"Shakespeare: Editions and Textual Studies in 2020" Not the Year's Work in English Studies

Only one major scholarly critical edition of Shakespeare appeared this year: Measure for Measure edited by A. R. Braunmuller and Robert N. Watson, which completes the Arden Shakespeare Third Series. According to their Preface, Braunmuller prepared the text of their edition and wrote almost all the commentary notes, and Watson wrote the Introduction and index. Others contributed Appendix One and the Casting Chart. Since they collectively take responsibility for everything in their edition, I will refer to Braunmuller and Watson when remarking upon their edition's choices.

    Braunmuller and Watson begin their Introduction (pp. 1-148) by observing that Measure for Measure is, with The Merchant of Venice, Troilus and Cressida, and All's Well that Ends Well, one of the four so-called problem plays that are formally comedies but with troubling endings. Measure for Measure was the last play of Shakespeare's comic period before he wrote his great tragedies Othello, King Lear, Macbeth, Antony and Cleopatra, Coriolanus, and Timon of Athens. Braunmuller and Watson start with a series of short sections of literary criticism with subheadings of "The Morals of Measure for Measure", "Mortality", "Substitution", "Sexuality, Law and Marriage", "Consent", "Catholicism, Protestantism and Puritanism", and "Love", each no more than seven pages long.

    Braunmuller and Watson note that the Elizabethan law and practice of marriage left quite a few ambiguities regarding promises and consent, and that with this play Shakespeare "ran a kind of stress-test on the joints of this structuring social institution" (p. 28). The play is concerned with male consent as well as female. Barnadine, for instance, will not consent to die and is part of an exploration of how consent is not the binary that we today tend to think it is; rather "it exists in degrees and contexts" (p. 32). With all its focus on people's redeemability and its rejection of the strict and puritan-like Angelo, the play seems distinctly sympathetic to Catholicism. But Braunmuller and Watson note that Mariana's tryst with Angelo has the trappings of the Annunciation -- the hortus conclusus and the Angel/Angelo presenting a key -- and that Catholics would be offended by the depiction of meddling friars.

    A rather longer section of the Introduction is concerned with "The Ending" (pp. 41-56) and is mainly a sketch of the choices of various theatrical productions. Then comes a substantial study of the "Characters" (pp. 56-94), which sounds rather old fashioned but in fact is refreshingly alert to how plays are understood by readers and playgoers as it works through the dramatis personae, describing each. Sometimes the tone is jarringly colloquial: "Duke Vincentio, once widely viewed as a wise, resourceful leader of his city, has slumped badly in the opinion polls over the past half-century" (p. 57).

    Samuel Johnson worried about just when the Duke learnt of Mariana's situation, since if it was before the Duke made Angelo his deputy than this reflects badly on him. Braunmuller and Watson find that the problem with applying a Foucauldian lens is that Michel Foucault was adamant that the important changes in socio-cultural and sexual forces that the play engages with emerged in the eighteenth century not the sixteenth and early seventeenth century. (Foucault was wrong.) Gary Taylor thinks that Shakespeare made the Duke genial but in revising the play Thomas Middleton made him sinister; Braunmuller and Watson disagree, finding that Middleton's changes did not much affect this character. A charitable view of the Duke is that unlike extreme Protestant Angelo and extreme Catholic Isabella, the Duke believes that all souls are redeemable and wants to teach his people that. 

    Braunmuller and Watson deduce that Kate Keepdown must have become a prostitute only after Lucio impregnated her (making her unmarryable) since ". . . if she had already been a sex-worker, he would hardly have been sure the offspring was his, and was unlikely to have 'promised her marriage' as Mistress Overdone reports he did" (p. 72). This makes sense, but I am not sure that Shakespeare expected us to reason this out and Lucio has enough moral defects even if we overlook this detail. The only misstep in this part of the book is the over-generalization that ". . . Puritans sought to suppress theatres . . ." (p. 82). In fact, as Margot Heinemann showed in Puritanism and Theatre 40 years ago, although the categories of 'Puritan' and 'antitheatricalist' overlapped somewhat there were plenty of Puritans who liked theatre and non-Puritans who hated it.

    In "Literary and Dramatic Sources" (pp. 95-111), Braunmuller and Watson sketch the main plot elements that draw on multiple sources, in particular sexual extortion, the disguised ruler, and the bed-trick. These sources include Cinthio, the pen-name of Giovanni Battista Giraldi, whose collection of Italian stories called Hecatommithi contains sources for Othello and Measure for Measure, and George Whetstone's Promos and Cassandra and prose work Heptameron of Civil Discourses. Shakespeare toned down the really outrageous outcome of Whetstone's play in which the victim actually submits to the sexual blackmail and then is forced to marry her blackmailer.

    Thomas Lupton's prose dialogue The Second Part of Too Good to Be True is a possible minor source, and Braunmuller and Watson also usefully look at some plays in Shakespeare's company's repertory that have plot parallels with Measure for Measure. They also find some matches between the play and King James's Basilicon Doron and his own life and habits. The first recorded performance of Measure for Measure was at Whitehall on 26 December 1604 (recorded for a payment to "Shaxberd") and stylistic evidence puts it probably after Hamlet, Twelfth Night, and Troilus and Cressida and hence after 1602. That Middleton adapted the play makes it hard to use topical allusions to date it. Plague closed theatres from March 1603 to April 1604, during which time Shakespeare presumably wrote the play. The only early edition is in the 1623 Folio collection of Shakespeare's plays.

    Braunmuller and Watson devote a little over five pages of their Introduction to Middleton's hand in the play. They sketch the work of John Jowett and Gary Taylor on Middleton's adaptation of the play, remarking that in the Oxford University Press Collected Middleton edition of 2007 ". . . Jowett (paradoxically) offers a conjectural reconstruction of Shakespeare's version before Middleton adapted it . . ." (p. 118). All other editions essentially reprint the Folio, so we have multiple modernizations of what the play became after Middleton adapted it. Jowett's intimate knowledge of what Middleton likely did was put to the illuminating task of undoing the adaptation.

    Braunmuller and Watson believe that there is really strong evidence for Middleton's hand in the 1623 Folio text only at 1.2.1-18, which evidence is that: i) the exchange is full of topical references that would have made no sense in 1603 but good sense in 1621; ii) this being an addition would explain Mistress Overdone later reacting with surprise to Pompey's news of Claudio's arrest that she herself previously delivered to Lucio and the two gentlemen; iii) the Folio page is heavily crowded, which is consistent with Mistress Overdone's delivery of the news of Claudio's arrest being Middleton's rewrite of the Shakespearian version in which Pompey delivers the news, which Shakespearian version was meant to be omitted; and iv) the word choices at the beginning of 1.2 are distinctly Middletonian.

    Braunmuller and Watson note that seventeenth-century performance texts -- they do not say which -- often omitted the parts, and undid the reorderings, that Jowett and Taylor attribute to Middleton. This suggests that Middleton's changes were thought not quite dramatically acceptable and perhaps even that some manuscript form of the pre-adaptation Shakespearian play was in circulation until the eighteenth-century. The awkward, unmetrical presence of soft oaths where strong oaths would be expected shows that the Folio version reflects alterations made in or after 1606 to meet that year's Act to Restrain the Abuses of Players.

    Braumuller and Watson slightly garble the implications of the play's act intervals, thinking that having four short breaks in a performance "became common in the indoor theatres around 1609 (partly to allow refreshing of the candles)" (p. 121). In fact, act intervals had always been observed in indoor theatres because these relied on candles and candles had to be tended. The King's Men got possession of the indoor Blackfriars theatre in 1608 and as part of the regularization of practices across it and their outdoor Globe theatre they began observing act intervals at both, even though these were unnecessary at the Globe because it did not use candles. Other companies began to copy them and use act intervals outdoors too, so that from 1609 act intervals became universal practice. Braunmuller and Watson find plausible Taylor's claim that the play was originally set in Ferrara rather than Vienna.

    There was no quarto edition of the play -- which would have been evidence of its popularity -- and no recorded performances between the one at court in 1604 and the Restoration. Braunmuller and Watson think that maybe the play was too experimentally radical and subversive to succeed until recent times, since it presents paradoxes that encourage scepticism towards moral prescriptions. In "The Play Onstage: From Prudery to Politics" (pp. 124-130) they give a compressed chronological stage history. There is a familiar pattern: the play was heavily adapted in the Restoration and then in the eighteenth-century the original, in the sense of Folio-faithful, script begins to return. The play was largely neglected in the nineteenth century and not until Peter Brook's 1950 production in Stratford-upon-Avon did it fully return to the repertory.

    In the final section of their Introduction, "Measure for Measure as Golden Meaning" (pp. 130-145), Braunmuller and Watson give a reading of the play's concern with finding living compromises between extremes and then they shift into a survey of recent productions that emphasized the play's claustrophobia. They suggest that the hinge of the play, "the instant that the pendulum starts to swing back from death towards life" (p. 139), is Barnadine's refusal to die because he has a hangover, which is a kind of affirmation of life.

    Braunmuller and Watson discuss the Duke's decision to let Isabella think that her brother Claudio lives and then to reveal that he does not via a quip: ". . . she asks, 'Hath yet the deputy sent my brother's pardon?', the Duke answers, 'He hath released him, Isabel, from the world; | His head is off and sent to Angelo'" (p. 141). With a pause after "released him, Isabel" this seems cruel, but Braunmuller and Watson suggest that the Duke might have learnt the cruel trick from Isabella herself. When Claudio in prison asks Isabella what news she brings him, she replies "Why, as all comforts are: most good" and then tells him "Yes, brother, you may live" before explaining why he cannot.

    And so to the text of the play itself. As usual, the comments here will be on particularly interesting cruxes that have exercised editors and on characterizing the general trend of the editing. Of the edition's 206 pages of play, 42 have no textual notes and as many again have only a single note recording the adoption of a preceding edition's stage direction. Thus it is clear that Braumuller and Watson have intervened little in the dialogue of the play as presented by the Folio.

    At 1.2.116-119 Braunmuller and Watson print "Thus can the demigod, Authority, | Make us pay down for our offence by weight. | The words of heaven -- on whom it will, it will, | On whom it will not, so; yet still 'tis just". The Folio has no punctuation after weight and there the idea seems to be that the words of heaven are what we are forced to pay down by weight. Paying down by weight seems to mean paying in full, without discount. C. J. Sisson suggested that paying down a bond makes even better sense and proposed the graphically plausible emendation of words to bonds. I cannot make sense of Braunmuller and Watson's full stop after weight since what follows seems to make no sense of "The words of heaven", but they say these "repeat and elaborate" the preceding "complaint".

    The Folio has Claudio tell Lucio that Isabella has "a prone and speechlesse dialect, | Such as moue men" (1.2.178-179) and Braunmuller and Watson do not emend this. The problem is prone, which does not give the right sense. Also, move seems to require a plural subject so that if we emend prone we should change it to a noun so that it and speechless dialect together form the two things that move men. Previous editors have suggested changing prone to the adjective prompt (which does not solve the disagreement-in-number problem), while Sisson offered the graphically plausible noun grace. At 1.3.21 the Duke says that he has "let slip" the city's harsh laws for fourteen years whereas earlier (1.2.163) Claudio said it had been nineteen years. Braunmuller and Watson assume that the inconsistency comes from an error in transmission or authorial indifference. I suppose it could be argued that without an explicit proclamation that laws will not be enforced, two people might simply disagree about when the latitude began.

    In the Folio, the Duke says that by giving Angelo the use of his name (as his deputy) to reimpose the application of strict laws, he (the Duke) manages to keep his nature (as distinct from his name) out of any slanderous imputation of wrongdoing: "my nature [is thus] neuer in the fight | To do in slander" (1.3.42-43). The problem is in slander where we would expect perhaps it slander with it referring back to my nature.  Braunmuller and Watson observe the problem with in but decline to fix it. James Orchard Halliwell proposed To do me slander which is graphically plausible and especially apt.

    Escalus reflects on how varied are the ways of sin and punishment in the Folio's "Some rise by sinne, and some by vertue fall: | Some run from brakes of Ice, and answere none" (2.1.38-39). The problem is brakes of Ice which Braunmuller and Watson leave unemended and modernize to breaks of ice, interpreting this as meaning that some people manage to avoid the dangerous cracks in the ice that they run across. They point out that a similar idea occurs in Troilus and Cressida: "The fool slides o'er the ice that you should break". In an article reviewed in YWES for 2014, Matthew Steggle found from searches of the Early English Books Online Text Creation Partnership (EEBO-TCP) dataset that in devotional literature notions such as brakes of sensuality and brakes of vanitie appear, giving credence to Nicholas Rowe's emendation to of Ice > of vice", meaning ensnaring thickets. (Rowe also emended from brakes > through brakes.)

    Explaining why he cannot postpone Claudio's execution, Angelo tells Isabella that the newly awoken law is like a prophet who "Lookes in a glasse that shewes what future euils | Either now, or by remissenesse, new conceiu'd" (2.2.99-100), which evils will be cut off by the law's strict action. Braunmuller and Watson leave the awkward Either now unemended without explaining what it means. Pope created the popular emendation now > new so that Angelo refers to future evils that are either new or newly conceived. At 2.2.152, Braunmuller and Watson have Isabella tell Angelo that she will bribe him "Not with fond sicles of the tested gold" where editors now modernize sicles (the Folio's Sickles) to shekels. Their explanation is that "Modernizing 'Sickles' /'sicles' to 'shekels' would unbalance the line's euphony, the sounds that move from fond to sicles to tested". I cannot detect the pleasing aural effect that they are referring to.

    Braunmuller and Watson's decision not to undo the censoring of all the religious oaths leads them to print the barely intelligible "Heaven in my mouth, | As if I did but only chew his name" (2.4.4-5). Most likely, the first word was originally God not Heaven since chewing heaven's name makes little sense. Braunmuller and Watson explain in a note that censorship turned God into Heaven but they muddy the waters by starting the note "Heaven i.e. God, the name of God". But heaven is not God nor the name of God and it appears here only as a clumsy substitution to meet the law against stage profanity.

    At 2.4.16-17, Braunmuller and Watson do not emend the Folio's having Angelo say "Let's write good Angell on the Deuills horne | 'Tis not the Deuills Crest", which is hard to make sense of.  For the New Oxford Shakespeare, Terri Bourus pointed out the parallel phrasing in As You Like It's "Take thou no scorn to wear the horn,| It was a crest ere thou wast born" in which, as here, ". . . 'horn' ends one verse-line, 'It' begins the next verse-line, and 'crest' occurs in the middle of that second line". Thus Shakespeare elsewhere equated the horn and the crest, which supports John Dover Wilson's suggested emendation of not > now.

    At 2.4.111, Braunmuller and Watson print "Ignomy in ransom and free pardon" using the Folio's form "Ignomie" (which was an accepted shortening in Shakespeare's time) rather than the Ignominy of the 1632 Second Folio. They defend this choice by remarking that "the archaic form is retained here for the metre", but it is not clear what they mean. Their version of the line is headless, lacking an unstressed syllable before Ig, and hence is no more metrically regular than if they had adopted Ignominy.

    Braunmuller and Watson have Isabella say that taken together Angelo's appearance and demeanour "follies doth enew" (3.1.92), meaning that they confine or restrain follies. The Folio has "follies doth emmew" and Thomas Keighley was the first to emend emmew > enew, which is a term from falconry. But Steggle found that despite the apparent French etymology this falconry term was often spelled emew so emendation is unnecessary.

    A much-discussed crux occurs at 3.1.95-98. The Folio has Claudio ask, after Isabella's description of Angelo's depravity, if she means "The prenzie, Angelo?" and she replies "Oh 'tis the cunning Liuerie of hell, | The damnest bodie to inuest, and couer | In prenzie gardes". The problem is the two occurrences of prenzie, which has no meaning, and it is compounded by the need to find a single alternative word that fits both occurrences' contexts and that might plausibly have been misread twice to produce prenzie. Braunmuller and Watson leave the problem unsolved and print prenzie both times. Popular emendations are princely and precise (in the sense of fastidious, puritanical), but editors who use these usually admit that they are not entirely convinced by them.

    In the Folio, the Duke says that the messenger knocking on the door "wounds th'vnsisting Posterne with these strokes" (4.2.87) and Braunmuller and Watson merely modernize to "wounds th'unsisting" even though the last of these is not a word. This really is abdicating the editor's responsibility to at least render the text in English. Past emendations for unsisting have included unresisting and unshifting and unlistening.

    At 4.4.24-25, the Folio has Angelo in soliloquy reassuring himself that nothing Isabella says to accuse him can be effective: "For my Authority beares of a credent bulke, | That no particular scandall once can touch". The problem is that beares of a credent is awkward and does not set up the structure needed for the consequence beginning with That. Braunmuller and Watson adopt Lewis Theobald's bears a credent bulk, but this makes the meter rough. For the New Oxford Shakespeare original-spelling edition, Bourus emended to beares a so credent which smoothes the meter and has the merit that the Folio's of a is an easy misreading of a so and so sets up the so . . . that structure needed for That in the second line.

    When the Duke says to Isabella "Give me your hand and say you will be mine" (5.1.492), there is only an indirect reference from Braunmuller and Watson (in a commentary note about Claudio being pardoned) to alert the reader that this is a marriage proposal and that the absence of a verbal reaction from Isabella is dramatically significant. Likewise at the closing lines of the play Braunmuller and Watson make no mention that this is a second marriage proposal and that again Isabella is given no words of response.

    The first of three appendices is a "Note on the Text" (pp. 361-374). Braunmuller and Watson sketch the history of the publication of the Folio and the compositor analysis of Measure for Measure, which was already much contested before Pervez Rizvi (in an article reviewed in NYWES for 2016) showed that virtually all compositor identification by spelling (which is essentially all Folio compositor identification) is unreliable: the methods just do not work. (Rizvi showed that he could construct arbitrary and random attributions of compositor stints and then find in support of these the same strength of evidence as the supposed evidence for the latest state of scholarship on compositor stints.) Braunmuller and Watson describe the minor problem of mislining of Measure for Measure in the Folio as a way into describing the casting-off of copy and the evidence from crowded and widely spaced pages. They also discuss the problem of amphibious lines, which they do not try to adjudicate on.

    Considering the copy for the Folio edition, Braunmuller and Watson suggest that the King's Men's scribe Ralph Crane copied either the current playbook or (less likely) Shakespeare's own draft of the play. They detail Crane's characteristic interferences including single-words-in-brackets. They find that the song that starts 4.1 could just about have been Shakespeare's own composition that the authors of Rollo Duke of Normandy later added a second stanza to, but Braunmuller and Watson consider this unlikely because the whole song seems dependent on a lyric printed in 1597.

    Braunmuller and Watson summarize the claim of Jowett and Taylor that Middleton adapted the play and that we have in the Folio the post-adaptation version. They consider the evidence mostly "circumstantial" and Middleton's presence "conjectural" (p. 371) and "an intriguing hypothesis rather than a fully proven hypothesis" (p. 372). Braunmuller and Watson remark that Jowett's edition of Measure for Measure for the Oxford Middleton and Bourus's for the New Oxford Shakespeare "venture to restore references to God and blasphemous oaths where their original presence seems justified -- a bold, if also subjective, editorial departure" (p. 373). In fact, Jowett did this for the 1986-87 Oxford Complete Works too so it was not quite the innovation that Braunmuller and Watson suggest.

    In a section on "This Edition", Braunmuller and Watson report that "Significant textual interventions of these early reprints and editions", meaning the Second, Third, and Fourth Folios and Rowe's 1709 editon, "are recorded in the commentary or textual notes in this edition" (p. 374). This is surprising phrasing. Usually one would record only those interventions that one thinks might be correct restorations of a lost Shakespearian reading, but perhaps that is what they mean to convey by the word "significant".

    Braunmuller and Watson summarize their editorial approach as "enquiring textual conservatism, sometimes in the face of editorial tradition" (p. 374) and they approvingly quote Samuel Johnson's wish "that we all explained more, and amended less" (p. 374). Then they provide a list of what they think are the most interesting textual puzzles. Appendix Two is a list of lineation notes that other editions put on the same page as the dialogue, alongside the commentary and textual notes. Appendix Three is a real "Casting Chart" that allocates roles to actors numbered 1 to 11 and boys numbered 1 to 3.

    Richard Knowles's two-volume King Lear for the New Variorum Shakespeare is a substantial work of condensation of scholarship, but it is not a critical edition of the play and will be noticed here only briefly. The "Plan of the Work" (pp. xv-xxxii) explains that it is based on the 1608 First Quarto (Q1) with lines that appear only in the 1623 Folio (F) edition imported from there, and that the textual notes record "significant departures" from those two authorities "in seventy-seven editions of the plays ranging in date from 1619 to 2000" (p. xv).

    Lines added from F are marked with asterisks and Q1-only passages are enclosed in half-brackets. Knowles does not indicate how he decides if a line is essentially the same the Q1 and F (in which case Q1's version is used), and this is not a trivial problem. Knowles considers Q1 the edition closest to Shakespeare's papers and that F "represents a later version incorporating changes made for various theatrical and artistic purposes" (p. xv). But if F represents a later version it is unclear on what basis Knowles imports its lines to Q1: does he do that only where he thinks Q1 lacks a line Shakespeare wrote in his authorial papers but which got omitted from Q1 by some error of transmission? Knowles does not say.

    Most importantly for our purposes, Knowles has not emended the text where it is in error: ". . . both Q and F lines presented here are in their original, uncorrected states, so that the commentary may proceed from the beginning towards later changes" (p. xv). For instance, Knowles prints Q1's meaningless "a dogge, so bade in office" where F has "a Dogg's obey'd in Office". Thus "The text is a modified diplomatic reprint . . ." (p. xvi) of Q1 and F. Even press variants are left in their uncorrected state, but "Minor typographical blemishes such as irregular spacing, printing space-types, and wrong-fount, damaged, turned, transposed, misprinted or clearly erroneous or missing letters or punctuation marks have been corrected, in most cases silently" except where the "anomaly is likely to have any bibliographical significance" (p. xvi).

    Although based on Q1, this edition still lets F dominate in certain matters. The Through Line Numbers (TLNs) in the right margin are F's, and "All Q prose has been relined as in F to enable the use of this F-based numbering system" (p. xvi). F's TLNs are used for numbering notes and making all cross-references. The running headers give the act-scene-line numbers from the 1974 Riverside edition, and hence also from Marvin Spevack's concordances to the Riverside. Knowles lists the 77 editions that are collated and given sigla, and a bibliography of other works cited in the commentary. Overall, then, this Variorum is an invaluable resource for quickly discovering who was first to provide a particular interpretation of a given word, phrase, or line, and as a record of what emendations previous editions made. But it is not a critical edition in its own right.

    In his discussion of "The Texts" (pp. 1042-1207), Knowles pays no attention to the demonstration by Pervez Rizvi (reviewed in NYWES for 2016, four years before this edition appeared) that the majority of work done on compositor identification is worthless. But he is scrupulous to describe the various arguments that have been offered for each topic under investigation and not to champion any particular ones. Knowles is known to oppose the theory that Shakespeare revised King Lear and that the substantive differences between Q1 and F are mostly due to his doing so; he has published a series of articles detailing his reasons for rejecting this idea. In the section of this edition about that topic ("Revisions and Other Changes", pp. 1174-1198) Knowles is scrupulously even-handed, conveying in detail and with copious quotation the arguments of the revisionists with whom he disagrees, and pointing the reader to the parts of the texts that they cite to support their argments, just as he points the reader to the bits that their opponents cite.

    The idea that F represents a revision of the script underlying Q1 was dominant until the early twentieth century. Knowles summarizes the major editors' views on the Q1/F differences from the eighteenth and nineteenth centuries. The one he approves of the most -- originating with Nicolaus Delius -- is that if Shakespeare set about revising the play he would not make the quite trivial differences we see between Q1 and F; he would have done a thoroughgoing revision. Knowles pays particular attention to the fact that F seems to add or remove or substitute words and phrases to improve the ragged verse of Q1, yet we know that towards the end of his career, when he is supposed by the revisionists to have revised King Lear, Shakespeare was using verse more freely than ever before. One possible revisionist response is that the roughness of the verse in Q1 in these cases might well be due to disruptions in transmission rather than Shakespeare writing unmetrical verse.

    The last of the Shakespeare editions to be noticed this year is David Lindley's The First Quarto of The Merry Wives of Windsor in The New Cambridge Shakespeare's Early Quartos series, which has interesting reflections on its copy text, the 1602 quarto (Q1). The series's General Editor Brian Gibbons provides a series Introduction (p. v) that summarizes the textual situation of Shakespeare, but claiming rather more than we truly know about certain matters. For instance Gibbons asserts that "The company of players required a manuscript fair copy of the play . . ." rather than the author's papers, whereas in fact some theatrical playbooks do seem to be based on the author's own  papers. Also, Gibbons remarks that "There are certain quartos which are abbreviated . . .". Using the verb abbreviate asserts that the scripts were once longer than they appear in the short quartos, but of course it is also possible that the short quartos represent the plays as originally written and that the longer versions represent expansions upon those base texts.

    In his Introduction, Lindley sketches the changing critical judgements about Q1 The Merry Wives of Windsor, from bad quarto to recent revaluation. Reporting on performances based on Q1, Lindley notes that the main plot came across successfully but the subplots -- especially the Caius/Evans rivalry and the horse-stealing incident -- made little or no sense. Moreover, the content of Q1 is in contradiction with its title page, which promises "the pleasing humours of Sir Hugh the Welsh knight, Justice Shallow and his wise cousin Master Slender. With the swaggering vein of Ancient Pistol and Corporal Nim". There is actually very little of Shallow, Pistol, and Nim in Q1. The title page of Q1 promises the text of a play as it was performed for the queen, which was a marketing advantage.

    We might thing that Shakespeare would not have written a Falstaff play after dramatising his rejection at the end of 2 Henry 4 and reactions to his death in Henry 5, but Giorgio Melchiori argued the opposite: The Merry Wives of Windsor makes up for his loss. Also, Nim first enters the Henriad in Henry 5 so if The Merry Wives of Windsor preceded that play the audience would not know who he is. (But so what? Why might Shakespeare not have introduced him in The Merry Wives of Windsor and then used him more extensively in Henry 5?).

    Dating the composition of the play via the allusion to the Order of Garter, which was celebrated in a feast in 1597, relies on two specific links: Lord Hunsdon (patron of Shakespeare's company) was elected to the Order at that feast (and so might have wanted a Garter play around that time), and so was Count Mompelgard (Duke of W rttemberg) whose name might be behind Q1's "cosen garmombles" -- in the Folio this is "Cozen Iermans" -- and the German subplot. But as Richard Dutton argued, The Merry Wives of Windsor is not a play that celebrates heroic knighthood and it would be unsuitable for a Order feast; he reckoned that the F-only courtly material was added for a court performance in 1604.

    The character named Brook in Q1 has the meaningless name Broom in F, spoiling puns about fords and overflowing rivers. If this difference reflects a renaming that was part of the alteration to Shakespeare's writing necessitated by the Oldcastle controversy of 1597, then why did Q1 print the form Brook? For Dutton, this means that the change happened later and hence Q1 represents an earlier stage of the play than F. At this point (p. 7n3) Lindley footnotes B. J. Sokol's essay "A Warwickshire Scandal" (reviewed in YWES for 2009), which found a court case of 1600 that strongly parallels the plot and phrasing of the play, making a date of composition of The Merry Wives of Windsor before 1600 impossible, but Lindley does not discuss the implications of this.

    Lindley considers a moment of apparent confusion in the play -- the entrance of Mistress Ford and Mistress Page in the first scene -- and shows that the Q1/F differences could be explained either of two ways. Either ". . . the Folio represents the first version and the Quarto records a slice of later improvised business, or else the Folio is tidying up the Quarto's confusion" (p. 9). The crux is that in Q1 Falstaff kisses "her" (which could be either of the women)  and Mistress Ford objects that he has made a mistake of some kind (it is not clear what), and in F (which has no entrance direction for the two women at this point) the kiss and reference to a mistake are absent. One way to make sense of Q1 is that Falstaff mistakes Mistress Page for Mistress Ford and kisses the former but calls her the name of the latter.

    Lindley summarizes the New Bibliographers' views regarding the determination of printer's copy from printed edition, and the scepticism about this that later came to dominance. The idea that Q1 represents a text cut down for performance (perhaps on tour) is undermined by the fact that it needs almost as many actors as F and that it is too short, since it can be played in about an hour and a quarter. The main problem is that several of the various explanations for Q1 -- it was the original version, it represents the play underlying F after it was cut down for performance, it was garbled and shortened by memorial reconstruction or shorthand transmission -- might be operating at once. 

    There is a lot of prose in Q1 that is typeset to look like verse. Someone taking shorthand notes at a performance would not know before a speech started whether it was going to be one or the other, and might write it like verse until becoming certain that it is prose. Alternatively, as Douglas Bruster suggested, the point of printing prose to look like verse was to fool the browsing book-buyer, since blank verse writing was deemed more valuable than prose. In some places the printing of prose as verse in Q1 seems like an attempt to stretch copy, but at others verse is printed as prose without there seeming to be a need to compress copy, so it looks like ". . . there was indeed some confusion in the layout of the copy itself" (p. 15). The text being so short, it is possible that the makers of Q1 wanted to bulk it out, hence sometimes setting prose as verse. They certainly used an unusually large type face.

