Donald Foster Uses high-powered Computer Tests to Search for Shakespeare's Hidden Hand. His Critics Challenge Him on Every Move.


By CALEB CRAIN

    He proves by algebra that Hamlet's grandson is Shakespeare's grandfather and that he himself is the ghost of his own father.

--James Joyce, Ulysses

"Downn, downn, unwuwy jades," Donald Foster intones, Elmer Fudd-like. The forty-eight-year-old Vassar English professor is reciting Shakespeare as delivered by his son at age three. "Down, down I come like glist'ring Phaethon, / Wanting the manage of unruly jades" were Richard II's original lines, which Foster used to quote as a joke whenever he descended the stairs at home. "He had picked this up and was throwing it back at me," Foster explains of his son's performance. "This is the way we all work; we just don't realize or acknowledge it."

Foster finds authors in texts. He reads patterns of vocabulary and style as the fingerprints a writer leaves on language. Whom you quote is one kind of linguistic fingerprint; a toddler who invokes unwuwy jades might well be the son of a Shakespeare professor, for example. Foster has built a career out of inferences like this. Or rather, a couple of careers, because Foster's work in attribution has led him down two different roads. As a Shakespearean, he claims to be the first scholar since the nineteenth century to have uncovered a genuine new work by the Bard: a rather plain funeral elegy for a young gentleman who died in a squabble over a horse. As a forensic linguist, on the other hand, Foster is wanted by the FBI and by police around the country. In 1996 he testified that Theodore Kaczynski wrote the Unabomber manifesto. On his desk today are four thick plastic binders of analysis for the Boulder police department, whose investigation into JonBenét Ramsey's murder centers on an anonymous ransom note.

The skills developed by Foster the Shakespearean have made Foster the forensic linguist very good at matching authors to their texts: In February 1996 he fingered Joe Klein as the author of Primary Colors. But the Shakespeare attribution that launched Foster's career is currently under attack. In Britain, Shakespeare dons find Foster's computer-assisted analysis an affront to their carefully trained aesthetic sensibilities. Meanwhile, in California, technophile Shakespeare amateurs find his text crunching flawed. If Foster's Shakespeare attribution--the cornerstone of his academic career--is undermined, will his forensic career also topple?

Foster admits he doesn't have much room for error. "All I need to do is get one attribution wrong ever, and it will discredit me not just as an expert witness in civil and criminal suits but also in the academy." He smiles nervously. "So far, I've got a fine record going."

Despite his gray hair and bald top, Foster does not look like a tenured professor. He's too wiry and vigilant. Dressed in jeans, a T-shirt with a hamburger-restaurant logo, and cross-trainer athletic shoes, he looks more like a student who has just pulled an all-nighter but is still eager to make a good impression in class. He speaks with the overcaution of a drunk driver who can see the police in his rearview mirror.

Raised in Chicago and California, Foster studied psychology and sociology as an undergraduate at Wheaton College in Illinois, the alma mater of the Reverend Billy Graham. After graduation, he backpacked around the world with his brother. Then he got married, sold the wedding presents, and backpacked for another year with his wife. At age thirty, after stints working in a bookstore and teaching in a high school, he enrolled in UC-Santa Barbara's English Ph.D. program.

Foster never intended to specialize in attribution. As a graduate student, he stumbled into it while solving an old puzzle. In 1609, the printer Thomas Thorpe had dedicated his edition of Shakespeare's sonnets to "W.H." Thorpe called W.H. the poems' "onlie begetter." Did "begetter" mean "author" or "inspirer"? Were W.H. the initials of the fair youth whom Shakespeare loved? For Oscar Wilde and hundreds of others, W.H. proved to be a homoerotic biographical tar baby. Foster dispatched the mess. In a prizewinning essay later published in PMLA, he neatly demonstrated that "W.H." was no more than a misprint of "W. SH." Shakespeare's full name, after all, was printed clearly on the title page.

But Foster buried one Shakespearean conundrum only to unearth another. While scanning a microfilm of other Thorpe publications, he discovered A Funerall Elegye in memory of the late Vertuous Maister William Peeter. Exit W.H., and enter W.S., this elegy's unknown author. At first read, W.S.'s style struck Foster as uncannily Shakespearean, so he invested a pocketful of dimes to make a printout of the poem, that would become the subject of his dissertation and then his first book.

William Peter (1582-1612) was a gentleman from Exeter who studied at Oxford for nearly a decade. He left school in 1608, married in 1609, and was murdered by his kinsman Edward Drew in January 1612. Drew had defaulted on a loan for a horse, and he was furious at Peter for finking about it to Drew's mother. After Drew, his brother, and Peter spent a long day of uneasy drinking together, Peter galloped toward home down a darkening road. Drew spurred his horse in pursuit, and just outside Exeter's city walls he stabbed the twenty-nine-year-old scholar through the back of the skull.

