University of Virginia Library

A FUNERALL ELEGYE . . . NOT . . . BY W.S. AFTER ALL
by
Jill Farringdon

In 1989, Donald W. Foster published his book Elegy by W.S.: A Study in Attribution, [1] an examination of the authorship of A Funerall Elegye In Memory of the Late Virtuous Master William Peter of Whipton near Exeter; by W.S., printed for Thomas Thorpe, 1612. A work of investigative literary scholarship, it received widespread interest for both its scope and methodology. Some who read it were persuaded that the initials "W.S." could well have belonged to William Shakespeare; in fact, so persuaded that the poem has been included in more than one edition of Shakespeare's collected works.[2] No doubt confidence in Foster's scholarship was enhanced when he was able to identify correctly the anonymous author of Primary Colors (Joe Klein), and went on to be engaged in legal cases as a forensic expert, including that of the Unabomber. Such popular fame made him a widely known and admired "expert" in attribution.

However, not everyone was persuaded. Debate over the Elegye's authorship continued in literary circles, though "debate" is a polite word for what has been described as a "battle" which raged in the pages of the Times Literary Supplement. [3] Academics took sides. Vituperation and venom and sarcasm flew right and left ("Has he read it?"). Foster's own riposte to his most powerful critic, the Shakespearean scholar and editor Brian Vickers, asserted "an inattention to facts that would not be tolerated in an undergraduate".[4] At the same time, one of Foster's supporters wrote that "Vickers should probably try to keep his foot out of his mouth".[5]

Now it seems that the "battle" is over. In the light of new work suggesting that the Elegye's author was John Ford, Shakespeare's near-contemporary, Foster has conceded that he was mistaken: he, too, now believes Ford to be the likely author,[6] and joins what seems likely to become a literary consensus. Is there any more to be said?

Well, yes. This is a classic case where recourse to an objective method could have made a major contribution to such an acrimonious debate. It is somewhat unseemly (if fun for some readers) for professional academics to hurl brickbats at each other in public, particularly when arguments over attributing an obscure seventeenth-century poem can seem to some modern readers like the mediaeval debates over how many angels can dance on the head of a pin. The purpose of this article, therefore, is not merely to confirm the consensus that now appears to be emerging (although it is worth noting, en passant, that authorship questions are rarely finally settled by consensus: Freud went to his grave believing that the plays of Shakespeare were written by the Earl of Oxford). Its purpose is to suggest that where literary debate exists, a non-literary technique of language analysis has the virtue of avoiding the pitfalls of serious error which, in this case, led to publishing mistakes that must be an embarrassment to the editors of the Shakespeare editions now including a poem likely to be by John Ford.

Furthermore, a comment expressed since Foster's recantation pinpoints the difficulty for the non-specialist who must rely on the opinion of "experts" in these terms: such attributions can come to seem only pointless exercises (or "games") by which anyone can "show" anything which may be later disproved by another game-player. Thus, the new consensus may seem mired at the personal level: "And how joyfully the anti-Don-Fosterites dance around it".[7] This sense of disillusion underscores the urgent need for independent witness.

For it really does not matter how strong or weak a case critics and scholars may make in terms of cogent literary analysis: these will be irrelevant where an objective procedure exists which does not depend on literary judgments, however well-argued and compelling those may be to fellow-academics. What has always been needed is an independent method, one which is capable of delivering a statistical probability in matters of attributing human utterance in general and which can then be applied to particular cases like this one. This would by-pass any need for the development of a statistical "Shaxicon" (Foster), or reliance on rhetorical figures of speech like, say, anadiplosis or polyptoton (Vickers).

This article proposes that use of the scientific technique of authorship attribution called "cusum analysis" made it possible, as long ago as 1997, to show that the poem is certainly by one writer (i.e. it is homogeneous), but that

it is distinguishable from the writing of William Shakespeare. Further, it can be shown that the short Dedication[8] to the Funerall Elegye is also homogeneous, but can be distinguished from samples of the writing of Shakespeare, including his own two Dedications to his poems Venus and Adonis and The Rape of Lucrece, usefully signed "William Shakespeare".[9] There is, however, a new and surprising finding: the Dedication signed "W.S." can be distinguished from samples of the Elegye which it prefaces, indicating that the author of the one is unlikely to be the author of the other.

In other words, this analysis proposes that the ascription "by W.S." on the title page and the Dedication signed "W.S." were never reliable indicators of authorship either of poem or of Dedication. In his book, Don Foster began by asking that same crucial question: "the first thing we must ask is whether the initials are not simply wrong."[10] He went on "The answer is almost certainly no". The answer of cusum analysis was and remains: almost certainly, yes.

