In the winter of 1469, Margaret Paston dictated an urgent letter to her son about violent men destroying her tenants’ crops. Like virtually all medieval women, Margaret could not write—every letter bearing her name was penned by male scribes who may have subconsciously altered her voice to match their expectations of how a woman should sound. It is assumed Margaret could read, however, which might have helped ensure her intended meaning came through.
Five centuries later, I’ve used forensic linguistics to detect these subtle manipulations, uncovering fascinating patterns that reveal when Margaret’s authentic voice breaks through versus when scribes imposed their own linguistic biases. The detective story unfolds through tiny grammatical clues hidden in pronoun choices and sentence structures.
The Medieval Voice Problem
Medieval letters present a unique challenge for historical linguists. In 15th-century England, literacy rates were devastatingly low—Norman Davis estimates that only 10% of men and 1% of women could even write their own names. This meant that powerful figures like Margaret Paston, despite managing vast estates and wielding considerable influence, relied entirely on scribes to record their words.
But here’s the crucial question: when a scribe writes “I think that we should proceed carefully,” are those Margaret’s exact words, or the scribe’s interpretation of what a respectable woman should say?
Letter of Margaret Paston, in the hand of James Gresham, to her husband, John Paston I, informing him of the assault on James Gloys. 19th May 1448 – British Library Add. MS 39848, fol. 2
The Paston family letters, spanning 1422-1509, offer an unprecedented window into this dilemma. This remarkable collection contains 422 documents totalling 245,000 words—the most complete medieval family correspondence to survive. Margaret appears as author of numerous letters, but analysis reveals they were written by at least five different hands: professional scribes like James Gloys and John Wykes, plus her own sons acting as secretaries.
The Linguistic Fingerprint Method
To detect scribal interference, I focused on unconscious linguistic choices—the grammatical patterns people use without thinking. Specifically, I analyzed two key variables that were rapidly changing in 15th-century English:
Personal Pronouns: The old Anglo-Saxon forms (hem, here) were being replaced by Scandinavian-influenced pronouns (them, their). This change happened gradually, creating a linguistic spectrum from conservative to innovative usage.
here and their change in periods adapted from Bergs (2005:105)
Relative Pronouns: Medieval writers could choose between traditional that, emerging which, and the elaborate the which. These choices often reflected gender, social class, and personal style.
If Margaret had a consistent linguistic personality, her letters should show similar patterns regardless of which scribe wrote them. If scribes were imposing their own styles, we’d see dramatic variations.
The Detection Algorithm
I manually coded every instance of these pronouns across eight representative letters spanning 1449-1469, comparing:
External scribes writing for Margaret vs. the same scribes writing for male family members
Margaret’s sons writing for her vs. the same sons writing their own letters
Patterns across time to track Margaret’s evolving voice
The results revealed a complex linguistic crime scene.
Margaret Paston’s use of personal pronouns across time adapted from Bergs (2005:112)
Evidence of Scribal Manipulation
Professional Scribes Adapting Their Style
James Gloys provides the clearest evidence of scribal accommodation. When writing for John Paston I in 1449, Gloys used the traditional hem six times (0.6% of text). But twenty years later, writing for Margaret, hem completely disappears while innovative they spikes to 1.8% of the text.
This dramatic shift suggests Gloys consciously or subconsciously adapted his language to match what he perceived as Margaret’s evolving linguistic identity. Remarkably, this change aligns with Margaret’s transformation from subordinate wife to independent widow managing family properties after her husband’s death in 1466.
Gender-Based Linguistic Expectations
The evidence becomes even more intriguing when comparing how scribes wrote for Margaret versus male family members. John Wykes, writing for Margaret in 1465, used the conservative form here seven times but completely avoided the innovative which. Yet writing for her son John II the same year, Wykes used which seven times and here only twice.
This pattern suggests scribes subconsciously feminized Margaret’s language, giving her more traditional forms they associated with women while reserving innovative constructions for male voices.
The Sons’ Dilemma
Margaret’s sons John II and John III present the most complex case. When writing as themselves, they used distinctly different linguistic patterns than when writing for their mother. John III, writing his own letter in 1461, used the traditional hem four times but never used it when writing for Margaret two years earlier.
This reversal is puzzling until we consider the social dynamics. The sons may have felt pressure to make their mother sound appropriately feminine and conservative, even when their own speech was more innovative. Alternatively, they might have been more faithful recorders of her actual conservative speech patterns.
The 1464 Linguistic Revolution
The most dramatic revelation comes from letters dating to around 1464, when Margaret’s husband was imprisoned. John II, writing for his mother that year, records her using highly innovative pronouns—six instances of they (1.2%) and four instances of hem (0.8%)—a striking mix of old and new forms.
