"Shakespeare: Editions and Textual Studies in 2015" Not the Year's Work in English Studies

One major critical edition of Shakespeare appeared this year: Sandra Clark and Pamela Mason's Macbeth for the Arden Shakespeare Third Series. Their Introduction is divided into six thematically organized sections--on the play as a Tragedy, its Contexts and Language, the theme of Time, the play's Sources, and the play on the Stage--and they largely confine the discussion of textual matters to their Appendix One. Clark and Mason evidently do not accept Gary Taylor's reconstruction of the original ordering of the scenes, for they see Macbeth reacting to the "show of kings" with "murderous rage, which he vents on the family of Macduff" (p. 6). Taylor, in the Companion to the 2007 Oxford Collected Middleton, gave good reasons for supposing that this murderous threat to the Macduff family originally belonged much earlier in the scene, as a reaction to the witches (whom Clark and Mason consistently call "Sisters") telling him to "beware Macduff".

    Clark and Mason accept the traditional date of composition of 1606, noting that the reference to "twofold balls and treble sceptres" must indicate the union of Scotland and England of 1603. A strong allusion to Macbeth appears in Francis Beaumont's Knight of the Burning Pestle which is usually dated 1607, and Clark and Mason accept claims of allusions to Macbeth in Thomas Tomkis's play Lingua (Stationers' Register entry 23 February 1607) and John Marston's Sophonisba (Stationers' Register entry 17 March 1606) but in the latter case they are not sure who borrowed from whom. If Macbeth is a Gunpowder Plot play, then they reckon it was composed after May 1606 when the last of the conspirators, Father Henry Garnet (alias Farmer) was executed, after which the term equivocation, used by Garnet, became current; the Porter of course alludes to a hanged "farmer". Despite some speculations, we do not know that Macbeth was first performed for any particular royal occasion and the first record is Simon Forman's in 1611. Other, less-strong, possible allusions to Macbeth in performance are considered and dismissed.

    Clark and Mason are not convinced that Shakespeare wrote Macbeth to please the new king: it can be read as highly ambivalent about topics on which James I had strong views. Other plays on Scottish history were performed shortly before this one and Shakespeare had Scottish characters in his other plays; King David of the Scots is an unattractive figure in Edward 3 and the Scots are rebels in the Henry 4 plays and spoken ill of in Henry 5. Clark and Mason briefly survey the anti-Scottishness evident in other plays produced shortly after 1603, including Ben Jonson, George Chapman, and John Marston's Eastward Ho!, Thomas Middleton's Michaelmas Term, and John Day's The Isle of Gulls. Clark and Mason trace the various historical accounts of Macbeth and consider the possibility that Shakespeare might have known them.

    At the start of the play, Shakespeare evokes Scotland with a few geographically and culturally specific terms, using none of the popular stereotypes; we do not know what the actors wore. It is a violent and desolate opening, however. Clark and Mason consider the mix of pagan warrior-ideal and Christianity in the personality of Macbeth and read his reference to a former time when if "the brains were out, the man would die" (3.4.77) as his nostalgia for "a time when killing had no repercussions" (p. 31). I would see rather more comic potential than philosophical reflection in this moment. An astute private observation from Richard Proudfoot is here repeated, that in the late Shakespeare there are quite a few "Attackers of their own country" like Malcolm in Macbeth: Cordelia in King Lear, Antony in Antony and Cleopatra, Timon in Timon of Athens, Coriolanus in Coriolanus, and Posthumus in The Winter's Tale (p. 32n1). The infusion of good Englishness into savage Scotland is made explicit at the end with Malcolm's introduction of the rank of Earl for the thanes that supported him. On the other hand, the sense at the end that everyhing is about to start again--with Donalbain overthrowing his brother Malcolm as Raphael Holinshed records--undercuts a reading in which the play is thought to flatter King James I.

    Clark and Mason note that medieval Scotland was an elective monarchy in which succession was often sideways (between brothers or cousins or uncle/nephew) but was turning to primogeniture in Duncan's time and that the play dramatizes the conflict between the two forms in the naming of Malcolm as Prince of Cumberland (the title of the heir apparent), which Macbeth treats as an additional impediment to his aims ("a step | On which I must fall down, or else o'er-leap", 1.4.48-49). Having thwarted primogeniture himself, Macbeth is then tormented by his own failure to pass the throne to a son. In the sources, Banquo helps Macbeth to seize the throne, but Shakespeare makes him blameless, presumably because James I considered Banquo his ancestor and the founder of the Stuarts' house. But this ancestor's child Fleance does not end up on the Scottish throne in Shakespeare's play.

    Turning to language, Clark and Mason credit Macbeth with introducing a series of expressions that are still popular, including "the milk of human kindness" (1.5.17) and "all our yesterdays" (5.5.21), and also with coining new words, some of which, however, are not Shakespeare's. He did not coin yeasty, unbend, unspeak, dauntless, darefull, and marrowless, all of which can be found in earlier books within the database Early English Books Online Text Creation Partnership (EEBO-TCP). For some reason, Clark and Mason report that Macbeth contains only "765 lines" (p. 44) but the true figure is over 2,000 (varying slightly depending on how we count them) and they think that Richard III has only "1,224 lines" when the true value is over 3,500. Clark and Mason note that as Macbeth gets sucked into the witches' view of the world he increasingly uses repetitions, rhyme and alliteration as they do. Moreover, the Macbeths' growing use of circumlocations--calling their planned murder "the deed" (1.7.14) and "this business" (1.7.31)--establishes a pattern of evasion that makes them suitable dupes of the witches' equivocation.

    In their discussion of time in the play, Clark and Mason note that everything rushes onwards towards the end of 4.1 in which Macbeth finds out that in the long run all his dashing about comes to nothing (although there is one more vindictive murder of the Macduffs in 4.2), and then comes the long, reflective scene 4.3 when we shift to Malcolm in England. Again, I would have thought that some consideration of the effects attributable to Middleton's adaptation would be relevant to this consideration of the dramatic pace. The final act moves repeatedly between the orderly preparations for Macbeth's deposition and his own frantic yearnings for it all to be over, including putting on his armour too early, and his pointless, anxious bustle. "Even his wife's death is outside of temporal order: 'She should have died hereafter'" (p. 69). 

    So why, Clark and Mason ask, do so many of the play's timings seem off, such as Macbeth's being said to know of Macduff's flight in 3.6 and then reacting with surprise when he learns of it in 4.1? And why does Macbeth boast of keeping spies in his thanes' homes in 3.4, yet he does not seem to have been on the throne long enough to achieve this level of surveillance? And how come Macbeth has beaten the Thane of Cawdor by 1.2 and yet in 1.3 says that the Thane of Cawdor is a "prosperous gentleman" (1.3.73)? Clark and Mason do not accept the explanation that adaptation of the play has created these contradictions. They speculate why Macbeth says, after finding that he is now not only Thane of Glamis but also Thane of Cawdor, "The greatest is behind", when he should mean that the greatest (being king) is yet to come (p. 72). I would have thought the answer is simply arithmetical: he means that two out of three--that is, the greater part of three--are now behind him. Clark and Mason make the point that Macbeth seems to want to have it both ways: to believe in a predestined future and to believe in his power to create the future he wants to enjoy. They do not, however, trace this contradiction to its source in the paradoxical nature of the witches, who make self-fulfilling prophecies in which the act of telling them to Macbeth as predictions creates the precondition--the stimulation of his ambition--that is sufficient for their own fulfilment.

    Amongst the play's sources, Holinshed's Chronicles is paramount, but Clark and Mason note that Shakespeare did not confine himself to the bits about Macbeth: he drew on its accounts of other Scottish kings for certain details. Clark and Mason list what Shakespeare takes from Holinshed and where he departs from this source. In Holinshed, Macduff goes to England to revenge the murder of his family, while Shakespeare has him there when he learns of it and thus has him feel guilty about leaving his family unprotected. In Holinshed, Duncan is an inept and unwarlike king who cannot satisfactorily rule his kingdom, but this is never suggested by Shakespeare. In Holinshed, Macbeth has a reasonable expectation as Duncan's cousin of being named as his heir, but in Shakespeare he does not. (This does not stop directors from importing this detail into their productions: for the Royal Shakespeare Company in 2004, Dominic Cooke had Macbeth (Greg Hicks) step forward expectantly as Duncan named his heir, as though he thought it would be himself not Malcolm.) Most importantly, in Holinshed Macbeth heads a faction (including Banquo) that deposes Duncan because he is incompetent and then Macbeth rules well for 10 years before turning into a tyrant. Some lesser sources are also considered by Clark and Mason, including George Buchanan's Latin work Rerum Scoticarum Historia (1582) and Seneca's tragedies generally.

    The performance history for Macbeth starts with Simon Forman's eyewitness account of the play at the Globe in 1611. Rather than give a full chronological survey, Clark and Mason offer "a selection of themes and topics that have proved significant in productions in England over a long period, up to the present day" (p. 97). From William Davenant's adaptation in the 1660s to David Garrick's first performance as Macbeth in 1744, and on to the mid-nineteenth, there was much adding of spectacle involving witches, who frequently sang. The witches were typically performed comically by male actors. From the early twentieth century it was generally accepted that the Hecate scenes are inauthentic and they were often omitted. More recently, the witches are given extra things to do, being silent presences throughout the play or taking part in such 'presentations' as the appearance of Banquo's ghost. Clark and Mason reflect that no actor has really made the title part his own.

    David Garrick famously added a dying speech for Macbeth, and a lot of actors who followed him--John Philip Kemble, William Macready, John Gielgud, Simon Russell Beale--also tried to add to the audience's sympathy for the character; but equally a number of other actors--Charles Macklin, Edmund Kean, Henry Irving, Donald Wolfit--reacted against this interpretation and played up the villainy. That Garrick felt he needed to add a dying speech for Macbeth (which John Philip Kemble and Edmund Kean then used) shows dissatisfaction with the ending that Shakespeare provided, and modern directors still cut down the attention given to the invading army near the end in order to sharpen the focus on the title character. There is a definite textual problem around Macbeth's death, which may with textual warrant be performed on stage or off, and also with the offstage decapitation. Just how satisfying a closure the ending  provides varies from production to production, but modern ones tend to end pessimistically. The visual record of the play starts with Nicholas Rowe's edition of 1709 depicting Thomas Betterton in modern dress in the title role, which is also how Garrick was dressed. Charles Macklin in 1773 tried to recreate the look of medieval Scotland. The Introduction ends abruptly here with no attempt to conclude anything.