Before proceeding to outline the results of analysis, it will be necessary to elucidate this technique, which is not as well known as it should be.

The Question of Authorship

Donald Foster's book is packed with detail of the sort that literary scholars and attribution studies have traditionally used to investigate authorship of anonymous literary texts: namely, external evidence, where it exists, and internal evidence involving tables of massive statistical data, vocabulary counts, the construction by Foster of his special "Shaxicon" based on Spevack's concordances to Shakespeare's plays, and verbal parallels of specific lines and allusions from the plays. With all these approaches to attribution, the literary mind is familiar. Sifting through and understanding such extensive detail requires focussed concentration, and is a task capable of being undertaken mainly by professionals who are in a position to weigh the evidence.

Even then, the use of numerical methods, especially when a computer is used, remains for many scholars a suspect intrusion into the traditional literary world: witness the remark by poet Peter Levi that "such [computerassisted] analysis is almost always complete rubbish", an attitude described by Stanley Wells in a review of Foster's book as "an extreme point of view".[11] Nevertheless, it is one which is also widespread: one of Vickers's objections to Foster's conclusions in his book was his "too great reliance on computerized statistics" (my italics), a method he found served "an atomistic notion of

style".[12] A Foster supporter, though, saw Foster's study as a statistically unimpeachable example of "Shakespeare's late style".[13] The point to note here is that each critic, whether for or against, assumed that Foster's investigation made claims about literary style. This point will be crucially important as elucidation of cusum analysis proceeds.

However, the first observation to make when coming to negative conclusions regarding Shakespeare as possible author of the Funerall Elegye, by the use of a quantitative method very different from Foster's, is to note the satisfaction which may arise in the hearts of all those hostile to the use of computers in literary studies. O joy! The "computing experts" cannot agree!

The Cusum Technique [14]

This would be a short-lived joy since cusum analysis could actually be carried out—if more slowly—using an abacus, or pencil and paper, instead of a computer. Nevertheless, as one who is literary-critical by both training and inclination, the present writer has a large measure of sympathy with the dilemma of literary sceptics towards numerical studies, although that sympathy may be qualified by the observation that the computer is only a mechanical aid, and that there can be no objection to counting as such. [15] Consider discussion of the "number" of questions in Macbeth or the "preponderance" of disease imagery in Hamlet.

Cusum analysis has been used since 1990, both for studies in literary attribution and also in a forensic setting. It is the invention of A. Q. Morton, Fellow of the Royal Society of Edinburgh and a retired Minister in the Church of Scotland, whose life-time research has been devoted to developing an objective, scientific method of authorship attribution, one capable of being independently verified.[16]

The virtue of the method is that it is an attributive measure applicable to utterance of all kinds irrespective of date of composition or genre, whether in speech or writing, and has been widely tested. Attribution studies have been made on a short story newly found and attributed (initially) to D. H. Lawrence; to new essays by Henry Fielding; to the disputed Famine Diary, by Gerald Keegan purported to have been written on board a Famine Ship in 1845; and most recently to a newly-found poem "The Barberry Tree", the conclusion of joint-authorship by Coleridge and Wordsworth being precisely

identical to that arrived at by literary scholarship. It has been used on the utterance of children, including a study of Helen Keller from her first year of language-use at the age of eight.[17] including the highest courts—the Appeal Courts in London and Dublin, the Central Criminal Court, Dublin, and the Old Bailey.[18] So how does it work? Basic to the method is that a sophisticated analysis of language for attribution purposes must be based on regular and recurrent usage which is very frequent while also being unconscious to the user. What has always been needed is a method simple in principle and reliable in results, and, in cusum analysis, such a method has become available. The identification of authorship has been found to lie mainly in the small function words, usually of two, three or four letters, with which sentences are structured. Obviously, spelling is conventional and has differed over time, but the syntactic features analysed remain remarkably stable.

In trying to understand why cusum analysis "works" with function words, it is reasonable to think about an individual's total vocabulary. As any child's reading scheme will confirm, this may be divided into sections for frequency of usage. Twenty-five per cent of the time, language use in English consists of the repetition of very few words. One common scheme gives a mere twelve words (a, and, he, I, in, is, it, of, that, the, to, was,) followed by another 20 words for up to 35 per cent of normal utterance.