This linguistic inconsistency might reflect Margaret’s psychological state during a crisis, but more likely reveals the scribe struggling to capture rapidly changing speech patterns. As Margaret assumed greater authority during her husband’s absence, her language may have genuinely shifted toward more assertive, innovative forms.
Cracking the Code: Margaret’s Authentic Voice
After analyzing thousands of words across multiple scribes and time periods, several patterns emerge that likely represent Margaret’s authentic linguistic personality:
Conservative Core: Margaret consistently preferred that over which in relative clauses—a pattern that appears across all scribes and suggests genuine personal preference rather than scribal imposition.
Gradual Innovation: Her pronoun usage shows steady evolution from conservative (hem, here) toward innovative (them, their) forms, but at a slower pace than contemporary men.
Crisis-Driven Change: The most innovative language appears in letters from 1464-1469, when Margaret wielded unprecedented authority as a widow managing legal battles and estate management.
Beyond Grammar: What This Reveals About Medieval Women
This linguistic detective work reveals something profound about medieval women’s voices. Margaret Paston wasn’t just passively dictating words—she was actively evolving her linguistic identity as her social role transformed.
The evidence suggests that, in private family correspondence, medieval women could push against linguistic constraints. While official documents might require formulaic language, Margaret’s letters show her experimenting with new forms and asserting authority through grammatical choices.
Most remarkably, the scribes—despite their gender and educational advantages—appear to have genuinely attempted to capture Margaret’s voice rather than completely overriding it. The linguistic variations we detect suggest they were listening carefully to her speech patterns, even when those patterns conflicted with their expectations of feminine language.
The Computational Medieval Feminist
This research demonstrates how modern computational linguistics can recover silenced historical voices. By treating grammatical patterns as linguistic fingerprints, we can detect subtle forms of historical bias and manipulation that traditional literary analysis might miss.
The implications extend far beyond the Paston family. Thousands of medieval documents attributed to illiterate authors await similar analysis. Each represents a potential voice waiting to be computationally recovered from centuries of scribal interference.
Margaret Paston’s story reveals that even in an age of systematic female subordination, women found ways to assert their linguistic agency. Her authentic voice—conservative yet evolving, traditional yet strategically innovative—emerges from the medieval manuscripts like a ghost finally given permission to speak.
The next time you read a historical document, consider: whose voice are you really hearing? In Margaret’s case, the answer is both more complex and more hopeful than you might expect.
Daniel Morris has a Master’s in Computational Linguistics from Lancaster University and works as an AI/ML Engineer at Reddit developing and researching in the world of LLMs.
By Daniel Morris
In the winter of 1469, Margaret Paston dictated an urgent letter to her son about violent men destroying her tenants’ crops. Like virtually all medieval women, Margaret could not write—every letter bearing her name was penned by male scribes who may have subconsciously altered her voice to match their expectations of how a woman should sound. It is assumed Margaret could read, however, which might have helped ensure her intended meaning came through.
Five centuries later, I’ve used forensic linguistics to detect these subtle manipulations, uncovering fascinating patterns that reveal when Margaret’s authentic voice breaks through versus when scribes imposed their own linguistic biases. The detective story unfolds through tiny grammatical clues hidden in pronoun choices and sentence structures.
The Medieval Voice Problem
Medieval letters present a unique challenge for historical linguists. In 15th-century England, literacy rates were devastatingly low—Norman Davis estimates that only 10% of men and 1% of women could even write their own names. This meant that powerful figures like Margaret Paston, despite managing vast estates and wielding considerable influence, relied entirely on scribes to record their words.
But here’s the crucial question: when a scribe writes “I think that we should proceed carefully,” are those Margaret’s exact words, or the scribe’s interpretation of what a respectable woman should say?
The Paston family letters, spanning 1422-1509, offer an unprecedented window into this dilemma. This remarkable collection contains 422 documents totalling 245,000 words—the most complete medieval family correspondence to survive. Margaret appears as author of numerous letters, but analysis reveals they were written by at least five different hands: professional scribes like James Gloys and John Wykes, plus her own sons acting as secretaries.
The Linguistic Fingerprint Method
To detect scribal interference, I focused on unconscious linguistic choices—the grammatical patterns people use without thinking. Specifically, I analyzed two key variables that were rapidly changing in 15th-century English:
Personal Pronouns: The old Anglo-Saxon forms (hem, here) were being replaced by Scandinavian-influenced pronouns (them, their). This change happened gradually, creating a linguistic spectrum from conservative to innovative usage.
Relative Pronouns: Medieval writers could choose between traditional that, emerging which, and the elaborate the which. These choices often reflected gender, social class, and personal style.