    And so to Clark and Mason's text and its supporting apparatus. The most obvious problem is in the form of the textual notes, which do not follow the layout described for them in the General Editors' Preface. There it is written that a textual note "comprises, first: line reference, reading adopted in the text and closing square bracket; then: abbreviated reference, in italic, to the earliest edition to adopt the accepted reading, italic semicolon and noteworthy alternative reading(s), each with abbreviated italic reference to its source" (p. xiv). In this edition, though, instead of always putting after the square bracket the authority for the reading adopted in the text (the lemma), Clark and Mason put one of three things. Either they put the form of the reading in the copy text (the First Folio) where it amounts to the same thing as the lemma, as in "1.1] Actus Primus. Scoena Prima.", or else they put the list of readings alternative to the lemma reading as in "1.1.1 again?] again Hanmer", or else they put an abbreviated reference (a siglum), as in "2.0.1 Duncan] Capell", that does indeed point to the edition (here, Edward Capell's) that was the first to use the lemma's reading. Thus, rather than being told that the lemma's reading first appeared in the First Folio this fact is assumed to be understood unless an alternative authority is indicated by the presence immediately after the square bracket of a siglum pointing to another edition. That is a reasonable enough note format if it is explained to the reader, but it is tough on those who are just starting to make sense of textual notes for this edition to silently depart in this way from the form that the General Editors prescribe and explain. As can be seen from the following analysis, the textual notes to this edition are in any case thoroughly unreliable.

    The problems start with the textual note at line 1.1.8. The lemma is "2 WITCH . . . Anon.]" and that points the reader to the body-text part-line "2 WITCH  Paddock calls. | 3 WITCH Anon.". This reading is attributed to "Singer", meaning (according to the "Abbreviations and References" list) the fourth edition of S. W. Singer's 10-volume complete plays edition published in 1826. But this is not the reading in Singer's edition: like the Folio, Singer gives this part-line to "All" rather than to 2 WITCH and 3 WITCH. What is worse, the note then goes on to give two alternate readings and gets them both wrong. For the Folio reading, instead of giving only its version of the snippet indicated by the lemma the note gives the next couple of lines too (wrongly implying that there is repetition in the Folio). The note also misrepresents the Folio's lineation, not recording its break after "anon:". Then the note gives the reading from Alexander Pope's edition of 1723 and this covers a different span of text again (the lemma plus one more line) and mispresents the punctuation, typographic styling, and spelling of Pope's text.

    At 1.2.0-1 there is an emendation marked in the body-text by square brackets to indicate within the opening stage direction that the king's first name is Duncan, and it is supported by a textual note telling the reader that Capell was the first to alter the stage direction in this way. Since this is not a substantive alteration it seems rather fussy to record this. On the same theme, there is a second textual note (which unaccountably has no lemma at all) that records Capell's choice (not adopted by this edition) to use "Dun." as the king's speech prefix; this is fussier still since it does not even affect the text of the present edition. At 1.3.71 Shakespeare's Macbeth says that his father's name was "Sinell", because that is what his source, Raphael Holinshed, reported (and so indeed did Holinshed's source, Boethius). Historically, this is an error since Macbeth's father's name was Finel and Clark and Mason emend to that name here, but they ought not to since once we start correcting Shakespeare's historical errors it is impossible to know where to stop. The point here is that Shakespeare intended the name to be Sinell and altering it takes us away from his intention.

    In the Folio (hereafter F), Ross refers to the king's pleasure at Macbeth's performance in "the Rebels fight" (1.3.92) and Clark and Mason rightly resist the emendation of fight to sight on the grounds that F makes perfect sense. F has Ross say of Duncan's receiving the news of Macbeth's success that "as thick as Tale | Can post with post" and Clark and Mason emend to "As thick as tale | Came post with post" (1.3.98-99). The emendation of can > came is unexceptional but retaining tale where editors since Rowe have emended to hail requires some explanation. It is not clear what Clark and Mason think Ross means. They report that tale can mean tally so that the image is of multiple news-bearing posts (couriers) arriving as fast as they can be counted, which makes sense. But they also seem to think that Ross is not to be taken literally but metaphorically and that the image refers not to the reporting of Macbeth's prowess but to that prowess itself: "Ross means that Macbeth killed men so rapidly that it was like posts (couriers) arriving so fast that they could not be counted". It is hard to imagine an audience comprehending either of these incompatible meanings. At 1.3.103, Clark and Mason mark as an emendation (with the Fourth Folio as their authority) their use of the verb to herald instead of to harrold (as F has it). But this is merely a modernization of spelling (the OED lists F's spelling as current in the period), not an emendation.

    Clark and Mason set the bar pretty high for emendation. They have Duncan ask "Is execution done on Cawdor? Or not | Those in commission yet returned?" (1.4.1-2), as F has it, rather than emending to ". . . Cawdor? Are not | Those . . ." on the assumption that the copy had "ar" (for are) and it was misread and capitalized by the compositor. They are right that are can be "understood after Or" but the assumption of error gives the simpler explanation. Clark and Mason follow F to make Macbeth say to Duncan "The service and the loyalty I owe, | In doing it, pays itself. Your highness' part" (1.4.22-23). The second of those lines was emended by Gary Taylor for the Oxford Collected Middleton to "In doing, pays itself . . ." on the assumption that the compositor interpolated the "it", but no mention of this is made by Clark and Mason. Taylor's edition does not appear in their list of "Editions of Macbeth Consulted" (p. 349), though it appeared eight years before theirs.

    Most bizarrely, Clark and Mason add the name "LADY" to a scene's opening stage direction to produce the unidiomatic "Enter Macbeth's wife LADY alone with a letter" (1.5.0). This seems to mistake the correct use of the honorific "Lady", treating it as if it could stand for a personal name in a sentence. That it can stand alone as a speech prefix--a choice this edition defends in Appendix 1--does not mean that it can stand alone in a sentence. (After all, it is hard to imagine a stage direction such as "Enter Lady Macbeth's husband LORD alone with a letter" being accepted as reasonable dramatic language.) Moreover, the explanatory note here misrepresents the emendation it is supposed to explain, claiming that "Throughout this edition the character commonly known as 'Lady Macbeth' is referred to as 'the Lady' or 'Macbeth's Lady' . . .". But that is not true: here she is not called either of those things. Throughout the edition there are examples of mere modernizations, such as then > than, 1.5.3, being treated as if they were emendations. Likewise for the same modernization at 1.5.37 and also weyward > weïrd at 1.5.8, humane > human at 1.5.17, high > hie at 1.5.25, and powre > pour at 1.5.26. There are many such further examples, including trivialities such as Offrings > offerings and I > ay.

    At 1.5.30 Clark and Mason retain the Folio's stage direction and speech prefix name "Messenger" for the character who comes in to tell Lady Macbeth that Duncan is on his way to visit her. This "Messenger" is clearly one of the Macbeths' men ("Our thane is coming") and his account of how he got his news is that one of his fellow servants of the couple ran on ahead of Macbeth ("had the speed of him") to bring the news to Lady Macbeth. The man giving this news can hardly also be the messenger who ran on ahead, and hence Capell was right to emend the name to "Attendant" although "Servant" would do as well. An emendation note credits A. R. Braunmuller's New Cambridge Shakespeare edition with putting the full stop at the end of Macbeth's line "We will speak further" (1.5.71). But an explanatory note contradicts the emendation note and says (quite rightly) that "Braunmuller punctuates with a dash at this point", not a full stop. The emendation note does not tell the reader what the copy text reading is at this point (which is a basic function of this kind of note), and in fact it is just a full stop so there is no emendation to record. At 1.6.5, Clark and Mason have a textual note and an explanatory note about their printing of the word mansionry instead of F's mansonry, but since manson is a recorded sixteenth-century spelling of the modern mansion this is simply modernization rather than correction.

    This edition's emendation notes' departures from the conventions laid out by the Arden General Editors make for some deeply puzzling notes. At first glance, "15] (In . . . double) Pope" (1.6.15) seems to lack a lemma and a record of the copy-text reading. In fact, it appears that by omitting the lemma the editors mean the whole of line 15 to be their lemma and we are to assume that the copy-text reading is the reading they adopt. These things granted, we seem to be told that here Pope put the whole line in brackets. This is a cryptic way of presenting the evidence, and in any case it is wrong since Pope idiosyncratically put a comma between double and the bracket that comes after it. Just why this is offered as a viable alternative reading is mysterious. Surely Clark and Mason know that ermites is an ordinary early modern spelling of the modern word hermits and that they need not record this as an emendation at 1.6.20. Likewise Ho-boyes > Hautboys at 1.7.0 and Curriors > couriers at 1.7.23.

    At 2.1.19, Clark and Mason invent an exit for Fleance so that Macbeth and Banquo may have their "intimate and confidential" conversation about the witches, but this creates a practical problem. Banquo seems to have given Fleance his sword to hold (2.1.4) and, according to Clark and Mason, is perhaps also carrying the torch that they need to see where they are going (2.1.0.1). Banquo would be particularly neglectful of his and Macbeth's safety if he let Fleance walk off with these necessities while they are outside at night, and in any case obedient children in Shakespeare tend not to leave their parents without permission or at least being remarked upon.

    Clark and Mason credit the 1869 Oxford Clarendon edition of the play by W. G. Clark and W. A. Wright with the innovation of halfe World > half-world (2.1.49) but in fact Clark and Wright's 1863-66 Cambridge-Macmillan edition was the first to adopt this reading. In the Folio, Macbeth enters and says "Who's there? what hoa?" at 2.2.9 and Lady Macbeth, who seems to hear this but not realize it is her husband speaking, worries that she has given Duncan's grooms insufficient sedative: "Alack, I am afraid they haue awak'd". If Macbeth enters to speak "Who's there? what hoa?", why does Lady Macbeth not recognize him by sight and voice? Some editors place Macbeth's entrance 'above' so that he is not immediately visible to his wife, and others have him speak 'within' the tiring house, but Clark and Mason just describe the problem and the various proposed solutions but implement none in their text. Where the Folio Macbeth refers to washing his hands in the ocean and thereby "Making the Greene one, Red" (2.2.6), Clark and Mason follow Samuel Johnson's emendation of punctuation to give "Making the green, one red". The two readings suggest different stress emphasis and different ways of parsing the clause, and each has its own poetic merits. It could be argued that this is one of those occasions where omitting the punctuation altogether--as several editors do--is desirable for keeping the interpretative and performative options open.