It may be asked how reliable such lists of "most frequent" words may be. Concordances of authors usually show that these twelve words appear at, or near the top, of the most-frequent-words lists. A comparison of the lists for Henry Fielding's novel Joseph Andrews (written in 1741) and for Dylan Thomas's Collected Poems (1950) yields a remarkable similarity. These two authors were writing very different kinds of work—novel and modern poetry—separated by two hundred and fifty years of England language usage; yet both revealed high-frequency lists of near identical words (the, and, of, in, I, a, to, you, my, is, that, he were among their top twelve; compare the twelve most-frequently used words in Shakespeare's total corpus: the, and, I, to, of, a, you, my, that, in, is, not—the overlap is obvious).

This surely confirms the usefulness of using these vocabulary items for recognising authorship. Half the time we speak, we are using over and over again the same function words; two-thirds of the time we speak, we are using a total of about two hundred and fifty words, which constitute only a tiny proportion of our total vocabulary. From the remaining thousands of words, which constitute only a tiny proportion of our total vocabulary. From the remaining thousands of words, we select those we need to convey content and meaning, or semantics.

Reading Cusum Graphs and Charts

The principle on which cusum analysis is based is thus extremely simple, being predominantly based on unconscious, habitual use of function words. The practise is within the grasp of any academic prepared to devote the necessary time and effort to learning it; but, to sound a cautionary note, this would be no rapid process with automatic results. Learning how to use the method with confidence takes about three to six months of practise on a wide variety of texts. By starting with the analysis of one's own utterance (as providing samples of known integrity, a crucial proviso), any researcher is enabled to gain confidence in the effectiveness of the method. The benefit of "tutorials" from experienced analysts is also important. Only then may one approach one's literary problem.

It should be noted here that there have been various attempts at critiques of cusum analysis: these have all failed in remarkably similar ways, namely, by misunderstandings of both principle and practise. One misunderstanding is the notion that the method is based on a single invariable language habit (counting words of 2 or 3 letters is the one usually selected), which is suitable for all writers/speakers and occurs in rigid proportions or ratios in each sentence—an obvious absurdity. There are, in fact, nine language-habit tests used on any sample under analysis to discover which one will be consistent for the writer; and these are counted in no simple proportional manner. Analysing for Authorship devotes a whole chapter to "Answering the Critics", and each critique is carefully examined there. Despite fairly widespread awareness of the few critical attempts, there is complete ignorance of the endorsement of the method's validity by the statistician responsible for writing the British Standard on cusums—who, further, developed a refinement which enhances the original method for the satisfaction of professional statisticians.[19]

One impression of the method's "unreliability", apparently widespread and also described in the book, deserves special mention here since it involves an interview with Morton by a television company who had asked for two samples of writing to be analysed for homogeneity. Given Morton's opinion that they were indistinguishable, the interviewer then dramatically revealed that they were by a man convicted of corporate crime, and by the Lord Chief Justice (Taylor). In actual fact, the "crook" was reading out a company report compiled by his department so that the sample was not his own utterance at all; moreover, the TV crew had got hold of a sample of writing by the wrong Lord Taylor![20] As always, the integrity of the text is everything. Given corrupt data, "wrong" answers will be inevitable: in computer jargon, "garbage in, garbage out". In Morton can be faulted, it is surely in accepting too readily samples whose origin he had not personally been able to verify. But to the viewing public, it apparently seemed Q.E.D. and remains so among some academics

in the attribution field to this day, even if it calls for a certain naïvety to believe that scientific validity can be decided by a ten-minute TV stunt. The old "I saw it in the newspapers" seems to have been superseded by the new "I saw it on TV".

Less well-remembered (or known at all) are the many blind tests where the technique has been successful, some of them required by the presiding judge in a court case. The most amazing example of these is perhaps the one where the "challenger", or person setting the test (Sir Kenneth Dover),[21] did the counting himself and passed over only the resulting numbers—not the Greek text—to Morton. Morton promptly discovered the inserted passage.

Properly used, cusum analysis is a useful tool. A major advantage of the method is that the results of analysing quite small samples can be visually demonstrated in graphic charts which may be understood by the non-expert (for example, a juror in a court case). As well as resistance to computers, there can also be an in-built resistance among literary professionals to looking at graphs. Yet a graph is only a way of presenting information. The more familiar we become with pictorial ways of interpretation, the easier it becomes to "read" the information in that form. How naturally we can now read, for example, television weather maps: "See those isobars packed in tightly together", says the TV Weatherperson, and we know we are in for a spell of high winds.

The sample of utterance under investigation is counted, by cumulative sum (hence, cusum), first for sentence length and the deviation of each sentence in the sequence from the average. The second step is to count, again by cumulative sum, some feature, usually called the "habit", of language-use within the sentence. The nine tests available to the analyst are based on function words, as already described. This is not the place to explore why such features have been found to work, although speculation is intriguing.[22] We need only point to the success of the method in analysing: natural utterance, both written and spoken; edited text; translated work; children's writing; dialect; and disputed utterance. Here we have an attributive measure of great sensitivity which is objective and which works across both time of authorial composition, and genre.

