In 1996, two Claremont McKenna College professors, Dr. Ward Elliott and Dr. Robert Valenza, published an article titled “And Then There Were None: Winnowing the Shakespeare Claimants.” They and their students applied 51 stylometric computer tests to the writings of 37 alternative authorship candidates, including Thomas Sackville based on his two Mirror for Magistrates poems. No claimant, including Sackville, was a match to Shakespeare. Why doesn’t their analysis rule Sackville out as the “true Shakespeare”?
Stylometric tests of Thomas Sackville’s Mirror poems, both written by 1563, cannot preclude the possibility that he wrote as Shakespeare in the 1590s and early 1600s. During the intervening decades, the English language underwent rapid changes, and several poetic fashions went in and out of style. Even if this were not the case, Elliott and Ward’s stylometric profile for Shakespeare is derived from a 32-play Shakespeare baseline. As they admit, only 14 of 51 stylometric tests derived from this baseline were reliably successful at distinguishing Shakespeare’s poetry from that of his peers. In the case of their “grade level” test, a function of word length and sentence length, there was no overlap at all between play and poem ranges. Among this subset of 14 poem tests, even Shakespeare’s poetry showed marked changes over his writing career for 3 of them, requiring Elliott and Ward to use two separate ranges of parameters, “early” and “late,” for those particular tests. I remain skeptical that it is possible to generate a meaningful stylometric comparison between Sackville’s two 1563 narrative poems and Shakespeare’s plays, given that Sackville had three intervening decades to mature as a poet and that narrative poetry is so different from dramatic dialogue.
Another difficulty with using stylometric tests to evaluate Sackville as an authorship candidate is that his only known play, Gorboduc, was co-authored with Thomas Norton. Although the 1565 unauthorized edition states that Norton wrote Acts 1, 2, and 3 and Sackville wrote Acts 4 and 5, there is not a clear stylistic break between the third and fourth acts. It seems more likely that Sackville’s stylistic fingerprints are intermixed with Norton’s throughout Gorboduc, making it impossible to generate an individual stylometric profile for Sackville based on this play alone.
Other aspects of Elliott and Ward’s research provide new support for the Sackville theory. In particular, they discovered that two plays and one poem from the Shakespeare Canon, Titus Andronicus, Henry VI Part Three, and A Lover’s Complaint, do not match Shakespeare’s stylometric profile. Elliott and Valenza refer to Titus Andronicus and Henry VI Part Three as “gross outliers,” each being rejected on six different Shakespeare tests as authentic, and accounting for a quarter of the total Shakespeare rejections. Because the stylistic difference was so pronounced, Elliott and Valenza were forced to eliminate these works from their baseline. “At the very least, the rejections mean that Shakespeare may have taken a year of two to settle into his mature style, as measured by our tests,” they write. Titus Andronicus, Henry VI Part Three, and A Lover’s Complaint are the very works which I have argued are the linguistic bridges between Thomas Sackville’s youthful style and Shakespeare’s mature style. It would be very interesting to see a stylometric comparison between Sackville’s Mirror poems and these three rejected Shakespeare works.
The stylometric findings which most surprised Elliott and Valenza concern the sections in The Two Noble Kinsmen and Henry VIII which are ascribed by modern Shakespeare scholars to Shakespeare, rather than to the playwright John Fletcher. While the Shakespeare sections “come much closer to fitting within Shakespeare’s range than the non-ascribed parts,” they “still don’t quite fit.” Shakespeare’s ascribed portion of The Two Noble Kinsmen was rejected on four Shakespeare tests, and his ascribed part of Henry VIII was rejected eleven times. In seeking to explain these findings, Elliott and Valenza speculate as follows:
They could indicate that the conventional allocations of text aren’t quite right, or that our tests aren’t quite right, being validated only for play-sized samples or perhaps being hypersensitive to small admixtures of non-Shakespeare even in sizable samples. Where could all those ’em’s in Shakespeare’s parts of TNK and H8 [The Two Noble Kinsmen and Henry VIII] have come from? Perhaps elements of both sets of explanations are involved.
From a Sackvillian perspective, not only I am not surprised that the “Shakespeare sections” in The Two Noble Kinsmen and Henry VIII were rejected by stylometric tests, I would have predicted this result. My theory holds that these are earlier plays by Thomas Sackville which William Shakespeare and John Fletcher co-adapted for the public theatre around 1613. William apparently used a light editing touch in his sections, preserving much of Sackville’s original language but inserting some “’em’s,” while Fletcher infused his own characteristic style into his sections.