A humanist doing a detailed reading of a single text and a humanist analyzing a body of texts are performing two very different activities. A researcher addressing published accounts of early modern witchcraft trials in England, for example, must hold about ninety texts in her or his mind in order to locate patterns across and among them. This particular pattern-finding behavior is the basic research function which John Unsworth includes in his taxonomy of “scholarly primitives” under the name “comparison” (Unsworth). Comparison is unique among the scholarly primitives in that it includes, by definition, more than one object of study; other primitives, such as annotation or representation, do not absolutely require more than one datum. Comparison’s uniqueness, however, involves more than its research context. Because it necessarily incorporates multiple objects within a single research focus, comparison demands a kind of study which literary critic Franco Moretti calls “distant reading.” In order to find overall patterns among texts, the texts themselves must “undergo ... a process of deliberate reduction and abstraction” (1). Distant reading is, Moretti argues, more than a research practice; it constitutes “a specific form of knowledge: fewer elements, hence a sharper sense of their overall interconnection. Shapes, relations, structures. Patterns” (94). This is the kind of knowledge one aspires to in the study of a body of related texts, such as published accounts of early modern English witchcraft trials.
Access to the forms of knowledge which comparison can provide, however, grows more difficult with any increase in the complexity or number of the texts involved. At some point, the reduction of elements necessary for pattern-finding produces a set of objects for study that is not an accurate and/or sufficient representation of the texts involved; getting a good view of the forest may mean losing sight of the trees. Building on a similar landscape metaphor, Stan Ruecker points out that large digital collections of texts present the humanities researcher with, essentially, a problem in prospect: the desire for the broadest possible scope for comparison needs to be balanced with the need for meaningful representation of the individual texts involved (3).
English witchcraft trial documents present particularly acute demands on this scholarly balancing act. Their web of witnesses, accusers and reporters; their geographical diversity; their deployment (and redeployment) of legal discourse; and their inclusion of supernatural evidence cheek-by-jowl with the most material kind of “proof” render even an individual report difficult to decode. Comparison among multiple reports needs to remain sensitive to these complex local nuances. More importantly, it must keep in view the unexpected slippages and elisions that typify the texts; the process of “reduction and abstraction” should not obscure the fact that accounts of English witchcraft trials may ignore, confuse, or even invert distinctions among semantic categories. The boundaries between incantation and prayer, animal and demon, accuser and defendant are fluid and dynamic. The term most central to the texts — “witch” — is perhaps the most slippery of all, resisting even contemporary efforts “to categorize, classify, and contain” (Uszkalo 170).
In this paper, we will address published accounts of early modern English witchcraft trials as an example of a relatively small digital corpus which nonetheless challenges the limits of prospect and representation. Ideally, the interface design for a digital collection should expedite and even expand the research activities of the humanist, offering enhanced opportunities to interact with the texts. We envision this interaction in terms of what Stephen Ramsay calls “algorithmic criticism”, which aims at “provoking thought and allowing insight” (173), enhancing the higher functions of analysis rather than simply supporting data collection. As our guide to interface development, we have employed the witch’s familiar, a figure which is among the slipperiest, the most difficult to grasp, in our target texts. Our aim was to produce a means of interaction that could not only accommodate the familiar — tracking it across documents while retaining its complex semantic role — but also imitate its fluidity and dynamism. The result is the Witchcraft in Early Modern England (WEME) project (www.kirsten.uszkalo.com/weme), which offers users opportunities to interact with trial records using multiple representations of individual terms and semi-supervised classification tools for manipulating these representations.
English witchcraft trial documents constitute a highly localized area of study, both historically and geographically. Witches had long existed in England. When they became part of the legal and literary record, they at last left a textual trace, which records just how real these legendary women were to their contemporaries. Their numbers were small; those legally executed in England, according to C. L’Estrange Ewen, totaled somewhat less than 1,000 (112). According to Gregory Durston, England experienced the least spectacular of the European witch hunts. As a disproportionate number of England’s witchcraft indictments and trials came from Essex and, to a lesser extent, Lancashire and Kent, “about 90 per cent of villages would not have produced a single hanged witch and many not a single formal allegation of witchcraft” (7). Despite the relatively small scale of England’s witch-hunt, there were two curious spikes in its witchcraft persecutions: the first in the mid-sixteenth century, and the second in the mid-seventeenth. The most infamous increase in witch-hunting activity occurred between 1640 and 1660, after a long lull, and is largely attributable to Mathew Hopkins, its self-styled Witch Finder General. This singular craze “resulted in the execution of several hundred witches in Essex, Suffolk, Norfolk and neighboring counties” (Durston 7). The work done in the courtroom during these periods of persecution did more than create stories for the printing presses; it reified the reality and power of witches and their familiars.