    For their reading of "Come in time." (2.3.5), Clark and Mason report in an explanatory note that editors have emended to "Come in time-server", which they reject because the phrase simply means that an arrival is timely. But in their collation note they report the alternative of putting a dash instead of a full stop after "time" without indicating what difference they think that makes, or why they prefer a full-stop to F's comma, and they misattribute this alternative reading to the 1986 Oxford Complete Works edition (which in fact uses an exclamation mark in place of the Folio's comma) rather than Nicholas Brooke's 1994 Oxford Shakespeare single volume edition of the play. Over and again Clark and Mason's textual notes misrepresent a prior edition. They report that the Oxford Complete Works edition of 1986-87 gives the stage direction "He opens the gate" (2.3.20) after the porter says "remember the porter" but in fact that edition puts this stage direction before that line. It is particularly odd that Clark and Mason should get this wrong, since they have a note explaining that "remember the porter" is "a request to Macduff and Lennox for a tip". Clark and Mason give no stage direction for the porter to unlock or open anything: he simply says "remember the porter" and then Macduff and Lennox enter. How could his line be addressed to them as a solicitation before they have entered? The Oxford Complete Works' staging--which Clark and Mason reject and misreport--makes possible their interpretation of the line: the porter opens the gate, says "remember the porter" (perhaps to Macduff and Lennox on the threshold), and then they enter.

    Clark and Mason have the Old Man report of Duncan's horses that "'Tis said they eat each other", where F has the spelling ". . . eate each other" (2.4.18). Clark and Mason's use of the present-tense eat is odd since Ross confirms that the action is past: "They did so" when he "looked" upon them. The past-tense of the verb to eat, spelt ate today, could be spelt eat or eate in Shakespeare's time and it is clear that he meant the past tense here since the action is obviously fatal, not ongoing. Clark and Mason acknowledge the possibility of using ate here, but treat that as an emendation rather than the modernization it really is. At 3.1.22, Clark and Mason have Macbeth say to Banquo that he would have liked his advice but since Banquo is busy "we'll take tomorrow", which is F's reading, rather than emending to "we'll talk tomorrow" as Malone did. This is setting the bar for emendation markedly high: while F just about makes sense (meaning they will take tomorrow as their day for discussion) the emendation assumes an easily made error and provides a much better sense. Indeed, Clark and Mason resist emendation even when metrical irregularity seems most clearly to point to error in transmission. The three witches speak a series of five headless iambic tetrameter lines and then the First Witch says, in F, "Toad, that vnder cold stone" (4.1.6). Editors since Nicholas Rowe have emended to repair the meter (for example by cold > coldest), as Clark and Mason record in their textual notes, but they leave this clearly defective line uncorrected.

    The conventions of the textual notes of the Arden3 series are abused in a fresh way at 4.3.160. The General Editors' Preface says that "Distinctive spellings of the base text . . . are enclosed in italic brackets". In support of Malcolm quite sensibly saying, of Ross, "I know him not" (the Second Folio's reading) where the First Folio has the nonsensical "I know him nor", the editors provide a note that reports "160 not] (nor)". By the conventions of the series, this asserts that nor is a distinctive spelling of not, when what needs to be recorded is the source of the correction of the error in the First Folio. The impression given is that the editors could not be bothered to find out and report which edition first made the nor > not correction.

    Clark and Mason emend F's "What Rubarb, Cyme, or what Purgatiue drugge" to "What rhubarb, senna, or what purgative drug" (5.3.55), using the Fourth Folio's emendation cyme > senna. But they say nothing about the metrical consequence of this alteration: just how is purgative to be spoken (disyllabically?) if the line is to be a regular iambic pentameter? Clark and Mason retain the Folio's reading so that Malcolm says of deserters from Macbeth "For where there is advantage to be given, | Both more and less have given him the revolt" (5.4.11-12). Capell proposed be given > be gone which makes much better sense, especially since, as C. J. Sisson pointed out, Malcolm goes on to say of those who remain with Macbeth that their "hearts are absent too". Capell's emendation supposes a plausible graphic error.

    This edition puts its conspectus of the textual situation of Macbeth into Appendix One, split into "Editing the Play" (by Mason) and "The Folio Text and its Integrity" (by Clark). Mason sketches the scholarship on the Folio text's production, including Charlton Hinman's work on the compositor stints. Concerning mislineation, Mason is sceptical of blame falling on the compositors, but seems not to know of Paul Werstine's important essay "Line Division in Shakespeare's Dramatic Verse: An Editorial Problem" (1984) which showed that it most often is compositorial. Mason is sympathetic to Richard Flatter's baseless argument that the Folio typographical layout may reflect authorial intention. She thinks that editors have been too keen to put in exclamation marks and notes that the Folio is sparing of them. Mason reports that Macbeth's wife is only ever called "Lady" or "Wife" and is always "Lady" in speech prefixes, and that Macduff's wife is consistently "Wife" in speech prefixes, and Mason seems to assume that this is an authorial choice: "Shakespeare's text emphasizes the marital relationship for both women and makes a simple but clear distinction benween them in the speech prefixes: Lady as against Wife" (p. 311). Perhaps so, but since the play had been much altered before it reached print that is a hard thing to judge. Moreover, giving her the speech prefix "LADY" is of course legitimate because speech prefixes are not written in sentences, but Mason applies her rule even to stage directions, which often are sentences, to produce (as we saw) the bizarre "Enter Macbeth's wife LADY alone with a letter" at the start of 1.5.

    Regarding stage directions, Mason complains that editors are interpreting the text when they add these to clarify the action, which is true, but she does not offer an alternative editorial procedure and simply highlights some moments where it matters what the editor chooses to do. A particular interesting example is the action that we must imagine accompanying Lady Macbeth's "Help me hence, ho", in response to which Macduff says "Look to the lady", repeated six lines later by Banquo (2.3.119-20, 126). Mason finds the common suggestion that she faints, or pretends to, part of the "accumulation of masculine editorial interference" in the text. "Of course she might faint", Mason writes, "but it is important to recognize that the text of the dialogue will work if she knocks over a table, tears her hair or even vomits" (p. 316). These alternatives raise their own problems: nobody has been told to bring a table on stage; hair-tearing seems excessive and is elsewhere explicitly spoken about when it happens (King John 3.4.45); and onstage vomiting is even more extreme and no other aristocratic woman in Shakespeare (or in any early modern drama?) does so. Crispinus, a man, vomits up difficult Latin words in Ben Jonson's Poetaster, but that action is symbolic not realistic.

    Regarding the Folio text's integrity, Clark sketches E. K. Chambers's opposition to disintegration, acknowledging that even he accepted that parts of Macbeth are not by Shakespeare. At around 2,500 lines it is only two-thirds the length of the other major tragedies. Clark's description of the problems of the text casts them as essentially subjective matters rather than signs of interpolation. She does not mention in her opening remarks (pp. 321-4) the clincher that a song demonstrably written after Shakespeare's death appears in the text; once interpolation is incontrovertibly shown to be present, it surely has to be considered a strong possibility as an explanation elsewhere in the text where it would fit facts that otherwise are hard to account for. Moreover, Clark simply ignores the considerable computational stylistic evidence for Middleton's writing in Macbeth, asserting that ". . . many of the questions about the status of the Folio text and the extent to which it may have been subject to adaptation cannot be resolved from the evidence currently available" (p. 325). Gary Taylor's work on this problem is briefly mentioned and John Jowett's is for the moment entirely ignored.

    Having rejected the argument that Middleton adapted Macbeth, Clark then turns to the songs that are strong evidence that he did. According to Clark, it is not that Macbeth (as we have it) contains material from Middleton's The Witch but that The Witch itself drew on Macbeth and she wonders if the songs were already in existence before either play was written. Clark gives the reasons why this is in fact quite implausible, not least, as Inga-Stina Ewbank observed, the fact that the songs are "inextricably connected" with their context in The Witch (p. 328). Oddly, the present reviewer is credited with "the idea that the interpolations were made by Middleton after Shakespeare's death" (p. 329n2); I agree with the idea because the evidence for it is so strong, but I can take no credit: the hard work was all done by Taylor while I was still an undergraduate. Clark acknowledges Roger Holdsworth's proof that the stage direction formula "Enter X meeting Y" (used in Macbeth 3.5) is uniquely Middletonian, but rather than accepting this as strong evidence that 3.5 is a Middletonian interpolation she writes that "this is not to say that the scene is totally redundant" (p. 331). Well, no, if it were redundant it would not have been a professional writer's successful interpolation; the question under discussion is not how good the writing is, but who produced it.

    Finally, Clark gets around to the recent computational stylistical work on the problem, but she gets an important detail wrong. In a footnote, Clark reports that Brian Vickers and Marcus Dahl present data showing that 3.5 has "Shakespearean diction throughout the scene" (p. 332n1) and that the present reviewer has "questioned the methodology used to support their claims for Shakespeare, which derive from an incomplete database of the Middleton canon". That is not quite the point. Vickers and  Dahl do not merely attempt to show that the scene's diction is Shakespearean but also, and crucially, that this diction is unMiddletonian. That is why the incompleteness of their database of Middleton works matters so much: they report that many words and phrases found in the scene are present in Shakespeare's works and absent from Middleton's work. But what the present reviewer has shown--and can most easily be seen in his contributions to the Authorship Companion to the New Oxford Shakespeare--is that those same words and phrases are in fact present in Middleton's work; Vickers and Dahl simply fail to find them because their Middleton database lacks a great deal of what he wrote. Despite this misreporting of the matter, Clark acknowledges of 3.5 that "it seems likely that Shakespeare was not its author" (p. 332), which is a crucial admission.

    Once co-authorship of a play is accepted, the editors' attitude towards its textual problems should be utterly transformed, but in Clark and Mason's edition it is not. They acknowledge the co-authorship but edit as if they did not. Having expressed scepticism throughout her account, Clark concludes that in fact in all probability "the songs in 3. 5 and 4 .1 were added to the first version of Macbeth at a later date . . . and then perhaps also the suspect passages in 4.1, were interpolations" (p. 334). But she does not accept that acknowledging interpolation in those places means that we should accept that it is likely present elsewhere in the play, nor that we should suppose that the play is short because Middleton cut a lot of it when doing his adaptation. Rather, ". . . the extent to which it [the Folio text] differs [from the play as Shakespeare wrote it] may well be very slight, and confined to 3 .5 and two passages in 4.1" (p. 336). This sounds like wishful thinking. The remaining appendices offer a reprint of Chambers's 1930 transcription of Simon Forman's eyewitness account of the play in performance at the Globe, an historical family tree, a list of the theatrical productions mentioned in the Introduction, and a casting chart that lists who is in each scene but offers nothing about doubling or any suggestion of cast size.