    Lindley considers particular moments in the action to see if Q1/F variation can be explained as Q1 representing a theatrical adaptation of the play represented by F. A key moment is when in Q1 Doctor Caius finds Simple in his closet. Simple came on an errand from Hugh Evans to get Mistress Quickly (Caius's servant) to help Slender marry Anne Page. But only in F does Simple mention Evans's role in this and only in F does Caius acknowledge that Evans sent Simple. Yet in both versions Caius immediately pens a challenge to Evans. The problem is that in Q1's version he has no reason to challenge Evans, since he has not been told by Simple that Evans sent him on this errand. (This problem is also discussed in an essay by Laurie Maguire and Emma Smith in Shakespeare Quarterly for 2021, to be reviewed in next year's NYWES.) This certainly looks like an error and is not easily explained either by Q1 representing an early version of the play nor Q1 representing a theatrical adaptation; it seems like sheer garbling.

    The Latin Lesson scene is not in F and around it F orders the scenes differently from Q1 and in a way that appears intended to maintain temporal logic. These two related differences suggest to Lindley that either the Latin Lesson was cut from the play (and the scenes around it reordered to make Q1's order) or the Latin Lesson was added to the play (and the scenes around it reordered to make F's order). Either way, this means that the Latin Lesson was not accidentally omitted to make Q1: this was an intentional theatrical alteration. Also, there is the F-only line "Well, I must of another errand to Sir Iohn Falstaff" that seems to Lindley a consequence of the reordering of scenes in F, which "strengthens the possibility that Q's arrangement is the 'original', F's the consequence of the substantial addition of the Latin lesson" (p. 18). Lindley also thinks that F's impossible time scheme is a consequence of the adding of the Latin Lesson.

    Q1 and F are both hopelessly muddled about the horse-stealing plot and about the colours worn by Anne Page for her elopement. It takes some considerable work using non-scripted clues to make Q1 intelligible in performance. Lindley finds that the possibility of Q1 being a cut-down version of the script behind F is rendered unlikely by the fact that the actors' cues are largely different in the two editions: in cutting speeches it makes sense to preserve the cues. (I would say that this is true only when the actors performing the cut-down version have already learnt the long version; but if a company were cutting for a revival, everyone would be learning his part afresh so this consideration would not apply.)

    Lindley surveys the various objections to Q1 The Merry Wives of Windsor being a memorial reconstruction of the play better represented by F, and then the recently revived theories about notetaking and shorthand transcription from live performance. Lindley reports that notetaking (not necessarily using shorthand) was commonly used to record important events such as the Frances Howard trial, and could easily have been applied to plays. Variation in the quality of transcription might reflect the varying audibility of the performers who were speaking, which could account for the variation in Q1's likeness to F.

    Lindley admits that while some scenes' variations (between Q1 and F) can be explained by notetaking, this does not prove that notetaking was used. And there are other scenes were the substantial verbal variation between Q1 and F cannot plausibly be explained by notetaking. Lindley concludes that ". . . we will never know with certainty exactly what the Quarto of Merry Wives represents" (p. 28). Most likely there was some revision of the play, but in addition Q1 and F are separated by what Jowett calls "disruptive textual transmission" (p. 28), probably involving notetakers. Lindley describes his edition as a reader's text, not a theatrical script, and reports that he has not fully collated Q1 against F since the whole point of making an edition of Q1 is that it is its own thing distinct from F.

    Only one monograph relevant to our topic was published this year: Alice Leonard's Error in Shakespeare: Shakespeare in Error. It is, in essence, a logically incoherent argument for not emending errors in Shakespeare. In a foreword to the book, Adam Smyth discusses the misprint on the title page of Peter Heylyn's 1639 edition of Microcosmos that gives the date as 1939. Smyth attributes this to "the compositor's fingers on the type, picking up a '6' from the case, placing it in the stick, but placing it in the wrong way around" (p. ix). In fact, this is more than we know, since a '9' might have been accidentally distributed into the '6' sort box in the type case and picked up by the compositor reaching for a '6'. Indeed, the latter is the more easily made error since although a compositor would not look at each piece of type to check it was the right one, he would feel for the nick that told him which way up it should go. Smyth's supposed error requires the compositor to have neglected this deeply ingrained habitual check.

    Smyth makes the case -- and says that Leonard does too -- for not editorially correcting errors in books. Smyth describes how his first published book gave his date of birth as 1946 instead of 1972 and reports his feeling the need to correct this while also seeing it as "a performance of [his] theme" (p. xii) of the instability of early modern texts. Eventually, Smyth reports, he began to feel that "in some way" this error "was right" (p. xii). Such an approach to error is, of course, self-defeating. If Smyth had instead written that he came to feel that the error about his date of birth "was not right", and the word "not" was accidentally omitted in the printing process, he would have no grounds for wanting his readers to know about the mistake. And if Smyth means what he says about not correcting errors, then we have no grounds for assuming that this did not actually happen. If Smyth is presenting his honest view of textual error, then his reader has no reason to take seriously anything he writes, since he has explicitly disavowed any commitment to the words published under his name matching his intentions.

    For Leonard, pointing out the errors in Shakespeare is a way to challenge the centuries-long view -- still "alive and well today" (p. 1) -- that Shakespeare is the greatest writer ever. It is not clear why she thinks pointing out his errors would undermine this view, nor why Leonard wants to. Leonard defends error by its origin in the notion of wandering, which has a positive aspect. Leonard finds Shakespeare showing the creative potential of error most forcefully in his early comedies and histories, not in the later tragedies, where error is fatal rather than fun. She argues that the speech-prefix ambiguities in Folio The Comedy of Errors "replicate the same confusion of twins as the play itself dramatizes" (p. 8). Logically that cannot be correct, since for a reader and playgoer entirely unaware of who is who the play contains no confusions about identity. The play's errors arise in the gap between our knowledge of who someone really is and who someone else on stage thinks they are; if the speech prefixes are entirely confusing, that gap disappears.

    Leonard surveys early modern ideas about metaphor as a dangerous departure from the literal meanings of words, at best to be tolerated only in moderation. She does not mention George Lakoff and Mark Johnson's landmark work Metaphors We Live By (1980), which showed that in fact we think in metaphors: there is no underlying literal sense of things for figurative language to depart from. Leonard offers a long discussion of Ben Jonson's playfulness about literalism and figurative language, but it is hard to see that this has anything to do with Shakespeare or anything to do with error: it is just standard Jonsonian literary criticism. Not until page 39 does Leonard get around to Shakespeare, and here the errors she discusses are largely those of characters mistaking one another's meaning.

    In her chapter "Error and the Mother Tongue" (pp. 67-116) Leonard considers the idea of English as a national language and debates about the assimilation into it of foreign words. There is some interesting literary criticism here on the misunderstandings of Mistress Quickly in The Merry Wives of Windsor and the Nurse in Romeo and Juliet, but nothing of relevance to this review. The same is true of the literary criticism in her chapter "Error and the Nation" (pp. 117-143). But Leonard's chapter "Error and the Text" (pp. 145-185) is directly and extensively of relevance to us.

    Leonard's weak grasp of the complexities of her topic are clear from her generalisation that "In the seventeenth century a new publication of Shakespeare was based on the most recent (rather than the first as is most usually the case today) . . ." (p. 146). This was indeed a common practice but by no means universal, as the simplest cases of Shakespeare quarto reprints should have told her. Leonard suggests that people in Shakespeare's time simply felt differently from us about what we call error, and they reacted differently to it.

    According to Leonard, ". . . the eighteenth century initiated the problem of error in Shakespeare" (p. 146). In fact, John Heminges and Henry Condell's remarks in the preliminaries to the 1623 Folio focus on the problem of error in Shakespeare, and it is remarkable that Leonard claims otherwise. Leonard gives a chronological tour of the eighteenth-century editions, describing how they perceived and sought to remove error from Shakespeare's texts. When she gets to the New Bibliography of the twentieth century, Leonard is evidently out of her depth, particularly when describing W W. Greg's essay "The Rationale of Copy Text".

    Greg's essay was concerned with cases where the authority for substantive readings and the authority for accidental (= incidental) readings was split between two documents. According to Leonard it "established an explicitly intentionalist approach and argued that it is the job of an editor to identify and remove corruptions introduced by compositors or scribes in order to reveal the 'ideal' work intended by the author" (p. 150). Actually, that had long been editors' general aim and Greg was engaging in a specific debate about how best to approach this ideal, which none of the New Bibliographers thought we could ever actually reach.

    Leonard's substantial over-generalisations include: "Like Theobald, Bowers assumes that there was [that is, there once existed] an authoritative, finished manuscript" of each Shakespeare play (p. 151). In fact, Bowers explicitly denied this, writing that the notion of a "final form Shakespeare intended to give his plays" needs elaboration to be intelligible. Bowers asked whether we mean the final form

For publication? That is surely impossible to know . . .  The final form before modification for acting? . . . [we have] something close in plays printed from foul papers. The final form he intended it to take for stage presentation? The plays printed from transcripts of prompt copy must give us something close to that" (Bowers "McKerrow's Editorial Principals for Shakespeare Reconsidered", pp. 316-7).

In this phrasing of "impossible to know . . . something close . . . something close" Bowers denies the assumption about definiteness and finality that Leonard attributes to him.

    Leonard attempts to give a brief history of the arguments against the New Bibliography that emerged from poststructuralism in the late-twentieth century, but in such a short account she necessarily offers only a flavour of the debates and cites only a fraction of the scholarly contributions. This drives Leonard to a teleological distortion, as if we only now really understand what is going on: "An editor now makes room for censorship, scribal confusion, authorial revision, theatrical adaptation or compositorial space-saving as possible alternatives to error" (p. 152). In truth, the high New Bibliographers acknowledged and took into account every one of those phenomena. But Leonard's claim does not make sense even on its own terms, since "scribal confusion" is not one of the "possible alternatives to error" but rather one of the causes of error.

    Leonard starts referring to "the author function in editorial theory" (p. 152) without indicating what this means nor mentioning Foucault as the inventor of the notion of an author function. This will surely leave some readers behind. Leonard then starts writing incoherently about "the demand for authorial certainty" (p. 153) and what she sees as the baleful influence of the editorial impulse to correct error, using The Comedy of Errors as her illustrative text. In essence, Leonard argues that we should not tidy up the Folio's erroneous speech prefixes for the two Antipholuses and Dromios because these productively mirror in the form of the script the errors of misrecognition made in the story itself.

    The obvious problem with Leonard's argument is that in order to enjoy the play we as readers need first to understand the action -- what is really happening -- in order to see and appreciate the comedy of the errors of misrecognition in the story. That is, we need the "of Ephesus" and "of Syracus" qualifiers after characters' names to stand in for what the theatre is able to do by using actors who are like each other in appearance but are not identical. We need to see the differences that the other characters cannot alongside the similarities they can. (Where one actor doubles both Antipholuses or both Dromios, the difference of appearance is manifested as difference in demeanour.)

    Quoting R. A. Foakes's remark in his Arden2 edition of The Comedy of Errors that the Folio's inconsistent speech prefixes produce "a nice confusion" (Foakes's p. xii), Leonard remarks that Foakes is using the word nice "perhaps in the sense of coincidence" (p. 154). There really is no sense of nice that means coincidence, and Foakes actually meant that the confusion was intricate and needs tactful handling. Leonard traces how the play's speech-prefix' confusion was dealt with in some Restoration promptbooks, and the general pattern is that it was not done well. There is clearly a desire to tidy up the problem, but the solutions are imperfect. In these failings Leonard diagnoses "greater toleration towards inconsistency and ambiguity" (p. 163) than we have today.

    Leonard makes the surprising assertion about the plot of The Comedy of Errors that Nell is a "kitchen maid who they [the two Dromios] have both known sexually" (p. 166). I cannot find where the play's script gives her that idea. Leonard's own phrasing is imprecise to the point of inducing confusion, as when she writes that someone owned a particular book "at least from 1698-1709" (p. 168). Does she mean that from some point between 1698 and 1709 onwards (thereafter) this person owned the book, or that they certainly owned the book for these twelve years and perhaps owned it for years either side of this span? This is the sort of thing that publishers should catch and correct in an author's writing, along with such things as the misspelling of Edmond Malone's name.

    Leonard next considers some annotations in reprints of the First Folio, again with no clear picture emerging other than that attempts at clarification exist alongside unresolved ambiguities and even new ones introduced by the annotator. Rowe's edition largely cleared up the confusions about character names in The Comedy of Errors, but even in Pope's edition Leonard finds the editor getting them muddled up at least once.

    Thomas Dekker's Satiromastix (1601) has an errata list of printing mistakes that it calls a "short Comedy of Errors" (Leonard p. 172), which Leonard is sure is an allusion to the play. The EEBO-TCP dataset shows no sign that comedy of errors was a stock phrase before Shakespeare's play, so Leonard seems to be right about this. Because Dekker is playful about his errata list, Leonard finds that he presents "error not as a thing to be discretely excised but as theatrically valuable and comically important" (p. 172). I suspect that here the word discretely (separately, individually) is just Leonard's mistake for the word discreetly (subtly, as to avoid embarrassment), but either way the claim is a substantial overstatement. If Dekker wanted to keep the errors in the book, he would not have inserted the errata list and invited the reader to correct the text with a pen.

    Leonard criticizes Charles Whitworth and John Jowett for advocating editorial intervention for the sake of respecting the author's intention and giving the reader clarity since, she argues, the problem of error that her book has been exploring: "requires a response beyond 'intention' and 'clarity'. These editorial imperatives are luminous vapours, for they mislead the reader into thinking that such things are both possible and preferable" (p. 173). Presumably, then, if Leonard's book makes it to a second edition and the publisher offers her a chance to clarify what she meant by "at least from 1698-1709" (p. 168), to spell Malone's name correctly, or change discretely, she will decline to do so, since her intention and the clarity of her writing are, by her lights, merely luminous vapours. Or, if she does correct her own work but maintains her view that we cannot correct Shakespeare's, then it will be incumbent upon her to explain why Whitworth and Jowett are wrong to want to afford Shakespeare the privilege that she allows herself. That is what I mean by the position taken by Leonard (and Smyth in his foreword) being fundamentally incoherent.

    Two journals published special issues on our topic this year. One was the journal Memoria di Shakespeare, for which Hugh Craig guest-edited an issue (Number 7) on Stylometry. In "Shakespeare on the Tree (2.0)" (pp. 1-41), Giuliano Pascucci explores the ways in which character n-grams -- that is, sequences of individual letters rather than whole words -- can be distinctive of authorship. Pascucci begins with the parallel between genetics and texts: both are strings of coded symbols and amenable to the same tools that find commonalities and differences between strings.

    In genetic trees, inheritance is largely vertical (from ancestor to descendant) but can occur horizontally when, say, a virus passes to a new host. An analogous horizontal inheritance in texts is when two or more authors collaboratively write a work. Pascucci explains how to read phylogenetic trees, which may or may not represent changes over time. (Some trees just represent likenesses between organisms at any one time.) When the tree diagram does not represent change over time, it is "unrooted and called a cladogram" (p. 5) and it is that kind, the synchronic tree, that Pascucci is concerned with. He gives a worked example of how the measurement of difference in genetic strings occurs and how this leads to the drawing of a tree. The remainder of Pascucci's essay is on technical matters about computation that are of interest to those engaged on this topic but yield no results that are relevant to this review, so it is here passed over in silence. 

    In "Computational Philology" (pp. 42-74), Jonathan P. Lamb uses Principal Component Analysis (PCA) to process the results of counting certain words in early modern plays in order to produce meaningful classifications. Lamb shows his plots of the First Principal Component and Second Principal Component (PC1 and PC2 in the jargon) for a PCA of counts of the 200 most common words in Shakespeare plays. Lamb talks the reader through some likenesses between plays that are evident in his PCA plot, and argues that this approach is a kind of philology.

    For one of his PCA visualizations, Lamb counts the occurrences of the most common 200 words in Shakespeare as found in speeches of more than 500 words labelled by character name. Lamb finds that the characters "break down imperfectly into quadrants" (p. 58): medieval and classical in one corner, those "who tell people what to do" in another, women in another, and low characters in the last. To make even this rough-and-ready distinction Lamb has to claim that the ghost of Hamlet's father is like the Fool in King Lear in that both tell people what to do, and that Camillo in The Winter's Tale, Rosencrantz in Hamlet, the Duke in Measure for Measure, and Rosalind in As You Like It, are all servant-like.

    I do not think such categorizations are likely to be received as unimpeachable, and Lamb's interpretations of his results are also unlikely to convince all readers. From one such test, Lamb finds an unexpected "resemblance between Juliet [from Romeo and Juliet] and Caliban [from The Tempest]" (p. 60). He remarks that "Drawing on decades of postcolonial readings of Caliban, we might ask, do the differences within these characters' similarities make it possible to interpret Juliet as a figure marked by whiteness?" (p. 60).

    In "Selfish Bastards? A Corpus-Based Approach to Illegitimacy in Early Modern Drama" (pp. 75-110), Jakob Ladegaard and Ross Deans Kristensen-McLachlan perform keyword analysis of plays containing bastards and uncover a surprising distinct historical and thematic trend across the decade. In essence, bastards get nicer. We know that having illegitimate children was punished in Shakespeare's time and that the children were assumed to inherit moral degeneracy from the degenerate process of their making, and their legal rights of inheritance were curtailed and they could not join trade guilds.

    This study considers 20 bastard characters in comedies, histories, and tragedies from 1588 to 1642 looking for "generic and historical patterns" (p. 77) and performs keyword analysis. The big difference turns out to be between tragedies (where bastards cause the trouble) and comedies (where they help resolve the trouble). Over time, bastards get more likeable in the drama.

    The investigators create a corpus (called the target) of all the words spoken by their 20 bastards and all the words spoken by the other characters in the same plays (called the reference). They look for the positive and negative keywords in target, those appearing at significantly higher or lower frequencies -- using the measures Log Likelihood and Log Ratio (neither of which they explain) -- in target than in reference. The positive keywords turn out to be more useful: the negative were most interesting in the avoidance of first person plural pronouns.

    The investigators explain how they chose their 19 plays with 20 bastards in, which they got in XML form from the Early Print project (for the non-Shakespeare plays) and from Folger Digital Texts (for the Shakespeare plays). They left out bastards who speak fewer than 500 words. This gave just over 47,000 words in target and just over 342,000 words in reference. They used the AntConc software to do the keyword analysis, excluding proper nouns. The investigators find that bastards refer to themselves a lot (using I, me, my, am, self) and avoid words of collective identification (we, our, us). They also refer a lot to the negative connotations of bastardy.

    Having considered all genres together, Ladegaard and Kristensen-McLachan next break the data down by genre. In tragedy, the bastards' avoidance of words of collective identification (we, our, us) is even more pronounced, as is the increased use of pejorative terms for bastardy and for the women whose illicit sex created the bastards, and of terms connected to violent rebellion. This seems to corroborate critical claims that, at least in tragedies, bastards tend to be rebels who want to be judged on their own terms and not on where they came from.

    In comedies, we still see the predominance of bastards' references to themselves, but the pejorative words for bastardy disappear and in their place come words for social relations; these changes are most pronounced in the later plays. (Don John is a typical tragic bastard in the comedy Much Ado About Nothing.) In histories, the patterns seen in tragedies and comedies disappear: the bastards' language is just not very much about being a bastard.

    Richard Brome's comedies loom large in this study's dataset and he seems to be deliberately writing in reaction to, and in contradiction of, the character type of the dangerous usurping bastard found in earlier plays. The investigators give a literary-critical reading of Brome's A Jovial Crew as a reaction to Shakespeare's King Lear. In Brome's play, bastardy is not only the way to individual social mobility but also to a wider social reconciliation across the classes. The trend started by Brome continues in Restoration drama. Historically, bastardy became less common from the 1590s on, but the investigators do not posit a link between this and fashions in play writing.

    The late Edward Pechter asks the important question "Does It Matter that Quantitative Analysis Cannot Deal with Theatrical Performance?" (pp. 111-137), and gives the answer that it does. Hugh Craig and Brett Greatley-Hirsch investigated how well authorship-attribution methods can detect the purportedly distinctive styles of different theatre companies of Shakespeare's time in a book reviewed in NYWES for 2017. Craig and Greatley-Hirsch acknowledged that their not finding theatre-company style by their methods might only be because company style resides in things their methods do not capture, like the manner of acting and unscripted business.

    According to Craig and Greatley-Hirsch there is too little information in the play texts for us to know of these things, but for Pechter there is in a sense too much information: all sorts of acting and business are warranted by the script and we do not know which happened. Pechter finds that Craig and Greatley-Hirsch too quickly dismissed the page-versus-stage problem, as if working on the words avoided it. The line "To be or not to be, that is the question" has as many options for emphasis as there are actors who have played it, and the meaning is different for each. The line does not tell you how it should be said, so the meaning is just not all there in the text as Craig and Greatley-Hirsch's approach seems to assume.

    Moreover, even silent reading puts a voice in the mind's ear and this voice gives differential emphasis to the words, so the problem that is most apparent on the stage, that the emphasis shapes the meaning, is also present on the page. For this reason, the language on the page is no more stable than that on the stage; we can return to the book later and find that it means differently. Pechter thinks that what humans do with language on the page and on the stage is one kind of thing and digital processing of the language is quite another kind of thing. The trouble is that there exist no terms to mediate between these two things: the thing of human understanding and meaning is utterly unlike the thing of machine-processed information.

    Craig and Greatley-Hirsch referred to the computer reading a text, but according to Pechter that is not what a computer does, and treating multiple occurrences of the same word as if they are all the same (as a computer does) is not what a reader does. Most fundamentally, the meaning of the term style in authorship attribution (which Craig and Greatley-Hirsch wish to go beyond) is not the same as the meaning of the term style in studies that go beyond authorship. The former is quantifiable but the latter is not, since it requires prior expectations that are subjective.

    Thus, to discover that the methods used to detect style in authorship attribution does not find playing company style does not tell us that playing company style did not exist, only that the methods used in authorship attribution will not find it. Company style may well have existed in the things not captured by these methods. Pechter lists four kinds of interesting theatre-related investigations that can proceed with quantitative methods: into rare words telling us about plays' chronologies, into hearing and not understanding hard words as a particular kind of theatre experience, into what proportion of plays were collaboratively written, and into data-mining the money records in Philip Henslowe's Diary. Pechter likens Craig and Greatley-Hirsch's approach to investigations into Artificial Intelligence, which Pechter considers doomed to fail.

    Roslyn L. Knutson considers the consequences of the new discoveries about authorship for her field of theatre historical study in "Christopher Marlowe and a Mashup of Stylometry and Theater History" (pp. 138-162). She starts with a summary of what has been pieced together about Marlowe's canon and biography in studies from the late-nineteenth century to the present. Her purpose is to show how knowledge of theatre history is impressed upon by the recent stylometric claims about Marlowe's authorship and co-authorship, particularly regarding Arden of Faversham, 1 Henry 6, and Edward 3.

    Marlowe arrived in London from Cambridge in 1587 and Knutson traces his addresses (inside and outside London) from then to his death in May 1593, and the writing and performing of his plays. Only Dido Queen of Carthage was published as a co-authored play. Scholarship has not been able to discern any standard unit for division of the labour of writing a play, whether by act, scene, or characters, or something else. Knutson argues that where stylometry seems to show us possible but surprising collaborations between dramatists we should not dismiss them as improbable just because they cut across known theatre-company allegiances. Possibly, such allegiances were looser than we used to think. Knutson welcomes the new evidence on Marlowe's collaborations with Shakespeare as they make him, Marlowe, more engaged in the theatre industry.

    Roger Holdsworth's contribution to this special issue is concisely conveyed by his title "The Tempest: Notes on Date and Text" (pp. 189-214). The Tempest was played at court on 1 November 1611 and presumably played publicly before then, but how much before? Holdsworth does not think that the play's dependence on the "True Repertory of the Wreck and Redemption of Sir Thomas Gates" is strong and turns instead to its likeness to Jonson's The Alchemist. But which came first, The Alchemist or The Tempest? Holdsworth cites parallels between the two plays that are also parallels between The Alchemist and earlier Jonson plays, such as the phrase use your authority, which first appeared in Jonson's Poetaster in 1601 and then in The Alchemist and then in The Tempest, and is found in no other Jacobean drama.

    The names Prospero and Stephano in The Tempest come from the quarto version of Jonson's Every Man in His Humour. We find in her [or his] fit first in Jonson's Sejanus's Fall and then in The Alchemist and then it occurs in The Tempest, and nowhere else. Also from The Alchemist and earlier Jonson's works (but nowhere else) is The Tempest's progression strange . . . stranger. Thus The Tempest follows The Alchemist. The Jonson Folio of 1616 gives 1610 as the year of performance of The Alchemist and the London theatres were closed from July to November in that year although we know it played in Oxford in September. Presumably Shakespeare saw The Alchemist in London when the theatres reopened in November 1610 so that makes December 1610 the earliest date for composition of The Tempest. For the recollection of seeing The Alchemist to be fresh in Shakespeare's mind, he would have to have finished The Tempest no later than early 1611.

    Alonso's exhortation "Play the men" was, EEBO-TCP shows, a common expression meaning to be valiant. He probably says it not to the mariners (who do not need to be told it) but to his fellow courtiers who are panicking. The Folio's "worke the peace of the present" makes no sense and the problem is the first the, which needs to be deleted. (The 1986-87 Oxford Complete Works was the first to implement this emendation.) Gonzalo's "I'll warrant him for drowning" means to guarantee he will not drown so it does not need emendation.

    Some of Holdsworth's next textual notes are glosses not discussions of emendation and those are omitted here. Of his sweet thoughts of Miranda, Ferdinand says while shifting logs: "But these sweet thoughts, doe even refresh my labours, | Most busie lest, when I doe it". Most editors emend busie lest to busilest (or similar) to mean most busily but Holdsworth prefers the Second Folio reading of busie least. His objection is that it is meaningless to say that his labour is most refreshed when he does his labour: it could hardly be refreshed when he is not doing it. The Second Folio reading allows Ferdinand to say the rather more suitable paradox that because of the sweet thoughts of Miranda he finds that when he is most busy at his labour he is least busy at it because his mind is on her not on the labour.

    Ferdinand remarks that in every woman he has previously liked there was a fault that fought with her best quality "And put it to the foile", which is usually glossed in relation to fencing foils (so, the fault challenged the quality). Holdsworth argues that the image is from wrestling, in which putting to the foil meant subjecting the opponent to a fall. Holdsworth has a long note on whether Ferdinand says "So rare a wondred Father, and a wise" or ". . . and a wife". He starts with a survey of what the Folio actually reads since this has been disputed: Jeanne Addison Roberts saw wife with a damaged f and Peter W. M. Blayney saw wise with an extra bit of inked matter causing the illusion of a damaged f.

    Holdsworth supports the case for wise on literary-critical grounds, including the rhyme with paradise in the next line, the fact that Miranda is not yet Ferdinand's wife, that a wife is impersonal and clumsy -- is it modified by wondered or not? -- and that Paradise was made without Eve present and Paradise at the end of time has no marriages in it. Moreover the construction 'a modifier X and a modifier' is found elsewhere in Shakespeare and in other and it is common for the two modifiers to alliterate as here. That wise men were wondrous and rare was a commonplace observation. Also, the likening of Prospero to God (in making a paradise) is strengthened if he is wise as God is.

    An entire special issue (33:2-3) of the journal ANQ: A Quarterly Journal of Short Articles, Notes and Reviews (formerly known as American Notes and Queries) was this year devoted to "Authorship Studies in Early Modern Drama and Literature" and guest edited by Darren Freebury-Jones. In his Introduction to the collection (pp. 112-121), Freebury-Jones recounts his experiences in learning to do authorship attribution, and he insists that the quantitative approach is not enough. "The most convincing contributions to attribution scholarship", he writes, "remain those that combine quantitative analysis with sensitive readings of play texts" (p. 112).

    Freebury-Jones gives a whistle-stop tour of some of the attribution approaches, allocating a few sentences each to Philip Timberlake (on feminine endings), Ants Oras (on pauses within a verse line), and Marina Tarlinskaja (on stress patterns) before moving on to a survey of studies in counting word-choices. Freebury-Jones scarcely mentions MacDonald P. Jackson's astonishing contribution to the field which, by my count, totals 76 publications, but he finds space to describe his own at length.

    This being the Introduction to a journal special issue, Freebury-Jones then describes the various essays; all are separately reviewed here. An odd note is sounded when Freebury-Jones reports that in his sole-authored essay in this special issue, "Unsound Deductions in Early Modern Attribution: The Case of Thomas Watson" reviewed below, he will show that the present reviewer "provides a misleading account of the 'mode' in mathematics" (p. 119). In fact, Freebury-Jones's essay "Unsound Deductions" contains no mention of the mathematical notion of 'mode'.

    Also apparently directed at the present reviewer is the baffling remark that ". . . practitioners of the Word Adjacency Networks method have, at the time of writing, yet to disclose their actual results. Readers, of course, should not accept authorship claims without seeing the actual results" (p. 119). The inventors of the Word Adjacency Networks method (including this reviewer) have published all their results -- and had done so when Freebury-Jones was writing in 2020 -- and moreover have published the source code for the software that implements the method together with a detailed guide for those wishing to implement the method. Pervez Rizvi has taken up this opportunity and replicated our results, as will be seen in a future NYWES review.