The creation of the Gutenberg press enabled the production of cheap pamphlets, creating a public space for scholars (and would-be scholars) to debate the details and plausibility of the witchcraft beliefs. Although witchcraft accusations were part of an oral-cultural landscape, the publication of witchcraft tracts created witches by creating controversy about them; the theoretical witch gave rise to the real one. A broad range of opinion informed these tracts. Johann Weyer, for example, believed in the Devil, but not in witches, seeing women accused of witchcraft as being sick, crazy, deceived (or all three), and in need of religious training (Clark 13). George Gifford argued that witches exist, but are not in themselves powerful nor numerous; Satan gives the witch any power she has. In Dæmonologie (1597), James VI of Scotland claimed that the Devil gives witches image magic, medicinal magic, and poisons with which to harm their enemies. At a much later date, Joseph Glanville wrote Saducismus triumphatus (1682) to prove the existence of witches and their real power.
The continued popularity of witchcraft tracts continued even when English witchcraft executions had dwindled to almost nothing, a circumstance which speaks to the tensions the witch aroused. The public remained sharply concerned about the women who had once spoken in their midst. Robin Briggs argues that “there is no clear link between those writers who advocated sweeping campaigns to eradicate witches and the messy reality of what were largely small-scale local persecutions” (262). The real-time influence of Glanville, or any of the authors mentioned above, is difficult to prove. As more texts were published about witchcraft, more scholars wrote in response. As greater critical interest developed, judges and witch hunters had an increasing arsenal of broadly written guidelines, outlining both what witches could do and why they should be punished. Ultimately, more women were discovered to be witches once witches were written about. Just as the witch was understood to create a reality with her words, so did her critics.
In the case of early English witches, researchers are still struggling to understand who these women were and what they represent. Women in early modern England did not suddenly appear as witches, scratching at their neighbors’ doors. Nor did interpersonal tensions between women spontaneously combust into trials and executions. A woman had to be seen and publicly labeled as a witch before being prosecuted for witchcraft and, as a result, appearing in print. She needed to receive the forthright endorsement of her community before she could even begin to be publicly persecuted. Her physical body needed to display, or adopt, the witch’s mark, and she had to be associated with a familiar who would do her magic for her, a feature which distinguishes the English witch from her Continental sisters.
The familiar offers a case study in the difficulty of making comparisons across and among English witchcraft trial documents. Its role went far beyond its ability to establish a witch’s identity. Familiars were traded as commodities between individuals or households. They might also be exchanged between neighbors as a symbol of initiation into a demonic community. Witches could compel a familiar to labor by providing it with food and shelter, but it was still capable of disobedience or outright rebellion, in which case the witch could try to discard it. The familiar was capable of changing its own form. More importantly, the familiar was the witch’s second “self,” acting for her and embodying her power; harming a familiar would injure the witch it served. Even the “witch’s mark,” which English courts recognized as proof positive of a witch’s identity, was the mark of the familiar. According to many contemporary trial reports, these unusual moles, marks, or nipples were sites where the witch had fed her familiar with her blood. The familiar offers little in the way of self-definition, bedeviling taxonomies which seek to classify it: is it a devil? a faerie? a traveling spirit? (Purkiss 84; Lecouteux 26). It is hard to pin down, yet it permeates English witchcraft texts — defining, connecting, transforming, transmitting and confounding.
