterns.

Given that inquisition records are founded on relational information, the search for social patterns connected to religious dissent and to the trial is a most natural focus. Systematically analysing human relationships and interactions, the building blocks of the social fabric, paves a way to exploring supra-individual social phenomena—such as collective identity, coordinated action, and religious belief and expression—in a way very different from more common (and more deterministic) top-down explanations of social realities: as wider patterns emerging from the local actions and decisions of particular human actors. Rather than attributes of clearly delineated social groups, dissident and inquisitorial beliefs, practices, identities, and institutions can be regarded as fluid, socially communicated, and negotiated.

An outlook that understands social, political, and religious phenomena as locally emergent and bound to small-scale interactions and relationships has often held great appeal to historians. It is an aspiration that echoes the intricate prosopographic social approach made famous by Lewis Namier and his followers, including K. B. McFarlane in his study of Lollard knights (Namier 1957; McFarlane 1972). But while such prosopographic studies have sometimes been critiqued (often somewhat unfairly) for overstating the influence of personal, pragmatic interests, social science has now provided a methodological toolbox—social network analysis—for studying local emergence in a way that can respect not only how individuals produce socially negotiated cultural features, but what these produce in turn. The quantitative techniques it provides also allow for this line of study to proceed to new levels of scale.

At its core, social network analysis allows us to explore and explain how connections between nodes (typically people), those nodes’ individual characteristics, and attributes shared by nodes (e.g., same/different gender, same/different place of residence) might be interrelated. Social network analysis found its way into history at the turn of the 1990s, and has since then inspired the lively, emerging field of historical network research.29 Social network analysis techniques can be readily applied to data encoded in the relational manner that we have suggested, both individually and across multiple texts. It thereby offers systematic outlooks on questions that historians have previously only been able to treat through suggestion. For instance: What factors (e.g., gender, kinship, occupation, spatial proximity, reciprocity, etc.) may have contributed to the observed shape of a network of dissident interactions (whether it be their recorded totality or just specific types)? What is the relative importance of these factors? What is the position of men and women, religious specialists vs. supporters, providers vs. users of gathering spaces, etc., in the network of dissident interactions portrayed in the different inquisition registers, and is the pattern consistent across multiple registers? Does the resulting structure of the networks correspond to the supposed goal of inquisitors to emphasise the organised nature of heresy? What were the recorded flows traversing these networks (e.g., information, beliefs, narratives, money, gifts, books, spiritual benefits, etc.), who engaged in them, and what do they suggest concerning the material, spiritual, and information economies that existed among those investigated?

While these questions on dissidence can and should take into account the available information on the interaction at trial, the defining context in which information was produced, social network analysis can also answer direct questions on the trial itself and the different biases potentially in operation: Were some categories of people targeted by inquisitors more than others,30 taking into account their preponderance in the records, the intensity with which information was sought on them, and the severity with which they were treated? Do inquisitors appear sensitive to certain types and quantities of social connections when investigating and punishing their subjects? How significant do such biases appear to have been in guiding their actions, in comparison to the weight of the crimes they actually record? Are there patterns in how defendants incriminate others, and if so, what are they? For instance, do they seem to protect their close kin more than other people? What other factors might influence those patterns, including the development of the investigation and path dependencies? How did inquisitors act upon the evidence they had gathered, and how, in turn, did defendants adapt their strategies as the trial progressed?

While the promise of social network analysis for the study of medieval religious dissent and inquisitorial trials has been observed (Ormerod and Roach 2004) and explored in inspiring initial studies (Timberlake-Newell 2012; Pouivet and Schulz 2014; Nieto-Isabel 2018; Rehr 2019), the challenge is to significantly scale up in terms of both coverage and depth of data collection. Another important goal is to proceed from descriptive statistics and exploratory visualisation to more advanced modelling techniques which enable specific hypotheses about dissidence, inquisition, and even some aspects of premodern social networks more generally to be tested. The statistical models for social networks have not yet been deployed within this field, nor indeed in historical network research more broadly. The combination of the multilateral approach we have described (on the one hand, studying the social networks charged with heresy; on the other, analysing the contexts and associated biases of the trial from a social, interactive perspective) and these now well-developed modelling techniques holds particular promise. We can even make some of the biases of our sources (such as the documented use of the previously mentioned coercive techniques and the originator of information, i.e., who reports what) a part of our models. This capability of looking at several potential factors at a time and across a large sample of cases, and thus observing whether the hypothesised effect of a factor holds true even when a potential confounding factor is controlled for, is a feature of quantitative research that historians have a lot to gain from.