The burden of A Funerall Elegye is to say that Peter deserved better than this sordid end. The tragedy of A Funerall Elegye is that in 578 lines it says little else. As W.S. himself explains, on the topic of William Peter, "the sum of all that can be said / Can be but said that `He was good.'" W.S. nonetheless used many more than three words to say as much. Ready for printing just nineteen days after Peter's murder, the elegy jogs along with a fluent meter and an unstrained rhyme that suggest a skilled, practiced poet. W.S. writes with Shakespeare's compact syntax and something of the polished grandeur of the late plays. But W.S.'s imagery is barren, and his character as a writer is prudish. It's possible, of course, that W.S. is a sensuous author who on this occasion has censored himself into decorous generalities so as not to offend the bereaved. But the poem has few fans.

Not even Foster argues that the elegy is good poetry. "I don't stay at home nights reading the funeral elegy for William Peter," he says. Attribution does not concern itself with aesthetic quality. But in beauty's absence, the pains Foster took in his Elegy by W.S.: A Study in Attribution are remarkable. In this book, Foster argued that W.S.'s language was Shakespearean in its verbal quirks--its unusual diction, frequency of certain words, choice of rhymes, high use of enjambment, pattern of word length, and habits of hyphenation. For his Shakespeare stats, Foster was able to turn to a computer-generated concordance prepared by Marvin Spevack in 1968. But for his non-Shakespeare cross sample, Foster had to count the old-fashioned way--by hand. "Nothing," he observed in a footnote late in the book, "could persuade me to undertake another study that would require the same manual labors."

Numerically, if not poetically, the elegy rewarded Foster's efforts, passing every test he devised. Nonetheless, he presented his results modestly. "That W.S.'s Funerall Elegye for William Peter was written by William Shakespeare is more than I know," Foster wrote, despite pages of data tables. He confessed that he thought W.S. was William Shakespeare, but he respectfully submitted that "it is a worldwide community of readers, not I, nor even W.S. himself, who will have the final word."

Behind the scenes, however, Foster was not so submissive. When Oxford University Press rejected his manuscript on the elegy, the anonymous reader's report noted that it was axiomatic that authorship could never be proved on internal textual evidence alone, and external evidence was lacking. Foster thought he recognized the reviewer's biting style, so he wrote back. "Not only can you tell something by internal evidence alone," Foster averred, "but if this isn't Sam Schoenbaum, it's a very good imitation."

Foster guessed right. "It was a cocky thing to do," he concedes today, "but it seemed necessary at the time." In what must be every academic's fantasy, Foster then proceeded to nail Oxford's second anonymous reader (Stanley Wells), as well as the author of a favorable report for Harvard University Press (G. Blakemore Evans). "I know who's written every report I've ever gotten," Foster boasts. "Sometimes I do them for friends."

But Foster's stunts failed to win him a big-name publisher. Attribution studies suffered a blow in 1985 when Gary Taylor wrongly assigned the ditty "Shall I die?" to Shakespeare, and Oxford then prematurely included it in the press's Complete Works, edited by Stanley Wells. A string of erroneous Shakespeare finds followed Taylor's, scaring reputable presses away from attribution studies in general. Among the most vehement critics of the Oxford mistake was Foster, a fresh Ph.D., who worried that the British gaffe would render his own book unpublishable and himself unhirable. Despite these fears, however, Foster was hired by Vassar in 1986.

When the University of Delaware Press finally released his book in 1989, Foster set attribution to one side. He had nothing to gain, he felt, by pushing W.S. any harder. As he puts it today, "All it would take is for someone to come up with a letter saying, `Dear John Peter [William's surviving brother], I hope you like my elegy. Love, Wally Smith,' and whoever endorses this attribution would be humiliated." Foster decided to pursue an anthology of early-modern women writers instead. He also started to tinker with an index of Shakespeare's rarest words, which he put on a computer disk and christened Shaxicon.

According to Ian Lancashire, a University of Toronto professor who edits the Early Modern English Dictionaries Database, there are three kinds of evidence that can be used to establish authorship: external, linguistic, and interpretive. A signed manuscript in an author's handwriting is external evidence. So is an author's name on a title page or a mention that links a writer to a text in a contemporary diary or letter.

Foster specializes in the second kind of evidence: linguistic. He hopes to find language habits that are distinctive enough to serve as an author's fingerprints. For example, in Elegy by W.S., Foster showed that W.S. uses "who" several times to refer to an inanimate object that has not been personified. In the early seventeenth century, almost no one but Shakespeare used this kind of "incongruent `who,'" so Foster argues that the quirk helps to identify W.S. as Shakespeare.