    Only one monograph on our topic was published this year: Zachary Lesser's Hamlet After Q1: An Uncanny History of the Shakespearean Text. Lesser starts with an account of the discovery of Q1 in 1823 and its first reprinting in 1825, and sketches the allusions to a Hamlet play that begin well before most people think Shakespeare wrote his: Thomas Nashe's "whole Hamlets, I should say handfulls of tragical speaches" (1589); Henslowe's diary entry for a Hamlet on 9 June 1594; Thomas Lodge's "looks as pale as the Visard of the ghost which cried so miserally at the Theator, like an oister wife, Hamlet, reuenge" (1596); "my name's Hamlet reuenge" in Thomas Dekker's Satiro-mastix (1602); a couple of post-1603 allusions to the phrase "Hamlet, revenge" (not spoken in Shakespeare's play); The Hystoria of Hamblet (1608) which is an English translation of François de Belleforest's French translation of the Hamlet story in Saxo Grammaticus's Gesta Danorum; and the play Der bestrafte Brudermord (Fratricide Punished) that turned up in Germany in 1779 and that also calls the king's advisor Corambus. So, we knew there was some kind of ur-Hamlet but that fact could be ignored until the discovery of Q1 gave it some kind of physical form. Lesser ends his Introduction by remarking that although he has tried to avoid worrying about the origins of Q1, Q2 and F, one cannot dodge those questions altogether.

    Lesser's Chapter One is on "As Originally Written by Shakespeare: Textual Bibliography and Textual Biography" (pp. 25-71). The lack of the last page did not seem especially catastrophic when the first exemplar of Q1 was discovered in 1823 since performances often ended with the death of Hamlet in any case. Early commentators seem not to have have found Q1 especially deficient and there was a long tradition of belief that some of the quartos reflected Shakespeare's first stabs at each play and that Shakespeare revised them. There was a general eighteenth-nineteenth century view that all the early editions are corrupt, and there was also some talk of stenography or long-hand aural transcription being involved in their transmission. The difficulty was in reconciling the evidence of Q1 with the prevailing Romantic view of authorship as a spontaneous organic process arising from the writer's artistic development rather than the writer's labour and practice. John Payne Collier was the first to claim that Q1 is distanced from Q2/F only by corruption in transmission (specifically, stenography), and this idea soon spread to explain other early quartos: a version's being early into print did not (as had been assumed) necessarily mean that it was early in the history of the play's genesis.

    Lesser reckons that it was because Hamlet had already achieved its totemic status in English Literature ("emblematic of Shakespeare's mature art and the peak of poetic achievement", p. 38) that when Q1 Hamlet was discovered it could not be assumed to be a first stab and the corruption-in-transmission theory had to be invented to account for it; thus for Lesser Q1 Hamlet lies behind the whole New Bibliographical tradition. Charles Knight, Collier's rival, however, continued to champion the authorial revision theory. Knight favoured the Folio over quartos because the latter are rare and obscure and he was a popularizer and did not believe in an ur-Hamlet: for him, all the early references are to Shakespeare's first stab at the play from the 1580s. Because of his forgeries, Collier was air-brushed out of New Bibliographical accounts of prior scholarship, so that Tycho Mommsen rather than Collier is given credit for the idea of an aural report underlying Q1 Hamlet and P. A. Daniel for Memorial Reconstruction underlying Q1 The Merry Wives of Windsor.

    As a result, the only two explanations for Q1 Hamlet that were entertained until the late nineteenth century were Knight's (early version) and Collier's (stenographic report) theories. W. G. Clark and W. A. Wright came up with a new theory when editing Hamlet in 1872: Q1 represents the old Hamlet play after Shakespeare had begun to retouch it (around 1602) but before he had finished doing so, which Lesser sees as an idea that it was impossible to think within the confines of Romantic conceptions of authorship. This is a "stratification" theory (p. 58). Lesser considers Q2's title-page boast ("Newly imprinted and enlarged to almost as much againe as it was, according to the true and perfect Coppie") and the various interpretations of it. Is Q2 longer than Q1 because Shakespeare revised the play, or simply because the printers of Q1 had access to better, fuller copy for it? For most New Bibliographers, Shakespeare himself was the agent for Q2's printers having better copy: he wanted a good text to force out the bad. Lesser tells the story of Q2's registration and printing and the perplexities arising from Nicholas Ling's involvement with both Q1 and Q2 if the latter was meant to replace the former. Lesser considers the possibility that, contrary to its title-page, Ling was not really the publisher of Q2--that is, he did not put the money up but someone (the King's men?) paid him to publish it--and the alternative possibility that Q1 quickly sold out so Ling did not mind publishing a second edition. A problem with the latter theory, Lesser acknowledges, is that putting out a long Q2 would be expensive--a much longer text to be set, slowly, from manuscript copy rather than quickly from a printed edition--and risky because of the unknown popularity of the new text. If Q1 sold out quickly why not just reprint it?

    Lesser points out that Q2's title-page seems to copy Q1's, down to the same style of hanging indentation for the description after the author's name, which as Lesser shows was something Valentine Simmes (Q1's printer) pioneered in secular books and it was normally reserved for biblical quotations on title-pages. Ling must have told Q2's printer Roberts to copy Q1's title-page in this regard. Lesser reckons that having Q2 mirror Q1 in this way enabled Ling to have it both ways if stocks of Q1 were not exhausted: he could sell Q2 to the more discerning buyers and pass off Q1 as if it were Q2 to the less discerning. Q2's title-page refers to its text being improved, and Lesser shows that this was something said mainly on title-pages that claimed Shakespeare as their author and seldom on anyone else's. But Q2 does not actually claim that Shakespeare has corrected the text, only that it is "enlarged", which "suggests Ling's desire to leave open an alternative interpretation, one that allowed Q1 to stand side by side with Q2 rather than being superseded by it" (p. 69). Only one previous play had editions that made this enlargement claim: The Spanish Tragedy (1602), which needed refreshing since it was known to be an old play. As Lesser points out, Hamlet too was a play from the 1580s that needed to be pitched as having been refreshed. Indeed, having Q1 and Q2 side-by-side in his shop would allow Ling to satisfy readers who wanted the original 1580-90s version and those who wanted the latest version.

    Lesser's Chapter Two is "Contrary Matters: The Power of the Gloss and the History of an Obscenity" (pp. 72-113). In Q1 the equivalent of the familiar line "Do you think I mean country matters" is rendered as "do you think I meant contrary matters". Q1 also omits Hamlet's remark that nothing is a fair thought to lie between a maid's legs. Some nineteenth-century critics took this to show that Q1 was Shakespeare's unsullied original and that Q2 and F represent the vulgarity of what actors did to Shakespeare's text. In fact, however, before the nineteenth century readers of Hamlet do not seem to have assumed that "country matters" was itself obscene. If Shakespeare meant to divide his audience by "country matters" into those who get the pun and those who do not then modern editions' glosses spoil that intention. By not having the word country, Q1 bolstered the nineteenth-century idea that Q2/F's "country matters" is obscene. 

    In the Valladolid Folio censored by someone from the Spanish Inquisition, lots of bawdy was struck out but not "country matters". The same is true of a Smock Alley promptbook from the 1670s and an edition representing the play as it appeared on the Restoration stage, both of which cut the bawdy around "country matters" but not the line itself. Samuel Johnson also seems not to have thought "country matters" to be a pun on cunt, although he thought it vulgar. Almost no one before Malone in 1790 seems to have heard cunt in Hamlet's country. It is hard to explain country (Q2/'F) > contrary (Q1) by the theory of Memorial Reconstruction underlying Q1, since if the pun was so obvious the reconstructor would have recalled it. In Q1/F Hamlet first asks if he can lie in Ophelia's lap (which is a sexual request) and only then, when she says no, does Hamlet explain that he meant the non-sexual upon her lap; Q2 omits the upon clarification. But where Q1's Hamlet first asks if he can put his head in Ophelia's lap and then clarifies that he innocently meant upon her lap, in F he mentions his head only as part of the clarification in an innocent head upon phrase: Q1's head in phrase is not innocent, but "may easily suggest cunnilingus" or else, with head as glans, might suggest phallic penetration (p. 110). If Q2 reflects authorial papers and F the promptbook, it looks as if in rehearsal Hamlet's double entendre (including the cunt-ry joke) had to be emphasized to make it clear, by the adding of the head upon clarification line. If Q1 reflects what actually got performed then the actors had to do even more--moving head to create the more explicit head in joke--to press the vulgar point, and apparently they still failed since they gave up trying to make the cunt-ry joke and changed country to contrary simply to get across that all of this was indeed a double entendre (pp. 109-12). 

    Chapter Three is "Enter the Ghost in his Night Gowne: Behind Gertrude's Bed" (pp. 114-156). In Q1, Hamlet's speech about not letting the clowns improvise ends up contradicting itself, since he goes on to complain that clowns often have stale catchphrases that are too well known and that they often fail to be funny improvisers, which is the opposite of the point he started with. For Lesser, this confusion about "jests [that] have become scripted" (p. 115) usefully stands for our collective problem with the relationship between text and performance, which he will illustrate with the long-standing critical objection that closet does not mean bedroom (for which the right word was chamber) so that Gertrude's bed should not be on stage for scene 3.4. In fact, critics were calling this a bedroom scene even before onstage beds for it became normal in the 1920s-30s. The objection to this as a twentieth-century anachronism is, according to Lesser, really about scholars policing meaning and asserting their authority over mediations of Shakespeare. Commentators since the early nineteenth-century had considered this a bedroom scene and the London Adelphi theatre production of 1899 (with Sarah Bernhardt as Hamlet) had a bed in it, and the reviewers mentioned it without objection.

    Thinking that 3.4 is a bedroom scene made perfect sense after Q1 was discovered--long before the work of Sigmund Freud, who usually gets the blame--since it calls for the ghost of Hamlet's father to enter in his nightgown. The immediate response of critics to Q1's nightgown was dismay at its adding bathos to the already uncomfortably comic scene that ends with Hamlet ludicrously lugging Polonius away. For 50 years after the discovery of Q1, theatre productions still had Hamlet Senior enter the closet scene in his armour. Some critics explained the nightgown attire as showing that this manifestation of the ghost was not real (as he had been on the battlements) but just a figment of Hamlet's mind. Henry Irving's landmark production of 1874 used the closet scene as its central marker for the entire play's domestic setting; a number of critics thereafter identified Gertrude's violation of this domestic harmony (by adultery, by remarriage) to be the cause of Hamlet's melancholy and saw her acceptance, at Hamlet's importuning, that this was a mistake as enabling Hamlet to move forward to avenge Claudius's part in it. In keeping with this melodramatic take--in which Gertrude is a fallen woman rescued by her son's insistence that she repent and reform--Irving cut more of what follows in order to get as quickly as possible to the final duel.