A similar proposal for a more systematic use of formal methods can be made for the spatial patterns of dissent and inquisitorial action. Taking Duby’s call for a ‘cartography of heresy’ in its most literal sense, the formalisation of the spatial relations made possible by the syntactic-semantic modelling of inquisition records allows us to address this task in a qualitatively new way, informed especially, if not exclusively, by geographic information science. This field offers analytical techniques that open up new perspectives on questions such as the geographic coverage of particular inquisitions, spatial accessibility biases in local inquisitions, the mobility of suspects (e.g., of dissident ministers in Cathar and Waldensian religious culture, of women vs. men, etc.), the differences in reported regional dissemination of specific heterodox beliefs, as well as the relation between dissident activities and various geographic and demographic factors. From a broader social history perspective, simply mapping the relations between people and places mentioned in these records (including, for instance, those expressed by notation of residence, origin, or even suggested by toponymic surnames) has significant value. Drawing on multiple datasets, we can meaningfully test hypotheses concerning the broader relationships between space, religion and repression: Do people from settlements of certain sizes seem disproportionately represented in inquisition records, and if so, what are the most likely reasons for this? What spatial patterns and factors are there in the diffusion of incriminations? What patterns do we see in the location types reported as places of dissident communication: private houses, specific indoor spaces potentially signalling concealment (cellarium, solarium, granarium), open spaces within settlements, open spaces outside settlements etc.? Do these patterns correlate better with region, the phase of inquisitorial activity, or the kind of dissident culture? What are the patterns of the suspects’ mobility, and what impact do these patterns have on dissident social networks? Of course, source-critical caveats apply here as well, such as the reach and focus of particular tribunals and the impact of documentary losses. Patterns may be evidence of genuine regularities in ground-level dissident activities, or may be artefacts of biases. But a more complete recording of spatial data will make judging this question easier, not harder. For instance, how can we theorise that inquisitors were imbued with the topos of the hidden nature of heresy, if we have not first ascertained whether they actually represented spaces signalling concealment any more than others?

If social and spatial analyses offer enticing possibilities, including for our understanding of the biases present in the records, it might be remarked that, on their own, they do not truly engage with the lessons of discourse analysis and the broader discussion on the repression of heresy through texts.31 In fact, however, the relational approach that lies at the heart of the statement-based data collection we propose in DISSINET seeks to create new engagement with both the role of language in the exercise of power (inquisitorial categories and topoi,32 stylistic choices, rhetorical framing of punishments, etc.) and the study of narrative sequence and modes of emplotment in inquisition records.

Recording data in the form of a comprehensive syntactic-semantic model of the text via CASTEMO itself provides a powerful framework on which to build. Such an approach does not have to smooth away the quality of expression, and this is part of the reason why our proposal for data collection has stressed a structure derived from natural language as well as the use of the original languages of our sources. At the simplest level, each coded relation can essentially be defined by the actual predicate used in the source, and they can thus be categorised semantically. Denser coding of the modifiers of both actants and predicates, again closely following the original wording, allows for further investigations founded on the terminology of relations (rather than just the basic reality they appear to represent). For instance, we might ask to what extent were inquisitorial procedure and punishment influenced by subtle distinctions of crime types—e.g., someone who ‘received’ (recepit) heretics vs. someone who ‘held’ (tenuit) heretics in their house vs. someone who was defined as a ‘host of heretics’ (hospes hereticorum)—or, allying with social and spatial analyses, to what extent offenders to whom different semantic categories were applied (e.g., heretici, credentes, fautores, ductores) played different roles within reported networks and enjoyed different geographical associations or mobility profiles. The connections between units of meaning (e.g., different concepts) and social networks can be further explored through socio-semantic network analysis, a field now in development in the social network analysis community (Basov and Brennecke 2017, pp. 87–112).