If authors leave traces of themselves in their texts, it's because the memory used to create language is what cognitive scientists call "procedural." That is, you don't remember how to make a sentence except by making one. You can edit a sentence after you've generated it, but you can't altogether control it while it's flowing out of you. As Lancashire writes, "During true composition, speakers are blind to the process that gives rise to the utterances they make."

Semiblind, actually. Any feature of language you pay attention to will yield to your conscious control. Thus, it's the patterns it wouldn't occur to you to alter that are most reliable as fingerprints. For instance, an author may use function words--short, all-purpose words like "and" and "but" whose frequency is unaffected by subject matter--in distinctive ratios. Function words are common. They tend to slip by the self-editing radar; you're unlikely to revise what you say or write in order to raise or lower your rates of "that" or "the." Function words have mathematical advantages too. Even a short text will contain enough of them for the sample to be statistically significant. And a count of them is immune to subjective influence; everyone can agree that the paragraph you are reading now contains seven instances of the word "of." The bulk of the tests in Foster's 1989 book measured features that were, like function words, "substylistic."

All this may sound scientific, but Lancashire nonetheless calls attribution by linguistic evidence a "science in waiting." Often, a substylistic feature that's distinctive and stable for one author varies randomly for another. It's as though every author has fingerprints, but while some have their fingerprints on the ends of their fingers, others have them on their noses or toes. In other cases, the substylistic marks typical of a given time period and genre may overstamp individual preferences. Or editors, proofreaders, and typesetters may have worked with such heavy hands as to smudge out the authorial quirks beneath.

To make a sound attribution, then, you need to gather many different kinds of linguistic proof. Foster believes that his Shaxicon database offers a new kind of evidence. The theory behind Shaxicon is that it's not only an author's common words that are beyond his conscious control. If studied as a set, his rare words, too, will mark him.

Shaxicon is an electronic index of words that appear fewer than twelve times in the Shakespeare canon. "Aligarta" is in Shaxicon; so are "family" and "real," because they're rare for Shakespeare if not rare in any wider sense. Foster's early use of Shaxicon showed, predictably enough, that a Shakespeare play tends to share the most rare words with plays written just before or after it. Sometimes, though, a new play has an oddly high level of vocabulary overlap with a much older play. Foster guessed that these data spikes were due to the older play's returning to the repertory of Shakespeare's theater company while the new play was being written. A new production jogged its words back into the playwright's active memory. On closer analysis, it turned out that when rare words from an older play returned they tended to come from one or two speaking parts. Foster thinks Shaxicon pinpoints the roles Shakespeare himself performed.

Foster also thinks Shaxicon can testify in favor of an attribution. In 1613 Shakespeare and John Fletcher collaborated to write Henry VIII. When Foster used Shaxicon to analyze the 1612 Funerall Elegye and the portions of Henry VIII attributed to Shakespeare, he discovered that the two samples contained rare words from Shakespeare's canonical plays at roughly the same rates. The same mind, with the same verbal memory, Foster inferred, must have composed both.

The third kind of evidence named by Ian Lancashire--interpretive evidence--is the wobbliest. In a 1988 book review, Foster chided Mark Dominik for reading the disputed seventeenth-century play The Birth of Merlin as Shakespearean autobiography. Dominik thought he saw in Merlin Shakespeare coming to terms with his neglected, angry children. If you knew Shakespeare had written Merlin, you might make this interpretation of it. But until you do know that, even the most plausible psychological reading risks begging the question. Responding to Dominik, Foster scorned the circular logic that would use a critic's interpretation as evidence in a case for attribution, calling it one of "the worst mistakes of past attributional studies."

Nowadays, though, interpretive evidence is the technique of Foster's chief ally. In 1991 Richard Abrams, an English professor at the University of Southern Maine at Portland, approached Foster at a conference to say he'd read Foster's book and found the arguments sound. Abrams was also struck by parallels between the Funerall Elegye and The Two Noble Kinsmen, a 1613 play by Shakespeare and Fletcher about two loving cousins who feud until one dies on horseback. Abrams began to read the elegy as if it were Shakespeare's. No longer a drab wannabe, the elegy soon became in Abrams' eyes "an intimate document from the poet's final years." Doubters were guilty of "emotional resistance." To explain its banality, Abrams read the elegy as a palinode. Like Prospero bidding farewell to rough magic at the end of The Tempest, W.S. is Shakespeare surrendering his dangerous poetic gift. Alcoholic melodrama had killed his friend, and so "W.S. blames imagination for its role in Peter's murder," Abrams wrote. Furthermore, since William Peter was a plain man, W.S. composed a plain poem to match. "Harmonizing his style with the dead man's," Adamas concluded, "W.S. scrupulously mortifies his own habitual play of fancy."