Take, for example, the following account from John Phillips’s The Examination and Confession of Certaine Wytches at Chensforde (1566). The Examination is an excellent source text, as it is the first account of witchcraft in England, and introduces the witch’s familiar (and one that changes shape at that) to the English literary scene. The document begins with Elizabeth Francis’s testimony that her grandmother, Eve of Hatfield, taught her to renounce God and give her blood to her familiar, Sathan. Sathan appeared as a cat, which she kept in a basket and fed with bread and milk. Francis used Sathan to increase her wealth (in sheep) and to try to get Andrew Byles to marry her. Sathan assured her he had the power to guarantee a marriage if she consented to let Byles “abuse her” first, which she did. Francis became pregnant, and he left her. She then lowered her expectations and used Sathan to satisfy her desire for retribution, instead; she could not make Byles love her, but she could make him and his unborn child disappear forever. Francis herself, according to the Examination, attributed Byles’s death and her own miscarriage to Sathan’s agency. She did eventually marry and later testified that Sathan subsequently turned into a toad in order to lame her husband and kill her child, demanding payment in the form of a drop of Francis’s blood.
Fifteen years later, Sathan was passed on to Francis’s sister, Mother Agnes Waterhouse, and continued to provide a variety of malefic services: destroying livestock, ruining beer and butter, and killing people. Waterhouse kept him on a warm bed of wool in a pot and fed him well in blood and chicken for his efforts. Her daughter, Joan, frightened by the sight of Sathan emerging (in the form of a huge dog) from the darkness under her mother’s bed, revealed to him that she wanted her little neighbor, Agnes Brown, punished for refusing to give her bread and cheese. Sathan soon after appeared to Brown as a “thing like a black dog with a face like an ape, a short tail, a chain, and a silver whistle (to her thinking) about his neck, and a pair of horns on his head” (sig. A4v). Next, Sathan arrived at Brown’s milk-house, carrying the key to its door in his mouth. He opened up the milk-house and took what he pleased.
Elizabeth Francis’s appearance in witchcraft tracts, however, was far from over. She reappears in A Detection of Damnable Driftes (1579), this time testifying against Elizabeth Lord for poisoning John Francis and bewitching Joan Roberts. She also denounced Mother Osborne, a local widow, as a witch:
[S]he hath a marke in the ende of one of her fingers like a pit, and an other marke uppon the outside of her right legge, whiche she thinketh to bee pluckt out by her Spirit [i.e., Sathan]: and that one Mother Waterhouse her owne sister (long since executed for Witchcrafte) had the self same markes, whithe she termeth (nippes). (sig. A5)
Francis, whose detailed confession seems to have secured her from persecution, here establishes herself as an authority on witches. She is able to recognize the physical indicators that signified a diabolic pact; they were the ones her own sister had. Although women often testified against each other in English witchcraft trials, Francis manages to use her own identity as a witch as the basis for arrogating the role of witch-finder. By means of Sathan, she traverses a path from witch-apprentice, to witch, to court-recognized witch-hunter.
A Detection of Damnable Driftes crucially illustrates the familiar’s role as a social connector, and its legal importance as a proof of a witch’s identity. However, it also establishes an evolving slippage between the subject identities of witch and familiar. Elleine Smithe, whose story follows Francis’s in A Detection, had a similar genealogy of witchcraft: she was the daughter of Alice Chaundeler, a woman executed for witchcraft. She was also under suspicion on her own account for the untimely death of Window Webbe’s daughter and the death of her estranged father. So when Elleine Smithe’s son begged at John Estwood’s door and was turned away, Estwood’s subsequent pains were attributed to Smith. Later that evening, Estwood thought he saw a rat run up the chimney and a toad fall down–a supernatural transformation. Estwood, seeing an opportunity to strike back at Smithe, grabbed the toad and “thrust it in the fire and so held it forcibly. It made the fire burn blue as azure, and the fire almost out; and at the burning thereof the said Elleine Smithe was in great pain and out of the quiet” (sig. A6). In these events, Smithe’s direct personal involvement is minimal. Instead, she acts and is acted upon through the body of her familiar. The rat-toad is not, however, simply Smithe’s proxy: its presence makes Smithe legally culpable for what transpired in Estwood’s home, as though she had been there in her own person. In terms of its agency and legal status, Smithe’s ‘self’ is thus partially constituted through the ‘self’ of her familiar.