The fact that CASTEMO preserves every twist and turn of source narratives also grants unique perspectives on the intertwining of inquisitorial, notarial, and deponent discourses. For instance, the similarities and divergences between narrative paths concerning multiple suspects can now be modelled systematically, allowing historians to gain focus on patterns that are arguably even more subtle than ‘excesses of speech’ or ‘surpluses’. By breaking down such textual sequences into a series of phases—coded patterns concerning types of confessed action, inquisitorial topics of questioning, or both, definable at various levels of summarisation—we can not only gain a picture of the breadth of discursive possibilities, but also model the apparent dependencies of these paths both on internal narrative features and other factors, including spatial and social markers and measures. This new understanding of narrative sequences can in turn guide a re-theorisation of the agencies, perhaps even voices, that drove them: the divinatio described by Bruschi can take place on a more secure, quantifiable footing, and, crucially, at every level of scale. Scalable reading indeed can be seen as a critical advantage of what statement-based data collection enables, complementing the ‘close’ reading of classical qualitative work and the ‘distant’ reading (cf. Moretti 2013) now offered by computational/quantitative text analysis techniques, which certainly also hold promise for the analysis of inquisition records.

Overall, the modelling and analysis of inquisition records as complex webs of relations with social, spatial, and discursive qualities can be seen as a new form of the ‘serial history’ promoted by some Annalists.33 It provides a way of drawing a sizeable body of data from inquisition registers and subjecting them to meaningful forms of quantitative analysis; but crucially, it does so in a way that is far less reductive of the complexities of the sources than other formalised approaches to historical study. Rather than strip-mining the already-quantitative elements from historical documents or reducing qualitative evidence to free-standing facts, both of which carry the attendant risk of severing details from their textual context and its vagaries, this approach, in which every semantic detail matters, amounts to a new, computer-assisted way of conducting source criticism in parallel with historical analysis. The biases which affect inquisitorial records remain critical to any understanding of these documents. But those biases were constructed from human interaction, from relations, and as far as they are evidenced within the text, this methodology provides a way both to formalise our understanding of them, and to cut through them without obliterating them. When drawing social and spatial maps of dissident communities, for instance, or looking at the discursive patterns used to describe them, we can analyse and compare information derived from trials held in particular places, by particular inquisitors, and transcribed by particular notaries, from particular phases of the proceedings, from depositions by particular witnesses, and even from particular types of statements. The daunting and seemingly impenetrable fog of complexity found within these sources is broken down into a mosaic of individual yet interactive elements, which can be analysed on their own, in small clusters, or in their totality by the techniques we have described.

7. Conclusions

In the study of premodern heresy trial records and their biases, much more has been at stake than the reliability of one genre alone. Indeed, it has served as a focal point for far-reaching reconsiderations of source criticism, the ‘linguistic turn’, the crisis of representation, the anti-inquisition heritage of the Enlightenment,34 the ways of engagement with our sources and with people of the past, even research ethics. These sources have also served as a laboratory for testing new approaches to history, such as those influenced by anthropology and by discourse analysis. The results have had implications for our understanding of history and its sources that reach far beyond the study of heresy and inquisition.

In spite of the attention paid to inquisition records, which have captivated historians as far back as the sixteenth century, and more systematically since the end of the nineteenth century, there are still many unexplored avenues through which research into dissidence and inquisition can be renewed and the historian’s craft enhanced. Systematic computer-assisted approaches to inquisition records capable of translating these records into large series of structured data offer multifaceted possibilities. It is now possible to both theorise and actualise systems that fully retain the features that point towards the particular social and cultural conditions from which the records derive: the conceptual nuances, the uncertainties, the trial contexts, the conflicting testimonies, the chains of information transmission, and the discursive patterns characteristic of our sources. Turning all these features—and not just what might be seen as the ‘positive’ data on dissidence—into structured data accessible to various techniques of analysis acknowledges not only the fruits of classical source criticism but also the postmodern reconfiguration of bias: the latter’s focus on the situatedness of knowledge, the power inscribed in language, the strategies of resistance, and the discursive production of transgression and even of the deponents’ ‘subjectivities’ can be approached with fresh eyes.

Much research is needed before the techniques discussed here will bear their full fruits. Nevertheless, the overall approach is one that will allow historians to treat documents as dissimilar as the depositions of Giacomo Bech and the sentence of Laurence from Montauban both in detail and as part of the same continuum, and to handle qualitative subtlety as well as quantitative bulk. Most critically, it represents a first step towards inscribing a new, computer-assisted variety of source criticism at the very heart of research in a systematic, transparent, and reproducible way.