Berkeley sonnet scholar Stephen Booth has caricatured Abrams's interpretation thus: "William Peter, a dull, plodding man who lived a dull, plodding life, was a good man. The dull, plodding Funerall Elegye is imitative of its dull, plodding subject, and is, if looked at in that light, a good poem." But Abrams's enthusiasm convinced Foster to dust off his case. Meanwhile, Foster's Shaxicon database was turning up new evidence.

To the Shakespeare Association of America in March 1995, and then to the Modern Language Association (MLA) in December of that year, Foster and Abrams presented the upgraded case for the Funerall Elegye. Though most of the evidence was nearly a decade old--such as W.S.'s use of the Shakespearean "incongruent who"--this time Foster pulled no punches. "A Funerall Elegye belongs hereafter with Shakespeare's poems and plays," Foster wrote in the published version of his talk. The Chicago Tribune covered the MLA panel. Then The New York Times put the story on its front page. Suddenly the Funerall Elegye was world news.

British scholars were aghast. First into the transatlantic breach was Stanley Wells, general editor of the Oxford Shakespeare and chairman of the Shakespeare Birthplace Trust. Writing in the Times Literary Supplement (TLS) in January 1996, Wells called the elegy "tedious in a very unShakespearian way." Nor has the poem grown on him since. Asked today about Abrams's notion that it represents a "turn from the theatrical," Wells dismisses the argument as "special pleading": "It's no compliment to somebody, is it, to write in a boring way about them when they're dead."

But Wells admits aesthetics is somewhat beside the point, conceding that "great poets have written bad poems." And in fact, most objections by Wells and other British scholars have been not stylistic but biographical. In the TLS and then in Shakespeare Studies, Wells reminded Foster of the "obstacles to belief" that Foster himself had listed in his 1989 book: (1) A friendship between Peter and Shakespeare was unlikely, because they were separated by education, social milieu, fifty-five miles, and a twenty-year age difference. (2) Shakespeare's brother died and was buried in Stratford just when the Funerall Elegye was being composed. (3) W.S. refers to his "days of youth," and in 1612 Shakespeare was forty-seven. (4) Although W.S. sympathizes with "she who those nine of years / Liv'd fellow to [Peter's] counsels and his bed," Peter and his wife had been married for only three years. W.S.'s ignorance of the facts suggests he may have been a hired pen--a commission that Shakespeare, acclaimed and only semiretired, was far from needing to accept.

Foster finds none of these objections insurmountable. Shakespeare and Peter both seem to have known the playwright John Ford, so a direct friendship between the two is not inconceivable. No one knows whether Shakespeare attended his brother's funeral, or even cared to. W.S. may mean "youth" metaphorically, or he may be referring to it as a bygone. The passages in question are tricky to interpret. The nine-year bedmate might be simple error, or it might refer to a mistress Peter kept while at Oxford. (Fellows were not then allowed to marry.)

More intractable and less palpable is the conviction British critics have that Foster and Abrams's Shakespeare is not theirs. Class is one signal that seems to them to be pointing the wrong way. William Peter strikes Oxford lecturer and tutor Katherine Duncan-Jones as not quite grand enough. Why would Shakespeare have praised someone for dressing unadornedly, speaking little, and deliberately cultivating social obscurity? For Duncan-Jones's Shakespeare, "churning out a long elegy for a younger son of a fairly modest Devon gentry family seems a very dingy project."

Another aspect of W.S. to irk the British is his piety. As Brian Vickers of Zurich's Centre for Renaissance Studies complains, W.S. spends fifty lines arguing "that William Peter died a violent death, but so did Jesus Christ, and so did the Christian martyrs, and they're all in heaven." Vickers believes this is too straightfaced for Shakespeare, who never refers to religion except glancingly, and then often in jest. To Vickers's charge of excessive piety, Foster retorts that the Funerall Elegye is unique among its kind for failing to offer the bereaved the traditional consolation of resurrection in Christ. Instead, W.S. calls the hope of reunion in the afterlife "the weak comfort of the hapless."

A third red flag for the British is sexuality. In a letter to the TLS, Duncan-Jones derided "some of the poem's champions" who made it "sound more exciting by suggesting that it reflects the speaker's homoerotic affection for the dead man." Wells calls the notion of homosexual eroticism in the elegy "pure fiction." But Foster and Abrams deny they've ever made such a claim. "Terribly overdone," is how Abrams refers to media innuendo about the elegist's sexual orientation. The issue did receive a lot of media play, perhaps because it was news to The New York Times that most of Shakespeare's sonnets are love poems to a boy. "I wish it were homoerotic," says Foster now. "Then it would have something. It's about as erotic as cold broccoli."