This complication of identities can extend beyond the familiar’s bodily boundaries. Smithe and the rat-toad experienced physical distress jointly, as though they were connected, suggesting that both identity and materiality become permeable between the two. Hurting the familiar is literally the same as hurting the witch. The familiar retains, however, a robust materiality of its own. The rat-toad is not a spirit-animal; in spite of its physical fluidity and ambiguity, it has a concrete form, one capable of being grasped, manipulated, and burned. Smithe’s son later confessed in court “that his mother did keepe three Spirites, whereof the one called by her greate Dicke, was enclosed in a wicker Bottle: The seconde named Little Dicke, was putte into a Leather Bottle: And the third termed Willet, she kepte in a Wolle Packe [i.e., tight bundle of fleeces]” (sig. A7). Though Smithe’s familiars go by the appellation “Spirites,” they are physical beings who can be confined by physical means. The hardy containers Smithe chooses for her familiars’ homes suggest, in fact, a dangerously active material presence.
Critic Frances Dolan pinpoints the peculiar challenge of early English witchcraft texts when she notes that witches and familiars actively dissolve surfaces of separation between “dangerous and familiar,” “self and other,” and “immaterial and material” (5, 184, 194). The familiar’s fluid identity and its tendency to appear across documents offer further examples of its ability to resist categories of separation. How can one make comparisons among these texts, when the boundaries of core terms like “familiar” are, even to a close reading, indistinct?
Existing retrieval interfaces are demonstrably unequal to the challenge, if only because of the amount of data involved. All of the texts mentioned above, and many others like them, are available in digital format from the Early English Books Online (EEBO) database, an invaluable resource for researchers studying the early modern period. Its current interface is constructed around keyword searches designed for retrieval functions. Searching “Sathan” produces 55,092 hits in 4,685 of the database’s records; a more complicated search string–“Sathan AND cat AND witch”–produces 18,519 hits in 377 records. In both cases, the amount of data returned is unmanageably large. The search results appear as a bibliography of relevant records, listed ten to a page: the keyword “Sathan” thus produces over 5,500 such pages of material. EEBO’s interface provides tools for weeding and/or sorting these results, or for refining them into a smaller subset. Its keyword searches are also flexible, offering multiple fields linked by Boolean connectors, so that it is possible to define a more specific initial search. The interface offers, however, no tools to facilitate distant reading of the listed texts — only ways to select fewer of them. Faced with all of EEBO’s references to “Sathan,” the user’s best resource seems to be: look at less.
More importantly, keyword searches obliterate the very qualities which best characterize the familiar: dynamism, fluidity, ambiguity. The keyword “Sathan” excludes altogether the mysterious white dog connected with Elizabeth Francis’s final recorded appearance, which may (or may not) have been Sathan in another guise (Detection sig. A4). The search term “animal” overlooks Sathan’s multiple metamorphoses into specific animals, such as cat, toad, and horned ape-dog. If a researcher is looking for a particular text string, a retrieval interface will faithfully locate it, but no text string can adequately represent Sathan’s remarkable tale, the idiosyncratic evolution of Sathan’s shape and role, or the connections between the witches and families mentioned. The aim of the WEME project is to overcome these limitations, actively assisting the researcher in finding the nuanced meaning, the slipping subjects, and the connections between and inside texts. Digitization of English witchcraft documents should make these aspects of the texts easier and faster to find and should let the researcher see across the corpus without reducing its meanings to spots on a flattened landscape.
Its ability to trouble the points of separation and interaction makes the familiar interface a useful metaphor for rethinking interface design for early modern textual study. If one accepts Gui Bonsiepe’s definition of “interface” as a tool which enables “purposeful action” (29), then the witch’s familiar is itself an interface, a tool through which she acted and interacted with her environment. Its ungovernable and often unpredictable qualities provided affordances for interactions of tremendous variety. The WEME Project is designed to provide its users with an interface modeled on affordances of the early English witch’s familiar. As the familiar allows for slippage between the witch and itself, the interface should be able to treat individual texts as both a part of and apart from existing collections. Further, the interface must consider not only the idiosyncratic nature of the corpus, but also the ways in which texts within it speak to each other, separating out an idea and illustrating how it interacts across texts. Both the information design and the interface tools of the WEME project aim at explicating the complex nature of the textual witness to the early modern English witches by foregrounding both the witch and her familiar as objects that are themselves forms of interface. In the same way that a familiar could and did transform into various kinds of creatures while still remaining a familiar, so our ideal interface would support algorithmic criticism and unpredictability, where no single representation of the material is presented as though it were a stable categorization. Instead, the interface should provide researchers with the means to observe directly what is unstable and multiply valent in the texts. Like the familiar, it should be mutable both in itself and in the content it presents.