    Surprisingly, the first article in Freebury-Jones's collection is not in fact on authorship, but it is nonetheless relevant to this review. In "Playbook Titles as Evidence of Copy Text Type" (pp. 122-126), Pervez Rizvi argues that the long and sometimes rambling titles we find on some editions of early modern plays come from privately circulating unauthorized transcripts (not from authors, playing companies, or publishers), which are the source for at least some of the bad quartos. Rizvi does not specify how many words a title must have to be counted as long by his reckoning, claiming that titles fall into two distinct groups: those "merely stating the name of the play" (p. 122) in a few words (short titles) and those that go on to describe the contents and/or plot (long titles)

    Only 17% of first editions of professional plays have long titles, and from this Rizvi concludes that publishers did not lengthen the titles of the plays they received else we would expect there to be more. I think this is faulty logic, since we have no baseline from which to measure how many plays had long titles when the publishers received them. After all, it may be that only 5% of plays had long titles when the publishers received them and the publishers added long titles to a further 12% of them. Rizvi then turns to just the plays from publishers who published at least five plays, and for their plays the proportion with long titles is only 14%, which Rizvi finds "in line with" (p. 123) his previous figure from all the plays. This is false corroboration: just throwing away some of the data and counting the same thing again does not put anything "in line with" anything else.

    Rizvi remarks that "One other way of testing this conclusion [that publishers and printers did not lengthen titles] is to look at the cases where a playbook title was changed in the course of printing. That is because when a change was made during the printing of a sheet, we can be much more confident that it was decided upon by the publisher or printer, rather than an agent such as the author or a scribe" (p. 123) Sure, the decision to make a change during the printing is the publisher's or printer's, but that does not mean that the publisher or printer is the source of the material added to or taken away from the edition.

    Rizvi pursues what follows from his claim: "If the titles on title pages were not written by the publisher (still less the printer) then . . ." (p. 124). He discounts the possibility that authors wrote long titles to help pitch their plays to acting companies (or their agents, such as Henslowe) since such professionals were "hardly going to be swayed by a long title that tried to sell the play to them" (p. 124). Likewise there would be no reason for a scribe to lengthen a play's title when making a copy to show the Master of the Revels. Rizvi leaps to a conclusion: "The purpose of long titles is clearly to market the play to punters . . ." (p. 124). But I can think of at least one other purpose: when written on the wrapper around a bundle of actors' parts in the company archive, a long title would be a useful aide-memoire to recall what a play is about and its key roles when considering whether to revive it.

    Having eliminated professional readers as the targets of long titles, Rizvi concludes that only a manuscript made for a private collector would have one, and since such a  manuscript would be worth making only if the play was unpublished -- else it were cheaper to buy the printed version -- there was nothing for the buyer to compare the manuscript with and hence no check on the manuscript's quality. 

    Rizvi notes that more than half of the bad quartos considered by Laurie Maguire in her book Shakespearean Suspect Texts (1996) have long titles. He speculates that privately circulating transcripts were more likely than theatre-professionals' manuscripts to be inaccurate, because they were not made by authorized means from good copies owned by the theatre companies but were made merely to satisfy public demand, and were more likely to use long titles to entice non-professional readers. Thus many bad quartos may well have been printed from such privately circulating unauthorized transcripts.

    In private communication to this reviewer, MacDonald P. Jackson has pointed out that Rizvi's analysis here may have date-of-composition as a confounding variable. Almost all of the plays considered by Maguire are from the period 1590-1610 and over three-quarters of the plays with long titles found by Rizvi are from this period too. Rather than long titles being associated with appearance in Maguire's list of suspect editions, it may be that both phenomena are related to the chronology rather than to each other.

    The contribution to this special issue by Ian H. De Jong and Eric Rasmussen is called "Non-English Words in The Spanish Tragedy" (pp. 133-142). It shows that the Additions to The Spanish Tragedy, first printed in the 1602 edition, use non-English words at a much lower rate than the rest of the play, and we do not know why. De Jong and Rasmussen counted the English and non-English words in Q1 (1592) and Q4 (1602) editions of Kyd's The Spanish Tragedy (and in particular the latter's Additions), leaving out stage directions, speech prefixes, and spoken names, and also leaving out the 'Englished' words of the play-within-the-play. They mean by non-English words those that had "not yet entered common English parlance, as delimited by the Oxford English Dictionary" (p. 134). For instance, cuisine is an English word now but in the 1590s is was solely a French word.

    The important things are what proportion of the writing is non-English words and whether "linguistic individuality emerges as a tool for characterization" (p. 135). The non-English words in the base text of The Spanish Tragedy (that is, the play without the Additions or the play-within-the-play) total 225 tokens out of a total token count of 21,320. This is about one word in 100. In the Additions the numbers are 4 out of 2635, so about one word in 650. Of the non-English words, Hieronimo speaks 61% and Lorenzo speaks 10%, and this tells De Jong and Rasmussen that ". . . the use of non-English words is a characterizing tool deployed by Kyd to emphasize something about Hieronimo and Lorenzo, perhaps associating their multilinguality with sophistication or intelligence" (p. 137).

    That the Additions do not use many non-English words seems significant to Da Jong and Rasmussen but they are coy about what it means. They remark on their "reluctance to draw conclusions about authorship based on the relative absence of non-English words from the Additions" but also that ". . . this stylistic disparity between base text and Additions, destabilizes critical consensus regarding the identity of the Additions author(s)" (p. 137). This seems like self-contradiction, since either they have discovered something about authorship of the Additions or they have not.

    The candidates for authoring the Additions -- Shakespeare, Thomas Heywood, Jonson -- could all use non-English words, so why did they not in the Additions? De Jong and Rasmussen sketch these three authors' knowledge and use of non-English languages, which shows they could do it. Perhaps, they ponder, the author(s) of the Additions was/were not especially familiar with the play being added to, as Hand D seems not to have been when writing additions to Sir Thomas More. (This is not necessarily a valid analogy, since Sir Thomas More was probably an unperformed play when Hand D added to it, whereas The Spanish Tragedy was, when its Additions were added, already famous on the stage, had been much imitated, and had already been published two or three times.)

    Perhaps, suggest Da Jong and Rasmussen, the writer(s) of the Additions was/were told to not make them seamlessly blend with what went before -- as they would if lots of non-English words were used -- because it was wanted for them to stand out as distinctive. Maybe the point of the Additions was merely their novelty and the fact that they enlarged the play, and it was assumed that they could be perfected later if needs be. An unaddressed matter that bears on how we understand what De Jong and Rasmussen have found is the question of whether Kyd's using about one non-English word for every 100 English ones was an anomalously high or low rate, compared to other dramatists. If checking the same metric in other plays shows that the base text of The Spanish Tragedy was unusually full of non-English words, then maybe the simple explanation for everything discussed here is that the Additions were not written by the anomalous Kyd.

    In 2005 Helmut Ilsemann published his surprising finding that before 1599 the most common length for a speech in a Shakespeare play was nine words and after 1599 this dropped to four words. In "Speech Lengths in Early Modern Plays" (pp. 143-147), Pervez Rizvi shows that the change was not quite as sudden as Ilsemann thought and offers a suggestion for why it happened. Rizvi looks at 500 plays by Shakespeare's contemporaries and finds that they pretty much all make this change in their writing habits around 1596 to 1601. Rizvi points out that the change was not quite so extreme as it might seem, since the percentage of speeches that were 8-9 words, at around 6.5%, was only ever somewhat higher than that of speeches that were 4 words, at around 3.5%. The former declined and the latter rose across the 1590s and around 1600 the latter overtook the former.

    Why did four-word speeches become the most common type, that is, the modal form? Four words is roughly half a verse line and Shakespeare was increasingly, across the 1590s, using shared verse lines. Rizvi shows that at least in Richard 2 the full verse lines are typically 7-9 words long, and mostly 8-9. We know from George T. Wright's and Marina Tarlinskaja's tables that Shakespeare's rate of usage of shared verse lines rose across the 1590s and indeed continued to rise across the 1600s. Rizvi admits that we do not have the corresponding data for other writers of the period, but proposes that probably theirs did too and that is why four-word speeches became modal for everyone.

    In his contribution to this special issue, "Authorship Attribution and the New Oxford Shakespeare: Some Facts and Misconceptions" (pp. 148-155), MacDonald P. Jackson points out that various attacks on the authorship attribution methods in this edition by its opponents Darren Freebury-Jones, Pervez Rizvi, Brian Vickers, and David B. Auerbach are undermined by their failure to acknowledge that these methods produce some attributions that the opponents agree with. Also, they simply misrepresent the methods.

    For instance, Jackson considers the Word Adjacency Network (WAN) method of authorship attribution, on which the present reviewer was a co-investigator, as applied to The Two Noble Kinsmen. The WAN results agree with Eugene Waith's 1989 Oxford Shakespeare edition of the play on the division of scenes between Shakespeare and John Fletcher on 21 out of the 22 scenes for which the WAN method gives a result. Vickers in his book Shakespeare, Co-Author agrees with Waith's division, which was made by wholly different methods.

    Likewise for Henry 8: on the scenes for which it gives clear results the WAN method divides the play the way that non-computational scholars have done. Likewise Timon of Athens: the WAN method largely agrees with an orthodoxy that Vickers agrees with. Likewise Titus Andronicus. Next Jackson turns to the work in Hugh Craig and Arthur Kinney's Shakespeare, Computers, and the Mystery of Authorship (reviewed in YWES for 2009) on Titus Andronicus, Timon of Athens, Henry 8, The Two Noble Kinsmen, Sir Thomas More, and Edward 3 -- variously employing Principal Component Analysis (PCA) and Zeta -- and points out that their results agree with Vickers's views on these plays.

    Opponents of the New Oxford Shakespeare have criticized the microattribution method used by Gary Taylor and others, but as Jackson shows it has produced results on Titus Andronicus, Pericles, 1 Henry 6, and Sir Thomas More that Vickers wholly or very largely agrees with. Of course, none of this proves that the WAN, PCA, Zeta, and microattribution methods work. But the fact that these methods tend to agree with one another and to agree with attributions that Vickers believes in has not been sufficiently acknowledged by opponents of the New Oxford Shakespeare.

    Jackson ends with an objection to the part of Auerbach's "Statistical Infelicities in The New Oxford Shakespeare Authorship Companion" (reviewed elsewhere in this NYWES) that discusses Rory Loughnane's investigation into the possibility of Middleton's hand in All's Well that Ends Well. The core of the problem is that Auerbach inaccurately describes what Loughnane's experiment actually consisted of, "betraying a lack of attentive reading of the NOS Authorship Companion, chapters 17-19" (p. 152). This, Jackson shows, is compounded by Auerbach's faulty understanding of the constraints and significance of the method from probability mathematics called Fisher's Exact Test and by Auerbach's overstating the importance to Loughnane's argument of a "mere coda to the actual argument" (p. 153) regarding a single line that Loughnane explored the possible re-attribution of.

    Freebury-Jones's contribution to his own special issue, called "Unsound Deductions in Early Modern Attribution: The Case of Thomas Watson" (pp. 164-171), is an argument that, contrary to Gary Taylor's claim, Thomas Watson did not write Arden of Faversham. Freebury-Jones begins with a summary of recent work in Shakespeare authorship attribution and his role in it, and makes some corrections of errors in his published claims.

    In particular, Freebury-Jones in his essay "The Limitations of Microattribution" (reviewed in NYWES for 2018) made claims about the rareness of the phrases my loving brother, calls on Christ, a pardon for, I give/grant it thee then, shall all by my will, and shall alter, as I hope they will that he now retracts. In my NYWES review of his essay I listed these phrases, in this order, as ones that I had found that Freebury-Jones had miscounted, so his presenting the corrections here with no mention of my review -- as if these were his own self-corrections -- is plagiarism of my work.

    In revaluating an essay by Gary Taylor and John V. Nance called "Imitation or Collaboration? Marlowe and the Early Shakespeare Canon" (reviewed in NYWES for 2015), Freebury-Jones starts doing some arithmetic with counts of hits that he does not explain the validity of. "We can", he writes, "adjust these results according to the word counts provided for these playwrights' dramatic corpora . . ." (p. 165). The problem that Freebury-Jones is attempting to address is that discovering that certain words or phrases in a work to be attributed are also present in a candidate author's works is not equally significant a finding for every candidate. If one candidate has left us ten times as much writing as anybody else, the chances of finding a certain word or phrase in that candidate's canon is greater simply because the canon is so much larger, and the significance of finding the word or phrase is less.

    It would be valid to simply scale hit-counts by the sizes of the corpora from which they come if we knew that the hit-counts for the words we are tracking rise linearly in proportion to corpus size. And indeed, this linear relationship does hold for the most common words in the language: a 100,000-word sample of any writer's work will contain almost exactly twice as many occurrences of the as a 50,000-word sample. But this linear relationship demonstrably does not hold for the less frequent words that Freebury-Jones discusses. That is, we cannot reliably predict from the number of its occurrences in a 50,000-word sample how often a rare word will appear in a 100,000-word sample by the same author. This difference between highly frequent and rare words would more usually be explained by a linguist as a difference between closed-set words, also known as function words (which are highly frequent in English), and open-set words, also known as lexical words (which are much less frequent.)

    Unaware that counts of rare words do not scale in a linear fashion as canons get larger, Freebury-Jones produces meaningless values: "If we divide Marlowe's total by the figure provided by Taylor (95,064 words) we find that he averages 0.00525961457 hits, while Kyd (54,004 words) averages 0.00210384583 hits" (pp. 165-166). These are not counts of hits but rather the numbers of hits per word in the author's canon. A better way to present them would be via their inverse, which is the number of words per hit, so that 0.00525961457 is one hit for about every 190 words. A good rule of thumb for investigators is that if you find yourself presenting numbers below one that have multiple zeroes after the decimal point and that you need to state to nine significant figures, you are presenting them wrongly. There is in any case no need to consider these numbers further since they arise from Freebury-Jones's mistaken treatment of a non-linear phenomenon as if it were linear, and they have no meaning.

    Freebury-Jones spends a further half page summarizing Rizvi's article "The Problem of Microattribution" (reviewed in NYWES for 2019) before turning to the topic of his title: the claim by Taylor that Thomas Watson wrote Arden of Faversham, made in the article "Finding 'Anonymous' in the Digital Archives: The Problem of Arden of Faversham" in an issue of the journal Digital Scholarship in the Humanities guest-edited by the present reviewer. Freebury-Jones spends another whole page, in an article that has only 6 pages in total, on previous studies of Arden of Faversham's authorship (mainly his own) before turning to his attempted refutation.

    Taylor used a sample extract of 274 words from Arden of Faversham and showed that its words and phrases turn up more often in the canon of Watson than in the canon of anybody else. For his refutation, Freebury-Jones cuts this 274-word sample down to 173 words, because that his how many Taylor used in a different paper, and counts the 173 words moving forwards from the beginning of the 274-word extract and then counts the 173 words backwards from the end of the 274-word extract. Freebury-Jones does no new analysis of his own: he just slices Taylor's data a different way to get different results. Even this does not make his point, since "Watson is the clear winner of the first 173-word sample . . ." (p. 168) and Kyd comes out on top only in the test when the 173 words are counted backwards from the end of Taylor's 274-word extract.

    Just why it is valid to count this way is not explained by Freebury-Jones and he omits to mention the obvious conclusion: by using a window as wide as 274 word Taylor brought in a set of Kyd matches that he could have eliminated if he had just taken the first 173 words instead. Taylor chose the 274-word window not because it suited his foregone conclusion that Watson wrote this part of the play, but because it forms a dramatic unit, from the beginning of Scene 10 of Arden of Faversham with "the joint entrance of Arden, Franklin, Alice, and Michael to the exit of Alice" at 10.34 (Taylor p. 859).

    The foregoing is the entirety of Freebury-Jones's quantitative refutation of Taylor's evidence for Watson's authorship of this part of Arden of Faversham. He then sketches some biographical facts about Watson that he thinks discourage the attribution of Arden of Faversham to him. Watson was apparently friends with Kyd (who quoted him in The Spanish Tragedy) and Freebury-Jones wonders if perhaps what ". . . Taylor has identified is a deeper intertextuality between these fellow writers works than has hitherto been noted . . ." (p. 169). Freebury-Jones ends with reminding his reader why he favours the attribution of Arden of Faversham to Kyd.

    Brian Vickers's contribution to Freebury-Jones's special issue is called "Kyd, Edward III, and 'The Shock of the New'" (pp. 172-188) and it argues that the essay "Shakespeare and Who? Aeschylus, Edward III and Thomas Kyd" by Gary Taylor, John V. Nance, and Keegan Cooper (reviewed in NYWES for 2017) is wrong about Kyd not writing the play Edward 3. Vickers begins by describing in detail how he went about searching for 3-grams and 4-grams in common between the known three Kyd plays The Spanish Tragedy, Solimon and Perseda, and Cornelia, and how he happened upon matches to 1 Henry 6 (which he attributes to Kyd and Thomas Nashe and Shakespeare) and Edward 3 (which he attributes to Kyd and Shakespeare).

    Then Vickers turns to what he claims is corroboration of his findings in the blog posting by Martin Mueller called "Vickers is Right about Kyd". To point the reader to Mueller's posting Vickers gives an online citation of his own website with an URL that at time of writing (June 2025) returns a "404 Error. Page not found". Mueller's posting can be found by web-searching its title because Darren Freebury-Jones has reprinted it on his website, and its content is reviewed below in the context of another of Vickers's publications this year that draws on it extensively. Vickers rather unfairly blames the "old-fashioned" Mueller for the difficulty one has in finding his work because he "Unfortunately . . . never published his endorsement of my findings" (p. 174).

    Next Vickers turns to corroboration of his work by Rizvi, who has four items in Vickers's bibliography. One of these is the article "The Problem of Microattribution" in the journal Digital Scholarship in the Humanities (reviewed in NYWES for 2019). One is an URL pointing to a short online essay called "Arden of Faversham and the Extended Kyd Canon" in which Rizvi finds that his evidence points to Kyd (not Marlowe) writing The Jew of Malta and Edward 2, perhaps because "They were room-mates, and Kyd was tortured because some of Marlowe s words were mistakenly thought to be his". In this essay, Rizvi also finds that Marlowe wrote Scenes 4 to 9 of Arden of Faversham.

    The other two citations of Rizvi's work by Vickers point to nothing at all. One is an URL that resolves to a place on Rizvi's Microsoft OneDrive account that returns a "404 Error. Page not found" and the other is an URL to whole of Rizvi's website called Collocations and N-Grams where, according to Vickers, the reader will find a document called "Shakespeare s Text. A Collection of Resources". This reviewer looked all around that website but could not find a document of that name. A third corroborator of his work claimed by Vickers is Albert C. Yang, who has two items in Vickers's bibliography. One is an article from 2003 that introduces Yang's method, and the other is the article "Validating the Enlarged Kyd Canon: A New Approach" that is reviewed below.

    Vickers characterizes those who have criticized his authorship-attribution scholarship, claiming that some of them have "felt threatened by the notion of an expanded Kyd canon" (p. 175). Vickers does not explain precisely what this threat might consist of. There are, in the United Kingdom at least, serious threats to the job security of those teaching and researching English Literature in universities, because every year fewer young people choose to study the subject, but it is hard to see that reattributing some obscure early modern plays might increase that threat.

    Vickers turns to the history of science to understand resistance to his attributions, including the history of how long it took for mosquito bites to be recognized as the vector for malarial infection. He quotes Max Planck's observation that new ideas spread not by overturning existing orthodoxies but by the holders of the old orthodoxies dying out. Vickers seems to empathize with misunderstood scientific pioneers who were "subjected to mockery and ridicule" (p. 176), some of it systematically orchestrated, as with "denials of ozone depletion and climate change" (p. 176).

    The purpose of Vickers's discursion on the history of science becomes clear when he turns to serious refutations of his scholarship, starting with MacDonald P. Jackson's essay "New Research on the Dramatic Canon of Thomas Kyd", which was reviewed in YWES for 2009. For Vickers, what Jackson really feared was "reputational loss . . . since he had been denying Kyd s authorship [of Arden of Faversham] for over 50 years and increasingly urging Shakespeare's" (p. 176). According to Vickers, Jackson's error was to exclude the possibility that Arden of Faversham shares more rare verbal forms with Shakespeare's The Taming of the Shrew and 2 Henry 6 than with Kyd's works not because Shakespeare wrote it but because Shakespeare was influenced by it.

    Vickers notes that Martin Wiggins dates Arden of Faversham to 1590, so it predates The Taming of the Shrew and 2 Henry 6 and he concludes that for this reason "Shakespeare cannot have been its [Arden of Faversham's] author" (p. 176). I cannot follow Vickers's logic in deciding that the dating settles the matter, since the three plays share the phrasings that Jackson identified regardless of the order in which they were written. Indeed, Vickers seems to explicitly advance an illogical approach to this matter, writing that "If play X was written after play Y it is likely to have been influenced by Y, not the other way around" (p. 176).

    This is obviously false in both directions. First, a play (say Jonson's Volpone standing as Vickers's X) may easily be written after another (say Shakespeare's The Two Gentlemen of Verona standing as Vickers's Y) without being influenced by the earlier one. Secondly, the reverse situation is not a matter of likelihood but certainty: if play X was written before play Y (that is, "the other way around") then play X cannot have been influenced by play Y. Vickers appears genuinely confused on this matter.

    Vickers believes that an accumulation of evidence is sufficient to establish authorship attribution: "I cited a selection of phrasal matches, too many and too detailed to be the result either of imitation or coincidence" (p. 176). In fact, we do not know how many matches between texts should count as so many that only shared authorship can explain them. Vickers's fatal methodological slip is that he first finds phrases shared by the text he wants to attribute and the works of the author he wants to attribute it to, and then eliminates those phrases also found in every other writer's works until he has a residue that are genuinely rare and are demonstrably shared between the text and the candidate.

    Jackson's point was that following Vickers's method we can generate the same abundance of matches between texts that cannot have a shared author, since any two sizeable bodies of writing will share a certain number of phrases that are not found elsewhere. This is a discovery about the nature of language in general, not about authorship. Vickers does not address this aspect of Jackson's critique of his work.

    Vickers turns to the New Oxford Shakespeare and what he considers to be a "personal crusade" (p. 177) by its General Editor Gary Taylor against him, which he characterizes as "misinformation" that he likens to the "oil companies' campaign to discredit global warming" (p. 177). Vickers considers in detail the essay by Taylor, Nance, and Cooper, which showed that counting the 2-grams, 3-grams, and 4-grams and collocations spanning up to 10 words that are shared by the Mariner's speech in Edward 3 and the plays of candidate authors from the same period rules out Kyd as the speech's author. Taylor, Nance, and Cooper also claimed that the Mariner's speech is indebted to a similar speech in Aeschylus's Greek-language play The Persians, which was unavailable in Latin or English at the time, ruling out Kyd as he did not read Greek.

    Vickers pays special attention to refuting this last claim, pointing to substantial differences between the speech in Edward 3 and the speech in The Persians. He also rejects the claim that Kyd's inability to translate Greek was unusual among dramatists: "The only author on Cooper, Taylor, and Nance's list [of candidate authors] who could translate from Greek was Thomas Watson, although he preferred to render it into Latin, the international language, rather than the vernacular" (p. 180). This claim Vickers supports with accounts of the minimal role of Greek language learning in Elizabethan schools and evidence of each candidate author's knowledge of Greek. By his account, almost no one associated with early modern theatre -- and particularly none of Taylor, Nance, and Cooper's candidate authors -- could read Greek.

    Vickers attempts to show that Lucan's Pharsalia (rather than Aeschylus's The Persians) is the true source for the Mariner's speech in Edward 3. He quotes the likenesses but does not detail any particular evidence that points incontrovertibly to this source, the closest being a link between Edward 3 ("Then might ye see") and Kyd's translation of Cornelia ("Some should you see"). Lastly Vickers turns to Taylor, Nance, and Cooper's use of the so-called microattribution technique in their essay. Vickers quotes matches between Edward 3 and The Spanish Tragedy and Cornelia that he finds to be proof of Kyd's authorship of Edward 3, but he does not consider whether the matches are especially rare nor how many such matches we should expect to find in two plays not by the same author. Taylor, Nance, and Cooper's approach, it should be remembered, was to count the matches and decide the matter quantitatively.

    The first of Vickers's matches is hope and fear followed by no more than three words and then an occurrence of each in Edward 3 and fears and hopes followed by no more than three words and then each in The Spanish Tragedy. This is indeed a rare collocation: the Early Print project's interface to the EEBO-TCP dataset indicates no other occurrences. But how often should we expect to find such rarities? Without knowing this, the value of such evidence is unknown. Vickers's next match is further followed by wrack with up to eight words allowed in between. This is not particularly rare so finding it in Edward 3 and a Kyd play should not surprise us.

    Likewise, Vickers's next example of day followed by one word followed by turn to followed by one word followed by night is not rare, and indeed appears in Shakespeare's Love's Labour's Lost as "day would turn to night". Likewise, purple followed by gore with up to eight words allowed in between is found in hundreds of texts. And so on with a further four cases. The flaw in Vickers's approach is that without quantification of how unlikely it is that we would find such matches in texts written by different authors -- or even just evidence of how rare they are in absolute terms -- we can conclude nothing from them.

    Vickers believes that we do not need the actual numbers since our common sense will do just as well. He insists that ". . . the co-occurrence of 'shiuer' and 'Launce' in three plays written within six years of each other, and in no other dramatic text dating before 1596, far exceeds the bounds of coincidence". (The three plays are The Spanish Tragedy, Cornelia and Edward 3.) But unless we quantify the bounds of coincidence we cannot say this. Anthony Munday used shivering lance in The Mirror of Mutability (published 1579) and Zelauto (published in 1580) and George Peele used it in The Honour of the Garter (published in 1593), and Richard Johnson used shivered lances in The Nine Worthies of London (published 1592) and The Famous History of the Seven Champions of Christendom (published 1596). Putting this adjective before this noun was not self-evidently a marker of Kyd's authorship.

    Vickers ends with a promise of further clinching evidence that he has collated elsewhere: "Searching the period 1580-94 I have expanded my earlier findings and have placed on my website a list of 205 unique matches with Kyd, or one every 6.5 lines. This result leaves no doubt that Kyd coauthored Edward III" (p. 185). Vickers tells his reader that his extra evidence is at <http://www.brianvickers.uk/?page_id=1752>. At the time of writing this review (June 2025) following this URL leads to a "404 Error. Page not found".

    Albert C. Yang thinks that Vickers is right to expand the Kyd canon so that it contains not only The Spanish Tragedy, Solimon and Perseda, and Cornelia (as many scholars believe) but also King Leir, Fair Em, and Arden of Faversham. Yang's essay "Validating the Enlarged Kyd canon: A New Approach" (pp. 189-197) uses an "information categorization  method" that considers "rank-order frequency of all the words" (p. 189) in Kyd's supposed canon as expanded by Vickers and in other canons. For an explanation of the method the reader is referred to "Yang, Peng, and Goldberger" but there is no item in the Works Cited list with those three authors so it is presumably one of the "Yang et al." pieces in that list that is meant.

    Yang's method attends to the "unique frequency for the preference" (p. 190) for every word in a text, from the most frequent to the least. Again in pointing the reader to where to learn more about the method, Yang writes references that do not have corresponding items in the Works Cited list: "Yang, Goldberger and Peng" and "Yang, Hseu, et al.". Presumably these too are items that the Works Cited list calls "Yang et al." and one would have to get hold of all four of them to figure out which two are actually being referenced at this point.

    To illustrate his method, Yang uses the ten most frequent words in Shakespeare's Cymbeline and The Winter's Tale and Fletcher's Bonduca, sorted into rank order from most frequently used to least. For the comparison of two plays, a scatterplot is drawn so that each point represents a word and its x coordinate is that word's rank order position in one play and its y coordinate is that word's rank order position in the other play. If the rank orders were identical for the two plays, the scatterplot points would form a perfect diagonal heading north-east, since the dots would be at position (1,1) for the word that is first in both plays' rank orders, and at (2,2) for the word that both plays have second in the rank order, and at (3,3) and (4,4) and so on.

    The more dispersed the points are, the more they deviate from the perfect diagonal, the more unalike the rank orders are. Yang does not at this point mention where he gets his texts from, but later we learn from a reference to a website ("Hylton 2019") that it is the Moby Shakespeare based on the 1864 Globe edition. This probably explains why Yang's counts for many words are higher than mine from the 1623 Folio and higher than mine from a modernized edition. Neither the Folio nor any modern edition has the many stage directions that the Globe editors invented to help explain the action to the readers. I suspect that his counts of I, which are fully 10% higher than mine, also include those found within roman numerals used in act and scene markers.

    At this point, Yang report that his method measures how far the actual scatterplot differs from a perfect diagonal, but he does not indicate how he measures those distances. There are various ways to measure the distances between a set of points and a line. For instance there is Euclidean distance, which is how the crow flies from point A to point B, and there is Manhattan distance which is how many blocks east and north a taxicab would have to drive to get from point A to point B; although the former might seem the obviously correct measure the latter is commonly used in the kind of mathematics Yang is employing.

    Yang explains that having measured the various distances between plays using his method he visualizes them as a "phylogenetic tree" (p. 193). Because his trees do not represent change over time, Yang's visualizations are in fact regular binary dendrograms in which time plays no part and the branches merely represent differences in the numbers he is producing. Yang's trees tend to put plays by the same author close together in the sense that getting from play A to play B by the same author takes fewer hops -- fewer moves up the branch containing A to a common ancestor node and then down to the branch containing B -- than are needed if A and B are by different authors.

