Putting Bard to Computer Test on Thorny Issue of Authorship

The man from Stratford, as William Shaksper (his spelling) is often called by those who not convinced he wrote Shakespeare’s works, would have been 424 years old last Saturday.

That rainy morning, by a perfect and timely coincidence, a group of computer researchers gathered in Bauer Hall at Claremont McKenna College gave a preliminary report on their project for using the computer to shed new light on The Authorship Question.

Like the blood test in matters of paternity, the computer appears unlikely to confirm who did write the Shakespeare canon. But the computer may, according to the project’s leader, be able to say with some conviction who did not write the plays and poems.

This would be no small achievement. There have been 58 claimants advanced as authors of the work, commencing with Francis Bacon.

Even after eliminating the most obvious non-candidates (like “all the Jesuits” and “all the Rosicrucians”), there remain several Elizabethan figures to whom attention can be paid. The Claremont group is concentrating its efforts on Christopher Marlowe, Francis Bacon and Edward de Vere, the 17th Earl of Oxford, who has emerged in recent years as the strongest of the claimants.

The computer analysts have acquired between 2 million and 3 million words of computer-readable Elizabethan texts, including all 884,000 words of Shakespeare, and thousands more from perhaps two dozen of the other claimants, plus additional thousands of words of apocrypha and anonymous works, some of which have been attributed wholly or partly to the Bard.

The Shakespeare Clinic, as the project is called, was conceived in 1986 by Prof. Ward Elliott, a political scientist from Claremont McKenna College, and Henry Krieger, a professor of mathematics at Harvey Mudd. Karen Kossuth, a linguistics professor from Pomona College, has become a third faculty adviser.

Elliott obtained funding for the project from the Sloan Foundation’s New Liberal Arts Program. The foundation’s interest in the Shakespeare study is evidently as a testing ground for wider applications of the computer to textual analysis.

The work is being done by a team of undergraduate and graduate students led by John Otsuki of Harvey Mudd and the Claremont Graduate School. (None of the researchers has a particular favorite in the Shakespeare Stakes; their passion is for the process.)

My knowledge of higher mathematics stops well short of calculating the discount when you buy gas for cash. The terms tossed about Saturday morning were like something out of Walter Mitty: the Efron/Thisted method, purged lexicons, Louis Ule’s Constat, Robert Ule’s Order and Alt, Rudman, Wordcruncher and Character Count programs. I kept thinking of Mitty’s line, “I’m afraid it’s coreopsis, and it’s moving in fast.”

But the notion is that every author has a literary signature--words that he uses more than others, uses rarely, uses not at all. If you can develop mathematical models--frequency graphs of the word play in large samples of known work--you can crunch the word frequencies in another body of work and see how the graphs compare.

I can see statisticians going pale with shock at the over-simplicity of that, but there you are. The Claremonters actually apply four different tests, not all equally indicative in their view. The correlations end up as figures like 0.014297. I am back with Fred Allen and his, “Let x equal my father’s signature.” But the numbers let the analysts come up with probabilities, as of rainfall on an April morning: a 20% chance that Marlowe was the author of a particular piece, for example.

Only hours before the Saturday report the researchers typed into the computer the short poems from a Huntington Library manuscript that the Oxford scholar Peter Levi last week speculated could be by Shakespeare.

Crunched against the Shakespeare numbers, the poem’s computer read-out could be interpreted, the team leader John Otsuki said, as a strong “could be.” The numbers, that is, did not suggest Shakespeare did not write the small but charming poems (“O, be not proud, though wise and fair / Beauty’s but earth, wit is but air.”)

The work at Claremont has really just begun. It has taken two years to assemble the machine-readable texts and to discover how to speed up the number-crunching. It already goes 50 to 100 times faster than it did initially. One of the plays can now be translated into mathematical data in an hour.

The first part of the chore has been to see if the computer analysis works at all--measuring a swatch of Marlowe against Marlowe to be sure that the numbers agree Marlowe wrote it. So far so good.

Whether in fact literary analysis by computer is valid is still not clear, which is why the Claremont project is fascinating and potentially significant whatever the final conclusions.

But what is also significant is that the project, co-sponsored by the local Shakespeare Authorship Roundtable, is another indication that the authorship question is being taken seriously.

James Lardner’s 20-page article in the April 11 New Yorker on “The Authorship Question” reports on last September’s mock trial in Washington before three Supreme Court justices on the case for Edward de Vere. The verdict was “Not proved,” but at least two of the three justices agreed that the question of the authorship was a fit subject for further research.

Lardner’s piece outlines the historical background with exemplary fairness. Despite the scorn of the orthodox Stratfordians, the question is still open.