WEME benefitted substantially from its role as a use-case in a much larger undertaking, headed by Dr. John Unsworth (University of Illinois, Urbana-Champaign) and Dr. Martin Mueller (Northwestern University) under the name Metadata Offer New Knowledge (MONK) (http://www.monkproject.org). It builds in part on existing digital scholarship in the field of early modern witchcraft. Cornell University Library, the Text Creation Project (TCP) and EEBO have generously given us permission to use their full-text digitized witchcraft documents in our research. The University of Exeter has produced digital maps of early modern England, which offer researchers fast, flexible access to contemporary geographic data. We are also indebted to the example of the Salem Witchcraft Documentary Archive and Transcription Project, presented online through a collaboration between the Danvers Archival Center and the University of Virginia. This project has garnered considerable international attention with its innovative use of animation and geographic information systems (GIS) (Losh 383). GIS is, put simply, “computer-linked large sets of data with spatial representation in maps” (Marchand and Winchel 175), and the Salem Project’s use of animated maps allows users greatly expanded affordances for interaction with available spatial and temporal data. It is this kind of innovation that WEME aims to emulate and exceed.
WEME provides an alternative to text-string searches in the form of structural, bibliographic, and semantic classifications which can track themes within and across documents. Designed using interpretive XML tagging and relational databases, its interpretative-level schema includes basic structural tags (e.g., paragraph, page), but also includes publication data and semantic tags (e.g., town, parish, assize, witch, familiar, witch’s mark). A single word, such as “Sathan,” is thus available as a regularized search term that would locate a full set of possibilities, since it would appear as a tag on other words like “animal,” “dog,” “toad,” “ape,” “cat,” “familiar,” etc., rather than just being a string of characters. There are currently approximately 120 interpretive tags in the schema, with 8 at the highest level: Witchcraft location, Relationship, Familiar, Demonic union, Maleficium, Magical conditions, Demonic involvement, and Counter magic. Tagging the documents according to this schema will also ensure that the dataset includes place and time alongside witch and familiar. Pattern analysis within this dataset can then track how the form, function, and understanding of “witch” and “familiar” as complex constructs metamorphose over time and space.
The second phase of this project is an implemented GIS framework over the dataset. The documents’ information about witches will be placed within a relational database which identifies biographical, temporal, and geospatial data. In our preliminary work in this area, we have mapped each witch to a given town and county within England based on GIS data supplied to us by the UK Data Archive. This research has already uncovered an unexpected disparity between the locations where witchcraft trials actually took place and locations where published tracts locate peaks in witch-hunting activity (Appendix 1).
Finally, the WEME project seeks to represent fully the complex variety of the familiar as a metaphor for interface. The WEME interface will create visual clusters of information on the screen, which can be organized in multiple ways, separately and in conjunction with one another. One proposed function, for example, called “Throwing Bones,” will allow the user to select a number of texts and run a clustering algorithm on them to see if and how they relate to one another. Ideally, this function will serve as the front end to an unsupervised classification function like those in the work of Dominic Forest and Jean-Guy Meunier, offering a semi-supervised semantic tool for algorithmic criticism of the corpora (Forest and Meunier 3.3). The themes appear as clusters of three-dimensional symbolic objects related to witchcraft, such as bones or cards, thus offering meaningful representations as well as prospect for the tasks of distant reading and comparison (Appendix 2). The user may interact with the texts simultaneously on multiple levels and allow the algorithmically clustered sets of texts to speak back, telling a kind of story which might otherwise remain obscure under the auspices of standard date, author, and keyword search terms. “The interface,” Gui Bonsiepe argues,
can have either positive or negative effects. . . . [I]t can make learning easier or more difficult. It can be fun and make us relaxed at handling information, or it can be boring and stressful. An interface can illuminate connections or leave them murky and opaque. It can open up possibilities for effective action or obstruct them (62).
The goal of the WEME Project is to open up the possibilities for effective action available in these remarkable documents, to an almost uncanny extent.
Bonsiepe, Gui . Interface: An Approach to Design. Trans. Eileen Martin. Ed. Dawn Barrett. Maastricht: Jan van Eyck Akademie P, 1999. Print.