    Irving thus popularized the idea that the closet scene takes place either in or adjacent to Gertrude's bedchamber, and hence when the 1899 production with Bernhardt as Hamlet used a bed, no one was surprised. In fact, Lesser finds eighteenth-century drawings of the closet scene that include a bed and points out that in Gesta Danorum by Saxo Grammaticus there seems to be one. Having Hamlet Senior in his night gown connotes that this is a bedroom scene without the players having to go to all the trouble of showing us a bed. But Q1 does not call the place that this scene takes place a closet (as Q2/F repeatedly do), so it becomes a bedroom (or near one) only when the Ghost enters. That Gertrude is still using her dead husband's bed has the same meaning as the Ghost's still having his armour: both these things ought to have been, by the laws of inheritance, passed to young Hamlet.

    Lesser's last chapter is "Conscience Makes Cowards: The Disintegration and Reintegration of Shakespeare" (pp. 157-206) and it addresses the play's most famous speech. Q1's version of the "To be" speech seems more straightforward and conventional in its theology than Q2/F's version. But what does Hamlet mean by conscience making cowards of us all? A number of late-nineteenth and early twentieth-century critics glossed conscience as consciousness, rather than as moral sense, which was the prevailing definition before then. Around the middle of the twentieth century this preference flipped back to conscience being the moral sense to avoid sin. Just how we understand conscience affects the meaning of the whole speech. If it is moral sense then Hamlet is just conventionally religious in saying that a God-given aversion to sin and/or a fear of divine judgment makes us not sin, but if it is consciousness than Hamlet is saying that thinking hard about things (being philosophical) makes us agnostic, unable to do anything, including committing sin. Taking conscience as consciousness is often said to be the Romantic reading of Hamlet's character (too thinky) but in fact the Romantics, so far as we can tell, understood conscience as moral sense. So, the late-nineteenth century assertion that conscience is consciousness is really a response to Q1 making explicit the opposite interpretation of conscience as moral sense.

    The disintegration of Hamlet that followed Clark and Wright's 1872 Clarendon edition's acceptance of Q1 as partly Shakespeare's own work and partly a play he inherited is seen by Lesser as providing the model of the whole disintegrationist movement. For the disintregationists, the old play provided the Hamlet of action and religious orthodoxy and Shakespeare added the Hamlet of deep thinking and agnosticism. The disintegrationists' impact on criticism was that they promoted historicism: things that seem strange in the drama can be explained by i) the textual history, and ii) the mores and expectations of the time it was written. The idea that Q1 represents an earlier version of the play that became Q2/F was scotched, according to Lesser, by G. I. Duthie's 1941 account of Q1 as an Memorial Reconstruction with no connection to the ur-Hamlet other than passages where the reconstructors patched their faulty memories with bits of that play. This revaluation achieved, there was no need to contrast Q1's sense of conscience as fear of judgment with Q2's sense of conscience as agnosticism from over-thinking, and critics were free again to say that Q2's sense of conscience is fear of judgment. Lesser reckons that, in fact, across all three texts conscience in the "To be" speech "is decidedly unlikely to carry any meaning other than the religious one" (p. 195), which sense it has in all the other uses of that word in these three texts.

    Q1 daringly complicates the familiar idea that those with bad consciences are cowards--fearing to die because fearing the judgment that will follow--which implies that those with clear consciences are fearless, by claiming that conscience makes us all cowards, bad and good alike. But since there is no Christian theology in the Q2/F version of this speech, how did pre-Q1 readers come to read conscience in the theological sense? From their annotations, it is clear that the early seventeenth-century reader of one of the Meisei University Folios understood the speech to be about killing Claudius, not about suicide, and to express radical agnosticism. This annotator also commonplaced the conventional Christian proverb "Conscience makes us cowards", and Lesser reckons that this shows that the Meisei reader was able to take the speech as containing contradictions without needing to resolve them. Lesser sees Q1's as the only coherent version of this speech; Q2/F are incoherent and disjointed, almost as if based on an Memorial Reconstruction of Q1, he jokes.

    In his conclusion (pp. 207-221), Lesser traces the difficulties that textual scholarship and criticism after the New Bibliography have in accounting for the Q1/Q2/F relationship, not wanting to tie them all together under the rubric of Shakespeare but not having the nerve to treat them as independent works either. He details the Arden3 Hamlet editors' lack of any strong conviction about the relationship between the early editions, and points out that taken seriously this principled lack of convictions would require us to treat all versions of the story, including the sources and the derivatives, as equally representative of Hamlet. Lesser declares that he began with a New Textualist bracketing off of the three texts so that he could reconsider their relationships without inheriting ideas of the dependence of one upon another, but he concludes that one cannot do that forever: ". . . the issue of textual origins can be productively deferred, but it will not ultimately be evaded" (p. 219).

    Two book-form collections of essays relevant to our topic were published this year. The most important is M. J. Kidnie and Sonia Massai's long collection Shakespeare and Textual Studies. Several of the essayists here appear not to have noticed that recent computational stylistics scholarship has given ample reason to reject the post-structuralist exhortation that authors should not be central to literary-dramatic study. It is certainly interesting from an historical point of view to find out what people wrote in the margins of their books, reacting to the author's words, or how they organized them by binding individual ones together, and not in order to reflect common authorship. But if we think that doing this kind of scholarship is part of an essential act of de-centering of the author, needed because traditional scholarship has over-emphasized the importance of authors, then we are mistaken. Traditional scholarship was right, it turns out, to assume that the author was the centre of authority and the most important determinant of what got written. The essays are generally short, mostly between 10 and 20 pages.

    In their "Introduction" (pp. 1-10) Kidnie and Massai seem to place the cart before the horse when they describe stationers as "agents, who, by investing time, money and their professional expertise in printing and publishing the first editions of Shakespeare's works, transformed Shakespeare from popular poet and playwright into a best-selling author" (p. 3). How could he have been a "popular poet", which presumably refers to Venus and Adonis and Lucrece in 1593-4, before the publication of those works? Surely it was the stationers who made him a popular poet in the first place. The editors treat as it they were one phenomenon "the momentous advent of poststructuralism and digital technologies in the late twentieth century", with one outcome: "the dictum that editors should now edit without the 'author' and without the 'work' as their guiding principles" (p. 10). I would have said that post-structuralism dates from the 1960s and was having a significant (harmful) effect on thinking about authorship and textuality in the 1980s whereas digital technologies had no significant impact until the 2000s.

    In "Playwriting in Shakespeares Time: Authorship, Collaboration, and Attribution" (pp. 1-26), Heather Hirschfeld sketches the influence of Roland Barthes and Michel Foucault on notions of authorship and how their views have intersected with studies that characterize the theatre industry as inherently collaborative. Hirschfield has her own model for what it was like to be an author: ". . . Shakespeare, like his fellow dramatists . . . [employed a] model of dramatist-as-pen, a model that stresses the playwright's identity with his writing instrument" (p. 17). Thinking this way helps us comprehend the notion of a writer's style as it was understood in Shakespeare's time, and the Poetomachia of 1599-1602 proves that writers did think that each has his own style that would be recognized when parodied. But how can that be so if we accept that co-authorship was normal? Hirschfield finds that Jeffrey Masten's Textual Intercourse (1997) thoroughly critiqued "the brand of 'disintegrationist' textual study that sought to determine who wrote which act, scene, or line" (p. 20-21) and showed that collaboration dispersed authority. Since then, computational analysis has shown that in fact authors' personal styles are not lost in collaboration, but Hirschfield offers an entirely unsupported caveat to this conclusion: "Of course, the stylometric discovery of Shakespeare's and other dramatists' signatures is as much the assumption of computer analysis as it is the conclusion, a hazard of circularity as well as presentist thinking" (p. 23). No reason for accepting this claim of circular logic is given: she just asserts it is obviously true ("Of course").

    Hirschfield also thinks that computational stylometry is hampered by use of printed texts, "potentially assigning to a playwright linguistic characteristics that could be traced to scribe or compositor" (pp. 23-24). Hirschfield does not notice that this argument cuts both ways. Why not suppose that the opposing claim, the claim that collaboration blends/disperses authority, is based on mistaking scribes'/compositors' characteristics for the authors'? The difference, of course, is that computational stylometry done well is a science and is able to test its own validity as a knowledge-generating system. We can do controlled experiments to figure out how reliable our tests are when applied to cases for which we already know the right answer. Moreover, and more fundamentally, difference in style is empirically demonstrable while sameness in style is not. That is, we can show that on a particular test a particular piece of writing breaks down into two parts. We have proven that it can be so broken. But demonstrating that no test we know of breaks a particular text into multiple parts does not mean that there is no such test. We have not proven that it cannot be broken. This is a fundamental distinction about knowledge that should inform these debates.

    Some distinct academic paranoia enters Hirschfield's argument about computational stylistics: it is done because "administrators" and "funding organizations" like "the de-humanizing, mechanizing, and economizing work of computerized number-crunching that turns style into machine readable coordinates" (p. 24). This remark is insulting to computational scholars of enormous critical sensitivity such as MacDonald P. Jackson. Hirschfield thinks that Douglas Bruster's work on Shakespeare's characteristic spellings turning up in the Additions to The Spanish Tragedy (reviewed in YWES for 2013) is an example of how "the mathematized and mechanized approaches to style [can] accommodate the contours of Shakespeare's theatrical milieu" (p. 24). But she does not actually subject Bruster's claims about penmanship and typesetting practice to any kind of critique: she just approves of the work because she likes its not being entirely computational/mathematical. Adding nothing to the topic, this essay just tells us that Hirschfield likes certain kinds of scholarship and is unable to distinguish between good and bad empirical work.