    Yang compares 37 Shakespeare plays to The Spanish Tragedy, Soliman and Perseda, King Leir, Fair Em, Arden of Faversham, and Cornelia (the supposed Kyd canon). Testing these 43 plays, Yang found that "all of the Kyd texts are arranged on a very different branch from Shakespeare s plays, suggesting two different authors for Shakespeare and Kyd's works" (p. 193). Since Yang only tested these 43 plays and no one else's there are in fact only two authorial branches in his phylogenetic tree. Kyd's plays are not on  "a very different branch from Shakespeare's" as he puts it, but rather are simply on the other branch. This tells us that the six non-Shakespearian plays appear different, on this test, from Shakespeare's plays and the only reason that they appear to form a group is that they all have this non-Shakespearian quality. This test does not show them grouping by their own likeness, one to another; it merely shows them being distant from Shakespeare.

    To establish that these six plays are alike in some feature other than being not like Shakespeare's plays -- a feature such as common authorship -- Yang would have to put other non-Shakespearian plays into the test too and show that the six supposedly Kydian plays do not test like those other non-Shakespearian plays either. Moreover, since we know that Yang got his Shakespeare plays from the 1864 Globe edition and got his other six plays from some other (undisclosed) source, which may or may not be modernized as the Globe edition is, his test may simply be showing this difference in provenance.

    Next Yang turns to Edward 3 and 1 Henry 6. He made up four texts comprising "Kyd's proposed share" of 1 Henry 6 (presumably he means the share as proposed by Vickers), "Kyd's share of Edward III" (as proposed by Vickers?), "Shakespeare's share of Edward III" (according to whom?), and "three scenes from Shakespeare's Henry VI Part One and Three for the purpose of control comparisons" (p. 193). Aside from the problem that Yang does not disclose which parts of these plays he used nor whose proposed authorial divisions he is relying on, there is the problem that 1 Henry 6 and 3 Henry 6 cannot serve as "control comparisons" since both are also disputed co-authored plays.

    As before, the visualization of this second test is meaningless because there are just two authors present and one of them is Shakespeare. This means that the two groups the data divide into might simply be Shakespeare and non-Shakespeare. This does not tell us that the plays supposedly by Kyd are like one another. To explore that, Yang would have to introduce multiple canons into his test and show that the supposedly Kydian plays form their own set in distinction from multiple other writers and not just (as his present work shows) in distinction from the style of Shakespeare.

    In the Authorship Companion to the New Oxford Shakespeare (co-edited by the present reviewer), Giuliano Pascucci presented a novel way to measure entropy in a digital text using what is known as the LZ77 algorithm (named for its inventors Abraham Lempel and Jacob Ziv) which compresses a digital text in what is called a lossless fashion. That is, it is always possible to reverse the process to recover the original text in perfect form. The measurement of entropy -- which is essentially the same as measuring compressibility -- has for some time been a standard tool in the field of authorship attribution.

    David B. Auerbach presents what he calls a critique of Pascucci's approach (pp. 192-203). Auerbach rightly points out that the LZ77 algorithm will find and act upon any kind of repetition in a text it is asked to compress, and not just repetitions that are words and phrases. That is, if LZ77 finds that a text repeatedly favours ending a word with a particular letter, say d, and starting the next word with another particular letter, say s, then it will reduce the string of three characters (d, space, s) to a single token in order to compress it.

    Auerbach presents this feature of LZ77 as if it were something Pascucci does not know, but in fact Pacscucci makes this very point in his contribution to the New Oxford Shakespeare Authorship Companion and he elaborates upon it in a further essay discussed above. Auerbach writes that "There is, at the least, a serious question as to whether such sub-lexical sequences should serve as stylistic markers of an author" (p. 201). Clearly Auerbach is ignorant of the fact that such sub-lexical sequences have indeed been successfully used by other studies in authorship attribution. The remainder of Auerbach's critique is a series of explanations of entropy and compression that are rather less clear and accurate than those given by Pascucci.

    In her contribution to this special issue, "Function Word Adjacency Networks and Early Modern Plays" (pp. 204-213), Ros Barber critiques the WAN method of authorship attribution, and in process she reveals that she comprehensively misunderstands what it is. Rizvi has objected to the use of function words in authorship attribution on the grounds that these words are not independent of the lexical words around them. Rizvi gave no reason why such a dependence might undermine an attempt to use function words for authorship attribution and implied that the inventors of the WANs method simply had not thought that such a dependence might exist. (They had thought of it and do not think it relevant.)

    Barber rejects the WANs inventors' response to Rizvi on this point, arguing that such a dependence would undermine the use of a Markov chain to model function word proximities, since the WANs inventors' response "does not resolve the issue of the necessary independence of function words from what has gone before, which is part of what defines what is known as a Markov chain, and is therefore essential to applying Claude Shannon's theory successfully" (p. 204). This, of course, is not what defines a Markov chain.

    Barber is apparently misunderstanding the requirement that in a Markov chain the probability of a transition from one state to another depends only on the current state not any prior ones. That these states might in reality depend on prior conditions outside of the chain has no bearing on the operation of the chain. A common teaching example is a Markov chain encoding the transitions of a set of traffic lights through the sequence of red to red-plus-amber to green to amber and back to red. Although these states are dependent on outside factors -- including a constant supply of electricity and lamp filaments that have not 'blown' -- the existence of those outside factors does not, pace Barber, mean that the sequence cannot be represented by a Markov chain.

    Barber rejects some of the 100 words used in the WANS method as not function words, in particular "words like bar, dare, given, enough or might" (p. 204). Linguists do not agree on just which are the function words and the WAN investigators have experimented with alternative lists and found that these make no significant difference to their results. Barber gives no reason for suspecting that the words she objects to somehow bias the test in favour of any particular candidate for the authorship of disputed works.

    Barber believes she has also found an arithmetical mistake in the WAN work. The problem is that the investigators at one point claim to be looking for the author for whom the relative entropy between their canon and the play in question is negative but "In practice", writes Barber (p. 204), they attribute the play to the author whose canon has the lowest relative entropy to that play. There is no contradiction here, since negative numbers are indeed lower than zero and lower than positive numbers. Barber's objection only makes sense when the reader realizes that she mistakenly believes that in mathematics the word lower means absolutely smaller without regard to sign, but it does not: -3 really is lower than +2.

    Thus Barber asserts that the inventors of the WAN method have "chosen to interpret its results via 'lowest wins' rather than 'negative scores' attribution" (p. 204), but in fact what she presents as alternative interpretations are the same thing. This fundamental misunderstanding of numbers takes Barber on a long and pointless series of objections to problems that are entirely in her imagination and not in the research she is discussing. She also fails to notice that procedures described in the validation stage of the method are not the same as -- and need not be the same as -- the procedures used in the application, so that one part of the process may produce differences that span the number zero (some negative, some positive) and others may produce differences that are all positive; in both cases the WAN inventors looked for the smallest (lowest) differences.

    Barber repeats Rizvi's mistaken objection (in an essay reviewed in NYWES for 2018) to the practice of moving all the data points of a result up or down by the same amount to facilitate easier comprehension of the differences between plays. Such a translation up or down the scale does nothing to affect the differences between the scores, which is what is at stake here. Barber presents a table comparing the results of the WAN method regarding the play 1 Henry 6 to Hugh Craig's independently derived results for that play. Both investigations concluded that Marlowe's writing is present in the play, but Barber misrepresents this by tabulating only the scene-based conclusions from Craig and comparing them with the WAN results. The WAN inventors gave a detailed account of where their method agrees and disagrees with Craig's results, and by looking only at "selected scenes" (as her table's caption puts it) Barber gives the false impression that the two studies fundamentally disagree.

    Barber explicitly repeats Rizvi's already refuted claim that subtracting a constant from each result of the experiments exaggerates the outcome. Barber's objection makes no more sense than Rizvi's because she clings to the idea that all the results are relative to zero in the sense that positive numbers show no clear attribution and only negative ones tell us anything. This again is simply Barber's misunderstanding of the notion of what counts as the lowest of a set of numbers.

    Barber also repeats Rizvi's already refuted claim that the validation method used by the WAN team was faulty, asserting that the 94 plays used to find the best words for discriminating between authors should not have also been the 94 plays used to validate the method. According to Barber, the WAN investigators should have set aside fully half the set of 94 plays, using the first 47 plays for finding which function words are most discriminating and the second 47 plays for validating the method.

    This is a false objection because as the WAN inventors already pointed out it makes no significant difference which function words are used: we could just pick the top 100 most-used words in English and the results would be about the same. The leave-one-out cross validation method used by the WAN inventors is perfectly acceptable in this context for assessing the accuracy of the method, since the constitution of the function word list does not substantively affect the results.

    Barber again returns to her mistaken distinction between negative and lowest numbers to present an entirely false table showing that the WAN method's accuracy rate is worse than the investigators claimed. Barber repeats Rizvi's objection that the WAN method does not count those transitions (from one function word to another within five words of it) that are entirely absent in the plays under consideration. This objection has already been answered by the WAN inventors pointing out that i) there are lots of things the WAN method does not count and that itself is not a valid critique since every method counts only some of the many things we might count), and ii) counting absent transitions merely privileges the threshold of zero occurrences, making a purely binary determination (zero or non-zero?) whereas the WAN methods instead uses the richer data arising when we count whether things happen once, twice, thrice, and so on.

    Of course, someone could build a classifier based on the absence of function word proximities, as Rizvi and Barber prefer, and we could then compare this classifier's power with that of the WAN method to see which is best. But simply pointing out that such another classifier might be built, as Barber does at length, is not of itself a critique of the WAN classifier. Sketching what such a classifier might attend to, Barber lists the function-word proximities for though to nothing that occur 13 times in Shakespeare's plays and not at all in Marlowe's, and it is clear that her criteria are different from those of the WAN method: she admits proximities much greater than the five-word window of the WAN method, and she admits proximities that span a change of speaker, as the WAN method does not.

    Using the function-word transitions that occur in Shakespeare but do not occur in Marlowe, Barber finds that canon size matters in the sense that the smaller Marlowe canon has less opportunity, as it were, to use many of the transitions found in the much larger Shakespeare canon. Indeed, this dependence on canon size is a good reason not to set the threshold for counting at zero but instead to look at those transitions that Shakespeare and Marlowe both use but to differing degrees, as the WAN method does.

    Remarkably, Barber then attributes to the WAN method the fault that applies only to her own method, insisting that the problem of canon size "illustrates starkly why disparities in dataset size and period need to be taken into consideration in any stylometric test" (p. 209). Barber objects to the WAN inventors' claim that transitions that are so rare that they occur not at all in some writers' canons are in fact very rare in all writing, offering as counter-evidence the many transitions she found in Shakespeare's canon but not Marlowe's. But Barber's counter-evidence transitions are numerous only because of her broader filter that admits words further apart (even spanning two speeches) than the filter of the WAN method.

    Barber thinks that canon size affects the WAN method even though this method counts only transitions that occur in the play being tested and the canon in question, and this is true but only up to a point. As the WAN inventors explained, and indeed repeatedly quantified, the accuracy of the method falls off as the texts being examined become small, hence the reliability of the method is poorer for individual scenes from a play than for whole acts or whole plays. But once a certain minimal canon size is met it is not true that, as Barber claims it is, that authors with large canons are treated differently from authors with small canons. Above a certain threshold the size of the texts makes no difference because the method measures not the frequencies of occurrences of the function words but their relative proximities one to another and these stop changing significantly once the texts are above a certain size.

    Barber reports that a table published by the WAN inventors -- Table 3 in their essay "Stylometric analysis of Early Modern English Plays" -- actually shows the effect she claims, but she is misreading the table. It does not show "Marlowe being furthest from Shakespeare stylistically, compared with all the other authors" (p. 210) since their relative entropies are 8.9 and 10.1 -- there are two numbers since it matters which author we put first and which second in the WAN method's non-commutative calculation -- and other distances in the table are greater. Looking along the table's Shakespeare row (that is, showing distance from Shakespeare's style) the Fletcher distance is 8.9, same as Marlowe's, and looking along the Marlowe row the Fletcher distance is 17.4, which is greater than the Marlowe-Shakespeare distance of 10.1. Whichever way we parse Barber's claim, the table simply does not support it. 

    According to Barber, "what is really being measured here is the greatest disparity in canon size" (p. 210) but the table does not bear this out either, since the George Chapman and Fletcher canons used are about the same size (13 and 15 plays respectively) and yield distances of 9.6 and 8.4, while the roughly equally sized canons of Chapman and Jonson (13 and 16 plays) yield distances of 5.8 and 5.4. Put another way, Chapman's canon of 13 plays being less than half the size of Shakespeare's at 28 plays (referring as ever to the plays tested) does not have the effect that Barber claims follows from a "disparity in canon size", since their distances are 4.7 and 4.8, the lowest numbers in the table.

    The WAN method is measuring real stylistic differences, not canon sizes, as indeed is already clear from the multiple validation runs that show the method having better success in blind attributions -- that is, when making attributions for plays for which we already know the answer -- than other methods in use. Barber quotes from the WAN inventors an explanation of one of their equations: they "assume that the combined length of the texts written by author [a] is long enough to guarantee a non-zero denominator for a given number of function words" (p. 120), but this quotation is inaccurate and thereby misleading.

    The inventors actually wrote "In Equation (3) we assume that the combined length of the texts written by author a_c is long enough to guarantee a non-zero denominator for a given number of function words |F|". Barber does not appear to understand the process of normalization that this equation performs and she quotes Rizvi's mistaken belief that the WAN inventors fudge their data by assuming that a function word is "followed by every other function word in equal proportion (p. 210). Rather, because the WAN method is concerned with the differences between the frequencies of transitions, the normalization step perfectly reasonably records that there is no measurable difference when there is no transition to record.

    Rather than accepting that Markov chains are a way of looking at certain phenomena, Barber believes that some phenomena actually are Markov chains and others are not: ". . .  just because the data 'can be interpreted' as a Markov chain, it does not mean it is a Markov chain" (p. 211). This is akin to claiming that just because the sum of the squared differences from their mean that is shown by a series of numbers can be understood as their variance this does not mean that this sum is their variance. That is, Barber is taking a mathematical method for making sense of the world and mistreating it as an assertion about the nature of reality.

    Instead of using a reliable introduction to the topic of Markov chains, such as the one on Wikipedia, Barber turns to the definition in the Oxford English Dictionary and finds that it relies on the notion of a stochastic process, which word she in turn looks up and finds that it concerns random probability distributions. For Barber, this reveals the weakness of the whole approach since "It s a considerable stretch to see the language of a play, even its function words, as 'randomly determined' . . ." (p. 211).

    In Information Theory it is widely accepted that language generation can be studied as a stochastic process that may be modelled by probability distributions, and indeed the practical successes of such services as Google's Translate tool and the impressive language-generation systems such as the Generative Pre-trained Transformer (GPT) prove that this modelling works. But it takes more than the Oxford English Dictionary entries for 'Markov process' and 'stochastic' -- Barber's only sources -- to make sense of work in this field.

    In the summary of her essay, Barber introduces a new objection she has not hitherto mentioned: that the WAN method overfits to its training data. She offers no evidence for this new claim, and if it were true the method would not achieve the high success rate it shows in the validation runs. Barber, of course, rejects these validation runs too and insists that all the method can do is attribute plays by canon size and hence Marlowe stands out because his canon is so small. Even without the validation runs, however, Barber's conclusion is clearly contradicted by the non-Marlovian results. The WAN inventors' experiments confirm Shakespeare's co-authors Peele in Titus Andronicus, Middleton in Timon of Athens, and Fletcher in Henry 8 and The Two Noble Kinsmen. Barber is silent on these results, which cannot be explained by relative canon sizes.

    The final contribution to this special issue of the journal ANQ is a short proposal by Martin Mueller for a series of practical experiments of natural language processing (and other techniques) being applied to randomized chunks of early modern plays in order to generate new knowledge in authorship attributions ("Who Wrote What: Is it Time for a Blind Experiment in Early Modern Drama?", pp. 214-216). Mueller's suggestions are reasonable.

    Outside of Freebury-Jones's special issue on authorship, the journal ANQ published three further essays that are relevant to this review. In the first, David B. Auerbach presents what he considers mathematical mistakes by the present reviewer and others ("Statistical Infelicities in The New Oxford Shakespeare Authorship Companion", ANQ: A Quarterly Journal of Short Articles, Notes and Reviews 33[2020] 28-31). He starts by quoting the present reviewer remarking that the ranges of various textual phenomena that we measure become smaller as our textual samples become smaller. Auerbach thinks this claim is mistaken and that in fact ". . . the law of large numbers dictates that a smaller sample should show less regularity" (p. 28).

    Auerbach is evidently thinking here of averages, which certainly do become more irregular as our samples get smaller. For instance, if I flip a coin 1000 times and get 700 heads and 300 tails -- for a mean average of 0.7 heads -- then I am entitled to conclude that the coin has a bias towards heads, since over as many as 1000 samples the mean average should converge on 0.5 heads. But if I flip it just ten times I should not be surprised if I get 7 heads and 3 tails -- again for a heads mean average of 0.7 heads -- because ten is too small a sample for me to have confidence in its average.

    But in Auerbach's quotations of the present author the topic is ranges not averages, and these do not obey this law. Consider the range of sentence-lengths used in Auerbach's essay. If we sample only the first page, the shortest sentence is "Here Egan seems to have it backward" (7 words) and the longest is "While I plan . . . in the proceedings" (37 words). Based on this first page alone, Auerbach's range on the measure of sentence-length is 7 to 37. But if we consider the entire essay of four pages (a larger sample) the range grows because although "Here Egan . . . it backward" is still the shortest sentence the longest sentence is now "Loughnane's choice of . . . Fisher's exact test" (60 words), making Auerbach's range now 7 to 60.

    Thus the smaller sample shows the smaller range, as I observed is a general principle with ranges. Auerbach spends a paragraph elaborating his point about averages without noticing that my topic was ranges not averages, despite the word range appearing six times in the 127 words of mine that he quotes. Auerbach is inattentive to others' word choices and careless in his own. He remarks that ". . . John Burrows and Hugh Craig contravene Egan's statistical confidence . . ." (p. 29). Rules and constraints can be contravened, but confidence cannot.

    Auerbach quotes the present reviewer remarking that multiple independent tests pointing to the same conclusion can buttress one another to give us increased confidence in a result and he then treats this as if it were a rule that Burrows and Craig break when they decide that some of their tests are not revealing as much as others. My remark that multiple independent tests can buttress one another does not imply that all tests are equally revealing and all must be used in concert. For example, Auerbach's sentence-length range of 7-60 is not a good marker of authorial style and an authorship attribution study of this essay would not be obliged to use this fact merely because we have uncovered it.

    Auerbach next misapplies his principle to studies of Arden of Faversham, finding that because in some tests some small samples of the play seem like either Kyd or Marlowe we cannot conclude that Kyd did not write a substantial proportion of the play. If we had reason to give all tests equal weight and if we trusted all to be perfect, he would have a point. But tests differ in their accuracy and none is perfect, so in the combining of results a preponderance of evidence is what matters. Nothing in Auerbach's essay justifies its title's claim to be discussing statistics until its last 400 words address Rory Loughnane's work on Middleton's hand in All's Well that Ends Well. As Jackson remarks in an essay discussed earlier in this review, Auerbach misrepresents Loughnane's work because he is inattentive to what Loughnane actually wrote.

    Pervez Rizvi offers a critique of MacDonald P. Jackson's authorship attribution method used in the essay "A Supplementary Lexical Test for Arden of Faversham" in the New Oxford Shakespeare Authorship Companion ("Small Samples and the Perils of Authorship Attribution for Acts and Scenes", ANQ: A Quarterly Journal of Short Articles, Notes and Reviews 33.i[2020] 32-3). Jackson relied on counting the occurrences of just five words that Shakespeare favoured -- gentle, answer, beseech, spoke, and tonight -- and four that he disfavoured -- yes, brave, sure, and hopes -- in plays from 1580 to 1600, using the Literature Online (LION) dataset.

    For each of a set of plays from all 137 of the extant plays from 1580 to 1600, Jackson simply counted the number of Shakespeare-favoured words in the play and divided this by the sum of the number of Shakespeare-favoured words and the number of Shakespeare-disfavoured word. Thus because Edmond Ironside has 14 occurrences of members of the set {gentle, answer, beseech, spoke, tonight}, in all possible spellings, and 23 occurrences of members of the set {yes, brave, sure, hopes} it scores 14 ÷ (14+23) which is 37.8%. The higher this metric, the more Shakespearian a play is.

    Putting the plays into rank order of highest to lowest on this metric, the top of the table is dominated by Shakespeare plays and below them all the other dramatists' plays. The lowest ranking for sole-authored Shakespeare plays is The Merry Wives of Windsor at position 35 out of 137. If we use this rank order for attribution to Shakespeare, then according to where we draw the line we can achieve different false-negative and false-positive outcomes. That is, if the bar is far down the list then all the Shakespeare plays appear above the bar (giving no false negatives where a Shakespeare play is classified as not-Shakespeare ) but at the cost of bringing in quite a few false positives (non-Shakespeare plays being classified as Shakespeare's).

    If we move the bar up the list, then the false negative rate rises (we exclude some actual Shakespeare plays from the Shakespeare category) but the false positive rate falls (we exclude some non-Shakespeare plays from the Shakespeare category). Jackson showed that if Arden of Faversham is split into Scenes 4-9 on one hand and the rest of the play on the other, the former is decisively in the Shakespeare region of the rank order and the latter is decisively in the not-Shakespeare region. Jackson then brought in the rare-phrases approach that he used in the book Determining the Shakespeare Canon (reviewed in YWES for 2014) of looking for phrases and collocations found in no more than five plays from 1580-1600. 

    Applying this method to his table derived from the first approach, Jackson showed that in a mixed collection of correctly-classified-as-Shakespeare and incorrectly-classified-as-Shakespeare plays (from the top of his table), the links to Arden of Faversham are exclusively with the correctly-classified-as-Shakespeare plays. In Jackson's table, plays by Kyd come right down the bottom as distinctly unlike Shakespeare and distinctly unlike Arden of Faversham. 

    Responding to all this, Rizvi thinks it embarrassing to Jackson's method that Peele's The Arraignment of Paris came out as the most Shakespearian of all the plays tested, being at the top of his rank-order table for Shakespeare-favoured-words versus Shakespeare-disfavoured-words, and that in this table the Scenes 4-9 of Arden of Faversham appeared more Shakespearian (that is, higher up the table) than the rest of Shakespeare. Also embarrassing, according to Rizvi, is that Robert Greene and Thomas Lodge's A Looking Glass for London made the top 10 in Jackson's list.

    The big mistake, according to Rizvi, was for Jackson to count the values for whole plays and then to assume that the method could be equally applied to acts and scenes. Instead of whole plays, Rizvi repeated Jackson's experiment but divided each play into segments of length 4647 words (the length of the chunk of Arden of Faversham from Scene 4 to Scene 9), and found that done this way instead of by whole plays 36 Shakespeare segments test as not-Shakespeare.

    "These thirty-six segments", Rizvi writes, "contain more than 187,000 words and are taken from the following plays: Hamlet, The Merry Wives of Windsor, As You Like It, The Comedy of Errors, The Merchant of Venice, Julius Caesar, The Taming of the Shrew, The Two Gentlemen of Verona, Henry 4, Part 1, Much Ado About Nothing, Henry IV, Part 2, Love's Labor's Lost, and Henry V" (p. 33). Thus, according to Rizvi, one cannot simply compare segments of a play (as Jackson did with Arden of Faversham 4-9) to whole plays. Jackson responds to this critique by Rizvi in an article noticed later in the present review.

    Lastly from ANQ this year is Thomas Merriam's engagement with the problem of how much significance we should attach to different authors having canons of different sizes when we are looking for their matches to words and phrases found in a text we want to attribute ("Phrasal Criteria for Authorship, with Consideration of Negative Checks", ANQ: A Quarterly Journal of Short Articles, Notes and Reviews 33[2020] 156-63). According to Merriam, if we accept Rizvi's weighting formula for this, used for the results given on his online dataset called Collocations and N-Grams then we get some surprising results for which plays shares most n-grams with which other plays. Indeed, as Merriam points out, it is not easy to find just what weighting formula Rizvi uses.

    Merriam claims that using Rizvi's dataset and formula, the play 1 Troublesome Reign tests more like Shakespeare's King John than it does 2 Troublesome Reign, and 1 Troublesome Reign tests more like King John than King John tests like The Two Gentlemen of Verona. Merriam gives Rizvi's weighting process for matched n-grams as "dividing the number of matched n-grams by the sum of the number of all words in the target and matching play or text" (p. 158). I believe that in fact Rizvi's weighting process is that he finds the number of n-grams common to Play X and the plays by Author Y and divides this number by the total of the number of words in Play X plus the number of words in the matching plays by Author Y. Confusion about just weighting formula Rizvi uses is understandable, since his website is not clear on this matter and he has changed the formula since the site was first published.

    Using his understanding of Rizvi's weighting technique, Merriam looks for plays sharing unique 4-grams with 2 Tamburlaine that are not later than 2 Tamburlaine, and the play with the most weighted matches is 1 Tamburlaine. Using these weightings, the next nearest-matching plays are Greene's Alphonsus, King of Arragon and then some others, and this second-best match and the even less-good matches have much lower-weighted matches, only one-seventh as strong as the weighted match for 2 Tamburlaine with 1 Tamburlaine. In other words, shared-authorship really stands out on this scale.

    Merriam takes the comparisons made in Freebury-Jones and Marcus Dahl's table "Marlowe Results in Martin Mueller's Database" (in their essay "The Limitations of Microattribution", reviewed in NYWES for 2018) and using the same graphic representation as he used for 2 Tamburlaine and 1 Tamburlaine he shows that if we start with 1 Tamburlaine then 2 Tamburlaine is the next closest match (about twice as strongly matched as the next match, which is The Taming of a Shrew). And that if we start with 2 Troublesome Reign the next nearest match is 1 Troublesome Reign with a likeness about three times stronger than that of the next closest match which is The Battle of Alcazar by Peele. Again, in each case Merriam confines himself to plays no later than the one to which he is trying to find matches.

    When The Taming of the Shrew is the text to which we are looking for the most weighted matches, the highest scoring play is The Taming of a Shrew and the second-nearest is Fedele and Fortuni, with one-seventh as high a score. Merriam mentions some long n-grams and collocations (nine words and more) shared between The Taming of the Shrew and The Taming of a Shrew. When King John is the text to which we are looking for the most weighted matches, the highest score play is 1 Troublesome Reign of King John and the score for that match is more than twice the score for the second closest match, which is Shakespeare's The Two Gentlemen of Verona.

    Merriam lists some long n-grams and collocations (eight words and more) shared between King John and 1 Troublesome Reign of King John. Merriam does not tease out the consequences of all these likenesses, but if Rizvi's method is reliable for authorship attribution then it would seem that Shakespeare's The Taming of the Shrew and the anonymous The Taming of a Shrew have a common author, and so do Shakespeare's King John and the anonymous two-parter The Troublesome Reign of King John.

    Of articles in journals other than ANQ this year, the most consequential is M. J. Kidnie's argument that the manuscript behind Q2 Hamlet was not simply authorial: it embodies authorial and non-authorial revisions and amplifications ("Playhouse Markings and the Revision of Hamlet", SQ 71[2020] 69-103). Kidnie recounts the ideas about why Q2 and Folio (F) Hamlet differ that were circulating in the 1970-1980s and started to be challenged in the 1980s-1990s, particularly regarding the diagnosis of the kind of manuscript, foul-papers or promptbook, underlying a printed edition. She discusses the stage direction "A florish of trumpets and 2. peeces goes of" in Q2 Hamlet (D1r) and reports "I have silently emended a straightforward printing error: the letter c was printed in place of the first e in 'peeces' (p. 74n32). This is peculiarly incorrect, since the reading is clearly peeces not pceces in the facsimile by Michael J. B. Allen and Kenneth Muir (which reproduces the Huntington Library exemplar) that she says she is using, and other exemplars agree.

    Kidnie points out that in Q2 a later stage direction of the same kind, "Drum, trumpet and shot. Florish, a peece goes off", duplicates the instruction (since shot and a peece goes off call for the same sound effect). This, she observes, is the kind of duplication that one would expect a book-keeper to produce by adding an annotation to what the author called for, since a book-keeper would want to ensure that the need for a sound effect was highly visible in the book. The Folio does not have this book-keeper's duplication, so here Q2 seems more theatrically minded than F, reversing the usual characterization of Q2 as authorial and F as theatrical. Kidnie acknowledges that what she has been arguing is "not new" (p. 75) and that Werstine in his monograph Early Modern Playhouse Manuscripts and the Editing of Shakespeare (reviewed in YWES for 2013) and John Jowett in the New Oxford Shakespeare anticipated her.

    Q2 appears to lack an antecedent for "Whose" in the lines "Whose whisper ore the worlds dyameter, | As leuell as the Cannon to his blanck, | Transports his poysned shot, may misse our Name, | And hit the woundlesse ayre". John Dover Wilson thought that this omission resulted from a theatrical cut marked in the manuscript underlying Q2 that was understood by Q2's compositor as the deletion of a part line (the part that provided the antecedent for "Whose") but was in fact meant to create a larger cut, of the antecedent and all of "Whose . . . ayre", and did indeed create the omission of these lines from F. Harold Jenkins also found signs that the manuscript underlying Q2 was marked up with theatrical cuts. Kidnie thinks that the New Oxford Shakespeare failed to follow through the implication of there perhaps being theatrical cuts in the manuscript underlying Q2 Hamlet because it preferred to attribute revision to the author.