Briggs, Robin . Witches & Neighbours: The Social and Cultural Context of European Witchcraft. New York: Viking, 1996. Print.
A Detection of Damnable Driftes, Practized by Three Witches. London, 1579. Print.
Clark, Stuart . Thinking with Demons: The Idea of Witchcraft in Early Modern Europe. Oxford: Clarendon, 1997. Print.
Dolan, Frances E . Dangerous Familiars: Representations of Domestic Crime in England, 1550-1700. Ithaca: Cornell UP, 1994. Print.
Durston, Gregory . Witchcraft and Witch Trials: A History of English Witchcraft and Its Legal Perspectives, 1542–1736. Chichester, Engl.: Barry Rose, 2000. Print.
Ewen, C. L’Estrange . Witch Hunting and Witch Trials. 1929. New York: Barnes & Noble, 1971. Print.
Forest, Dominic and Jean-Guy Meunier
. “NUMEXCO: A Text Mining Approach to Thematic Analysis of a Philosophical Corpus.” CH Working Papers A.32 (2005). Print; rpt. 1 June 2008.
Glanville, Joseph . Saducismus triumphatus. London, 1682. Print.
James VI of Scotland . Dæmonologie. Edinburgh, 1597. Print.
Lecouteux, Claude . Witches, Werewolves and Fairies: Shapeshifters and Astral Doubles in the Middle Ages. Rochester, VT: Inner Traditions, 2003. Print.
Losh, Elizabeth . “Reading Room(s): Building a National Archive in Digital Spaces and Physical Places.” Literary and Linguistic Computing 19.3 (2004): 373–384. Print.
Marchand, Michael E. and Richard Winchell . “Tribal Implementation of GIS: A Case Study of Planning Applications with the Colville Confederated Tribes.” American Indian Culture and Research Journal 16.4 (1992): 175–183. Print.
McLafferty, Sara . “Women and GIS: Geospatial Technologies and Feminist Geographies.” Cartographica 40.4 (Winter 2005): 37–45. Print.
Moretti, Franco . Graphs, Maps, Trees: Abstract Models for a Literary History. London: Verso, 2005. Print.
Phillips, John . The Examination and Confession of Certaine Wytches at Chensforde. London, 1566. Print.
Purkiss, Diane . “Sounds of Silence: Fairies and Incest in Scottish Witchcraft Stories.” Languages of Witchcraft: Narrative, Ideology and Meaning in Early Modern Culture. Ed. Stuart Clark. New York: St. Martin’s Press, 2001. 217–236. Print.
Ramsay, Stephen . “Toward an Algorithmic Criticism.” Literary and Linguistic Computing 18.2 (2003): 167–174. Print.
Ruecker, Stan . “Experimental Interfaces Involving Visual Grouping During Browsing.” Partnership: The Canadian Journal of Library and Information Practice and Research 1.1 (2006). Print; rpt. 17 Nov. 2008. Web <http://gir.uoguelph.ca/index.php/perj/article/view/142/177>.
Thomas, Keith . Religion and the Decline of Magic: Studies in popular beliefs in 16th and 17th century England. 1971. Harmondsworth: Penguin, 1991. Print.
Unsworth, John . “Scholarly Primitives: What Methods Do Humanities Researchers Have in Common, and How Might Our Tools Reflect This?” Symposium on Humanities Computing: Formal Methods, Experimental Practice. King’s College, London, UK. 13 May 2000. Address; rpt. 2 November 2006. Web. <http://www.iath.virginia.edu/~jmu2m/Kings.5-00/primitives.html>.
Uszkalo, Kirsten C . “Dark Sisters: Witches and Prophets in Early Modern England.” Diss. University of Alberta, 2006.
─── and Darren James Harkness . “The Witches in Early Modern England Project.” Text Encoding Initiative Annual Member’s Meeting. Victoria, BC. 27–28 October 2006. Address.
 In what he called an “incomplete search,” Keith Thomas claims he was able to add “over 120 witch trials and 22 executions” to Ewen’s numbers, but was “unable to improve on [Ewan’s] overall estimate” (fn3, 450).
 Accusations of witchcraft dwindled over time until, finally, the 1734 Witchcraft Act ended the legal persecution of witches in England.