Investigating Crime in the Vector Space of Early Modern Women’s Writing

Investigating Crime in the Vector Space of Early Modern Women’s Writing

by Lara Rose, PhD Student: Literature, Northeastern University

This post is part of the Women Writers Project sub-project that explores the intersection of text encoding and text analysis using word embedding models. For brief explanations and early explorations, please see this post by Elizabeth Polcha, and this post by Jonathan Fitzgerald.

“When a woman killed her husband, it was petty treason—a crime against the nation—When a man killed his wife, it was only murder.”
-Professor Marina Leslie, on gender and crime in the early modern period

My weeks in the 1990s were outlined by watching the antics of Calista Flockhart in Ally McBeal while tucked at the foot of my parent’s bed. If I was at a friend’s house, our conversations were punctuated by the sound (I know you know it) of the scene transitions of Law & Order. Now, visiting my parents always involves trying to follow the plot of the CSI: Somewhere that’s on in the background or watching too many episodes of Drop Dead Diva on Netflix while I should be studying. I am obsessed with televised courtroom and crime dramas.

I actually didn’t realize until writing this post that a life filled with crime-story-consumption undoubtedly influenced my choice to write my M.A. thesis on the fiery courtroom drama of Anne Hutchinson. I was enthralled by her voice, sharp within the court records of the seventeenth-century Antinomian Controversy, like the Puritan version of Ally McBeal (though Ally was never cast out of Boston for heresy). Oddly, I don’t read crime fiction, and I’ve never cracked open a spine that said “John Grisham,” but I’m in the minority on that one—Crime/Mystery Fiction was the second-most popular genre in 2014, grossing $728.2 million in book sales the U.S. People love reading about crime, and they have for centuries.1

Curiosity about the historical obsession with crime and punishment sparked my entrance into the word vector space of the Women Writers Project. Particularly, I am interested in the ways crime and legality might have gendered associations within the corpus. As a basic query, I used the variations on the analogy function to look for words that were located near “crime” and “woman” but not near “man.” The crimes that appeared seem mostly related to sexual morality or to reputation: coquetting, perjury, theft, sex, indecency. For masculine crimes, the results seemed more general or related to loyalty (to self, to god, to the state). The list of words near to “man” and “crime” included crimes like treason and disloyalty and the name of a Greek general executed for conspiracy. Otherwise, the “masculine” list comprises more general concepts like sin, misdeeds, or vengeance. (You can see a table below of the full results and the code I used to get them).

My current level of knowledge within the world of crime or legal jargon is based mostly on modern popular culture, so I opted to ask for help from an expert: Professor Marina Leslie, who specializes in crime in early modern literature. When I heard she is currently finishing a book manuscript studying the criminalization of female productive and reproductive labor in 17th-century English print culture, I knew I wanted to reach out to her with the intention of generating a more robust vocabulary on which I could perform the word search function of the word2vec package.

Professor Leslie was not surprised to see sexual words or words relating to a moral reputation associated with women. She also reminded me that “coquetting” hadn’t gained ground until the eighteenth century—so it would have stemmed from the later part of our corpus.2 Beyond that, though, she expressed initial surprise that the basic findings did not have stronger negative connotations. Ultimately, she suggested that our data (women writers writing about women) is not likely to be indicative of the way women were typically written about in the early modern period.

As one example, she posited that many of our writers are likely defending against accusations against women made by male authors in texts. In fact, this comment reminded me that, in the research I am doing for WWP’s Intertextuality Networks Research Project, we have already uncovered several references to the work Arraignment of Lewd, Idle, Forward, Unconstant Women—a publication about which our women writers were furious.

The presumed differences between how authors of different genders might write about women prompted Professor Leslie to wonder how the word vector analyses would look different between the corpus of the WWP and another corpus like Early English Books Online.3 Along the same lines, we also discussed how the Women Writers Online text base may be too “upscale” for exploring crime in the early modern period. Our women writers are educated and, often, noble—thus perhaps less likely to engage in discussions that would have abounded in the popular cony-catching pamphlets (Coney-catching is Elizabethan slang for theft through trickery—so you can think of these pamphlets like the sixteenth-century ‘zine version of True Crime).