The last claim is one which has occasioned a degree of scepticism, and it is as well to outline the obstacles which must be overcome and habits of mind which must be set aside before the technique can gain confident acceptance by literary scholars. The cusum technique has nothing to do with style or literary value. It is purely quantitative, not qualitative (remember now those assumptions that statistics were analysing style?). Therefore, such procedure as comes naturally to the literary critic, that is sensitivity to tone, image, rhythm; comparison of like with like in terms of genre or date of composition; the difference between poetry and prose—all these must be set aside to turn instead to a study of language by measurable units, and that

unit normally the sentence. Whatever the soundness of the literary/stylistic judgments brought to bear on the Elegye by Professor Vickers or others, such judgments will remain interpretative rather than objective—indeed, a "game" to some readers. That is why it is worth drawing attention to a method which asserted five years ago that the Elegye had nothing to do with Shakespeare.

Attributing Shakespeare

There was no reason why Shakespeare should be an exception to the method. My own investigation, into analysing first Shakespearean samples and then samples of Christopher Marlowe, led to a submission for the Calvin Hoffman prize for attribution.[23]

Shakespeare proved to be homogeneous by more than one test, two language habits within the sentences of the chosen samples being found to be highly consistent: that of using three and four letter words plus words starting with a vowel (qs34lw+ivw) and a second consistent habit, that of using 2, 3 and 4 letter words (qs234lw).

Samples were chosen from Shakespeare's writing, from early and late work (using modern texts), to show consistency over time: thus, the ending of The Tempest (38 sentences) and the opening of Venus and Adonis (25 sentences) were chosen. When considering a disputed text—for example, in a legal setting—the first requirement is a sample of utterance indisputably that of the subject of the enquiry. In the case of Shakespeare, we have little personal utterance except for the short signed dedications to his two poems, Venus and Adonis (VA) and The Rape of Lucrece (RL): these may be regarded as authentic Shakespeare. The two dedications together comprise only eight sentences, yet the advantage of cusum analysis is its ability to show consistency in short samples.[24] This sample of the two dedications proved a useful discriminator. One of the sentences was in fact an anomaly: "What I have done is yours; what I have to do is yours; being part in all I have, devoted yours". Although cusum analysis has nothing to do with style, this departs so far from natural utterance as to be an exaggerated rhetorical flourish, a courtly bow in language—and may well be typical of Shakespeare in literary terms, which is of no relevance here. It should be stressed that in a longer sample of sentences, it would become virtually invisible; but in this short combined sample it slightly disturbed the consistent march of the graphlines, so was removed to make a clearer illustration.

In analysis, cusum graphs are drawn first of the sentence length (qsld), and then of the habit (e.g. qs234lw). These are produced on transparencies. It is then possible to slide the "habit"-transparency over the sentence-length transparency to see whether the two graph-lines track each other closely, or

even coincide. If they do, the sample may safely be assumed to be homogeneous, the work of one writer. If the two graphs separate, the sample may be safely assumed to the "mixed" utterance, or non-homogeneous.

However, for purposes of presentation, instead of using transparencies the two graphs can be produced as one, to make a cusum "chart": this shows one graph-line superimposed on the other. Figure 1 shows two graphs superimposed

as a "chart" for the cusum analysis of Shakespeare's two Dedications (VA and RL combined). The habit shown is one of his two discriminating habits, the use of three and four letter words plus words starting with a vowel (qs34lw+ivw). The two lines clearly track each other here, a consistent habit running throughout, so that a single author is indicated, as we knew to be the case.

Having shown that Shakespeare was no exception to the method, the authentic Shakespearean utterance was analysed with samples from The Tempest and Venus and Adonis. The Dedications blended smoothly with the selected samples and revealed no separation, the two graph-lines tracking each other throughout to show a consistent habit at work, confirming homogeneity for the two combined samples (see figs. 2, 3).

As already reported in the opening paragraph of this article, cusum analysis of the Funeral Elegye yielded the unsurprising results that the poem and its Dedication were both homogeneous. Each was written by one writer (and the many relevant charts produced are available for inspection). Was that writer Shakespeare?

Firstly, in a combined analysis of the Dedication "by W.S." and the two Dedications by Shakespeare, the two lines clearly separate, indicating two writers at work (see fig. 4). Then, this result of mixed authorship was repeated in subsequent analyses with samples from Venus and Adonis and The Tempest: the inconsistency with Shakespeare's writing was confirmed. Whenever either samples from the Elegye, or else its Dedication, were analysed in combination with samples of Shakespeare, the two-graph lines separated markedly, indicating not single but mixed authorship (see figs. 5, 6).