The American team may be playing a bit coy on this issue. At the 1995 MLA convention, the University of Tulsa's Lars Engle did speculate whether William Peter might be the fair youth of the Sonnets. "He has the right first name," Engle noted, "and is the right age." And if William Peter were Shakespeare's love, it might explain W.S.'s frequent references to a mysterious scandal in his past. W.S.'s "sadder taste of knowing shame" might then be the "vulgar scandal stamped upon my brow" in Shakespeare's Sonnet 112.

To clinch their cases, the British critics have put forward other candidates for the elegy's authorship. In doing so, they most often counter Foster's linguistic evidence with external evidence. They rarely answer Foster's argument in the language in which he made it. Wells even insinuates that this is perhaps as it should be. "Most people who make a profession of the study of literature do so because they have an artistic rather than a scientific bent," Wells wrote in his first attack. "They are, to put it simply, better at English than at maths." Does Wells think computer-assisted analysis is of any use to literary study? "I don't think it's reached the stage yet of scientific sophistication where it's likely to be accepted even by those who do understand the math," he replies.

So the British are digging in their heels. Although in the United States, the Norton, Riverside, and Harper editions of Shakespeare all currently feature the Funerall Elegye, it isn't in the Oxford. If anyone were to try to put it there, "I should certainly resist it," says Wells. Duncan-Jones is now editing the Arden edition of Shakespeare's poems, and although she might choose to print the elegy as apocryphal, she feels no obligation to. After all, as she puts it, "Seldom has such a bad poem been so readily available."

When it comes to a contest between British Luddites and American scientists, we know who wins. In the long run, humanists armed with educated impressions tend to lose their campaigns against technicians armed with silicon and the ability to count. And so Donald Foster's most serious opponents in this debate are probably not the establishment British curmudgeons but a few computer-adept amateurs in California.

Patience and faith are prerequisites for authorship studies, and the Claremont Shakespeare Clinic began with a wacky but highly durable creed: that the works of William Shakespeare were written by someone else. Like his father before him, Ward Elliott, professor of government at Claremont McKenna College, believed "Shakespeare" was a hoax. The real author of the plays and poems, he thought, was Edward de Vere, the earl of Oxford. Elliott started the Claremont Shakespeare Clinic in 1987 to test his father's anti-Stratfordianism with computers.

What inspired Elliott was a clever study by two statisticians, Ronald Thisted and Bradley Efron. In 1976 Thisted and Efron tried to estimate the size of Shakespeare's vocabulary by asking "How many new words would Shakespeare use if he were to write another play?" Mathematically speaking, this question resembled another, which had already been answered: If a butterfly collector has already trapped x different species, what is the likelihood he will catch a new species on his next expedition? When Gary Taylor unveiled "Shall I die?" in 1985, Thisted and Efron had a chance to test their theory. They calculated that, given the poem's length, if it were Shakespeare's it should contain 6.97 words never before used by Shakespeare. In fact "Shall I die?" had nine new words, so the mathematicians gave it their blessing.

Curious, Elliott called Thisted to ask if tests like this could resolve the larger Shakespeare-de Vere controversy. Thisted said yes. So Elliott recruited Robert Valenza from Claremont's math department, solicited the Sloan Foundation for funds, and set up the Claremont Shakespeare Clinic. According to Elliott, Donald Foster served as an early adviser. When the clinic was compiling a pure Shakespeare baseline, Foster told Elliott which texts to set aside because their authorship was in dispute. "He steered us away from God knows how many pitfalls," says Elliott.

Cadres of undergraduates initiated and executed many of the tests, but Elliott and Valenza compiled their results. In 1991 the clinic's first report was published in the journal Computers and the Humanities (known to its text-geek fans as CHum). Seven tests ruled out every poet whom the clinic had suspected of being the true Shakespeare--including Elliott's favorite, deVere.

CHum printed the clinic's second report

in 1996. No longer doubting the identity of Shakespeare, Elliott and Valenza shifted their focus, hoping their tests could help define the hazy borders of the Shakespeare canon. The 1996 report was impressive, offering fifty-one tests of Shakespeare-play authorship and fourteen of Shakespeare-poem authorship. Rather affably, Elliott does not present himself as an expert on attribution. His tests, he concedes, are better at determining that someone has not written a text than that they have. "Imperfect tests are like Cinderella's slipper," Elliott quips. "Only in a fairy tale can you prove that you are Cinderella because you and she both wear a size-five slipper." The most the test can say is you might be. But if you wear a size eight, then, sorry: The slipper does rule you out.