If Margaret had a consistent linguistic personality, her letters should show similar patterns regardless of which scribe wrote them. If scribes were imposing their own styles, we’d see dramatic variations.
The Detection Algorithm
I manually coded every instance of these pronouns across eight representative letters spanning 1449-1469, comparing:
The results revealed a complex linguistic crime scene.
Evidence of Scribal Manipulation
Professional Scribes Adapting Their Style
James Gloys provides the clearest evidence of scribal accommodation. When writing for John Paston I in 1449, Gloys used the traditional hem six times (0.6% of text). But twenty years later, writing for Margaret, hem completely disappears while innovative they spikes to 1.8% of the text.
This dramatic shift suggests Gloys consciously or subconsciously adapted his language to match what he perceived as Margaret’s evolving linguistic identity. Remarkably, this change aligns with Margaret’s transformation from subordinate wife to independent widow managing family properties after her husband’s death in 1466.
Gender-Based Linguistic Expectations
The evidence becomes even more intriguing when comparing how scribes wrote for Margaret versus male family members. John Wykes, writing for Margaret in 1465, used the conservative form here seven times but completely avoided the innovative which. Yet writing for her son John II the same year, Wykes used which seven times and here only twice.
This pattern suggests scribes subconsciously feminized Margaret’s language, giving her more traditional forms they associated with women while reserving innovative constructions for male voices.
The Sons’ Dilemma
Margaret’s sons John II and John III present the most complex case. When writing as themselves, they used distinctly different linguistic patterns than when writing for their mother. John III, writing his own letter in 1461, used the traditional hem four times but never used it when writing for Margaret two years earlier.
This reversal is puzzling until we consider the social dynamics. The sons may have felt pressure to make their mother sound appropriately feminine and conservative, even when their own speech was more innovative. Alternatively, they might have been more faithful recorders of her actual conservative speech patterns.
The 1464 Linguistic Revolution
The most dramatic revelation comes from letters dating to around 1464, when Margaret’s husband was imprisoned. John II, writing for his mother that year, records her using highly innovative pronouns—six instances of they (1.2%) and four instances of hem (0.8%)—a striking mix of old and new forms.
This linguistic inconsistency might reflect Margaret’s psychological state during a crisis, but more likely reveals the scribe struggling to capture rapidly changing speech patterns. As Margaret assumed greater authority during her husband’s absence, her language may have genuinely shifted toward more assertive, innovative forms.
Cracking the Code: Margaret’s Authentic Voice
After analyzing thousands of words across multiple scribes and time periods, several patterns emerge that likely represent Margaret’s authentic linguistic personality:
Conservative Core: Margaret consistently preferred that over which in relative clauses—a pattern that appears across all scribes and suggests genuine personal preference rather than scribal imposition.
Gradual Innovation: Her pronoun usage shows steady evolution from conservative (hem, here) toward innovative (them, their) forms, but at a slower pace than contemporary men.
Crisis-Driven Change: The most innovative language appears in letters from 1464-1469, when Margaret wielded unprecedented authority as a widow managing legal battles and estate management.
Beyond Grammar: What This Reveals About Medieval Women
This linguistic detective work reveals something profound about medieval women’s voices. Margaret Paston wasn’t just passively dictating words—she was actively evolving her linguistic identity as her social role transformed.
The evidence suggests that, in private family correspondence, medieval women could push against linguistic constraints. While official documents might require formulaic language, Margaret’s letters show her experimenting with new forms and asserting authority through grammatical choices.
Most remarkably, the scribes—despite their gender and educational advantages—appear to have genuinely attempted to capture Margaret’s voice rather than completely overriding it. The linguistic variations we detect suggest they were listening carefully to her speech patterns, even when those patterns conflicted with their expectations of feminine language.
The Computational Medieval Feminist
This research demonstrates how modern computational linguistics can recover silenced historical voices. By treating grammatical patterns as linguistic fingerprints, we can detect subtle forms of historical bias and manipulation that traditional literary analysis might miss.
The implications extend far beyond the Paston family. Thousands of medieval documents attributed to illiterate authors await similar analysis. Each represents a potential voice waiting to be computationally recovered from centuries of scribal interference.
Margaret Paston’s story reveals that even in an age of systematic female subordination, women found ways to assert their linguistic agency. Her authentic voice—conservative yet evolving, traditional yet strategically innovative—emerges from the medieval manuscripts like a ghost finally given permission to speak.
The next time you read a historical document, consider: whose voice are you really hearing? In Margaret’s case, the answer is both more complex and more hopeful than you might expect.
Daniel Morris has a Master’s in Computational Linguistics from Lancaster University and works as an AI/ML Engineer at Reddit developing and researching in the world of LLMs.
Subscribe to Medievalverse
Related Posts