    Also not an advance on what we know, but at least restating his earlier, sound empirical scholarship is Paul Werstine's essay "Ralph Crane and Edward Knight: Professional Scribe and King's Men's Bookkeeper" (pp. 27-37). Werstine begins with a sketch of the two men, with Crane coming out as the better scribe because he was more careful and understood the difference between prose and verse. Werstine then sketches Crane's characteristic habits when copying plays: consistently dividing into acts and scenes, applying massed entries, treating dialogue in a way "lavish with italics, capitals, and punctuation of all kinds, including parentheses, apostrophes, and hyphens" (p. 29), and some distinctive uses of apostrophes for elision and hyphens for compounding (detailed by Werstine). T. H. Howard-Hill was able to distinguish use of Crane copy in the Shakespeare Folio by counting (computationally) the density of punctuation. (Strangely, Werstine and this volume's Works Cited list date Howard-Hill's seminal work on this, his monograph Ralph Crane and Some Shakespeare First Folio Comedies, to 1974 when in fact it was published in 1972.) Other work that claims to detect Crane copy behind printed works is less convincing than Howard-Hill's, and Werstine rather tediously details it. Crane, like Knight, was more liberal in altering stage directions than dialogue. Werstine ends his account of Crane with some statistics showing that generally his copying was accurate regarding substantives. He turns to Knight and states the known facts of  biography and scribal habits. 

    In "Shakespeare's 'Strayng' Manuscripts" (pp. 39-53) James Purkis summarizes the current orthodoxy that no one tried to regularize manuscripts used in the theatre, of which Purkis reckons 18 survive. Purkis remains here publicly sceptical that Hand D of Sir Thomas More is Shakespeare's, but he seems conflicted because he also wants to draw conclusions about Shakespeare from the evidence of that manuscript in order to contradict others' uses of it. Purkis considers the common editorial reassignment of the line from Folio Othello "This Lodouico is a proper man" (4.3.34) from Desdemona to Emilia, which E. A. J. Honigmann supported by observing that Shakespeare's speech prefixes are omitted or added later in Hand D so misalignment here is not unreasonable. Purkis objects that although there is misalignment of speech prefixes in Hand D, and they do seem written in later after the dialogue was done, ". . . at no point does his tendency to allow speech headings to roam upwards on the page leave a speaker unclear" (p. 42).

    Purkis reckons that the dramatic characters in Hand D seem at first unlike the same persons we saw earlier in the play not because Shakespeare lacked familiarity with the play's persons but rather because of "the revisers' collective project of toning down the danger of the rebellion" (p. 44). Purkis compares the rebellion in Sir Thomas More with that in 2 Henry 6 without showing any awareness of the scholarship demonstrating Marlowe's hand in the Jack Cade's scenes of the latter play. Looking closely at the reassignment of speeches to characters in Sir Thomas More rebellion-quelling scene, Purkis finds that other scholars have overstated the problems and that the scene is generally consonant with  preceding characterizations of the individual citizens and unlike the characterization of Cade's followers. This section ends inconclusively and it never becomes apparent what conclusions Purkis wants us to draw from his comparison with 2 Henry 6. Next Purkis turns to the manuscript of The Second Maiden's Tragedy simply because it was the manuscript "most proximate to Shakespeare after More . . . [in being] prepared for performance by Shakespeare's company, the King's Men, in 1611" (pp. 47-48). This does not seem especially proximate. The supposed connection gives Purkis an opportunity to tell the reader everything he knows about this manuscript and its writer's powers of characterization. And then he does the same for The Captives and Sir John Van Olden Barnavelt. All that seems to emerge from this survey is the discovery that characterization changes when manuscripts are altered and that practical considerations, such as casting limitations, might predominate, and that to accommodate them is not necessarily to depart from the dramatists' wider intention.

    Massai's own contribution to her collection is "The Mixed Fortunes of Shakespeare in Print" (pp. 57-68) in which she asks the pertinent question of why first editions of Shakespeare suddenly fell off so sharply after 1600. Her short answer is that from 1600 books of plays by children's companies--which restarted in 1599 (St Paul's) and 1600 (Blackfriars)--began to out-compete books of plays by adult companies. Massai has found that the paratexts of boys company printed plays differ from those of adult companies in that they much more often include dedications and addresses to the reader, and these generally refer to literary/readerly rather than theatrical/playgoerly concerns. An important methodological point here, however, is that Massai does not offer us statistics for this claim, only some sample quotations that illustrate it; it is not clear if she has counted anything or is just giving us her impression.

    As so often in these matters, the exceptional case is Jonson, whose prefatory material for the printed editions of his Chamberlain's/King's men's plays Every Man Out of His Humour and Volpone made them seem "more like the printed versions of plays originally staged by the children" (p. 65). The trouble here is that Massai has to make Jonson exceptional; otherwise this is evidence against her thesis in showing an adult company's plays doing the thing she says is peculiar to the boys companies. The same problem recurs with Massai's next example: Dekker's The Whore of Babylon for Prince Henry's men (as the Admiral's men became after 1603) was published in an edition of 1606 whose paratexts were "nevertheless aligned with new modes of playwriting" (p. 65) as set by the boys companies. Massai wonders whether Shakespeare did not get many new first editions in the early 1600s because he did not write about his art and his plays "may not have appealed either to those readers who were drawn to the new, literary self-conscious style of the children's playbooks" (p. 66). Hamlet and King Lear were reworkings of old plays, not in the new style. His old plays still sold fabulously well, of course, but Massai reckons that in the first years of the new century Shakespeare was becoming distinctly old fashioned.

    In "'To London all'?: Mapping Shakespeare in Print, 1591-1598" Helen Smith considers how little we know about who made and who bought the early editions of Shakespeare. The links are often quite tenuous and uncertain: "[Richard] Field's ... father, Henry, was a tanner, producing leather that might be used both in the printing house (for tools) and for bookbinding" (pp. 75-76). Which works got bound with what is treated by Smith as evidence of how the early modern readers categorized their collections. Regarding the differences between Q1 and Q2 Romeo and Juliet, Smith remarks that the former "omits elements of the play that refer to the sonnet tradition" while in the latter "the verse prologue is amended to form a sonnet" (p. 85). That "amended" is more than we know: Q1's not-quite-a-sonnet (two quatrains, two couplets) could simply reflect corruption of a text that got more accurately printed in Q2.

    Alan B. Farmer's "Shakespeare as Leading Playwright in Print, 1598-1608/09" (pp. 87-104) sketches Shakespeare's phenomenal success in print in 1598-1609 and then turns to the unusually high occurrences of the words "augmented" and "corrected" on his title-pages: is this a sign of his active involvement in publishing? Six of his plays were printed in 1598-1609 with revision/correction claims on the title-page: Q2 Love's Labour's Lost (1598), Q3 1 Henry 4 (1599), Q3 Richard 3 (1602), Q3 Romeo and Juliet (1599), Q2 Hamlet (1604-5), Q4 Richard 2 (1608). Including reprints there were 10 editions of these six plays in 1598-1609 claiming revision/correction. But are these publishers' exaggerations or true claims? Farmer categorizes three kinds of "editorial pledge" (Massai's phrase) found on title-pages: i) that the book was reliably printed from a good source, ii) that the book contains additions (but without calling them new), and iii) that the work improves on a previous edition by including new additions. By far the most common kind of claim is that the work is revised. Farmer thinks it cavilling to wonder whether a "new additions" claim followed by an authorship attribution claim (as in "With the new additions of the Parliament Sceane . . . By William Shakespeare", Q4 Richard 2) was meant to claim that the author wrote the new additions: of course it meant that. (Okay, but might not this phrasing be deliberately intended to fudge the issue of just who wrote them?)

    It turns out that, overall, title-page claims of revision are rare--20% of all non-Shakespearian reprints have them--whereas most, 56%, of Shakespearian reprints have them; the contrast is starker still if we confine ourselves to editions that advertise the name of the dramatist. So, "... almost no other dramatists were advertised as having revised their printed plays" (p. 98). The two exceptions up to Shakespeare's death in 1616 were John Marston and John Webster for one reprint of one play each, while for Shakespeare it was 10 editions of four plays. After Shakespeare's death the pattern did not change: only Thomas Middleton and William Rowley for one play, and Thomas Heywood (for two plays) got named as revising authors on their reprints up to 1640. The same pattern holds for non-dramatic publications: these were not nearly so often advertised as revised by the author as Shakespeare's plays were. Thus Shakespeare was marketed to readers as "the author most concerned about the printing of his works in accurate texts" (p. 100). Perhaps the publishers were lying, but the trouble with this assumption is that most of Shakespeare's publishers made this claim much less often in their other editions and the rest only made it when it was demonstrably (to our satisfaction today) true.

    Farmer's conclusion is that Shakespeare did get involved in the publication of his plays, as a reviser/corrector. He remarks that as Sonia Massai has shown--in Shakespeare and the Rise of the Editor (2007), reviewed in YWES for 2007--revising authors appear to have been no more rigorous than what she calls the "annotating readers" (not presshop correctors) who were employed to provide improved copy for reprints; the resulting editions were still full of errors. So, argues Farmer, the fact that the allegedly revised reprints of Shakespeare are still full of errors is not strong evidence that he did not revise them. After all, Marston's alterations reflected in the second edition of Parasitaster (1606) do not rid the text of many small errors. Farmer reckons that the likeliest explanation for the claims about revision/correction in early Shakespeare reprints is that they were true.

    The single most important discovery in the collection comes in Zachary Lesser and Peter Stallybrass's "Shakespeare between Pamphlet and Book, 1608-1619" (pp. 105-133), and it is that the Pavier Quartos of 1619 were not a single-author collection: Thomas Heywood's A Woman Killed with Kindness was bound into the set at the point of sale. The 1608 King Lear was the first of Shakespeare's editions to put his name before the title on the title page, and in big type. Like the 1607 Volpone, it gave the playwright's name and called the play "his", distinguishing that word typographically. But unlike Volpone, which said nothing about the contents being a play or about anyone but Jonson's being involved in it, the 1608 King Lear names the chronicle history source, the dramatic characters in it, the patron, the acting company, and the venue. The practice of using "his" in this way did not last. 

    When John Smethwick published Q3 Romeo and Juliet in 1609 he took over the Q2 title page's claim that the text was "corrected, augmented, and amended" but he did not add the dramatist's name even though it was normal to do so by then. In other words, the habits of reprinting verbatim trumped the fashion of adding the author's name. The same year, Shakespeare's Sonnets was published with his name in large capitals on the title-page; the publisher was Thomas Thorpe, who had put Jonson's name in large capitals on the 1607 Volpone edition. But this marketing technique did not create a publishing success. Sonnets was not reprinted, while Shakespeare's Venus and Adonis was published in 1593 without Shakespeare's name on the title page and it was reprinted many times thereafter, each time without Shakespeare's name on the title page.