With his usual amiable goofiness, Elliott ended his clinic's 1996 report by denying that his was the last word, "even as the dust settles on what looks like a saturation-bombing of tests." It would turn out to be anything but the last word. W.S.'s Funerall Elegye had flunked two out of seven tests in the Shakespeare clinic's 1991 report. In the 1996 report, it suspiciously flunked five out of fourteen. Foster's British opponents, including Wells and Vickers, began to cite the Claremont results. An unhappy Foster had a few words to add.

Over the years, Elliott and the Shakespeare clinic had exasperated Foster. Despite Foster's advice, "such matters as editorial consistency in the copytexts, generic and chronological controls, and normalization of the data were shrugged off, year after year, as inconvenient," Foster would later write. Good relations between Foster and the Californians finally broke down at a UCLA conference on attribution in February 1996.

The conference was meant to be a victory celebration of sorts for Foster's Funerall Elegye. But during a question-and-answer period, Elliott stood up in the audience. Flanked by two assistants ("Marlovians," Foster grumbles), Elliott passed out thick, chart-filled copies of his clinic's latest report--with its critical assessment of Foster's Elegye attribution. It was a breach of academic decorum that alienated some of the scholars present. Foster's patience snapped. Soon after, he would cease to correspond with the Claremont team and arrange with the editors of CHum to publish a rebuttal in the same 1996 issue that featured Elliott and Valenza's work. Elliott now calls the conference "a total disaster from my point of view."

Foster's 1996 rebuttal slammed Elliott and Valenza as irresponsible meddlers who threatened to turn attribution studies into "a playground for fringe theories having no historical or computational validity." Elliott and Valenza's reply, published in Shakespeare Quarterly in 1997, painted Foster as a bully and a hoaxer, "trying to win too many points with bluster that he could not win honestly on the facts."

CHum is due to publish another response by Elliott and Valenza, with yet another rebuttal by Foster, sometime in 1998. Both Elliott and Foster "have been very, very trying," sighs CHum editor Nancy Ide, a colleague of Foster's at Vassar. She admits, however, that the issue of CHum with the Foster-Elliott feud has been the journal's best seller in years.

The UCLA conference coincided with a crucial moment in Foster's career. On the plane to California, he read a new roman à clef about the 1992 Clinton campaign, Primary Colors. Attracted by press coverage from Foster's Funerall Elegye attribution, New York magazine was pestering him to sleuth out the novel's anonymous author. The puzzle intrigued Foster, though as yet he had no leads.

A week later, a sleep-deprived Foster solved the mystery of Primary Colors' authorship. When New York magazine hesitated to let him write the piece himself--fearing perhaps that he would write like an academic--Foster threatened to sell the story to The New York Observer. New York capitulated, and the piece Foster turned in was witty, confident, and playfully nasty about not only the linguistic tics but also the apparent sexual and racial prejudices of the author he had unmasked. "The book struck me in some respects as homophobic," Foster says, "so I sort of tweaked him a little bit by saying his own sexual identity is unclear in the text."

Joe Klein was not pleased. When CBS News flew Foster down the Hudson River in a helicopter for an interview, Klein called his friend Dan Rather and, Foster surmises, told him to "piss on this story." Dan Rather seems to have obliged. The story ran, but Rather spun it in Klein's favor. Klein's employers at CBS and Newsweek would stand by his denial of authorship until July, when handwriting analysis confirmed that Foster had been correct from the start and Klein caved in. Until then, Foster says, "it was horrible. I went through six months of listening to people say, 'Well, you were wrong about Primary Colors. Why should we believe you about this Funerall Elegye?'" Foster's frustration during this trial in the wilderness may have sharpened the edge of the critique he wrote of Elliott and Valenza.

Elliott and Valenza joke that to a "Whitmanesque sensitivity," their analyses may "seem more like taking butterflies and grinding them up to find matching nucleotides than like looking up at them in perfect silence, like the stars." At the risk of grinding a few butterflies in the stomachs of Lingua Franca readers, here's a glimpse of the debate between the Claremont team and Foster.

Of the fourteen 1996 Claremont tests that were valid for poems, the Funerall Elegye flunked five.

(1) Claremont's "grade level" test fed Shakespearean texts to a piece of commercial software that rated their difficulty according to word and sentence lengths. Elliott and Valenza found that the Funerall Elegye has a higher difficulty level than other Shakespeare works. But word and sentence length often say more about a text's editor than its author, Foster noted, and the Claremont texts were edited inconsistently. Also, the Funerall Elegye, a long poem without breaks, would naturally tend to have longer sentences than plays, with their dramatic back-and-forth, or even than other Shakespeare poems, which are divided into stanzas. Elliott and Valenza conceded the point.