    Kidnie argues that if non-authorial cuts were marked on the manuscript used to make Q2 Hamlet "then the authority for revision is likely more mixed than scholars have yet entertained" (p. 79). Kidnie's plan is to work through Q2 Hamlet using the premise that its copy was a "playhouse manuscript that received theatrical attention of some sort" (p. 79). Cuts within speeches -- that is where Q2 has the matter and F does not -- can begin mid-line and end mid-line and Kidnie shows a pair of them in Hamlet's berating of his mother where the part-lines left behind join up to make new complete iambic pentameters. She also considers moments where a cut to the version represented in Q2 in order to make the version represented in F requires a small bit of mending (such as the adding of a few new words) to smooth the meter. As Kidnie points out, this last consideration alone precludes editorial conflation of Q2 and F, since the smoothing material found in F would not exist were it not for the cut that motivated its creation.

    Kidnie works through some cuts and joins in manuscript playbooks but I cannot see to what purpose except to show that we cannot easily apportion authorial or bookkeeper/theatrical agency to such things. Kidnie rightly points out that if untidy manuscripts like these reached the printshop the resulting printed editions would contain mysterious duplications and confusions that we would be hard put to resolve. She is again right that "These few examples also indicate how hard it sometimes is to untangle cancellations made by an author from those of the bookkeeper, even when one has access to the manuscript" (p. 87). In manuscripts, we have the aid of distinguishing different inks and hands that of course is lost in print.

    Because the process of making cuts can produce part lines, Kidnie goes looking in Q2 Hamlet for part lines that might be the relics of cuts marked in the underlying manuscript that we cannot otherwise see in Q2. Kidnie finds evidence of this in F's "Assume a Vertue, if you haue it not, refraine to night", where Q2 has four-and-a-half-lines between "not" and "refraine". Because F's line is an iambic heptameter and because "refraine to night" is not needed for sense, she ponders some possible causes by which F's reading arose accidentally in the cutting of Q2's version, and concludes that ". . . F s over-length line presents an unambiguous instance of excision producing what amounts to a part-line within a long speech" (p. 89). I would say that iambic heptameters are within Shakespeare's normal metrical range so there is no problem to be solved here: the cutting of the Q2 version to make the F version may have given us an F reading that Shakespeare fully intended.

    Kidnie has other examples regarding lineation differences between Q2 and F that coincide with material being present in Q2 and absent in F, but I cannot see what point she is trying to make from them. Often she seems to be describing what we find in the editions without offering hypotheses to explain their differences. In Kidnie's accounting for readings in the early editions and the possible evidence of cuts being marked in the manuscript underlying Q2, I count eight occurrences of her using the word might in just pages 92-93 (there are 58 across the essay). Necessarily, when preceded by the word might almost any assertion is true.

    Kidnie asserts that editors "necessarily" speculate about the manuscripts behind early printings "in order to undertake even the most basic tasks, such as prioritizing one printing of Hamlet over another as their choice of copy text" (p. 95). She is not right that this is an absolute necessity, since one could choose the copy text for a modern edition by some other criterion, as the New Oxford Shakespeare did in choosing the longest early edition. But Kidnie is right that "Conceptions of the manuscripts behind Q2 and F Hamlet . . . continue to matter a great deal . . ." (p. 97) because the emendation of errors requires that we have some theory of printer's copy. Kidnie's example of this is the erroneous Exit for Ophelia after her only soliloquy "O what a noble mind is here o'erthrown!" in some exemplars of Q2, which was either added or removed during stop-press correction. 

    Kidnie proposes that where we find in F "Oh most pernicious woman! | Oh Villaine, Villaine, smiling damned Villaine! | My Tables, my Tables; meet it is I set it downe" the middle line "O villaine, villaine, smiling damned villaine" is a bridge added to the manuscript underlying Q2 to replace a longer run of lines that were cut. Q2's reading is substantively the same except that "My tables" occurs once not twice, which becomes important for Kidnie's argument. I suspect that I am not entirely following Kidnie's argument in all this, since it seems to me that the case she makes for how useful that middle line is could just as easily be offered as a case for its being present all along.

    This line "O villaine, villaine, smiling damned villaine" is, Kidnie writes, "extremely useful, since Hamlet has just decried his mother as 'pernicious'" and since 'villain' was an insult only when used against men -- "applied to women, it implied affection" (p. 100n102) -- and "It allows a listener to realize that Hamlet's mind shifts to Claudius" (p. 100). In support of the hypothesis, according to Kidnie, is the fact that immediately after the supposed cut (which got replaced with "Oh Villaine, Villaine, smiling damned Villaine!") comes the Folio's metrically rough "My Tables, my Tables; meet it is I set it downe" where Q2 has the smoother "My tables, meet it is I set it downe".

    Kidnie supposes that "My tables" appeared in the margin of the manuscript underlying Q2 "to indicate (I take it) the insertion s positioning" (p. 101), that is, to show where in the body text "Oh Villaine, Villaine, smiling damned Villaine!" needed to be inserted in order to replace the lines that were cut. Q2's compositor understood this and used the marginal "My tables" to show himself where to insert "Oh Villaine, Villaine, smiling damned Villaine!", but "The scribe at whatever remove behind F builds the whole of the marginalia into his text, doubling up 'My tables'" (p. 101). This sort of hypothesis is impossible to prove or disprove, and judgement rests on whether one thinks the metrical disruption is sufficient to justify such a complex solution. I would have thought that "My Tables, my Tables; meet it is I set it downe" is metrically acceptable as an iambic heptameter with an omitted syllable at the caesura after the first "My tables".

    Kidnie's conclusion is that revision need not have existed only between early manuscript versions. Early editions that we used to take as entirely authorial, such as Q2 Hamlet, may have been printed from manuscripts embodying the complex tinkering by multiple agents that plays were subject to. Indeed, since the marks used to record what should be cut could leave visible the material to be cut, a single manuscript might embody at once multiple versions of a play depending on which excision marks the reader chooses to obey and which to ignore.

    According to Marina Tarlinskaja, metrical tests confirm the recent claims that Shakespeare contributed to the play Arden of Faversham and wrote the Additions to Kyd's The Spanish Tragedy that first appeared in the 1602 fourth edition ("Shakespeare in Arden of Faversham and the Additions to The Spanish Tragedy: Versification Analysis", Early Modern Studies 5[2016] 175-200). After a brief survey of the history of authorship attributions for Arden of Faversham, Tarlinskaja begins to explain her metrical analysis method, starting with the alternation of weak (W) and strong (S) stresses in a standard iambic pentameter, which she represents by the string "W S W S W S W S W S". Surprisingly, nowhere in her essay does Tarlinskaja mention the idea of metrical feet, of which this line has five.

    Tarlinskaja gives as her first example of deviation from this pattern the line "Seems to reject him tho' she grant Pray'r" by Alexander Pope. Tarlinskaja applies boldface, italics, and underline styling to the words "Seems to" and "tho'" but she does not indicate if these three forms of styling are meant to convey three different facets of meter or simply reinforce one another for additional emphasis. This matters because she later uses each of these forms of styling on its own and sometimes uses two of the three forms without the third, and it is impossible to tell what she means by any of this. Later she intermittently uses capital letters to indicate stress.

    On the central matter of how we tell when a metrical norm is being violated Tarlinskaja is also silent just where the reader needs an explanation. The line quoted from Pope, has, she asserts the existence of "an extrametrical stress on syllabic position 1 and two missing stresses on S, on positions 2 and 6", hence she depicts it as "Seems to reject him, tho she grants his Pray'r". For Tarlinskaja this deviation from the norm is self-evident and she gives no explanation why we must perceive the line this way. The unanswered question is why seems must be stressed and to and tho' not stressed?

    Tarlinskaja is clearly applying a set of rules -- perhaps including one requiring that monosyllables such as to and tho' must not be emphasized -- that she assumes the reader already knows and agrees with. This omission of her stress rules makes her argument essentially unintelligible. An additional terminological problem is that she calls the first deviation "an extrametrical stress on syllabic position 1" (p. 177) and most writers on meter reserve this term extrametrical for additional syllables and feet in a line, not for stress falling where we do not expect it to.

    Next Tarlinskaja introduces the notion of "metrical words", claiming that the line from Pope should be divided like this: "Seems to reject him, | tho she grants | his Pray'r". Her explanation is that "Metrical words contain a dictionary word or their groups whose stress falls on a metrically strong syllabic position" (p. 177). In that sentence, I cannot find the antecedent for their so I do not know what a metrical word is. Tarlinskaja goes on: ". . . Pope's line contains only three metrical words, because only three stresses fall on strong metrical positions 4, 8, and 10" (p. 177). That is, she finds just three stresses falling in the right places in the line, which we may represent in a way that Tarlinskaja does not: "seems to reJECT him tho' she GRANTS his PRAY'R".

    It seems that Tarlinskaja's rule for a metrical word is not, as she puts it, something that contains an occurrence of stress falling on a strong syllabic position (that is, an S position) but rather something that ends with a stress falling on a strong syllabic position. But perhaps I have misunderstood her; if so I do not think I am to blame as her explanations will, I think, baffle most readers. These problems emerge in just the second paragraph of Tarlinskaja's account of her method and this reader's confusion grew thereafter.

    A central problem in all this is how we decide whether a monosyllabic word should be stressed or not. (Polysyllabic words generally have an agreed stress so there is no ambiguity: everyone says POet not poET.) Tarlinskaja attempts to tell the reader how she stresses monosyllabic words: ". . . we conventionally divide monosyllables into three categories: predominantly stressed (lexical words; e.g., nouns, verbs, such as talk, ride, swell, as well as adjectives and adverbs), predominantly unstressed (grammatical words such as articles, prepositions, and conjunctions), and ambivalent, sometimes stressed and at other times unstressed (such as personal, demonstrative, and possessive pronouns)" (p. 178).

    This is a start, but the reader then wants to know how to decide the predominant cases and the ambivalent ones. What are the rules, exactly? Tarlinskaja gives us just one of them: "Personal pronouns, for example, are considered always unstressed on W positions, while on S positions they are considered unstressed if they are adjacent to their syntactic partner, and stressed if they are separated from the syntactic partner by a phrase" (p. 178). With the examples she offers, this makes sense. So why not then give all the rules? Since the meaning of the entire essay depends on the many unstated rules of stressing, there is no point in proceeding with a review of it: to this reviewer it is simply incomprehensible.

    In an article reviewed in NYWES for 2016 called "The Use of Spellings for Compositor Attribution in the First Folio", Pervez Rizvi showed that virtually all the scholarship that attempted to identify the stints of different compositor working on that book is invalid. He has now done essentially the same for the less widely known scholarship about which typecases those compositors used ("Typecase Attributions for the Shakespeare First Folio", Script and Print: Bulletin of the Bibliographical Society of Australia and New Zealand 44[2020] 135-77). In his book The Printing and Proof-Reading of the First Folio of Shakespeare, Charlton Hinman determined compositor stints (in Volume 1) almost entirely on the basis of spelling and devoted almost all of Volume 2 to type recurrence, which enabled his typecase attributions.

    To print the roughly nine hundred pages of the Folio it was necessary to reuse the same collection of type many times, since no printer had enough type to keep it all set up (that is, 'in type') at the same time. In principle each piece of type holding a given letter was identical to all the others for that letter in the same font -- a font being all the type cast from one set of moulds --  so it is normally impossible to tell which particular piece of type was used for each instance of a letter in a printed book. But Hinman found that where a piece of type was damaged in a distinctive way, it is possible to track that particular piece's reuse across a large book such as the Folio.

    Because the same piece of type cannot be in two places at once, wherever we find one being reused we can be sure that the pages it appears in were not set up in type at the same time. That is, the type for one of the pages in which it appears (or at least the type for part of that page) must have been distributed back into the typecase before the other page was typeset. This throws light on the order in which pages were set in type and makes certain patterns of working more likely than others. If we think that at any one time the printer was working solely on completing one book, then certain ways of working -- certain orders in which the necessary tasks were completed -- would have been so much more efficient than others that it is nearly certain that they were followed.

    Rizvi finds that Hinman's typecase attributions are as unreliable as his compositor attributions, for two reasons. The first is that he did not take into account the effect of concurrent printing of other books in the same printshop, and the second is that he assumed that Compositor A exclusively used case x and Compositor B exclusively used case y and that a third typecase z was sometimes used. Rizvi shows that type-recurrence evidence is compatible with assumptions other than these. He takes as correct Hinman's claims about the recurrences of particular pieces of type, and his deduction of the order in which the formes were set, based partly on skeleton forme reuse but corroborated by type recurrence. All compositor identifications Rizvi treats as suspect.

    Rizvi spots small slips in Hinman's counting of his own evidence: 619 (not Hinman's 615) distinct types occur 13,121 (not Hinman's 13,075) times across the book. Rizvi finds Hinman repeatedly combining inferences (say, a compositor attribution) with pure speculation (who distributed a particular forme or page) such as "Ax distributed i4 r and set i2" (Rizvi's p. 140). Hinman was aware that sometimes he introduced an anomaly into his system where ". . . a type is distributed into one typecase but is next seen on a page set from a different typecase" (p. 140), and overall Rizvi found 238 examples of this in Hinman's book

    Hinman did not take enough account of the concurrent printing of other books in the same printshop alongside the Folio, although he was aware of it. Rizvi points out that while compositor attributions for a particular page can be made by looking just at that page and using, say, the spelling evidence in it, ". . . typecase attributions are progressive: they cascade from one page to the next" (p. 142). This is because if we find type recurrence between Folio pages A and page B that are near each other then we are committed by definition to saying that the type from page A was distributed into the typecase from which page B was set, which may or may not be the typecase from which page A was set. Thus the attributions of case use for one page affect those of another.

    Concurrent printing disrupts the patterns of reuse that follow from this dependency. If the typecases used to set pages A and B were in the intervening period used for some other book -- on which job the type did not go back into the case it came from but rather got swapped between cases -- then it can appear from the perspective of our work on Folio pages A and B that some pieces of type have magically leapt from one case into another. Or rather, and worse, we will not know that this has happened behind our backs, as it were, and in ignorance of this we will make subsequent inferences based on what we see in subsequent pages and what we wrongly believe we know about which pieces of type are in which cases. Rizvi details more complex ways in which concurrent printing can invalidate inferences made on the assumption that the type in the Folio was being used solely to print the Folio.

    To prove that concurrent printing really would undermine everything Hinman did, Rizvi peforms a thought experiment in which The Winter's Tale was omitted from the Folio in order to see what Hinman would have concluded, using his methods, if it were in fact not part of the Folio but a work of concurrent printing. He finds that with The Winter's Tale removed, a significant number of anomalies are introduced into Hinman's set of inferences as they currently stand for the other 35 plays. Of course, Hinman would not have these anomalies to deal with if he had been working with a Folio that lacked The Winter's Tale: he would likely have made different inferences to suit the 35-play Folio since he was seeking to minimize anomalies.

    In fact, Rizvi shows, Hinman would have been able to interpret the evidence for a 35-play Folio by introducing only a small number of additional anomalies (245 instead of 235), so there would be nothing to raise his suspicion that we have manually excised one whole play. Concurrent printing would have the same effect: it would leave us with nothing suspicious in the Folio to alert us that a process had occurred that could make all subsequent case attributions faulty. The necessary conclusion is that the possibility of concurrent printing is, on its own, enough to make all of Hinman's typecase attributions unreliable.

    Concurrent printing makes invalid inferences that use type recurrence to deduce the order in which the formes were printed and inferences about how much type was in the typecases and how much was left standing (that is, set up as pages). Rizvi decides to set aside all these problems and to consider Hinman's typecase attributions on the assumption that his inferred order of the setting of formes is correct. Hinman unjustifiably assumed that each compositor had one typecase he usually worked at, and thereby he linked compositor identifications with typecase identification: if he knew which compositor set something he felt he knew which typecase it was set from, and vice versa.

    If Hinman had derived his compositor identifications solely from spellings and the typecase identifications solely from type recurrence and had then found a correlation between the two, his claim that each man had his own case would be reasonable. But in fact Hinman repeatedly let one kind of identification help him with the other. In particular, he started with a hypothetical set of compositor identifications and then turned to type recurrence to see if he could interpret it in a way that would give each man his own case. What he should have done is interpret the type recurrence as best he could, making typecase attributions independently of what he had decided about compositor attributions.

    Rizvi shows that for quire mm Hinman demonstrably introduced inconsistencies in his attribution of typecases to pages in order to preserve his axiom that each compositor used his own case. Rizvi then interprets the type recurrence evidence for quire mm without thinking about the compositor stints and is able to come up with a simpler, less inconsistent set of typecase attributions, in the sense of less often needing to split a page into subsections that got treated differently, although at the cost of marginally increasing the number of anomalies, in the sense of recurrent type popping up where we do not expect it to and type being distributed into a different case from the one it was set from.

    Why did Hinman hypothesize a third case z just for use in a few quires when he already had cases x and y in use throughout the Folio? His first reason was that when the type from quire I recurred on page K2v it was almost all from just one page, I1r. Rizvi points out that this is not as strange as it seems and it happens for 30 pages of the Folio. The matter is not just one of random chance, since in each sort box newly distributed type is placed on top of type that was already in there and is the first to be reused on the next page that is set.

    Another reason for the third typecase was that Hinman had already decided on the basis of spelling that K2v was set by Compositor A and hence from case x, but type that had recently been distributed into case x did not appear on K2v so he invented a new typecase to explain this. Also, having invented case z and attributed pages to it, it was inevitable that no piece of type from case y would reappear where Hinman was expecting them to, and as Rizvi points out this is circular logic. If we do not invent case z and attribute its pages to case y then pieces of type from y do recur where Hinman expects them.

    Yet another reason that Hinman gave for inventing case z is type shortage as evidenced by forced substitutions on K2v such as 'vv' for 'w', which told Hinman that ". . . the case from which it was set was far from full, as if this case contained only the types just distributed into it from page I1^r"(Hinman 2:400 quoted on Rizvi p. 156). But as Rizvi points out, the unusually high demand for 'w' on K2v would have exhausted any typecase's supply of this letter, so we do not need to invent a new but depleted typecase to explain the substitutions. Exacerbating the problem, pages with unusually high numbers of 'w' were in press or waiting to be distributed at this point. The same consideration governs the substitution of B for H on K2v: several pages with a lot of letters H were not yet distributed when K2v was set.

    Setting aside Hinman's explicit reasons for inferring the presence of case z, can the type-recurrence evidence be accounted for without it? Rizvi tries to reassign all of Hinman's case z attributions to either case x or case y. He manages it with just an additional three anomalies (as defined above) and producing a much simpler pattern of setting and distribution, eliminating or reducing in several cases the need to subdivide a page into smaller units treated differently.

    Where a forme of pages a and b is distributed into one typecase that is then used to set a forme of pages d and e, we might expect the distinctive type from pages a and b to be found across pages d and e rather than, say, page a's distinctive type all turning up in page d and page b's distinctive types all turning up in page e. But in fact since either page a or page b was distributed into the typecase before the other was -- that is, they could not simultaneously distribute type from two pages into one case -- either a's or b's distinctive type is going to be on top of the other page's distinctive type in each sort box, and when that typecase is next used to set page d or page e that page is going to get that on-the-top type first. Such page-to-page transfer of distinctive types is possible and Hinman's work shows it happening quite a lot. Rizvi's new typecase attributions do not add to the number of times this unexpected phenomenon occurs.

    So why did Hinman not adopt the typecase attributions that Rizvi found? It is hard not to conclude that it was because doing so would undermine his axiom that each compositor had his own case. Rizvi ends by critiquing Paul Werstine's essay "Cases and Compositors in the Shakespeare First Folio Comedies" (Studies in Bibliography 35 (1982): 206-34), which is the only post-Hinman scholarship on Folio typecase attributions prior to Rizvi's. Werstine assumed that when a page is distributed into a typecase the pieces of type from either column of that page are equally likely to recur on the next page set from that typecase, but this is wrong. Formes were distributed from the outer edges in, so the first columns distributed were the a column of the verso page and the b column of the recto page and then the two remaining middle columns were distributed. This order put type from those middle columns on top in the sort boxes and this type was the first to be used for the next typesetting done from that case.

    Rizvi shows that unreasonable assumptions about probability -- in particular, attributing to positive actions by the compositors things that can easily happen by chance -- led Werstine to overfit his theory to the evidence. Werstine also took too little account of how concurrent printing could produce the evidence that he interpreted as "simultaneous composition of different formes" (Werstine p. 213). Werstine's chief conclusion of interest to editors was that because of simultaneous composition of different formes within the quire at once, copy for the whole of a quire had to be cast off in advance. And thus we should look for the stretching or compressing of copy in the second half of a quire. Previously we thought only the copy for the first half of a quire had to be cast off because the second half was set in reading order and hence the second half could not suffer from stretching or compressing. Rizvi shows that there are no good reasons to accept Werstine's conclusion about this.

    Brian Vickers has another essay arguing that Thomas Kyd was the sole author of the anonymously published play Arden of Faversham, couched as a response to MacDonald P. Jackson's claim that Vickers's authorship attribution method does not work ("Kyd, Shakespeare, and Arden of Faversham: A (Belated) Reply to MacDonald Jackson", ROMRD 56-57[2020] 105-34). Vickers starts by reviewing the history of attributions of Arden of Faversham and identifying as especially perspicacious the ones that give it wholly to Kyd. Vickers promises that evidence in support of the attribtion is on his website at <www.brianvickers.uk/?page_id=808> but at the time of writing (June 2025) this page returns a "404. Not found". He also promises that a copy of his 2008 essay in the Times Literary Supplement (reviewed in YWES for 2008) is at <www.brianvickers.uk/wp-content/uploads/2016/05/Vickers-Thomas-Kyd-The-Secret-Sharer.pdf>. It is not and this URL produces the same "404. Not found" error.

    Vickers spends considerable space surveying his own previous publications on Kyd's authorship of Arden of Faversham. He takes some of the example matches between accepted Kyd works and Arden of Faversham that he has offered before and puts them into their dramatic contexts. He leads the reader through his reasons for thinking that, despite Jackson's objections, the wider contexts show that the matches are conceptually as well as verbally alike. The result is necessarily subjective, although Vickers asserts that "subjective aesthetic judgments . . . [are] out of place in modern attribution studies" (p. 109).

    Vickers's examples of close verbal matches between known Kyd plays and Arden of Faversham start with some whole phrases in common but soon peter out to such things as the rest [four other words] it is [one other word] bad matching with the rest [one other word] tis [one other word] bad (p. 110). The value of this evidence depends on how often we should expect any large text to contain such a loose verbal match. (It is loose in the sense that we allow any word to appear where I have marked other words appearing in a match.) Vickers offers no remarks on whether his matches are rare phrases or common ones, and he mixes examples of the two kinds indiscriminately. Vickers concludes his survey of phrases in Kyd's Solimon and Perseda matching similar ones in Arden of Faversham by claiming that "The quantity and quality of verbal matches" (p. 111) clinches the case.

    As noted above in our review of his essay "Kyd, Edward III, and 'The Shock of the New'", Vickers's method is first to find the phrases shared between two texts -- one already securely attributed and one to be attributed -- and then eliminate the phrases that he finds are present throughout the language to leave the few that are not common. In the present essay Vickers turns next to Jackson's demonstration that this procedure sets up what Jackson calls a one-horse race and that it is possible to repeat the operation with a different candidate and get just as many matches for the text to be attributed, since it is a fact of language that any two substantial bodies of writing will necessarily contain a certain number of phrases that no other texts contain.

    In response to Jackson's argument that Vickers's method should give other dramatists a chance to compete for attribution, Vickers explains: ". . . over my scholarly career I have acquired enough knowledge of Elizabethan drama to be able to exclude those dramatists as having had significantly different linguistic and stylistic characteristics, not to mention dramaturgy" (p. 114). Vickers thinks his scholarly experience entitles him to pick the winner before the race starts, so there is no point entering other runners. He explicitly asserts that there is no need to actually run races for other candidates: "I am confident that individual searches in any of the other dramatists that Jackson mentioned -- 'Marlowe, Shakespeare, Peele, Greene, Lodge, Porter, Wilson, and others' -- will find no-one who can shake the high probability that Kyd wrote Arden of Faversham" (p. 114).

    Vickers could justify his confidence by actually running the experiments needed to prove his claim, but he chooses not to. Vickers offers a list of reasons for not running such an experiment, such as the fact in respect of Marlowe's candidacy that "The diction of Tamburlaine the Great, both parts, is distinct from most Elizabethan plays and, indeed, from anything he subsequently wrote. The chances of it sharing 'unique matches' with Arden are remote" and his other plays Dido, The Massacre at Paris, and Doctor Faustus "have highly unreliable texts" (p. 114).

    Marlowe's Edward 2 echoes a lot of other plays of its time, and Vickers thinks that this "negates Jackson's argument that the links between Arden of Faversham and Kyd's plays are commonplace features of that period in Elizabethan drama" (p. 115). I cannot see the logic in this. Where there are parallels between plays that tell against Vickers's claims, he adduces evidence about dating to show that Marlowe was borrowing from Kyd rather than the other way around or rather than Marlowe writing the drama that Vickers wants to attribute to Kyd. But the dating of plays under discussion (particularly in the 1580s) is considerably more uncertain than Vickers allows.

    Where Arden of Faversham has an abundance of parallels with Shakespeare plays, Vickers wishes away this evidence by invoking the idea that "Shakespeare remembered the play well, from public performances" (p. 116). Combining inferred dates of authorship with known dates of publication, Vickers attempts to show that parallels between Arden of Faversham and Shakespeare's works must be caused by Shakespeare borrowing from Kyd rather than the other way around. At this point Vickers seems to forget his own point about the possibility of a dramatist remembering another's play from public performances and he relies on dates of publication to infer which borrowings are impossible.

    Vickers objects that in his critique of the attribution of Arden of Faversham to Kyd, Jackson cherry-picked the evidence to dismiss it as inconclusive and failed to address the sheer quantity of corroborating evidence: there are "some forty matching collocations, too many and too detailed to be the result either of imitation or coincidence" (p. 117). But Vickers nowhere shows that 40 is too many matches to be explained by imitation or coincidence; he merely asserts that it is. This is particularly a problem where the matches are imprecise, in the sense that we allow one or other text to vary the wording slightly and to have differing numbers of words intervening between the words that are the same. Unless we quantify this imprecision we cannot begin to judge how alike two texts are and how likely it is that two texts by different authors will nonetheless match according to this level of looseness in correspondence.

    To counter Jackson's statistical demonstration that the evidence Vickers has found is not significant, Vickers points to some blog postings by Martin Mueller. The most impressively titled of these is "Vickers is Right about Kyd". Vickers quotes Mueller's work selectively, leaving out the bits that undermine confidence in that title. In particular, Mueller reported that this own understanding of the test he was applying using Linear Discriminant Analysis (LDA) was imperfect: "The math is well beyond me", Mueller wrote. In fact, Mueller demonstrably misused LDA, thinking that the confidence metric it produces shows how confident we can be about the specific proposed attributions on which the discrimination is made, when in fact the confidence metric in LDA reflects how well the data presented to it can be discriminated, that is, how well it can be divided into distinct categories. This is not the same thing.

    Moreover, results reported by Mueller (passed over silently by Vickers here) include some that are quite absurd, such as "Discriminant analysis thinks that there is a 92% chance that John Lyly wrote Love's Labor's Lost as opposed to an 8% chance for Shakespeare" (Muller "Vickers is Right about Kyd"). To close the quantitative discussion, Vickers cites the work of Albert C. Yang that corroborates his, which work is reviewed above in this round-up. Next Vickers turns to literary-critical matters, arguing that some distinctive patterns of dramaturgy are common to Arden of Faversham and Kyd's plays.

    For his part, Jackson responds to Pervez Rizvi's critique ("Small Samples and the Perils of Authorship Attribution for Acts and Scenes", reviewed above) of his contribution to the New Oxford Shakespeare Authorship Companion ("Reconsidering a Lexical Test of Arden of Faversham: A Response to a Critique", Shakespeare 16[2020] 182-7). This kind of apparently instant response is possible because Rizvi's critique was available in what is called the Advance Access form of the journal ANQ, mounted online but without final page numbering, a couple of years before publication of the journal issue to which it is finally assigned.

    Jackson acknowledges that in his rank-order table of most-Shakespearian plays (based on preferential use of the words answer, beseech, gentle, spoke, and tonight and avoidance of brave, hopes, sure, and yes), Peele's The Arraignment of Paris came out top and Greene and Lodge's A Looking Glass for London was sixth, ahead of most Shakespeare plays. Jackson defends his approach by saying that no method is perfect and it is the agreement amongst multiple methods that gives reliability. Jackson accepts too that Rizvi is right in general to assert that counts derived from whole plays cannot simply be compared to counts derive from individual acts and scenes.

    Jackson describes Rizvi's experiment in which he broke all the plays into 4647-word segments to match the 4647 words in Scenes 4-9 of Arden of Faversham. Rizvi gave Jackson the raw data of his experiment and Jackson here shows the top 15 segments in the rank-order table and 14 out of 15 of them are segments from Shakespeare plays. And the single segment that is Arden of Faversham Scenes 4-9, which was already suspected of being by Shakespeare, tests just as highly as the other segments in that top 15. Jackson's point is that the anomaly of a segment from Peele's The Arraignment of Paris coming top of the table should not concern us since no one proposed in advance of the experiment that Peele's play is Shakespearian. But independent evidence already did point to Arden of Faversham Scenes 4-9 being Shakespeare's, so their being found to top the table in this experiment is significant.