    Lesser and Stallybrass turn next to the Pavier Quartos of 1619, which W. W. Greg determined was a set of 10 plays in 9 volumes--The Contention of York and Lancaster and Richard Duke of York were combined as The Whole Contention--intended as a collected works edition. The collectedness is most strongly witnessed in the continuous signatures of The Whole Contention and Pericles; the rest do not. But the Folger Library's copy which does indeed have them all together in a seventeenth-century binding does not put Pericles after The Whole Contention, nor did a Folger copy in an eighteenth-century binding, which added an eleventh play, Thomas Heywood's A Woman Killed with Kindness. Other sets of Pavier Quartos are bound in other orders too--and another shows signs of having had Heywood's A Woman Killed with Kindness bound in with it--and none of them puts Pericles after The Whole Contention. So, was it buyers or the publishers, the Jaggards, who put these things together as collections?

    Lesser and Stallybrass have evidence that Heywood's play was part of the collection as it was first offered in the bookshops. The evidence comes from the stab-stitching that provided a temporary binding for some short books up to the point of sale. Other books were fully bound before sale and they will not have stab-stitching holes. The Pavier Quartos are found in both forms, showing that some were sold as individual single-play editions (and have stab-stitching holes) and some were sold as collections (and lack stab-stitching holes). A significant number (six of the 17 examined) of the exemplars of Heywood's A Woman Killed with Kindness, more than a third, do not have stab-stitching holes, so these ones were originally sold bound in with other plays, and Lesser and Stallybrass reasonably enough assume that they were bound in with other Pavier Quartos.

    Like Massai in Shakespeare and the Rise of the Editor (reviewed in YWES for 2007), Lesser and Stallybrass reckon that William Jaggard and not Pavier was the prime mover in the creation of this collection, but they think the question of why he used false imprints "remains obscure" (p. 131). Heywood's play was the only play that the Jaggards had sole-published before 1619 and Lesser and Stallybrass speculate that putting it into a Shakespeare collection was their way of making the collection look unlike something created by a publisher. Perhaps putting in The Yorkshire Tragedy and Sir John Oldcastle was part of the same strategy of disparateness rather than, as has always been assumed, part of a strategy of making these plays look as though they are by Shakespeare.

    In "The Canonization of Shakespeare in Print, 1623" (pp. 133-146), Emma Smith focusses on the Shakespearian dramatis personae lists of characters that appear in the Folio and have "no proven provenance in the theatre" (p. 133). Seven Folio plays have these lists: The Tempest, The Winter's Tale, Othello, The Two Gentlemen of Verona, Measure for Measure, 2 Henry 4, and Timon of Athens. Some are relatively complete, others (The Winter's Tale and Timon of Athens) omit significant speaking parts. Smith thinks that these lists are "designed explicitly for readers" because "theatrical texts do not include them and they have no practical place in the processes of performance organized around actors rather than roles" (p. 135). I am not so sure. If the documentary 'parts' embodied doubling choices, so that there was one 'part' for each actor and it recorded his multiple roles, then Smith would be right that lists of characters would have little utility. But with just one surviving actor's 'part' to go on--Edward Alleyn's for the title role in Robert Greene's Orlando Furioso--we simply do not know if 'parts' for lesser roles that were routinely doubled (as title roles were not) were organized one-per-actor rather than one-per-character. If each 'part' was for one character then the Folio character lists could easily be derived from a list of such 'parts', as for example might be written on a wrapper used to hold them all together in storage.

    Smith notices that if the cast lists were printed to fill a space at the end of the play, it is odd that several Folio plays with big spaces at the end do not have such a list; she also notes that it is strange that the list for Measure for Measure gives the personal name of the Duke, Vincentio, which cannot be derived from the printed script. (That surely is evidence against her claim that the lists have "no practical place in the processes of performance".) The Folio casts lists do not look like the work of Ralph Crane that we know of from his manuscript of Middleton's The Witch, which uses two columns and preserves categorizations of gender and class that the Folio lists do not. Smith considers the claim of Gary Taylor and John Jowett that Middleton's adaptation of Measure for Measure explains the discrepancies between its script and cast list. She finds that it makes the cast list reflect the pre-adaptation version of the play in some respects--keeping Vincentio's name that was abandoned to make the play sound less Italian, and omitting the Justice whose appearance in 2.1 seems like an afterthought--while also reflecting the post-adaptation version of the play in other respects, such as giving the Bawd the personal name Overdone, making Lucio a "fantastique", and using the Middletonian phrase "Two other like Gentlemen". For Smith this constitutes a contradiction and hence "suggestions about the revision" cannot "help us resolve the provenance of the character list", although she acknowledges that it is itself evidence of that revision, being "a document of textual flux" (p 139).

    Smith remarks that the Folio's placement of the cast lists at the end of the play is odd, since other play editions printed them on the left side of the first opening where the drama starts (typically the reverse of the title page). From 1616 to 1625, 30% of play quartos had such lists, and classical and closet drama was much more likely than other genres to have them, suggesting that they were a marker of high status, literary publication. Edward Blount's pre-Folio classical drama editions routinely included cast lists. We might think the Folio's use of cast lists, then, an attempt to give it high-status associations, but if so why use them for only seven plays? The term "actor" was itself in transition at this time, as it took over from the word "player", and could mean the character or the performer or (in the cast list in the 1623 edition of John Webster's The Duchess of Malfi) both. Smith's solution to the puzzle is that the Folio is a hybrid thing: it tries to appear literary and lavish in its preliminaries but it is theatrical and economic in its main text, which does not try to resolve difficulties that readers might encounter in making sense of documents from the playhouse.

    Laura Estill's "Commonplacing Readers" (pp. 149-162) is a survey of how people copied out bits of Shakespeare into their manuscript commonplace books. An opportunity is missed to compare the manuscript readings with those of the early printed texts to speculate on whether in performance the actors departed from the script that we know and/or whether in the act of copying the commonplacing writer altered the wording. Quite often the commonplacing was intended to adapt Shakespeare's words for use by the commonplacer, especially in wooing. Similarly, Jean-Christophe Mayer's "Annotating and Transcribing for the Theatre: Shakespeare's Early Modern Reader-Revisers at Work" (pp. 162-176) considers theatre annotations around the Restoration and manuscript books based on printed editions, in particular Edward Dering's Henry 4 conflation. These are interesting essays but outside our remit here.

    In "Shakespeare and the Collection: Reading beyond Readers' Marks" (pp. 177-195), Jeffrey Todd Knight laments the habit of libraries of breaking up sammelbands, including ones having Shakespeare's works in them, to isolate and then literally clean them to make individual books; we thereby lose evidence of the history of the books' use. There arises something of a contradiction between essays here. Knight writes that "Books were for the most part issued unbound in the early part of the period, leaving the binding arrangements to the buyer and sometimes the retailer" (p. 181n18), while Lesser and Stallybrass earlier wrote that ". . . longer books were generally bound up at the time of initial sale, not, as has usually been claimed, by the purchaser after sale" (p. 127). The contradiction of course depends on where we draw the boundaries of "early part of the period" and "longer books". Likewise Knight quotes Peter Stallybrass and Robert Chartier's contention, writing in 2007, that "The texts that Shakespeare wrote were published as pamphlets, not bound books". Stallybrass's essay with Lesser earlier in this volume somewhat corrects this view, showing that the Pavier Quartos were intended to be sold in bound and unbound form at the same time. The volume editors might usefully have put Knight and Lesser and Stallybrass into dialogue with each other to make them more carefully define their boundaries so that the apparent contradiction between their views could be either eliminated or else openly acknowledged as a difference of opinion. Knight's essay ends with a couple of case studies of what David Garrick and John Philip Kemble did with their Shakespeare books and draws parallels with their acting styles: the former theatrically innovative and alert to contingency--he let his books get scribbled over by others--and the latter theatrically conservative and obsessed with stability: he had his books inlaid, "set off from history in an immaculate paper frame" (p. 193).

    Alan Galey's "Encoding as Editing as Reading" (pp. 196-211) is about how XML transcription and encoding is itself a form of close reading, as Galey learnt  when encoding for Internet Shakespeare Editions two decades ago. He rather strays into how to teach encoding as an interpretive practice. Galey argues that attending to the tricky problems of interpretation-in-encoding is a valuable way that otherwise results-oriented project work in Digital Humanities might throw up humanistic, rather than simply capitalistic, insights. An example of the benefit of pausing and considering interpretative possibilities in what Galey calls "high-friction texts" (p. 210), ones that make the work difficult, is the encoding of the early modern word proud because it is not simply a matter of choosing between prov'd  and modern proud if Randall McLeod is right that both words were active in the early modern mind in response to seeing this run of five pieces of type impressed in a book.

    The low-point of the collection is W. B. Worthen's largely incomprehensible "Shax the App" (pp. 212-229). Typical of its unintelligible writing is the sentence "Troubling this apparently constitutive dichotomy of dramatic theatre, Moss's play enforces an oscillation between the unmarked and the marked, the real and the aesthetic; it also enforces an exchange between spectator and performer as well, recalling Jacques Ranciere's critique of the 'oppositions--viewing/knowing, appearance/reality, activity/passivity'--constructing the spectator as passive across a range of social institutions, institutions which 'specifically define a distribution of the sensible, an a priori distribution of the positions and capacities and incapacities attached to these positions'" (p. 213). The essay seems to be about some postmodern postcards that were published with certain narrative words highlighted on them and with instructions on how to fold the postcard to make a tube to look through. More tedious still are Worthen's lengthy descriptions of some Shakespeare-related apps for smartphones and tablets. A picture or two would have saved the reader from sentences such as "In the tiles on the homepage of The Tempest for iPad, 'the tempest' occupies the central square, flanked on the left by 'resources' and 'experts', on the right by 'preferences' and 'bookshelf, and above and below by 'workshop' and 'about'" (p. 218) and "The play text scrolls vertically, and a range of commands appear across the menu bar at the top of the screen . . ." (p. 219). This level of descriptive detail goes on for several more pages, before Worthen turns to some other apps. From his descriptions, nothing intellectually new is achieved by any of these apps: they are just--as all apps ultimately are--miniature web-browsers that work with just one website.