(2) Elliott and Valenza give near top billing to a series of tests they call bundles of badges (BoBs). Adapting a term first coined by Foster, Elliott and Valenza defined a badge as a word Shakespeare used more often than his contemporaries--sometimes only slightly more often. A fluke is a word he used less often. Elliott and Valenza grouped badges and flukes together in single tests. The test BoB5 counted the badges "the," "is," "to," "you," "he," "his," "your," "we," "him," "as," and "an" against the flukes "a," "sir," "I," "now," "I'll," "'tis," "all," "come," "her," and "she." The odd-sounding result: the elegy had too many of these badges and too few of the flukes. Its diction for these common words was too pronouncedly Shakespearean to be Shakespeare's.

Foster ingeniously disentangled the tests that Elliott and Valenza had bundled together in BoB5. BoB5 worked on most Shakespeare texts, Foster argued, because it selected for "you" and "your," thus accurately matching Shakespeare's usual preference for the formal "you" over the more personal "thou." In a similar way, by testing for "he," "his," and "him," and against "her" and "she," BoB5 judged according to the gender ratio common in Shakespeare plays: more male characters than female. The Funerall Elegye, however, exceeds even Shakespeare's usual preference for "he"s over "she"s, rendering its BoB5 score too high. But Foster pointed out that you shouldn't expect gender parity in a man's elegy for a man. To apply BoB5 to the elegy was to ask the rather inappropriate question, "If it's by Shakespeare, why does this funeral poem not have more women in it?" Yielding to Foster here too, the Claremont team rescinded the Funerall Elegye's BoB5 flunk.

(3) and (4) Microphrase tests measure how often syllables that would be stressed in normal speech lose their stress because of their placement in iambic verse. In practice, the tests are fiendishly difficult to administer. Elliott cheerfully admits that "a duffer like me" consistently undercounts the microphrases by 5 to 10 percent compared to the tests' inventor. Foster is incredulous that Elliott stands by tests he admits he can't adequately administer. But Elliott notes that Foster's test for incongruent "who" also requires judgment calls that may vary from tester to tester.

(5) When they compared the frequency of the word "no" in Shakespeare to the combined frequencies of "no" and "not," Elliott and Valenza found a distinctive ratio. The Funerall Elegye came out low--several "no"s too few--on this test. In response, Foster argues that "no" is more common in dramatic dialogue than in a nondramatic form such as an elegy, so one shouldn't find this result surprising. But in fact Shakespeare's nondramatic verse contains a slightly higher rate of "no"s than his plays do, so the Claremont team stands by this test.

In their 1997 Shakespeare Quarterly essay, Elliott and Valenza added a few brand-new tests that the Funerall Elegye flunks. Foster brushes them off: "I think I could probably find a lot of other phrases that the Funerall Elegye has too many of, too." However, the Shakespeare Quarterly piece also revisited an earlier doubt that Foster does take seriously. In a 1991 review of Foster's book, MacDonald P. Jackson, an English professor at the University of Auckland, New Zealand, noted that, although Foster identified nine common words "that never deviate in the plays by more than a third from their respective mean frequencies," Foster chose to test only five of these basic function words on the Funerall Elegye. When Jackson calculated rates for the remaining four--"by," "in," "to," and "with"--the Funerall Elegye flunked them all. Jackson diplomatically raised the question of "a degree of unconscious bias in the selection and application of the tests."

Foster's ally Abrams attacked Jackson in the fall 1995 Shakespeare Newsletter. Abrams argued that the Funerall Elegye's use of "by," "in," and "to" was not too high but rather continued an upward trend. "With" looks too low until you see it as extending a downward trend. To defend himself, Jackson countered that Shakespeare's poems alone were too scant a sample to justify Abrams's suggestion of trends, which data from the plays did not support.

Asked about the exchange today, Abrams is apologetic. "I feel bad about not letting Mac off the hook," he says. Disinclined to renew the battle, Abrams suggests the elegy's case can withstand the four prepositions' negative testimony. "Don should have included those [in Elegy by W. S.] and given the dismal news," Abrams says. But Foster is not ready to concede that the prepositions work against him. A few days after I ask about them, Foster hands me three pages of recalculated stats. They resurrect Abrams's hypothesis of prepositional trends. For his part, Jackson is planning "a comprehensive statistical analysis" of thirty or so function words in the Funerall Elegye, Shakespeare, and other writers.

As attribution problems go, W.S.'s Funerall Elegye is a tough case. To appreciate how tough, consider the field's one great success: Mosteller and Wallace's 1964 attribution of unsigned Federalist Papers. There were only two candidates for authorship (Madison and Hamilton), and long samples were available for comparison that matched the disputed texts in genre (political essay) and date (1787-1788). By contrast, any number of people might be W.S., and Shakespeare never wrote a signed funeral elegy. The texts from the right time period, 1612-1613, are the wrong genre completely: collaborative drama. It should be no surprise, then, that although Foster is itching to leave the elegy behind him by writing about it one last time--"I will be declaring victory and getting out"--W.S.'s case is not yet closed.