Analysis by samples

Obviously, it is only possible to show as illustrations a few of the results obtained from the process of analysis. These are intended to show how the method works and to indicate the outcome of the tests. There may be some unease that what is being reported is based on samples of authors only. But

FIGURE 4. Shakespeare's dedications to Venus and Adonis and The Rape of Lucrece followed by the Elegye dedication.

FIGURE 5. Twenty-four sentences from Venus and Adonis followed by 20 sentences from the Elegye.

if these consistently give the same result, whether it be that the samples are indistinguishable or distinguishable, it is fair to ask how many samples would become "satisfactory" to the doubtful? Ten—twenty—fifty? There has to be a point at which the accumulated evidence becomes overwhelmingly persuasive. Readers disinclined to trust sampling should ask themselves whether

FIGURE 6. Fifteen sentences from the Elegye followed by 30 sentences from The Tempest.

it is mere chance or co-incidence that the two Dedications which were signed by William Shakespeare can instantly be shown to be consistent; and then, whether it is equally co-incidental that they should happen to be indistinguishable from samples from his first poem and last play but distinguishable from samples by other writers, like Christopher Marlowe and John Ford.

It would be disquieting, though, for literary scholarship if this investigation were to seem to set itself up in opposition to traditional methods. As Morton has written, "Cusum analysis is not some isolated touchstone of authenticity".[25] In fact, cusum analysis works best when able to support an attribution made by literary scholars, and this it has done many times. In addition to literary attributions already cited, Donald Foster's conclusion with regard to the authorship of "Twas the Night Before Christmas"[26] is entirely supported by my cusum investigation of that poem's authorship.

Naturally, most important for a positive attribution of the Funerall Elegye is the virtual consensus that it was John Ford who wrote the poem. This provided a new putative author to work on. Cusum analysis showed first that Ford's writing was consistent by the habit of using words of 2, 3 or 4 letters (qs234lw). Then, samples of Ford were shown to be consistent with samples of the Elegye (fig. 7). Four dedications by Ford can be also shown to be perfectly consistent with each other and with the poem (fig. 8). In both

FIGURE 7. Combined sample of 42 sentences from John Ford's 'Tis Pity She's a Whore with 20 sentences from the Elegye inserted.

FIGURE 8. Twenty sentences from the Elegye followed by John Ford's dedications to The Lover's Melancholy, Perkin Warbeck, Love's Sacrifice, and The Lady's Trial.

these charts, there is a smooth tracking of the graph-lines and no separation: Ford's writing and that of the Elegye's author are indistinguishable.

The Dedication to the Elegye

We have seen in figure 4 that Shakespeare's two dedications and that of the Elegye separate markedly, indicating mixed authorship. What was utterly unexpected was that analysis which combined samples of the poem with its Dedication (fig. 9), and Ford's signed dedications with it (figs. 10, 11), also gave the result of mixed utterance: dedications which we know are by Ford give a consistent result, while the dedication "by W.S." is not consistent with them. Samples of the poem were consistent with each other, but not with its dedication "by W.S.". The two graph-lines for these analyses can be seen diverging widely, indicating mix of authors.[27]

Key Questions

Is the poem A Funerall Elegye by W.S. by one writer? Yes: analysis of lengthy samples revealed homogeneity.

Can an authentic sample of Shakespeare, his two signed dedications, be shown to be consistent with other samples by him, but to separate from samples by another writer? Yes: all analyses with both Christopher Marlowe and John Ford resulted in separation, visible proof of mixed authorship.

Is the Dedication to the Elegye consistent with Shakespeare's own two dedications to Venus and Adonis and The Rape of Lucrece? No. The Dedication

FIGURE 10. Combined analysis of John Ford's dedication to The Lady's Trial (6 sentences) with the Elegye dedication (5 sentences) following.

FIGURE 11. Four dedications by John Ford (to The Lover's Melancholy, Perkin War- beck, Love's Sacrifice, and The Lady's Trial) followed by the Elegye dedication.

to the Elegye was clearly distinguishable from Shakespeare's signed dedications.

Is the Dedication to A Funerall Elegye written by the same person who wrote the poem? Cusum analysis has shown that it was distinguishable from the author.

Are samples of John Ford consistent with samples of the Elegye? Yes: they were found to be indistinguishable.

The conclusion, therefore, is that cusum analysis has shown Shakespeare to be extremely unlikely as author of A Funerall Elegye and supports the consensus that John Ford is the likely author (though not of its Dedication).

Notes