However, Professor Leslie brainstormed how we could also use the education of the female authors to our advantage. Could we gain insight into issues of religious and political crimes, or of women’s lived experiences of civil war, the Protestant Reformation. What does the gendered work look like of establishing oneself on the “correct” side of a political divide? Lucy Hutchinson and Margaret Cavendish lived on opposite sides of the political spectrum—can the vector space help us see how politics might change the writer’s ideas of what is considered criminal?

After Professor Leslie and I discussed initial findings and possibilities and limitations specific to our data, we delved into the notion of how to increase the search functionality of word vectors. She reminded me that, no matter the search term, it will be important to generate a list of grammatical and historical spelling variations of each query term, as well as vaguely associated terms: search “crime” and “criminal,” “murder” and “murther,” “blood” and “bloody.” For research in the early modern period, she even suggested using cant dictionaries (street dialect for thieves) as a starting place. That way, I would be able to search for specific crimes. Gender historians already know that crime was certainly gendered in the early modern period, and that there were professional classes of criminals that tended to be composed of mostly women or people who identify as women. Off the top of her head, Professor Leslie listed doxy, bawdy, bawd (a female pimp), or launderers as crimes or criminal profession for femme people. On the other hand, she said that men or people who identify as male would be peddlers, pickpockets, rogues, or would masquerade as sailors.

In general, just as I suggested in my initial findings, Professor Leslie agreed that women’s crime is likely to be coded sexually—relating to intercourse or inter-personal relationships (prostitution, bawdiness, infanticide, murder of husband, chastity or unchasteness). One of the most interesting crime tidbits she shared was a 1624 law that changed the way the legal system treated the death of a child or infant. If an unmarried woman became pregnant but the child did not survive, there was a legal assumption of guilt (rather than the “innocent until proven guilty” burden of proof in the United States). This means that legal authorities assumed that an unmarried woman would have murdered her bastard child. By and large, beginning in 1624, the death of a child was largely considered to be infanticide—a capital murder. For a woman to prove her innocence, she needed two witnesses. Midwives, though, who were most likely to know the cause of death, were not trusted due to their associations with witchcraft.

The mention of witchcraft spurred a fast-moving and passionate discussion (as witchcraft tends to do). Rather than writing more about the many veins the Professor Leslie illuminated as possibilities in the Women Writers Online corpus, I’ll list them here in bullet points. Comment below on which thread you think I should pick up next!

  • Crime and poverty : the “poor laws” enacted by Queen Elizabeth in the early modern period. Professor Leslie explained that these laws made parishes responsible for their poor people, which bred resentment and often led to officials enacting extreme measures to chase impoverished people into other parishes.
  • Gender-neutral crime: What is at the intersection of women and thief?
  • Can we find ways to tie this research to issues of gender with significant modern implications, like rape and sexual violence or assault?
  • Could we use the vector space to explore the space of early modern crime? Where does crime happen? (Several centuries ago, coffee houses were not the clean, well-lit places we all go to study.)
  • What about issues of slander and libel? As Taylor Swift’s new album reminds us, Reputation is everything.
  • And speaking of reputation, what about that original non-compliant woman, Eve?
Table of Initial Findings for Vector Space Query into Gender & Crime
Category Feminine Associations: closest_to(~’crime’+’woman’-‘man’,40) Masculine Associations: closest_to(~’crime’+’man’-‘woman’,40)
Legal Terms accused












Positive Connotations innocent






sory [sorry]

Negative Connotations shame











injustice pollute

Specific Crimes or Accusations coquetting














Neutral, General, or Unknown child













  1. That is, if we can trust this information I found via the internet:
  2. My next research question will be: how does the vector space change over time?
  3. Another possibility for a research question! Look for the list at the end of this post for many more possibilities.

Leave a Reply

Your email address will not be published. Required fields are marked *