The Unabomber's case, on the other hand, is closed. Early on a Monday evening in April, Foster gives a lecture for Vassar students in Lathrop House, where he and his family are in residence. The title: "Publish [Me] or Perish! The Works of Theodore Kaczynski." Before the talk, Foster confides that he flirted with the idea of delivering it in a hooded sweatshirt and goggly sunglasses, as in the sketch released by the FBI when the Unabomber was still at large.

About twenty-five students and half a dozen faculty members show up for cider, cookies, and Kaczynski. Foster tells them how in December 1996 Kaczynski's defense team asked Foster if he would help discredit the FBI's textual analysis of the Unabomber manifesto--the basis for the search warrant that let the government into Kaczynski's Montana cabin. On the World Wide Web, Foster found both the Unabomber manifesto and a 1971 Kaczynski essay given to the FBI by Kaczynski's brother, David. After comparing the two documents, he declined the defense's offer and testified instead for the prosecution. Using the Internet and on-line databases as enormous, contemporaneous samples of non-Kaczynski prose, Foster showed that Kaczynski and the Unabomber shared distinctive vocabulary, rare collocations of common words, echoes of the same source materials, irregular hyphenations and spellings, and a preference for the "corrected" cliché: "You can't eat your cake and have it, too." Soon requests for Foster's forensic skills came pouring in.

Though he doesn't mention it at his lecture, the work brings Foster up to $250 an hour. He's now advising in about half a dozen major cases, including those involving the Olympic Park bombing and JonBenét Ramsey. But no amount of money would tempt him to leave academia, he says. And when it comes to his budding courtroom career, that may be just as well. Without his academic position, Foster might not have the same entrée to the courtroom. As a category of evidence, forensic linguistics does not enjoy the legal cachet of fingerprints and eyewitness testimony. For each case, in a procedure known as a Frye hearing, Foster must show he is a recognized expert in a recognized field of scholarship.

How legitimate a field is forensic linguistics? How reliable is its evidence? In Foster's opinion, "forensic linguistics is about where DNA evidence was a few years ago." That may overstate the case. Foster's is the only name on the FBI's referral list for authorship queries. The FBI's James Fitzgerald can't think of anyone, besides himself and Foster, who does criminal attribution. Bethany K. Dumas, editor of the on-line journal Language in the Judicial Process, explains that some subfields of forensic linguistics (such as structural linguistics, which can determine whether the structure of a sentence in a contract is ambiguous) may have reached a legal standard of proof but authorship is not one of them. "It's a discovery tool," Dumas says. In Kaczynski's case, she points out, Foster's testimony supported a claim of probable cause for a search warrant; it didn't convict.

Bruce Fraser, a linguist at Boston University, agrees. Attribution study can "make a good case," he says. "I don't know if it can ever be beyond a reasonable doubt." Foster knows that the academy and the courtroom decide about truth differently, but he doesn't dwell on it. "One doesn't want to be wrong," he concedes, with an uncharacteristically impersonal pronoun, "especially in capital cases."

Tonight at Vassar, however, Foster speaks with an authority that sweeps through both worlds--academic and criminal. "All felons, poets, and other liars, when they speak or write, have a certain degree of slippage between what they intend and what they write," Foster tells his audience. The challenge, Foster says, is to teach the FBI how to read deconstructively.

"One hesitates to say literary texts create felons," Foster says, but he goes on to describe Kaczynski as a man who loved literature and, like many lonely people, saw it as a kind of haven. Kaczynski believed that his parents, by pushing him into math, deprived him of creative expression. As early as 1967, he aspired to write fiction. He admired Somerset Maugham and Joseph Conrad so intensely that traces of their styles helped Foster tag him as the Unabomber. Kaczynski, Foster says, was a thwarted writer who killed to win readers for his message--namely, that technology was ruining our culture.

It's poetic justice that Kaczynski's most attentive reader helped convict him. Foster may have also brought about Kaczynski's worst nightmare: If literature is a sacred haven, then Foster has invited the serpent of number-crunching technology inside to sample its choicest fruit--to chew through Shakespeare's words as if they were so many numbers.

Caleb Crain's essay on Charles Brockden Brown will appear in A Center of Wonder: The Body in Early America (Cornell, forthcoming). His article "Pleasure Principles: Queer Theorists and Gay Journalists Wrestle Over the Politics of Sex" was published in the October 1997 LF.


Copyright © 1998 Lingua Franca, Inc. All rights reserved.