    Also on the topic of Arden of Faversham, Gary Taylor argues that Thomas Watson was its main author and Shakespeare a minor co-author ("Shakespeare, Arden of Faversham, and Four Forgotten Playwrights", RES 71[2020] 867-95). There are four men known to have written tragedies, because their contemporaries said they did, that have left us "no known surviving full-length plays" (p. 868): Thomas Achelley, Michael Drayton, Richard Hathwaye, and Thomas Watson. Taylor takes them in turn.

    Henslowe's Diary records Richard Hathwaye getting five pounds for a play about King Arthur in 1598, and he recurs in the Diary as co-author on collaborative plays until 1603. Francis Meres mentions him as one of those "best for comedy" in Palladis Tamia in 1598. Arden of Faversham was a tragedy not a comedy and written in 1588-1590, and Hathwaye did not distinguish himself as a literary author then and his only certainly attributed literary output was a short poem prefacing the anthology Bel-vedere, or the Garden of the Muses in 1600 and it is not very good. Shakespeare's contributing a small part to a tragedy by such a man in 1588-90 does not match Shakespeare's other dramatic activity at this time.

    In 1598 Meres praised Michael Drayton as among the best "tragic poets" and well before that he was widely admired, and Bel-vedere and England's Parnassus (the same year) included hundreds of excepts of his works. But Arden of Faversham must have been written by an experienced professional dramatist and Drayton was not that in the late 1580s; not until the late 1590s was he collaboratively writing plays for Henslowe. Drayton was praised for not having knowledge of the very things -- swaggering tavern-and-brothel language -- needed to create Arden of Faversham's characters Black Will and Shakebag.

    In his A Knight's Conjuring of 1607, Dekker imagined the great actor John Bentley in Elysium with a group of professional dramatists including "learned Watson, industrious Kyd, [and] ingenious Achelley" who wrote for Bentley. Since Bentley died in 1585 this puts Watson and Achelley writing plays for Bentley and presumably for the company he worked for, the Queen's Men, from its foundation in 1583 to his death in 1585. In Palladis Tamia, Meres named Watson amongst the best for tragedy and among the commercial playwrights. Others also confirmed that Watson was a commercial playwright. Watson was older than Achelley and far more widely praised by his contemporaries, who wrote commendatory poems to adorn his publications: he is the likelier figure for Shakespeare to have collaborated with. 

    Taylor constructs a literary-critical argument that what we have of Watson's writing makes him a plausible candidate for authoring Arden of Faversham, since his works have some of the play's concerns and his life gave him the interests that the play's main dramatist shows. The play has an unusually high number of references to places in London (not in the sources) and gets wrong Kentish place names: the dramatist knew London. Unlike Shakespeare and Marlowe, Watson was a native Londoner and from what we know of his life he was acquainted with its seamier side.

    To supplement his earlier quantitative argument that rare n-grams found in one part of Arden of Faversham are disproportionately also found in Watson's works (made in an article reviewed in NYWES for 2019), Taylor offers this bit of Arden of Faversham:

See how the Howrs, the gardeants of heaven's gate,
Have by their toil removed the darksome clouds,
That Sol may well discern the trampled pace
Wherein he wont to guide his golden car.
The season fits. Come, Franklin, let's away.

Taylor points out that this is like a bit of Watson's 1591 royal entertainment Elvetham:

Behold where all the Graces, virtue's maids,
And light-foot Howrs, the guardians of heaven's gate,
With joined forces do remove those blocks
Which Envy laid in Majesty's highway.

As Taylor observes, "In both, mythological Howrs work together (their toil/joined forces) to remove plural objects (clouds/blocks) obstructing the path (trampled pace/highway) used by a sovereign being (Sol/Queen Elizabeth) who in Arden is explicitly, and in Elvetham actually, riding in a special horse-drawn coach/chariot" (p. 886).

    In Arden of Faversham there is a distinct pattern of stage directions beginning with "Here . . ." as in "Here enters" and then followed by a stage direction beginning "Then . . .", and this pattern is also found (once) in Watson's Elvetham. Arden of Faversham repeatedly associates Edward Seymour, who was Lord Protector and Duke of Somerset, with Arden and his friend Franklin, which is not historically accurate and serves no purpose in the narrative. But the historical Seymour's son (also called Edward) was the man for whom Watson wrote the royal entertainment Elvetham. Taylor has some more minor bits of circumstantial connexion between Watson and aspects of the play.

    Taylor has found a previously undetected source for Arden of Faversham in Ovid's Remedia Amoris in which a man wracked by sexual jealousy of a rival for his wife's affections is counseled by a close friend to leave his home and bury himself in his business to forget about it. In Arden of Faversham this close-friend role is played by Franklin, who does not exist in the sources, where Arden is a wittol instead of a man wracked by sexual jealousy. Ovid's Remedia Amoris was not well known in the late-sixteenth century and had no English translation, but Watson knew it and referred to it elsewhere. Watson's preferred poetic form was the 18-line sonnet of three sixains and Arden of Faversham has 18 scenes followed by an 18-line epilogue in Watson's preferred rhyme scheme of ababcc. "This is a kind of private, self-referential joke . . ." (p. 893), according to Taylor.

    Edward Pechter argues that quantitative analysis of literary works is helpful for classification tasks such as authorship attribution but is not suitable for literary criticism ("Does Digital Technology Advance Our Understanding of Literary History?", Archiv für das Studium der neueren Sprachen und Literaturen 257 Band 172[2020] 312-41). Pechter endorses Franco Moretti and Oleg Sobchuk's objection to the use of straight trend lines running through clouds of data points: these assume that things change in a steady fashion and that those points found a long way from the line are just random deviations, not evidence of non-linear change.

    This is a reasonable objection, but of course we can also derive more complex non-linear trend lines and measure if these better fit the data. The danger that investigators seek to avoid by keeping their lines simple is that of so-called overfitting. It is possible to derive a complex wiggly lines that follows every wheel and turn of the data, to which this line is thus almost perfectly fitted. But following the data so closely risks overlooking general trends that would be better represented by simple lines, from which we assume that particular points deviate because the data contains inherent random variations. Complex lines that fit the complexity of the known data almost perfectly (rather than finding coarse, general trends in the known data) are unlikely to give accurate predictions for the points that lie between the known data points (by the process called interpolation) or for the points that lie before and after the known data points (by the process called extrapolation).

    Next Pechter turns to Hugh Craig and Brett Greatley-Hirsch's book Style, Computers, and Early Modern Drama: Beyond Authorship (reviewed in NYWES for 2017), and its claim that the distinctive style that each early modern theatre company is often supposed to have had cannot be detected by the methods we use to detect authorship. For Pechter this was a misapplication of the technology: "How do you represent a diffuse and multi-faceted phenomenon like acting-company style in terms of quantifiable linguistic data?" (p. 320).

    Pechter generalizes that "Attribution scholarship depends on function words -- auxiliaries (has/hath), relative pronouns (who/which), interjections (tush/push), prepositions (betwixt/between), and the like" (p. 321). Indeed some of it does, but there are plenty of approaches that use words that appear at a lower frequency than function words. Pechter repeats his over-generalization: "In attribution scholarship, style is defined strictly by the incidence of function words generated subconsciously in a biologically determined process" (p. 323). This really is not the only way that authorship attribution is done.

    To justify their claim for a change in the style of dramatic language during the decades around 1600, Craig and Greatley-Hirsch use the case of that falling markedly in frequency of use, down from 0.45% to 0.10% of all words. Pechter points out that in a play of 10,000 words this means down from 45 thats to 10 thats, which as he points out may not be noticeable at all to readers or playgoers. Indeed, how can any function word's change in usage be noticeable since they are almost by definition too common to be noticeable?

    When they turn to literary-critical interpretation of a contrast between a line in an early play using that and a line in a late play avoiding that, Craig and Greatley-Hirsch simply claim more effect than this word has power to exert in these examples. Pechter objects that meaning is contextual so that not every that has the same weight as every other that, and if your quantitative analysis is not taking context into account it is treating language as evidence that speaks for itself, which it does not. Pechter also finds telling slippages in Craig and Greatley-Hirsch's rhetoric suggesting that it is not clear whether they start with the data and see what it shows them or start with an hypothesis and see whether the data fit it.

    The remainder of Pechter's essay finds essentially the same flaws in other recent works of quantitative analysis of literature. A new idea introduced in the conclusion is that the out-thereness that quantitative analysis invokes -- claiming to find real evidence that is actually, quantifiable in the text(s) and at large rather than just being some local and subjectively claimed pseudo-evidence adduced by the critic -- has its own counterpart out-thereness in traditional, pre-digital literary criticism: the wider community of readers who value this critic and not that one. That is, for Pechter there is something rigorous about the mass-scale collective judgements of the critics.

    Tania Demetriou asks whether we can date Shakespeare's Hamlet using Gabriel Harvey's mention of it in an annotation to his copy of a 1598 edition of Geoffrey Chaucer ("Tendre Cropps and Flourishing Metricians: Gabriel Harvey's Chaucer", RES 71[2020] 19-43). Harvey wrote in his copy of Chaucer's works about "Excellent matter of emulation for Spencer, Constable, France, Watson, Daniel, Warner, Chapman, Syluester, Shakespeare, & the rest of owr florishing metricians". Does the last phrase there govern the whole list (are they all flourishing metricians?) or is just Shakespeare (or perhaps Shakespeare and Silvester) meant? Demetriou searched EEBO-TCP using the Lancaster University CQPweb interface to find "lists of items followed by 'and the rest of'" (p. 27).

    Demetriou gives as her search term "(_{SUBST} _{$}){3,7} (_{SUBST}){0,2} and the rest of'" but does not talk the reader through how the bits before the phrase "and the rest of" find lists but not other bodies of writing, which is a shame because the logic of this would be interesting to hear. Since Spenser and Watson were certainly dead before Harvey bought his book, it is hard to see how they were flourishing unless we put a strained interpretation on that word, such as its meaning that they were popular, but we do not need to do this since Demetriou has shown that only Shakespeare and perhaps Silverster are said to be flourishing.

    Harvey also refers to "Owens new Epigrams". John Owens's Epigrammata was published in 1606, but Harvey could have read them in manuscript much earlier. Harvey wrote that "The Earl of Essex much commendes Albions England", which sounds as if the Earl (executed in February 1601) is still alive. But Harvey had a habit of writing about dead people as if they were alive. What of the idea that Harvey wrote his note in phases over several years, so that we cannot take it as a whole and use it to date Hamlet? Demetriou critiques the weak arguments of Michael Hirrel that this happened. Finally, Demetriou surveys Harvey's annotating habits in general and shows that in the case of this Chaucer volume he made all the notes more or less at the same time while giving the book a single reading, rather than returning to it several times over several years. Frustratingly for our purposes, Demetriou does not give a definitive answer either way to the question of whether Harvey's annotation can be used to date Hamlet.

    Faith D. Acker reckons that John Benson's 1640 edition of Shakespeare's poems is better than it is usually given credit for ("Manuscript Precedents for Editorial Practices in John Benson's Poems: Written By Wil. Shake-Speare. Gent.", SQ 71[2020] 1-24). Acker shows that the kinds of things for which Benson's edition is routinely condemned -- lack of textual fidelity, bowdlerizing the content, regrouping poems thematically and so decontextualizing them -- are just what we also see in manuscript transmission of Shakespeare's poetry in the early seventeenth century.

    Acker shows that in manuscript collections and in Benson's edition, such rearrangements and omissions emphasize certain aspects of each poem and de-emphasize others, and that there is artistic value in this essentially critical process of putting poems in juxtaposition and hence in conversation with one another. Indeed, Benson's edition of Shakespeare's poetry is more like a manuscript miscellany than like other printed editions of poetry.

    Towards the end of the essay Acker deals with Benson's notorious changing of Shakespeare's pronouns from male to female in Sonnet 101, which conceals the homosexuality. But as Margreta De Grazia pointed out, elsewhere in Benson's collection, indeed in its first poem's first line, the male pronouns are unchanged. Acker argues that where Benson's edition has female pronouns in place of Shakespeare's male ones, these "if not the errors of an unpracticed typesetter, likely reflect Benson s consistent use of manuscript practices to make his text more generic" (p. 22). That is, because heterosexuality is more common than homosexuality, the poetry is more generic if heterosexual.

    As part of giving the Shakespearian achievement of Israel Gollancz a "second look" (p. 221), Gordon McMullan proposes a new solution to an old crux ("Cormorant Shylock", SQ 71[2020] 221-41). Gollancz suggested that Shylock's name might come from shallach, the Biblical Hebrew word for a cormorant, since as a usurer Shylock was like that greedy evil bird. Shakespeare's point in doing this was to show Shylock transcending this stereotype and becoming "a genuinely tragic figure" (p. 235).

    When Shylock first meets Antonio he calls him a "fawning publican" and McMullan suggests that this second word is a compositorial misreading of "pellican" (p. 237) since later he will have his chest open to bleed for his childlike Bassanio and since ". . . the pelican, head bowed over its breast to peck out blood for its children to feed on, might quite easily appear (to someone unaware of the bird's mythical self-sacrificial tendencies and not trained in the finer points of Christological symbolism) like a creature that is 'fawning'" (p. 237). Pelicans were also thought to be rather depressive, as Antonio is. The cormorant and pelican were treated as two sides of one binary, as are Shylock and Antonio.

    Two book-form collections of essays appeared that are largely relevant to this review. In each, essays that are not relevant will be silently passed over. The first was edited by Rory Loughnane and Andrew J. Power (Early Shakespeare, 1588-1594). In their Introduction to the collection, Loughnane and Power note that there survives just one document about Shakespeare between 1585 (the christening of his twins) and the 1592 Groatsworth of Wit allusion: John Shakespeare's bill of complaint against John Lambert about property in Wilmcote, which names William as his son. Shakespeare's 25-year writing career is about 10 years longer than the average of his dramatist peers. 1588 is the starting point for this collection's consideration of early Shakespeare, this being also the year of the Armada and Richard Tarlton's death, and 1594 is the ending point because the Chamberlain's Men company was formed that year.

    The resumption of playing in mid-1594 followed a long period of theatre closure that began mid-1592. Malone's essay on the order in which the Shakespeare plays were written was first published in Samuel Johnson and George Steevens's 10-volume edition of 1778 and one of his criteria for judging a play's date of composition was its degree of "under-development, and immaturity" (Loughnane and Power's words, p. 6). Loughnane and Power understand the "Johannes factotum" remark in Groatsworth of Wit as perhaps, inadvertently, a compliment: we now admire Shakespeare for his range across genres and modes. 

    In "Shakespeare and the Idea of Early Authorship" (pp. 21-53), Loughnane lists some literary works whose authors were surprisingly young, but all we have for young Shakespeare is perhaps the "Hate-away" sonnet 145. In all we have only 32 professional London plays from the 1580s (12 of them by Marlowe or Lyly) and another 36 are named in the Lost Plays Database. 20 of the surviving 36 plays of the 1580s are from 1588-89 and these include three Shakespeare plays, so presumably he wrote earlier plays that are lost like much else that is lost for 1580-1587. Otherwise, we have to suppose that in the three surviving plays we have all of Shakespeare's earliest writing, which is inherently unlikely. 

    Loughnane's concluding thought is that Shakespeare's plays from 1588 to 1594 -- The Two Gentlemen of Verona, Arden of Faversham, Titus Andronicus, 2 Henry 6, 3 Henry 6, The Taming of the Shrew, Edward 3, Richard 3, The Comedy of Errors, and Love's Labour's Lost -- are three comedies, four histories, and three tragedies, showing the same range as in the rest of his career. Or we could call Titus Andronicus and Arden of Faversham histories and make it 2:6:2. And five of these 10 are co-authored: Arden of Faversham, Titus Andronicus, 2 Henry 6, 3 Henry 6, and Edward 3. And three of these five -- Titus Andronicus, 2 Henry 6, 3 Henry 6 -- appeared under Shakespeare's name alone, so maybe Groatsworth of Wit is right and Shakespeare's collaborators did object to his taking credit for their work. Overall, this early work of Shakespeare is "a study in ambition" (p. 48).

    In "Collaboration and Shakespeare's Early Career" (pp. 54-75), Will Sharpe notes that the Groatsworth of Wit allusion to the play tells us that 3 Henry 6 (and therefore its predecessor 2 Henry 6) had appeared on stage by September 1592, since that is when Groatsworth was entered into the Stationers' Register. The remark that Shakespeare was an "upstart crow beautified with our feathers" has been understood as a complaint that Shakespeare took over and put out as his own other people's work, as Loughnane thinks he did with Titus Andronicus, which was not an active collaboration, and that Shakespeare was an actor turned dramatist. 1594 seems to be when Shakespeare stopped collaborating and he did not start again until Timon of Athens in 1605. Sharpe concludes by reporting some competing chronologically ordered lists of Shakespeare's pre-1594 plays and then offering one of his own.

    MacDonald P. Jackson's essay "Shakespeare's Early Verse Style: Titus Andronicus, Venus and Adonis, Arden of Faversham" (pp. 102-120) is about authorship attribution. Jackson sets out to show that Titus Andronicus 2.3.10-50, Arden of Faversham 6.6-34, and Venus and Adonis 830-899 "must be by a single author" (p. 102). Jackson surveys his previous work on the authorship of Titus Andronicus. Turning to Tamora's evocation of the pleasures of the countryside at Titus Andronicus 2.3.10-29, Jackson notes a series of phrases that are found elsewhere only in Shakespeare's plays: make . . . boast (also in Henry 5 and Much Ado About Nothing), snake . . . rolled . . . chequered (also in the Shakespearian part of 2 Henry 6), cooling wind (also in The Merchant of Venice as wind cooling), under . . . sweet shade (also in Henry 5 and as sweetest shade in the Shakespearian part of 2 Henry 6 and as sweeter shade in the Shakespearian part of 3 Henry 6).

    Jackson finds most convincing the parallelism of Tamora's remarks on the echoing musicality of hounds in the hunt with those found in A Midsummer Night's Dream, The Taming of the Shrew, and Venus and Adonis 692-696, and also Tamora's remark on the nurse's song (also in Venus and Adonis 974). In Aaron's reply to Tamora we find the collocation of fatal and adder that is unique to this play and Venus and Adonis 878-879. Jackson tabulates a series of 16 verbal and conceptual likenesses of Titus Andronicus 2.3.8-36 and Venus and Adonis 830-831. Jackson finds that the part of Venus and Adonis he has been quoting from, lines 830-881, overlaps with the section in lines 855-899 that has a large cluster of parallel phrasings with Arden's account of his dream in Arden of Faversham and with no other plays.

    In detailing these parallels, Jackson is essentially drawing out the raw data from his book Determining the Shakespeare Canon reviewed in YWES for 2014. The rare phrases Jackson lists find 11 unique parallels to Shakespeare plays and 9 to all the rest of the plays of the period, and all 9 of those are to works by different authors; that is, no one person has even two such links yet Shakespeare has 11. Jackson explains that he had not noticed before that the bit of Venus and Adonis that Arden of Faversham has lots of links to is also the bit of Venus and Adonis that Titus Andronicus has lots of links to. Jackson lists another 16 verbal and conceptual links between Arden of Faversham 6.14-34 and Venus and Adonis 855-899. The simplest explanation for all this is the common authorship of Titus Andronicus, Venus and Adonis, and this part (Act 3) of Arden of Faversham.

    Since Venus and Adonis (published 1593) and Titus Andronicus (published 1594) were both unavailable in print until after Arden of Faversham was published (in 1592), the Arden of Faversham author cannot have been influenced by either them in print. And even if he was influenced by Titus Andronicus in performance why would he be influenced by exactly that bit of Titus Andronicus that the author of Titus Andronicus (Shakespeare) echoed in his own Venus and Adonis? The Arden of Faversham author reading Venus and Adonis in manuscript is inherently unlikely, and even if he did we must ask again why would he choose to be influenced by just that bit of Shakespeare's Venus and Adonis that Shakespeare drew from in his Titus Andronicus? And if Shakespeare was influenced by someone else's Arden of Faversham Act 3 then he would have to have been influenced by it twice, once in Titus Andronicus and once in Venus and Adonis, and furthermore (for this non-Shakespearian Arden of Faversham hypothesis to stand) the rest of Arden of Faversham Act 3 would have to be merely coincidentally like Shakespeare's writing when tested by other wholly independent means as described in the New Oxford Shakespeare Authorship Companion.

    Once we exclude Act 1 of Titus Andronicus as by Peele and Act 1 of 1 Henry 6 as by Nashe, we are left with a set of early Shakespeare plays that are evenly superb in their verse, whereas previous scholars found early Shakespeare's verse a bit uneven because they counted those two acts as his. Jackson acknowledges that there are verbal parallels between Kyd's Solimon and Perseda and Arden of Faversham, but notes that in these ". . . the phrase in Arden is apt to interact with other words in the immediate context that draw out its latent metaphorical content, whereas the same phrase in Soliman and Perseda remains inert" (p. 114). The speeches in Titus Andronicus, Venus and Adonis, and Arden of Faversham that Jackson is concerned with all engage in detail with hunting terminology, which is a recurrent topic of early Shakespeare.

    Also about Arden of Faversham is Terri Bourus's "Arden of Faversham, Richard Burbage, and the Early Shakespeare Canon" (pp. 200-219). Wiggins has claimed that the author of Arden of Faversham must have been a talented and enthusiastic amateur, not a professional, on account of its stage directions being written from an audience member's perspective, as with recurrent use of the phrasing Here enters. But as Bourus points out, scholars have found this phrasing in stage directions from other plays and in any case it causes the performers no problems. Moreover, a script's stage directions might have been rewritten by someone other than the dramatist, as we think happened with Ralph Crane's transcripts of Shakespeare's plays.

    Wiggins also thinks that the role of Alice Arden is too long for any boy actor to manage, but as others have pointed out there are comparably long roles, measured in terms of lines or more usefully of words, such as Juliet in Romeo and Juliet and Cleopatra in Antony and Cleopatra. Noticeably, in the apparently Shakespearian section of Arden of Faversham (Scenes 4-9), Alice has much less to say than she does in the rest of the play. But Bourus does not think that this is because Shakespeare knew that a boy actor must not be overtaxed and that whoever wrote the rest of the play did not. Rather, according to Bourus, Alice has little to say here because after having a lot to do in the first scene the actor is given a rest in Scenes 2-7 in order to be ready for his large part in Scene 8.

    Bourus also notes that in Scenes 1 and 8 Alice is frequently speaking every second speech, which as Matthew Vadnais pointed out lowers the cognitive load required for learning cues: the actor knows that whenever someone else stops speaking, he is next to speak. (Bourus cites this from Vadnais's 2022 PhD thesis, but he also explains it in articles reviewed in YWES for 2013 and NYWES for 2018 and NYWES for 2019.) This seems to Bourus like the mark of a professional. She considers also, but does not place too much weight upon, the possibility that the part of Alice was written for the exceptionally talented Richard Burbage when he was around 19 years old. Bourus ends by considering the long speech in Scene 8 where Alice repents of her conversion to goodness and goes back to her adultery and murder plot with Mosby, including its remarkable stage action of tearing pages out of a prayer book.

    John Jowett's contribution to the collection is called "The Origins of Richard Duke of York" (pp. 235-260). Instead of being, as we used to think, a debased derivative of the play that was later published as Folio 3 Henry 6, might the 1595 octavo Richard Duke of York actually be "the original collaborative play" (p. 235) that Shakespeare took over to make his 3 Henry 6? Jowett aims to show that there was a collaborative play underlying the 1595 octavo and the Folio text, and that the latter contains Shakespeare's revisions, but also that the 1595 octavo is an unreliable derivative text. If we accept that the Folio 3 Henry 6 is a collaboratively written text then the 1595 octavo must be too, since it has bits that are essentially identical to bits of Folio 3 Henry 6 that we think are by Shakespeare and it has other bits that are essentially identical to bits of Folio 3 Henry 6 that we think are by Marlowe.

    A key passage for the claim of faulty transmission affecting the 1595 octavo is the complaints of Richard and Clarence in 4.1 about the marriages the new king Edward 4 has arranged for his wife's relatives when the brides would better have suited Richard and Clarence themselves. In the 1595 octavo the marriages they complain of are not merely historically wrong but also historically absurd. Among the names that are wrong in the 1595 octavo are "Bonfield" where the Folio has the correct "Bonuill". In early printed books the name "Bonfield" appears nowhere except its many occurrences in the play George a Greene, which was written in the early 1590s around the same time as the original version of 3 Henry 6, so the obvious inference is that the 1595 octavo of 3 Henry 6 got this name by corruption from George a Greene.

    Shakespeare's play has Richard and Clarence be quite clear that their disappointments in marriage are strong motivations for their rebellions against Edward 4, and the 1595 octavo does not just get wrong the unimportant details (people's names), but also obscures almost entirely that the brothers have this resentment at the new king. Also, the octavo names the unworthy recipients of the new king's favour but does not make the essential point, made in the Folio, that these favoured people are the new king's new wife's relatives. Despite this omission, the octavo has the Queen react to Richard and Clarence's resentment of her, even though they have not expressed that resentment of her. The lines in the octavo that create these problems are also metrically rough, making them suspicious.

    These flaws make it hard to see how the octavo version could have been revised to produce the Folio version. Extreme metrical irregularity elsewhere in the octavo where the corresponding lines in the Folio are metrically regular happens even in the Shakespeare portions -- and Shakespeare was never bad at meter -- which also suggests corruption in the octavo. The metrical irregularity coincides with the absence in the octavo of words and phrases present in the Folio, as if the octavo had just remnants of the Folio. But the octavo version of the play cannot be merely a corruption of the Folio version, since the octavo has lines not in the Folio, and noticeably these are almost all entirely metrically regular and make good sense. This strongly suggests "two separate stages of textual evolution, the first involving demetrifying paraphrase, the second involving metrical replacement of missing passages" (p. 246).

    The octavo-only passages are not connected with any one of the presumed authors of the play. Jowett explores some verbal links between these octavo-only passages and Peele's work and finds it possible that Peele "mended a deficient text in anticipation of the printed edition" (p. 249) but he notes that authorial attribution work by Hugh Craig and John Burrows shows that Peele is not generally the third hand in the play after Shakespeare and Marlowe. The Peele plays that these octavo-only passages have verbal links with were predominantly published in 1593-94, and this seems to be the key: somebody put these octavo-only passages into the play for the purpose of publication, and the obvious point of doing that would be "to patch a deficient text" (p. 255).

    The octavo and Folio are particularly discrepant in Act 4 where whole scenes are in a different order, and Jowett argues that the order of scenes in the octavo is especially awkward and wrong. Overall, the picture is of the octavo version of the play being derived from the Folio version by disruptive transmission of some kind, but Jowett is distinctly agnostic on what kind of transmission that might have been. That the disruption does not follow authorial stints strengthens the case for this being a matter of disruptive transmission.

    John V. Nance's contribution to the collection is called "Early Shakespeare and the Authorship of The Taming of the Shrew" (pp. 261-283). For centuries scholars have found The Taming of the Shrew to be suspiciously deviant from Shakespeare's usual metrical habits. Nance's approach to the authorship of the play is to take two 173-word extracts from the parts of the play commonly held to have been written by Shakespeare (1.40-61 and 16.134-154) and two 173-word extracts from a part of the play widely suspected of being by someone else (3.1-23 and 3.1.23-47). Using LION and EEBO-TCP, Nance extracts every 2-gram, 3-gram, 4-gram and "collocations (disjunctive word combinations)" (p. 267) found in these four 173-word extracts and looks for them in all their variant spellings and forms in other plays first performed in 1576-1594, isolating the examples that appear in exactly one other play.

    Where a 2-gram or 3-gram from The Taming of the Shrew finds no matches with any other play, Nance looks for "discontinuous constructions" of those 2-grams and 3-grams (p. 267). (I would have liked Nance's phrases "disjunctive word combinations" and "discontinuous constructions" to be precisely defined as I am not quite clear what they mean.) In a second test, Nance turns to all books published in 1576-1594 (rather than just plays) and seeks to isolate words and phrases used exactly once in this entire set of printed books. This gives a chance for authors who wrote a lot of things other than plays, such as Nashe, to be detected as possible candidates for writing part of The Taming of the Shrew.

    In the first test, starting with The Taming of the Shrew 1.40-61 (the Lord's instructions for tricking Sly), Nance found 49 phrases that also appear in just one other play from 1576-1594 and of these 9 were in plays by Shakespeare, 4 were in plays by Marlowe, 4 were from The Taming of a Shrew, 4 were in plays by Peele, 3 were in plays by Lyly, 2 were in plays by George Whetstone, and the rest were in plays for which we do not have a securely attributed author. From this Nance concludes that Shakespeare is "the likeliest candidate for the authorship" (p. 269).

    An obvious question not addressed by Nance is whether Shakespeare has more plays first-performed in the period 1576-1594 than the others. Looking at the online Database of Early English Playbooks (DEEP) it seems not, although the answer changes slightly according to whether you use DEEP's dates from Harbage's Annals or those from Wiggins's Catalogue. Using Annals, Shakespeare has 6 plays by 1594, Marlowe has 6 plays, Peele has 7 plays, Lyly has 8 plays, and Whetstone has 2 plays. Using Wiggins's Catalogue, Shakespeare has 5 plays, Marlowe has 6 plays, Peele has 7 plays, Lyly has 8 plays, and Whetstone has 2 plays. Either way, Shakespeare's dominance of Nance's hit list is not easily explained by his predominance in the dataset.

    The same picture emerges when Nance switches to looking at plays having more than one matching phrase: the top of the rank order is The Taming of a Shrew, The Two Gentlemen of Verona, Richard 3, and Titus Andronicus. In the second test, again using The Taming of the Shrew 1.40-61 and considering all books from 1576-1594, Nance finds matches in works by famous prose authors who did not write plays (the likes of John Calvin and Francesco Guicciardini) and also Anthony Munday, who did. Nance explains Munday's matches by saying that his "massive non-dramatic canon" (p. 270) causes his over-representation in the results, and in any case he gets no more matches than John Calvin, an impossible candidate.

    Nance moves to his second passage, The Taming of the Shrew 16.134-154, which is Katherine's final speech of apparent submission to male authority. In his first test, which is matches with just one other play from 1576-1594, this extract has 19 matches with Shakespeare plays, more than three times the number of matches with any one else's plays, and in a rank order of plays with most such unique matches Shakespeare plays dominate the top of the list. Essentially the same thing happens with the second test looking for matches with all books (rather than just plays): Shakespeare, and in particular his Lucrece, has by far the most matches.

    Nance turns to passages in The Taming of the Shrew that have been suspected of being by someone else. First there is 3.1-23, and in his first test Nance finds 60 unique matches to other plays and of these 20 are to Marlowe plays, 5 to Peele, 3 to Shakespeare, 3 to Lyly, and even smaller numbers to other writers. Marlowe clearly predominates in phrases shared with this part of The Taming of the Shrew. In a rank order of plays with the most of these links, Edward 2, Doctor Faustus, The Jew of Malta, and 2 Tamburlaine are at the top of the list. In Nance's second test, against all books rather than just plays, Marlowe has 3 matches, and so does Munday, and so does Guicciardini, and Petrarch has 2. But of course Munday and Guicciardini have vast non-dramatic canons.

    The last passage tested by Nance is The Taming of the Shrew 3.24-47 (the very next section of 173 words after the previous passage) and again Marlowe seems to be the author, having the most links and his plays being at the top of the rank order of plays by number of links. On Nance's second test (links to all books not just plays), Munday has 3, Raphael Holinshed has 3, Philip Sidney has 2, Gabriel Harvey has 2, John Foxe has 2, Thomas North's translation of Plutarch has 2, and Whetstone has 2. The obvious explanation for these links to non-dramatic texts is that each of these authors has a large non-dramatic canon. It is starting to look like The Taming of the Shrew was collaboratively written and hence we should not include it in the set of sole-authored-well-attributed plays that we use in other authorship studies. According to the title page of its 1594 quarto, The Taming of a Shrew was a Pembroke's Men's play, and so too were other early Shakespeare collaborative plays Titus Andronicus, 2 Henry 6, and 3 Henry 6.

    The last essay in the book is Gary Taylor's "Who Read What When?" (pp. 284-301). The mention of the ship called the Nonpareil in the play Edward 3 seems to come from Petruccio Ubaldini's account of the Spanish Armada published in 1590, but how can we tell whether the author of Edward 3 actually read Ubaldini or the ship had merely become well known after 1588 because everyone was talking about the Armada? Taylor shows that the phrase were . . . to give them way appears nearby in Ubaldini and also nearby in Edward 3 and nowhere else in any book published in 1583-94, so the author of Edward 3 must have read Ubaldini, since this phrase would not travel by word-of-mouth.

    Using searches in EEBO-TCP for rare phrases, Taylor is able to show that the 1594 quarto of The Contention of York and Lancaster (but not Folio 2 Henry 6) has certain phrases from the prose chronicles -- Richard Grafton's, Edward Hall's, and Raphael Holinshed's -- found nowhere else in a published book, and this is hard to explain if the quarto was based on a memorial reconstruction. I agree with all Taylor's claims about rare phrases except this one: "Aside from the chronicles, Contention is the only early example identified by EEBO-TCP of 'like a' before 'governed by'" (p. 292) in books "up to 1595" (p. 300n29). In fact, EarlyPrint.org finds "in Florence he liued like a Prince yet so gouerned by wisedome as he neuer exceeded the bounds of ciuill modestie" in Niccolò Machiavelli's The Florentine History published in 1595. Taylor not finding this is explained either by ProQuest's EEBO-TCP's search engine being not as good as EarlyPrint's or else by the EEBO-TCP dataset (that both use) having expanded since Taylor did his searching in 2017.

    John Dover Wilson showed that Folio 2 Henry 6 has phrasing that is unusually close to that of Marlowe's Edward 2, and Taylor shows that the likeness is even more pronounced in quarto The Contention of York and Lancaster. The standard explanation is that the Pembroke's Men performers who made the memorial reconstruction that underlies The Contention of York and Lancaster were recalling bits of Marlowe's play that was also in the Pembroke's repertory. But Taylor's searches of EEBO-TCP show that both The Contention of York and Lancaster and Edward 2 are at these points using phrasing from the prose chronicles that is not found elsewhere. The simplest explanation is that Marlowe read the prose chronicles as he wrote his part of The Contention of York and Lancaster and reused the same phrasing when he wrote Edward 2.

    Also against the idea of the 1594 quarto of The Contention of York and Lancaster being a memorial reconstruction is the fact that it begins with a 17-line sentence, which is hard to achieve. This is something most writers could not do, but Shakespeare and Marlowe could; it is not at all what we would expect from a simplifying process of recall. Where before the parallels between The Contention of York and Lancaster and Marlowe's plays were used as evidence of memorial reconstruction lying behind the 1594 quarto of The Contention of York and Lancaster-- because of an actor recalling bits of Marlowe's plays he had performed in -- the strong evidence for Marlowe being a collaborator with Shakespeare on the three Henry 6 plays makes it more likely that Marlowe was simply the author of the bit of The Contention of York and Lancaster that has the match.

    The second book-form collection of essays this year was edited by Tiffany Stern: Rethinking Theatrical Documents in Shakespeare's England. In her Introduction, Stern makes the surprising claim that ". . . masques were sometimes printed in advance of performance, and were often intended to be read during it" (p. 4) and it is disappointing that she cites no authority for this. In "Writing a Play with Robert Daborne" (pp. 17-32), Lucy Munro speculates about what Daborne was reading as source material for his plays and who lent him the books. By the early 1610s, 20 pounds was the standard price paid to the dramatists for a new play. Munro has a new bit of evidence showing Daborne being in the theatre industry as early as 1606 rather than the 1608 we previously thought (p. 22).

    Munro thinks that Daborne revised Machievelli and the Devil after the actors began to learn it, taking the reference to "they have now in parts" (Article 81 in Henslowe Papers) to mean in the form of cue-scripts. A large part of Munro's essay is speculative and it is presented in an unusual style as a piece of creative writing purporting to be Daborne's "self-help manual for the aspiring Jacobean playwright" (p. 18).

    Holger Schott Syme's "A Sharers' Repertory" (pp. 33-51) also speculates, about "what we might know if the Diary of an actor-sharer such as John Heminges or Thomas Downton had survived" (p. 34). Syme's main point is that no one actor always had the largest role in every play for one company. Syme notes that quite a few of the name roles in early modern drama are not the roles with the most words: in Othello, Othello says less than Iago; in Volpone, Volpone says less than Mosca; and in Jonson's The Alchemist, Subtle says less than Face.

    Some of the roles we know Burbage played were large but others were not: Hieronymo in The Spanish Tragedy and King Lear are not particularly long. So other members of the company must also have been capable of large roles such as Iago and Face. Syme tries to make the same case for Edward Alleyn but has not enough hard evidence to establish the point since the roles we know of really are dominant in their plays, being the name parts in Doctor Faustus, Tamburlaine, Orlando Furioso, and The Jew of Malta.

    A role may be the longest role in a play without actually being terribly long, so that "Muly Mahamet, although the longest part in Peele's Battle of Alcazar, only speaks around 17 per cent of the text" (p. 39). Using the evidence of backstage plots, Syme is able to determine how many scenes an actor is in but can only guess at how much the actors had to say in their scenes, since the plots do not record dialogue. From the slight evidence of the plots, Syme's major claim is that Burbage and Alleyn would sometimes take fairly minor roles in their companies' plays, letting others take the lead. Scott McMillin claimed that when Alleyn stopped acting in 1597-1600 the Admiral's Men's new plays did not have large main roles, but Syme thinks that too many plays are lost for that to be a secure conclusion and a couple of the ones we can be reasonably sure about -- An Englishman for My Money and Lust's Dominion -- do have long leading roles.

    Burbage and Alleyn taking small roles in some plays would allow them a rest from playing the big roles, and Syme suggests that we should assume that repertory planning took this need into account. Confirming that plays that were star vehicles were rare is Syme's count of what percentage of surviving plays have a lead role that accounts for more than 30% of the text: it is just 8% of plays (33 plays). And these few plays are not concentrated in any one company's repertory at one time: they are spread across the decades. Perhaps unsurprisingly, Syme reports that big lead roles tend to be tragedies while comedies have their parts more evenly distributed, as for ensemble delivery.

    Syme ends by considering the eighteenth-century theatre and showing that David Garrick took minor roles between his major ones and did not perform in all his company's shows. In particular, he avoided playing different major roles on successive nights. If Garrick could not manage it -- and he said that the strain was too much -- then we should not assume that an Alleyn or a Burbage could manage it.

    James J. Marino's contribution to the collection is "Parts and the Playscript: Seven Questions" (pp. 52-67). According to Marino, we have to think of revision, done for instance for provincial touring, in terms of the actors' parts. Any change was possible in theory, but in practice changing the middles of speeches was a lot easier than changing the ends, since (except for the final speech in a scene) the last words of a speech are the next speaker's cue.

    Marino finds a moment in The Contention of York and Lancaster / 2 Henry 6 where the differences between the versions underlying the quarto and Folio editions are particularly hard to explain by the former being based on a memorial reconstruction of the latter because the supposed reconstructor, the actor playing Warwick, "would have to forget whether or not it was his turn to speak" (p. 56). This is because in the quarto one of York's speeches is followed by one of Warwick's, whereas in the Folio York's speech is followed by one of Salisbury's. Marino reckons that the actor playing Warwick would know his cues so well that this is impossible, but I would have thought that in patching together a play from memory anything would be acceptable so long as it made dramatic sense.

     Marino's account of just what the memorial reconstructor must have thought (p. 56) does not entirely make sense to me so I cannot fully understand what his objection is. Marino offer seven questions that he thinks a new model of early modern playtexts would have to incorporate answers to if it were to properly integrate what we know about working from cue-scripts. As far as I can tell, Marino's target is the theory of memorial reconstruction, but since almost no one believes in it any more this seems like a straw man argument. Indeed, Marino's citations of those he disagrees are no more recent than 20 years ago.

    Marino's questions focus on the materiality of the cue-script and the practicalities of making revisions, but throughout he seems to be assuming that to revise a play the cue-scripts would be altered and that the existing actors would have to relearn their parts. For instance he writes that ". . . the changes to 2.2 of 2 Henry VI would require all three actors, York, Warwick, and Salisbury, to learn new cues" (p. 59). But surely revision might take place between productions and involve different actors and possibly the making of new cue-scripts. Marino seems never to entertain this possibility. He imagines play scripts constantly in flux rather than revision being done all at one time.

    Marino overlooks the objection that revision of a licensed play entailed getting the new version relicensed, for a fee, by the Master of the Revels. This is reason enough to suppose that early modern companies were not endlessly revising their plays. Marino is wrong to think that "Early modern plays were . . . divided into semi-independent components amenable to independent revision or replacement" (p. 60); that is not how the state censor saw it.

    Marino assumes that actors were utterly dependent on hearing their cues so that "A player speaking a speech that was no longer in the script could seriously derail a scene. . . . the play might grind to a halt. The only thing worse than not giving the cue is giving a cue no one else on stage can answer" (p. 61). But in fact, we see today that actors do not entirely depend on their cues since they know the general logic of the scene and can improvise around a mistake; professionals do not let the play grind to a halt for want of a cue.

    A distinctive effect that adherents of the memorial reconstruction theory pointed to was that when the Host enters in the Q1 text of The Merry Wives of Windsor the script becomes like the Folio version of the play, and when he exits Q1 drifts away from the Folio. This Marino calls the Mine Host effect. We see the same thing but to a lesser degree in the differences between Q1 and Q2 Romeo and Juliet, in which the Nurse causes the Mine Host effect. Marino suggests that a localized revision to the parts of just Romeo, Mercutio, and Benvolio had been undertaken, so that the entrance of the Nurse "forces a return to the older version of the script" (p. 63).

    In his "Undocumented: Improvisation, Rehearsal, and the Clown" (pp. 68-88), Richard Preiss starts by characterizing the Clown's role and freedom to improvise. He ponders whether and how certain improvised lines got into the written script of a play. Preiss acknowledges that "improvisation circumvented state censorship" (p. 76) since the censor is meant to have read the whole script before agreeing to licence it, but he does not believe that this meant there was necessarily little improvisation happening.

    Preiss thinks that the conversation in Much Ado About Nothing in which Dogberry tells the Watch how to deal with the various disturbances that might occur during their shift (from drunks, thieves, wailing babies and so on) "is clearly an improvisational 'pocket' in the play" (p. 79). He supposes that "Hamlet's advice to the players to 'let . . . your clowns speak no more than is set down' may very well have been a laugh-line" (p. 81), since audiences knew that clowns' lines were mostly not written down. Preiss's promise to explore just how and why improvised lines got into a document that was printed as the play script is not fulfilled by his essay.

    Sonia Massai and Heidi Craig contribute an essay called "Rethinking Prologues and Epilogues on Page and Stage" (pp. 91-110). There are 622 plays in Greg's Bibliography of English Printed Drama and a third of them have prologues and 22% have epilogues. For some reason, the 1590s saw the lowest rates of the use of prologues and epilogues and the 1630s and 1640s the highest. Massai and Craig speculate on some reasons for this, such as ". . . the closure of the theatres ushered in a new era of dramatic self-consciousness, when the rise in printed prologues and epilogues reflected a wider increase in dramatic paratexts" (p. 95).

    I have trouble following Massai and Craig's argument referring to prologues and epilogues on the stage and those on the page, since it is not clear where they are getting the evidence for the former if not from the latter. For instance, I do not know what they mean by "the differences between the rates of inclusion of, and attitudes towards, theatrical and printed paratexts" (p. 94). Inclusion of what in what, exactly? Massai and Craig rightly claim that prologues and epilogues could be written merely to accompany a published edition, but they do not disclose how they detect that this has been done.

    Throughout this essay I have trouble following the authors' logic, such as this about the decline in rates of publishing prologues and epilogues in the 1590s: "Given that this drop is limited to a single ten-year period, it seems unreasonable to assume that dramatically fewer prologues and epilogues were spoken on stage in the 1590s than from 1600 onwards . . ." (p. 97). Why? Is it in principle unlikely for something to happen for just one ten-year period? Massai and Craig's writing would have benefitted from the book's editor's help, since she writes with admirable lucidity on topics such as this.

    On slight evidence from the 1566 morality play The Repentance of Mary Magdalene, Massai and Craig decide that "The drop in the number of prologues and epilogues in the 1590s should therefore be ascribed not to a waning of popularity on stage but to what types of prologues and epilogues were deemed to be popular with readers at different times during the entire period under discussion here" (p. 98). It is not clear why a prologue from 1566 should lead to this conclusion.

    Massai and Craig are of course right that one printer might decide to include prologues and epilogues and another might not, each judging differently these elements' effectiveness in selling books. Thus editions of Francis Beaumont and John Fletcher plays published before Humphrey Moseley's 1647 Folio collection mostly do not include a prologue or epilogue, but plays in the 1647 Folio mostly do. Massai and Craig reason that the earlier plays probably were performed with prologues and epilogues and the difference is just in the rates at which these parts got included in the printing process. Most likely, the 1647 Folio included prologues and epilogues to "reproduce the theatrical experience for reading audiences" (p. 99) since they could no longer get it in the theatre.

    The prologue to Wily Beguiled contains a metatheatrical joke in which the Prologue reads out the stage board title as Spectrum and then a Juggler changes it to read Wily Beguiled. This joke works better in the theatre than in a printed edition, since the latter has a definitive title page to tell us what the play is called. Massai and Craig explore the evidence for prologues and epilogues attached to plays printed in the 1640s perhaps being recycled from other plays or even written just for the purpose of publication. They also explore some moments when a prologue or epilogue being recycled from one play in order to be printed with another that it presumably had not accompanied in performance gave the printer the opportunity to select one that picks up new resonance in the new context, or could be revised to do so.

    Sometimes epilogues and prologues were reprinted in second and subsequent editions of a play, and sometimes not, and sometimes more literary paratexts such as arguments were added. Massai and Craig's wider point is that the paratextual materials took on a life of their own in print regardless of what was happening to the play on the stage: print had become its own medium independent of the theatre.

    To start "Title- and Scene-Boards: The Largest, Shortest Documents" (pp. 111-127), Matthew Steggle lists the plays that involve title boards. Jonson's Cynthia's Revels and Poetaster each has one, since a character reads it aloud, and so does anonymous's Wily Beguiled, and so does Beaumont's The Knight of the Burning Pestle, as does Brome's The City Wit and James Shirley's Rosania. Steggle writes that "Examples could be multiplied, many times over . . ." (p. 112) and I wish he had named them, since I thought that the foregoing list was all of the ones we have from roughly 500 professional plays surviving from the period.

    Steggle adds to the evidence of title and scene boards by looking at the six surviving plays of William Percy, while acknowledging that we have no reason to think that these were ever professionally performed. Steggle notes that Robert Fludd's book Microcosmi Historia (1619) has a picture that might represent an actual early modern theatre and which has a "framed signboard" (p. 118) on which is written "Theatre Orbi" (Latin for Theatre of the World or Theatre of the Zodiac). I would say that this is either a label for the picture rather than part of the picture, or else if it is meant to be a feature of the theatre then it is the theatre's name, not a title or scene board.

    Steggle offers as evidence some Inigo Jones sketches for stage designs for court entertainments, but of course these are not necessarily any use for telling us what happened in the professional theatres. He has more tenuous evidence: the title pages of William Alabaster's play Roxanna and Nathaniel Richards's play Messalina, both of which have a picture of a theatre stage and above it a bordered rectangular panel containing the play title. But in both cases the rectangular panel is not part of the picture of the stage. Steggle considers what effects the use of title and scene boards might have had, including in plays for which we have no reason to think they were used, but in my view this is pointless since he has not established their use outside of certain specialist works.

    Sarah Wall-Randell considers the use of books as properties in "What Is a Staged Book? Books as 'Actors' in the Early Modern English Theatre" (pp. 128-151). Because the early modern stage used no sets, Wall-Randell thinks that "the objects actors hold take on greater prominence" (p. 129). I would have thought that one could argue precisely the opposite: that because there were no sets audiences used different habits of mind about mimesis altogether. Alan Dessen and Leslie Thomson's dictionary of stage directions counts 130 examples of books used on stage from 1580 to 1642, whereas Wiggins, who includes property books implied by dialogue, counts 128 plays using property books from 1580 to 1623. Henslowe's properties list contains no books.

    Wall-Randell fills the empty space in our records with speculation, including that the theatre company kept a small library of books for dramatists to use as sources and actors to use as properties. She asks "Was the Bible that Sir Hugh Evans reads while he waits to duel with Caius in The Merry Wives of Windsor 3.1 the same prop as the poisoned Bible Julia kisses in Act 4, Scene 2 of Webster's Duchess of Malfi, a play first performed by the same company about fifteen years later, in 1612-13?" (p. 133). Even if we knew the answer to this I am not sure we would care: the reuse would have no dramatic significance.

    Wall-Randell wonders what the Qu'ran used in Tamburlaine looked like, and the closest she can get is a wall-hanging at Hardwick Hall in which one is depicted. Wall-Randell follows the bible-kissing scene in Heywood's If You Know Not Me to its source in a pageant for the coronation of Elizabeth in 1559, which itself was inspired by the title page of the 1557 Geneva Bible. The size of the bible held by Elizabeth on stage in If You Know Not Me would indicate what kind of bible it is: a small book would suggest the Geneva Bible, meant for personal devotional use, while a big book would suggest the Bishops' Bible made to be read aloud in church.

    Wall-Randell decides on balance that it has to be a small book because of how it is used, and likewise another couple of books used in the play. From the biblical words quoted from the property play, Wall-Randell tries to figure out which edition it is. But I cannot see why Heywood must be quoting from the property book used: he could be quoting from his favourite bible or from memory, no matter what property book was used.

    Claire M. L. Bourne's contribution to the collection is "Typography After Performance" (pp. 192-215). She points out that when Ferdinand reads Don Armado's letter about catching Costard with Jaquenetta in the first scene of Love's Labour's Lost, the quarto edition prints the words in italics to show the change in register (that is, Ferdinand putting on Don Armado's voice). Bourne reads meaning into the use of punctuation for Costard's interruptions of Ferdinand's reading of the letter, such as "The periods after the king's demands for silence [top of page B1r] signal that, at these moments, Costard is not interrupting" (p. 201), but instead is just adding to what Ferdinand says to make a new sentence, whereas when Costard interrupts Ferdinand the last word Ferdinand says is followed by a colon. That is, colons indicate interruptions.

    Other interruptions of Ferdinand by Costard are indicated with the body of the letter by Costard's speech prefix and short speech appearing within round brackets in the body of the letter, so "(Clow. Mee?)" and "(Clown. Still mee.)" and "(Clow. O mee)", which is something Bourne has found in no other printed playbook. Bourne makes much of the use of a fleuron and pilcrows in the typesetting of quarto Love's Labour's Lost. In manuscript practice, fleurons were used to distinguish commentary from main text and pilcrows to show the start of a new poem.

    Thus ". . . the most abstract of typographic symbols teach readers to hear the poems in different vocal registers and to see their recitation as capsule performances inside a larger theatrical event" (p. 206). Bourne finishes with some examples of the strategic use of italic and roman type to convey to readers the dynamics of performance, including that "Actual actors perform in roman type. Fictive actors -- characters performing within the fiction -- perform in italics" (p. 209).

    Stern's own contribution to her collection is "Shakespeare the Balladmonger?" (pp. 216-237). I am unclear what Stern means when she claims that Shakespeare wrote songs as "'gifts' for the audience of specific plays and that seem to have been intended, at least from 1603, when his company became The King's Men, to be set to music by specific, often royal, composers" (p. 217). In support of this claim Stern cites two pages in the Arden3 The Tempest (where the play's music is discussed, but not these specific claims), and also page 149 in her Documents of Performance in Early Modern England (2009), which page offers no evidence for these suggestions other than that some court composers' settings of King's Men's plays' songs survive.

    Stern draws some tenuous parallels between ballads as a print genre and plays, including their title-page phrasing using words such as lamentable and true and that they relied upon performance for sales (p. 218). The tenuousness of all the connexions is apparent in Stern's phrasing such as "may be", "it seems", and "are we . . . ?" (p. 218). The ballad Get You Hence sold by Autolycus in The Winter's Tale may or may not have been a real ballad, and Stern admits that her evidence base is a "few records in the Stationers' Registers" (p. 219), which give a far from complete picture.

    The most that Stern can show is that at times the meanings of Shakespeare's uses of snatches of ballads are enhanced if the audience knows the full contents of those ballads. Stern assumes that Trinculo's reference to the invisible Ariel (who is making music) as "the picture of Nobody" is a reference to the ballad Nobody's Counsel to Choose a Wife and the play Nobody and Somebody, both published by John Trundle. This may well be, although Trinculo's line is still found to be witty by modern audiences who do not appreciate this topical allusion.

    Early modern plays often ended with a jig, "typically set to one or several ballads" (p. 227); Stern gives no reference for this assertion. There is vague phrasing here in respect of key claims, for instance that ". . . the 'wise fool' Robert Armin, meanwhile, is often linked with the ballad-writers Thomas Deloney, Philip Stubbs and William Elderton" (p. 227). But linked by whom and why? Stern gives a footnote, but it is simply to a page in Nashe's Strange News with no explanation of how what we find there is to be understood.

    Stern writes that the ballad of "Romeo and Juliett" is lost but "was presumably inspired by Shakespeare's the bullard of Romeo and Juliet, in performance at the time" (p. 227). I do not know what Stern means by a bullard here and none of the OED's definitions fits the context. Stern speculates that ballads were used to advertise plays and plays helped sell the ballads but there just is no hard evidence; it is another mere possibility.

    Next Stern turns to "the heart of this chapter, and also its most speculative aspect" (p. 230). There is some slight evidence that ballad-sellers hung around the theatre entrances. Shakespeare's Cleopatra fears that in captivity "scald rhymers [will] / Ballad us out o' tune" and Stern remarks that this "only makes sense if such a ballad is available, ideally where the very squeaking Cleopatra is performing" (p. 232). But I do not see this necessity, since Cleopatra's fear might be demonstrably groundless precisely because the boy actor playing her is not squeaking out these words.

    Likewise I do not share Stern's view that in A Midsummer Night's Dream Bottom's reference to the ballad he will write indicates that the play ended with that ballad. Perhaps it did, but we do not know this. Likewise Falstaff's suggestion that he will make a ballad of his capture of Coleville in 2 Henry 4. Stern does not deny that more or less everything she has suggested here is speculation.

    To end a collection that is marked by its speculative responses to our documentary poverty, Roslyn L. Knutson and David McInnis present an essay called "Lost Documents, Absent Documents, Forged Documents" (pp. 238-259). They argue that there are degrees of lostness because a play was not "a whole entity" (Stern's phrase quoted p. 241) but rather was a patchwork of parts, and some of these could be lost and others not. As an example of a play for which not everything survives, Knutson and McInnis offer Titus Andronicus, for which although we have multiple early editions we do not have "a backstage plot, actors' parts, a playbill, a title or scene board" (p. 242).

    This seems an impossibly high threshold for completeness, since we have all those documents for no play in the entire 2500-year recorded history of drama and indeed we have no surviving examples of any kind of playbill, or title board, or scene board for any early modern play. Sometimes extensive cataloguing just confirms what we know we do not know. The Records of Early English Drama (REED) project has extensively documented regional playing across Britain in Shakespeare's time, but ". . . not a single REED item supplies the title of a London based commercial play that was also performed in the country" (p. 249).

    Outside of whole collections that are relevant to this review, just two book-chapter essays were published this year. In the first, David Fuller argues that King Lear should be presented to readers in a form that conflates the 1608 edition (Q1) with the 1623 Folio (F) edition ("Rescuing Shakespeare: King Lear in its Textual Contexts", in Caruso, ed. The Life of Texts: Evidence in Textual Production, Transmission and Reception, pp. 154-75). On the topic of how we judge large-scale variants between editions, Fuller considers the abdication/deposition episode in Richard 2 and decides that ". . . the play must always have contained the scene . . ." (p. 156) and therefore its absence in the first three editions must be censorship. There is in fact an entirely plausible argument, advanced by David M. Bergeron, that the abdication/deposition episode was added to the play after it was first written.

    After a brief survey of the history of ideas about the early editions from the New Bibliography to the present, Fuller rightly concludes that there are two reasons now given why an editor of a play that exists in multiple early versions must confine herself to making an edition of one or other version rather than combine them: 1) Shakespeare himself revised the play, causing the multiple versions, so merging them would efface his artistic labour, and 2) we do not know why the early editions are so different, so we cannot just merge them (p. 161). He is just not convinced by these arguments.

    Fuller's defence of the conflation of quarto and Folio King Lear is that thereby ". . . the play does not, as F does, lose a major part of one of its most original and significant scenes, the mock trial of Goneril (3.6)" and that ". . . play does not lose other important material, including a confrontation between Goneril and Albany (4.2)". Fuller finds these "vital" and "vivid" and "important" parts of the play (p. 162). He proceeds in this fashion to criticize the revision hypothesis on literary-critical grounds and defends the artistic merit of the bits that Shakespeare is supposed to have cut. Fuller notes that theatre and screen directors too tend to retain from Q1 what is absent from the Folio, particularly the mock trial scene.

    Fuller is dogmatic in his judgements. Where Q1 has "Lear. Who is it that can tell me who I am? | Lear's shadow? I would learn that", he remarks that "Q cannot be right by itself: it has Lear both answer his own question ('Lear's shadow'), and respond to the answer he has given himself ('I would learn that')" (p. 169). But of course this kind of self-directed enquiry and colloquy is not uncommon in Shakespeare, as with Richard 3's "Is there a murderer here? No. Yes, I am" or Hamlet's "Am I a coward? | Who calls me villain . . . ?" Thus Fuller's certainty that Q1 is wrong seems excessive.

    This objection applies across the essay: Fuller's arguments rely on the reader agreeing with him about just what must constitute corruption in one or other early edition. Similarly, Fuller finds Q1's version of the death of Lear impossible, since he first utters the "extended death groan", meaning "O, O, O, O", and then Q1 "keeps him alive for two more lines, so that after Edgar's exclamation, 'He faints', he speaks again to desire his own death -- 'Break, heart, I prithee break'" (p. 172). By contrast, F gives "Break, heart, I prithee break" to Kent as he observes Lear's end. For Fuller this is self-evidently a case of textual corruption in Q1: "Is it really possible to imagine that Shakespeare wrote this line for Lear and in revision reassigned it to Kent? It seems that to sustain their view revision theorists will persuade themselves of anything" (p. 172).

    When Fuller finds an editor doing what he approves of, "it is surely correct" (p. 173). As with the words, Fuller is dogmatic about the verse in Q1, which he sees as irredeemably corrupt, and he treats those who do not agree with him as simply unable to hear the problem: "Anybody to whom this is not a matter of the first importance should not be editing Shakespeare" (pp. 173-174). Of course, those who disagree with Fuller need not be indifferent to verse; they might simply disagree that Q1 is so bad.

    Molly G. Yarn's topic is student editions of Shakespeare ("Katharine Lee Bates and Women's Editions of Shakespeare for Students", in Wayne, ed. Women's Labour and the History of the Book in Early Modern England, pp. 187-203). She finds that Fredson Bowers overstated how bad and unimportant student editions were just because they recycled some out-of-copyright prior texts. Katharine Lee Bates was a late-nineteenh century American poet and professor of English at Wellesley Colleage who wrote the words to the popular patriotic song "America the Beautiful". Education for ordinary women began earlier in the US than in Britain, with female literacy in 1850 reaching 90% there when it was 55% in Britain. US novelists were overwhelmingly women. Bates edited three student editions of Shakespeare: The Merchant of Venice, A Midsummer Night's Dream, and As You Like It.

    Single-play editions were rare until the nineteenth-century British and American education systems provided a market for them: schools were not teaching all the plays and students wanted to buy just the one(s) they were studying. This led to the starting of series such as the Arden Shakespeare in 1899, for which a single general editor oversaw a team, each member of which edited one play. This brought women into editing Shakespeare.

    Bates collated quartos and the Folio for her edition of The Merchant of Venice, but took the view that deciding between competing readings was the work of the student reader not the editor. Indeed, she wrote textual notes that asked the reader to judge for herself, putting "Which is better?" after recording a variant reading. Yarn comments that "Bates thus created a democratic, open text in which her own opinions are subordinated to the goal of enhancing the student's experience" (p. 194). Yarn quotes Mary Cowden Clarke on the desirability of excluding textual notes because they are "mere vehicles for abuse, spite, and arrogance . . . written for the sole purpose . . . of proving that other editors are wrong" (p. 194). Yarn ends with a reflection that academia is "a system that still overvalues the traditionally masculine work of editing" (p. 199).

    So to Notes and Queries. MacDonald P. Jackson published three contributions, in the first of which he offers reasons for thinking that Kyd did not write Arden of Faversham ("How Many Playwrights Wrote Arden of Faversham? The Samuel Rowley Connection", N&Q 265[2020] 223-30). His first aim is to show that the middle of Arden of Faversham is stylistically unlike the rest of the play. Jackson previously noted that the three locutions "tush, seeing as a synonym for 'since' or 'considering that' and 'Ay, but'" (p. 225) are rare in Shakespeare and never appear in Arden of Faversham Scenes 4-9 but occur 41 times in the other scenes of Arden of Faversham.

    To this he now adds and so farewell as something said by an exiting character six times in Arden of Faversham but never in Scenes 4-9. These four locutions occur 31 times in Arden of Faversham and never in Scenes 4-9, and also in The Spanish Tragedy 13 times, Solimon and Perseda 12 times, and Cornelia 0 times. The figures for the new plays that Vickers wants to add to the Kyd canon are: Fair Em 1, King Leir 0, Edward 3 1, and 1 Henry 6 7. So, the new additions are unlike the securely attributed Kyd plays in this regard.

    Jackson returns to some neglected research by Karl P. Wentersdorf that detected a series of phrases preferred by Samuel Rowley and found in Arden of Faversham, including let me alone and I am content . . . for this once. Before Wentersdorf, E. H. C. Oliphant had noted that Rowley's play When You See Me You Know Me has a character called Black Will and he speaks much like the Black will in Arden of Faversham. The parallels are:

I am as safe as in a sanctuary . . . petty robberies . . . the stews from whence I had my quarterage . . . There's not a whore . . . opens shop before I have my weekly tribute (When You See Me You Know Me)

petty robberies I have done . . . I can come unto no sanctuary . . . Take this but for thy quarterage; | Such yearly tribute will I answer thee . . . The bawdy houses have paid me tribute; there durst not a whore set up unless she have agreed with me first for op'ning of her shop windows (Arden of Faversham)

There's not a sword and buckler man in England nor Europe but has a taste of my manhood . . . this is my sconce, my castle, my citadel (When You See Me You Know Me)

a buckler in a skilful hand is as good as a castle; nay 'tis better than a sconce, for I have tried it (Arden of Faversham)

There are some lesser parallels of character likenesses and short and fairly commonplace phrasings. The latter gain significance because although they are common phrases they occur together in clusters in Rowley's writing so that within a few dozen lines several of them will co-occur.

    Jackson has also discovered a new rare phrase shared by Arden of Faversham and Rowley's When You See Me You Know Me and found nowhere else in EEBO-TCP: God's dear lady. Black Will in Arden of Faversham refers to his having fought in Boulogne, captured from the French by Henry 8 in 1544, so this puts Arden of Faversham's Black Will in the same period as When You See Me's Black Will, although the real historical Arden was killed in 1551 after Henry 8 had died. The Rowleyan tags that Jackson has uncovered are rare in the supposed Kyd plays: in all of The Spanish Tragedy, Solimon and Perseda, Cornelia, Fair Em, King Leir, Edward 3, and 1 Henry 6 there are 28 instances of these Rowley tags, compared to 33 in just Arden of Faversham alone. Jackson does not think he has uncovered enough evidence to say that Samuel Rowley wrote the non-Shakespeare parts of Arden of Faversham, but he does think it is pretty damning against the theory that Kyd did.

    Jackson's second contribution to Notes and Queries shows that there really is something bad about bad quartos ("'Omnes' Stage Directions in Three Suspect Shakespeare Texts", N&Q 265[2020] 237-9). There are 27 examples of the error Exit omnes (where Exeunt omnes would be the correct Latin) in Q1 Henry 5 and Q1 The Merry Wives of Windsor, and none elsewhere in Shakespeare except for a single case in Folio Antony and Cleopatra. Jackson looked for this mistake in EEBO-TCP texts of plays, and found 41 instances in 11 non-Shakespearean plays, and of these 41 a whole 33 come from just 7 play texts listed as suspect in Laurie Maguire's Shakespearean Suspect Texts: "twelve in 1 Hieronimo (1605), nine in Thomas Lord Cromwell (1602), six in The Fair Maid of Bristow (1605), three in The True Tragedy of Richard the Third (1594), and one in each of Sir Thomas Wyatt (1607), A Knack to Know an Honest Man (1596), and The Taming of a Shrew (1594)" (p. 238).

    Across the 1623 Folio there are 28 instances of Exeunt omnes and, aside from the 1595 octavo of Richard Duke of York that has 17, in the Shakespeare quartos there are just 8, so Shakespeare does not seem to have been particularly fond of omnes at all. Thus it looks like the high numbers of Exit omnes in Q1 Henry 5 and Q1 The Merry Wives of Windsor and the high number of Exeunt omnes in the octavo of Richard Duke of York are due to abnormal transmission rather than their being early drafts of the plays in question.

    Jackson's third note concerns Albert C. Yang's essay reviewed above ("The Enlarged Kyd Canon: Comments on a Recent Study", N&Q 265[2020] 218-20). Jackson points out that Yang tested six supposed Kyd plays again the 36 Shakespeare Folio plays and found them falling into two groups. This does not mean that the non-Shakespeare group (the supposed Kyd group) are alike nor that they are all by Kyd. To conclude that, Yang would have to include plays by other authors and find that the supposed Kyd plays stick together when tested against such a broad field. Jackson also complains that Yang's tables of numbers do not match his dendrograms.

    Darren Freebury-Jones's concern is the data from Pervez Rizvi's Collocations and N-Grams dataset for the plays that share most phrases with the six plays that the new Vickers and Freebury-Jones edition attributes to Kyd: The Spanish Tragedy, Solimon and Perseda, Cornelia, King Leir, Arden of Faversham, and Fair Em ("Unique Phrases and the Canon of Thomas Kyd", N&Q 265[2020] 220-3). Freebury-Jones glosses over the detail of the weighting formula by which Rizvi aims to adjust for the fact that different authors have differently sized canons: ". . . the raw figures [for phrasal matches] are divided by composite word counts" (p. 220). But what is put together to make that composite? Freebury-Jones does not say.

    Freebury-Jones recounts an experiment that Rizvi undertook, giving no citation for where we can read it other than that it is on Rizvi's website. Freebury-Jones does his own reslicing of Rizvi's data, but his description of his methodology is barely comprehensible. From what I can make of it, Freebury-Jones appears to think that if amongst the list of 12 early modern plays with the most phrasal matches to a putative Kyd play there is one other putative Kyd play, then this shows shared authorship. He maintains this even when the two plays -- that is the one being tested for likeness to others and the one among the top-12 plays found -- are both ones he is trying to add to the accepted Kyd canon.

    Freebury-Jones thus finds when trying to attribute Arden of Faversham this way that among the top 12 plays with the most shared phrasal matches is Fair Em, and for him this counts as evidence supporting his attribution of Arden of Faversham and Fair Em to Kyd because also among that top 12 plays matching Arden of Faversham is Kyd's Solimon and Perseda. There are multiple ways to count phrasal matches, since we might count two plays sharing the phrase hear a tongue shriller as one match of a four-word phrase or as two matches of three-word phrases: one match for hear a tongue and one match for a tongue shriller.

    This enables Freebury-Jones to reslice all the data multiple ways, looking for outcomes in which his Kyd attributions seem to be confirmed. In all this data-slicing, Freebury-Jones considers that he has a confirming result so long as a Kyd play occurs within the top 12 rank order of matching plays, which is an exceptionally low threshold. And even with this low threshold, Freebury-Jones finds that King Leir still does not test like Kyd.

    Freebury-Jones believes that Kyd wrote all of Arden of Faversham and he sets out to counter the idea that Shakespeare wrote Act 3 (Scenes 4 to 9) by seeing what proportion of the phrasal matches that he finds between Arden of Faversham and Kyd plays fall into the one-fifth of Arden of Faversham that other people think is by Shakespeare. Freebury-Jones comments: "As the supposed Shakespeare scenes amount to around one-fifth of the whole play, a percentage of 20 for total unique matches [with Kyd plays] co-occurring with Scenes 4 to 9 would indicate an even distribution of Kyd matches in Arden of Faversham" (p. 223).

    In fact Freebury-Jones does not get 20% of his matches with Kyd falling within Act 3. Instead he gets "The Spanish Tragedy 50 per cent; Cornelia 33 per cent; King Leir 29 per cent; Fair Em 25 per cent; [and] Solimon and Perseda 14 per cent" when counting phrases one way and "Cornelia 50 per cent; King Leir 33 per cent; Fair Em 22 per cent; The Spanish Tragedy 20 per cent; [and] Solimon and Perseda 9 per cent" when counting phrases another way.

    An obvious problem with these results is that plays that we all agree are by Kyd are at opposite ends of the spectrum. In both lists, Kyd is at the top and bottom of the list: The Spanish Tragedy and Solimon and Perseda in the first list and Cornelia and Solimon and Perseda in the second list. This occurs because the percentages are much higher than expected for one Kyd play (50% for The Spanish Tragedy and Cornelia) and much lower than expected for another Kyd play (14% and 9% for Solimon and Perseda).

    The experiment then is a failure. That is, Freebury-Jones's hypothesis is that Kyd wrote Act 3 of Arden of Faversham just as he wrote the rest of it, and thus he predicts that about 20% of the matches of Kyd plays with Arden of Faversham would fall within the 20% of the play that is Act 3. But his results do not show this: known Kyd plays The Spanish Tragedy and Cornelia match to Act 3 of Arden of Faversham more than twice as often as they should (50% against 20%) and the known Kyd play Solimon and Perseda matches less than half as often as it should (9% against 20%).

    Despite getting these results, Freebury-Jones declares the experiment an almost unqualified success: "We can see that in every instance except results for Solimon and Perseda, the scenes Jackson would excise from the rest of Arden of Faversham appear to be no less Kydian than the remainder of the play" (p. 223). A more reasonable conclusion is that since the results do not match the prediction he made for them, his method is not detecting authorship at all.

    Two more contributions to Notes and Queries were concerned with co-authorship attribution, both by Thomas Merriam. His study of Henry 8 argues that we can divide its scene 4.2 into Shakespearian and Fletcherian parts ("One or Two Katherines in Henry VIII?", N&Q 265[2020] 267-71). Changes in the character of Queen Katherine across Henry 8 are usually attributed to her responding to what happens to her, but Merriam thinks they can also be explained by the two co-authors, Shakespeare and Fletcher, thinking somewhat differently about her. Scene 4.2 is particularly interesting since Katherine shows a sharp contrast in her behaviour in it: the "apotheosis of Katherine in the dream vision (IV.ii.82) and her petulant and self-pitying decline (IV.ii.83-173)" (p. 268).

    Merriam divided scene 4.2 into 25-word segments and then counted in each the number of occurrences of the known Shakespeare-favoured words being, did, doth, each, hath, hence, in, itself, might, of, rather, that, the, to, very, which, who, and with, and of the known Fletcher-favoured words again, all, are, can, dare, do, may, more, must, now, only, still, sure, there, these, too, ye, and yet. These are preferences detected by Hugh Craig.

    Across scene 4.2 Merriam calculated the result of dividing the count for the Fletcher set by the total count for both sets and then produced a Cumulative Sum chart in which the line descends in those parts of the scene where Shakespeare's style predominates and ascends in those parts of the scene where Fletcher's style predominates. The graphs shows a descent (hence Fletcher's writing) up to line 82 and then an ascent (hence Shakespeare's writing) from there to the end of the scene. Merriam works through the two halves of the scene, explaining just how the treatment of Katherine is Shakespearian in the first and Fletcherian in the second.

    Merriam's second note finds further evidence for his belief that Shakespeare's Henry 5 was co-authored ("Verse and Prose in Henry V Continued", N&Q 265[2020] 239-41). In a note reviewed in NYWES for 2017, Merriam showed that the verse and prose of Henry 5 test differently for verbal likeness with 1 Tamburlaine, 2 Tamburlaine, Hamlet, King Lear, Othello, Romeo and Juliet, and The Winter's Tale. The result was that the verse of Henry 5 is like Marlowe's work and the prose parts are like Shakespeare's work. (Merriam does not mention here how he did the testing back then: but it was by counting most-frequent word bigrams.)

    In this new note, Merriam separates the prose and verse parts of 31 Shakespeare plays, including Henry 5, plus the undivided texts of 1 Tamburlaine and 2 Tamburlaine, counting the 119 most-frequent single words. For each play the 119 data points are then reduced to two by Principal Component Analysis and plotted. The result shows the verse of Henry 5 falling near to the two Marlowe plays and far from all the Shakespeare verse and prose segments.

    Débora Cibele de Benedetto E. Silva and Régis Augustus Bars Closel find that Peele's play David and Bathsheba is a source for Shakespeare and Peele's Titus Andronicus ("Two New Sources for Titus Andronicus", N&Q 265[2020] 230-2). They reckon that before co-writing Titus Andronicus, Peele wrote David and Bathsheba, which has some similar character names such as Thamar (like Tamora) and Amnon (like Aaron) and similar themes of lust, rape, and revenge. (We do not actually know that David and Bathsheba came first: the possible dates of first production are 1584-94 for both plays, according to Wiggins's Catalogue.)

    David and Bathsheba has a scene in which a would-be rapist is told to take the victim to another place to rape her, as Aaron tells Chiron and Demetrius about raping Lavinia, and this is not in the source Ovid. (But perhaps this is just because plays have to take a rape off stage, both for decency and realism.) Silva and Closel cite some verbal parallels too, but they strike me as weak: "'Now Thamar' vs. 'So Tamora'; [and] 'Now, Amnon, loose those loving knots of blood' vs. 'Then, Aaron, arm thy heart and fit thy thoughts'" (p. 231).

    Another parallel is that "Both Thamar . . . and Lavinia . . . meet a family member after being ravished" (p. 231), whereas in Ovid ". . . Philomela is imprisoned in the hut where Tereus attacked her" (p. 231). Thus the rape scene in David and Bathsheba is a source for the rape scene in Titus Andronicus. Silva and Closel find some lesser plot parallels that link the revengeful banquet scene in Titus Andronicus with Peele's earlier play too, and even weaker ones for other scenes.

    Brett Jones has found something that he admits might be only a coincidence ("A Curious Synchronicity of Dates: How a Hypothetical Reconstruction of the Stage History of Two of Shakespeare's Comedies Mirrors the Performance History of Lost Plays Recorded in Philip Henslowe's Diary", N&Q 265[2020] 232-4). He starts by claiming that the 1598 quarto of Love's Labour's Lost was "the first play quarto to identify Shakespeare as the playwright on its title page" (p. 232). This is more than we know, since the same year editions of Richard 2, and Richard 3 also did this and we do not know which came first.

    On the assumption that Love's Labour's Lost followed closely after The Comedy of Errors, and got its masque of the Muscovites from the Russian ambassadors' visit in the Gesta Grayorum, Jones lays out a hypothetical stage history in which the Lord Chamberlain's Men were performing The Comedy of Errors in late 1594 and Love's Labour's Lost in early 1595 and Love's Labour's Lost was revised and revived between 1595 and 1597, hence its 1598 edition's title page's reference to its being "Newly corrected and augmented".

    Weirdly, this matches what was happening with the Lord Admiral's Men over at the Rose. Henslowe's Diary shows that "the greasyon comody" played there from November 1594 to May 1595 and "the frenshe Comodey" played there  (as "ne") in February 1595. Something called "a frenshe comody" was then played (again as "ne") at the Rose in April 1597, so maybe this was the revised version that the Love's Labour's Lost title page refers to. Perhaps both playhouses (Theatre and Rose) had plays on the same themes at the same time, since we know that in other ways they imitated each other's repertories.

    James H. Runsdorf looks at the usually claimed sources for Shakespeare's Sonnet 18, which begins "Shall I compare thee to a summer's day?" ("'Shall I Compare Thee':? Sonnet 18 and Robert Greene's Menaphon", N&Q 265[2020] 234-7). Finding them not close enough, he offers a bit from Greene's Menaphon:

Shall I compare her forme unto the spheare
Whence Sun-bright Venus vaunts her silver shine?
Ah more than that by just compare is thine,
Whose Christall lookes the cloudie heavens doe cleare

Runsdorf then offers some lesser parallels between Greene's Menaphon and Shakespeare's writing.

    Lukas Erne has news about the text and performance of Hamlet ("A Newly Discovered Copy of the Fifth Quarto of Hamlet (1637,) with a Performance Record of Hamlet in 1664", N&Q 265[2020] 243-4). His first sentence captures the essence of it: "A hitherto unknown copy of the fifth quarto of Hamlet (1637) contains information about a performance of the play in 1664". The book was found at the Herzog August Library in Wolfenbüttel, Lower Saxony. Its title page has a handwritten annotation "Acted a[t] the the / atre in Lincolns / Inn- Fielde the 4 / of 9ber. 1664", where "9ber" means November.

    Matthew Baynham notes that in the first scene of Othello, Iago's "I am not what I am" blasphemously inverts God's "I am that I am" spoken to Moses in Exodus 3.14 ("Iago and the Denial of Peter", N&Q 265[2020] 246-8). Later, Othello calls to Emilia "You, mistress, | That have the office opposite to St Peter | And keeps the gate of hell". Iago likens himself to Saint Peter by saying to Othello "My Lord, you know I love you", which echoes Peter's thrice-stated "thou knowest that I love thee" (John 21.15-17). Baynham connects these three parts of Othello via Peter's denial of his discipleship: "Then saith the damsel that kept the door unto Peter, 'Art not thou also one of this man's disciples?' He saith, 'I am not'" (John 18.17).

    Tom Bishop finds a source for the moment near the end of King Lear when Edgar leaves his father under a tree to go and fight on the side of Lear and Cordelia against the British ("A Virgilian Echo in King Lear V.ii?", N&Q 265[2020] 248-50). He soon returns with the news of his side's defeat and has to cajole his father to fly with him. Bishop reckons that this is influenced by Aeneas's trying to get his father Anchises out of burning Troy in Virgil's Aeneid. The parallel is strengthened by Anchises, like Gloucester, being blind, as he is in the commentary on Aeneid by Servius the Grammarian of the late medieval period. Bishop explores some ways in which Edgar and Aeneas are alike in overcoming tragedy and in founding a new line of monarchs and a new state.

    Addressing a famous crux, John Jowett reckons that in Coriolanus Martius should say not "wolvish tongue" but "moonish tongue" ("'Wooluish Tongue' and Shakespeare's Togas", N&Q 265[2020] 251-5). It certainly is odd for Martius to say that he stands "in Wooluish tongue" to beg the citizens' approval. The word tongue is often emended to toge, referring to the cloak of humility that he wears, but Jowett thinks the problem is Wooluish. In fact, the word toge was virtually unknown in the English of Shakespeare's time and toga was regarded as a Latin word. In his four Roman plays, Shakespeare does not use the word toga.

    The 1622 quarto (Q1) of Othello refers to "the toged Consuls" where F reads "the Tongued Consuls". In the context, which is Iago disdaining those who can talk only theoretically of war and have not experienced it in practice, tongued is the better reading. Jowett notes that for the New Oxford Shakespeare's edition of Othello, Gary Taylor adopted togaed, based on Q1's toged, on the grounds that tongued cannot be right since Shakespeare elsewhere always precedes this word with an adjective such as shrill or long. But Jowett points out that elsewhere in Othello Shakespeare uses hearted without a preceding modifier although everywhere else he always gives it a preceding modifier, so this forms a precedent for his using tongued without a modifier.

    In Othello, Shakespeare appears to have used signory (an Anglicization of the Italian signoria in his source) interchangeably with Consuls to mean the Venetian ruling assembly, a talking shop, hence "My services which I have done the signory | Shall out-tongue his complaints". This makes tongued Consuls more likely than togaed Consuls. How could tongued get accidentally turned into toged? It is possible for a macron to be overlooked, turning tōged (an abbreviation of tongued) into toged but it is hard to go the other way, imagining a macron where none exists.

    In context in Coriolanus and Othello, "duplicitous spoken language" (p. 253) is the point, so in both the word tongue[d] is the more likely correct reading. Indeed, Coriolanus is full of references to the human tongue, and if we suppose that this word is correct within Wooluish tongue the problem to be solved is Wooluish. Jowett suggests that this is a mistake for Moonish, since M > W and n > u errors are easily made, and the l could come from "An intrusive descender from a letter in the previous line or an accidental pen-stroke or a mis-formed letter" (p. 254). Shakespeare used moonish once elsewhere, in the moonish youth that Rosaland says she will impersonate in As You Like It. This context of deceptive impersonation matches that in Coriolanus, where Martius is trying to get the citizens' approval.

    John Klause reckons that in writing Cymbeline, Shakespeare was influenced by the Jesuit Robert Southwell ("Robert Southwell in Cymbeline", N&Q 265[2020] 255-9). He finds a series of borrowings from Southwell's writing in Cymbeline, including the Soothsayer calling the Roman emperor "our princely eagle" as Southwell did the Blessed Virgin in his poem "On the Assumption of Our Lady", and these are the only two uses of this three-word phrase in EEBO-TCP.

    This is the strongest connexion Klause has found and the others are all fairly commonplace, such as Cymbeline's "Mount, eagle" being like the Geneva bible's "eagle mounts" (Job 39.30), and he gives a list of parallels between Cymbeline and Southwell's Epistle of Comfort. Some items on this list are close parallels, including "As chimney-sweepers, come to dust" in Cymbeline being like "the dust of our grave . . . the chimney of our flesh" in Southwell, but others are tenuous, such as Cymbeline's "No exorciser harm thee" being like Southwell's "drive the devils out of the possessed".

    Both works use the name Cadwal, both have a beheading in Wales, and both feature mountaineers. The significance of Klause's finds of this kind is hard to determine, since on their own many of the parallels are loose enough to be coincidental and we do not have the means to quantify how many loose parallels we should expect to find between Cymbeline and Southwell's poetry if Shakespeare had not in fact read and been influenced by him.

    Klause ends with two striking unique parallels. The first is "the fits and stirs of 's mind" in Cymbeline with "the fits and corrosives of the mind" in Southwell's Epistle of Comfort. Klause claims that in the EEBO-TCP dataset these two are the only occurrences of the fits and followed by of and mind although he does not state how many words are allowed between the three key terms. I agree, finding that even allowing up to 14 words in between the terms these are the only two texts that are hit; at 15 words a text from 1696 also matches. Klause's second claimed unique parallel is workmanship followed by and value and for this I find a third text beyond Cymbeline and Southwell: Nicholas French's The Unkind Deserter of Loyal Men and True Friends of 1676 also has workmanship and value. Klause is of course entitled to ignore this much later text as irrelevant to the relationship between Southwell's and Shakespeare's writings.

    Perry McPartland has two successive notes this year. In the first he argues that the roles of Mamillius and Perdita were written for doubling in the first performances of The Winter's Tale ("'This Seeming Lady and Her Brother': Further Remarks on the Doubling of Perdita with Mamillius in The Winter's Tale", N&Q 265[2020] 259-63). Just before Florizel and Perdita arrive in Leontes's court, Paulina painfully reminds Leontes of the loss of Mamillius. At this point, bringing Mamillius's likeness back in the form of the character of Perdita anticipates and matches the bringing back of Hermione later. The word print for what a parent does in creating a child is used by Paulina when showing the newborn Perdita to Leontes before he banishes her -- "Although the print be little, the whole matter | And copy of the father" -- and is used by Leontes when Florizel and Perdita turn up in his court as Leontes says to Florizel "Your mother was most true to wedlock, Prince, | For she did print your royal father off".

    McPartland makes some literary-critical arguments for the recognition scene of Florizel and Perdita in Leontes's court being enhanced by the Mamillius-Perdita doubling. The Lord who bursts in to bring the news of Leontes's arrival in Sicilia in pursuit of his son refers to Perdita as "this seeming lady" and McPartland suggests that as well as suggesting that she only seems aristocratic he is implicitly suggesting that she only seems female, being in fact the player of Mamillius and being male.

    McPartland's second note is on the same play ("'Quit Presently the Chapel': A Note on Setting in the Final Scene of The Winter's Tale", N&Q 265[2020] 263-5). The location for the final scene of The Winter's Tale is introduced in passing, with Paulina's instruction to "Quit presently the chapel" to those who do not want to see the miracle of the statue of Hermione moving. The only location mentioned before was the gallery that the party had passed through and which the play's audience might have thought they were still in. The belated revealing of the scene's location produces a strong dramatic effect, since we suddenly have to revaluate the things that turn out to be Catholic gestures, such as having works of art in a church and Perdita kneeling to receive a blessing. And when the statue seems to move, this must have seemed like the hoaxes of Catholic statues miraculously moving.

    Lastly for this review, Daniel Kaczyński sets out to explain a puzzling exchange in the Folio text of The Tempest ("Does 'The Watch of [Gonzalo's] Wit' Strike 'One' or 'On' (The Tempest II.i.13-16)?", N&Q 265[2020] 265-7). Sebastian and Antonio mock their older companion Gonzago: 

   Seb. Looke, hee's winding vp the watch of his wit,
By and by it will strike.
   Gon. Sir.
   Seb. One: Tell.

The word tell has been glossed as meaning that Sebastian is telling Antonio to count the chimes that he is marking with, first of all, One. But Kaczynski objects that tell never means that: we tell the time by counting all the strokes. It has been suggested that tell is a stage direction that got mistaken for dialogue, but as a stage direction this one would be anomalously cryptic. Kaczynski thinks the likelier explanation is that one is a spelling of on, which makes sense if Sebastian is telling Gonzalo to continue (as in go on) or is remarking to Antonio that Gonzalo has (with the word Sir) gone on. Kaczynski acknowledges that he was anticipated in this explanations by the editions of Nicholas Rowe, Alexander Pope, and Thomas Hanmer emending One > On. He might have mentioned that they were all preceded by the Second, Third, and Fourth Folios doing this.

Books reviewed

Caruso, Carlo, ed. The Life of Texts: Evidence in Textual Production, Transmission and Reception. Bloomsbury [2019]. 272 pp., £99.19, ISBN 978-1350039056

Leonard, Alice Error in Shakespeare: Shakespeare in Error, Palgrave Shakespeare Studies. Palgrave Macmillan [2020]. 216pp., £69.99, ISBN 978-3030351793

Loughnane, Rory and Andrew J. Power, eds. Early Shakespeare, 1588-1594. Cambridge University Press [2020]. 336 pp., £81.99, ISBN 978-1108495240

Shakespeare, William King Lear, ed. Richard Knowles, The New Variorum Edition of Shakespeare. The Modern Language Association of America [2020]. 2000 pp., £211.85, ISBN 978-1603290913

Shakespeare, William Measure for Measure, ed. A. R. Braunmuller and Robert N. Watson, The Arden Shakespeare. Bloomsbury [2020]. 440 pp., £10.11, ISBN 978-1904271437

Shakespeare, William The First Quarto of The Merry Wives of Windsor, ed. David Lindley, The New Cambridge Shakespeare: The Early Quartos. Cambridge University Press [2020]. 92 pp., £73, ISBN 978-1107358492

Stern, Tiffany, ed. Rethinking Theatrical Documents in Shakespeare's England, The Arden Shakespeare. Bloomsbury [2020]. 304 pp., £21.15, ISBN 978-1350051348

Wayne, Valerie, ed. Women's Labour and the History of the Book in Early Modern England, The Arden Shakespeare. Bloomsbury [2020]. 336 pp., £22.39, ISBN 978-1350246638