    In sharp contrast to Worthen's essay is Peter Holland's informative account of "Theatre Editions" (pp. 233-248). He starts with the 1676 edition of Hamlet as performed at the Duke of York's Theatre, which is thoroughly edited to modernize the expression (but does not tell the reader that), probably because it inherited its text from William Davenant's 1661 production. Holland compares this to the 2009 play edition that accompanied Michael Grandage's Donmar Warehouse production at Wyndham's Theatre of Hamlet starring Jude Law, which reattributes some of the opening scene's lines so that Horatio, who has been away from Elsinore, is not required (as he is in Shakespeare's script) to tell people who live there just what is going on. The Duke of York's Theatre edition claims to represent the play as currently (in 1676) performed there, but in fact its cast list cannot be right as we know that at least one actor named in it had stopped performing before then. Things get even messier, Holland observes, when you think of all the versions of the script that get generated in the theatre itself as working documents. Holland gives an account of John Bell's theatrical editions of plays "as now performed" in the late eighteenth-century London theatres, and the detailed performance notes in them by Francis Gentleman. Holland ends with a sketch of some modern editions that try to comment on how the scripts might be performed. An uncharacteristic slip is that Holland dates Michael Pennington's book Hamlet: A User's Guide a couple of decades too early at "1964".

    In "Editing Shakespeare by Pictures: Illustrated Editions" (pp. 249-268), Keir Elam argues that illustrated editions are as important as non-illustrated, and that "it is possible ... to edit Shakespeare by pictures". Elam gives a history of the illustrated edition and his argument is that although of course a performance happens in space and time, a picture can at least (unlike the text) represent the space of performance. Illustrated Shakespeares often pick a key scene as "an interpretative frame for the reading of the text as a whole" (p. 252). Elam amply demonstrates his point with a reading of two eighteenth-century illustrated editions of The Tempest published by Jacob Tonson, one showing the opening storm, emphasizing the disruptiveness of the play's opening action, and one showing Miranda and Ferdinand playing chess, emphasizing the harmonic resolution at the end. As Elam remarks of the differences between them, "The choice of the final scene reverses the frontispiece's hermeneutic relationship with the play" (p. 254). Indeed, and neither can encapsulate the total effect of reading and seeing the play, which consists of conflict and resolution. Charles Knight's nineteenth-century edition introduced the innovation of pictures to explicate particular lines, such as Prospero's threat to Ferdinand "I'll manacle thy neck and feet together" (1.2.464), which Elam calls "foot-pictures", by analogy with foot-notes (p. 261).

    Andrew Murphy's "Format and Readerships" (pp. 269-284) starts with miniature editions, the smallest just 35mm by 50mm, as a way into a discussion of the significance of format. The first two editions of Venus and Adonis were in quarto but the subsequent reprints were all octavo and most followed the quarto text page-for-page so they took exactly half the number of sheets of paper and half the press time. From 1598 the Stationers' Company limited the price per sheet that books could be sold for, so the octavos of Shakespeare's poems were cheaper than the quartos. Lucrece too made the transition from quarto to octavo in reprints, which although not done page-for-page were still made of fewer sheets and hence were heaper to produce. But with the exception of Richard Duke of York in 1595, Shakespeare's individual plays were always printed in quarto and it is not clear why: did the play publishers miss a chance to sell a lot more copies spotted by the poetry publishers?

    Murphy glances at the innovative Pavier Quartos project to offer a collection, acknowledging Lesser and Stallybrass's chapter in the present volume and saying no more than that the project was ahead of its time. Whereas Shakespeare's poems got cheaper across the seventeenth century, the plays effectively got more expensive since for half of them you had to buy a whole set in the Folio. The trend towards expensive, elite Shakespeare continued into the eighteenth century with the major editions, but was challenged by some attempts to offer cheap editions that could be bought play-by-play and then bound into a complete works volume. Murphy traces the small-format publication of Shakespeare in eighteenth-century Ireland, in association with Dublin's Smock Alley Theatre, which also challenged the London publishing cartel by exporting its cheaper books to England. Next Murphy traces the relationship between cheap publication, partly driven by technological changes in printing, and rising national literacy in the nineteenth century. Then he sketches the 20th century publishing of Shakespeare.

    Leah S. Marcus's "A Man Who Needs No Introduction" (pp. 285-299) considers the emergence of the introductory essay in eighteenth-century editions of Shakespeare, with special attention to The Tempest. Interestingly, editions produced in the British colonies and former colonies tended to be more critical, or at least suspicious, of Prospero than the ones produced in Britain, and to stress the colonist's great duty of care. Kenneth Deighton's editions of the plays for a Macmillan's series called English Classics for Indian University Students which began in 1888 "had a format very similar to" the first series Arden that followed it (p. 291), which Arden series Marcus reckons "was born in part out of the experience of editing Shakespeare for the Raj" and hence bears "the imprint of colonial experience" (p. 293). Marcus makes the idle speculation that the particular decorative border that tops The Tempest in the Folio was chosen because it also appeared "in other volumes about travel and exploration, most significantly Richard Hakluyt's Principal Navigations, which included the source text for Caliban's god, Setebos" (p. 295). This, she suggests, might show that the Folio's creaters employed a "strategy" that recurred later in twentieth-century editions of The Tempest of promising their readers "a vicarious voyage of exploration" (p. 295). Well, no, because as her footnote acknowledges "The same ornamental border appears three more times in the 1623 folio (Hinman 1996): at the head of the prefatory letter to the brothers Pembroke (sig. A2r), at the head of the 'Catalogve' of plays (sig. A6r) and heading the epilogue to Henry IV, Part 2 (sig. gg8r)" (p. 295n2). A rather unscholarly note is sounded by Marcus's alteration of the Folio's "CATALOGVE" to "Catalogve" without changing the "V" to "u".

    In "Emendation and the Editorial Reconfiguration of Shakespeare" (pp. 300-313) Lukas Erne sketches both sides of the unediting debate. Some people want to emend the received text more thoroughly than has been done before, others want the exact opposite: to go back to the earliest editions. As an illustration of when emendation seems inevitable and right, Erne uses Andrew Murphy's example of "aud" where "and" seems the required meaning, in which example Murphy proposed that "a turned 'n'" best explains the apparent "u" (p. 306). Erne calls this an instance of "foul case", which in fact is something different again: a letter "u" being mistakenly distributed back into the typecase's sortbox for the letter "n" and subsequently picked up and (without being closely examined by the compositor) used where, as in "and", an "n" is wanted. According to Erne, the difference in attitude between interventionist editors, for which Gary Taylor and John Jowett stand as prime instances, and uneditors is like the difference between early modern Catholics (who felt that the word of God had to be mediated to the laity) and early modern Protestants (who felt that the laity should have direct unmediated access to the word).

    John Jowett's "Full Pricks and Great P's: Spellings, Punctuation, Accidentals" (pp. 317-331) surveys the efforts to standardize orthography in Shakespeare's time and illustrates the normalizations towards house style that were typical of printing houses and which took their output away from the author's habits as embodied in the manuscript copy. Jowett illustrates by contrasting Q1 Much Ado About Nothing (1600) with its Folio counterpart that there is absolutely no truth in the still-persistent myth ("as argued . . . by Don Weingust", p. 324) that the Folio reflects Shakespeare's personal punctuation. But when Crane transcripts were used to make Folio texts, his idiosyncratic habits seem to have been largely respected; otherwise we would be unable to detect his role in preparing Folio copy, so it seems that the compositors were more faithful to the accidentals of manuscript copy that looked carefully punctuated, which (to judge by Hand D of Sir Thomas More) Shakespeare's own papers did not.

    Anyone who objects to modernization of Shakespeare's spelling and punctuation on principle ought to at least reflect that, as Jowett shows with his examples from quarto and Folio Much Ado About Nothing, the process was already in motion amongst the early editions: the Folio texts modernize Shakespeare for their readers. Jowett does not accept as linguistically valid the strict substantive/accidental distinction that W. W. Greg made, but does see this distinction as necessary to the pragmatic work of editing and points out that it appears as early as Joseph Moxon's seventeenth-century guide to the art of printing. Jowett closes with remarks on the principles that should apply in making an original spelling edition today, arguing that all emendation--of what Greg called substantives and of accidentals--alters meaning and should not be distinguished as substantive and accidental. An original spelling edition should of course emend substantive readings where they are in error but should aim to alter punctuation only in "the exceptional cases where the text fails to maintain its own standards of consistency" (p. 331). Thereafter, if desired, the act of modenization can be applied to the spelling and punctuation without any further emendation. This principle is put in practice in the New Oxford Shakespeare, of which Jowett and the present reviewer are two of the General Editors.

    Alan C. Dessen's "Divided Shakespeare: Configuring Acts and Scenes" (pp. 332-341) starts with a brief survey of theories about act intervals in relation to the indoor and outdoor theatres, and some of the difficulties in deciding just what counts as a scene, especially in the drama of the 1560s and 1570s. Dessen speculates about the effect of particular choices, and at length about just where the room called Jerusalem is at the end of 2 Henry 4. Henry asks where is the place where he "first did swoon" (4.3.362) and, in Dessen's staging, this place is visible as a distinct area of the stage, where perhaps a sick chair has been left to remind the audience of that first swoon. In this staging, Henry asks to be taken to Jerusalem to die but rather "the king is seen taken off in a direction away from the area now" (p. 338) and the audience does not know if he ever makes it to even this reduced Jerusalem before he dies. Dessen regrets that editors often close down the possibilities in their choices of scene breaks and he worries too that departing from traditional scene breaks, even when there is good dramatic reason to do so, makes for difficulties in teaching a class in which the scripts all vary in scene numbering, even if the editors try--as several have--to include two numbering systems at once. Dessen predicts that this difficulty will disappear when everyone is using an electronic device instead of paper, since these can be made to incorporate in one text the multiple divisioning schemes of different editions. I would think the solution is even simpler than that, since with electronic editions all the students can be brought to the same moment by being told, say, "Everyone go to the first occurrence of 'Jerusalem' in the script". Who needs act, scene, and line numbers if you have an electronic text?

    In "Shakespeare's Strange Tongues: Editors and the 'Foreign' Voice in Shakespearean Drama" (pp. 342-357 ), Matthew Dimmock considers the various kinds of strange tongue that the plays refer to, including the babbling of creatures such as Caliban, but focusses in particular on the speaking of foreign languages. Dimmock calls Queen Katherine's reference to Latin as a "strange tongue" in All is True (3.1.44) a part of "Shakespeare's drama" (p. 343), and says of her lack of a discernible Spanish accent that "Shakespeare chooses on this occasion not to accentuate the strangeness of a strange character with a strange tongue" (p. 348). He makes no menttion of John Fletcher, who in fact wrote the scene he is referring to. Dimmock's relating of this reference to wider religious debates about Latin as an international language and about the correct role of the vernacular in matters spiritual is very learned, but it is a shame that in a discussion of differing voices Dimmock is silent on the highly germane topic of who actually wrote the material. Likewise in his ensuing discussion of the fear of foreigness in the Jack Cade scenes of The Contention of York and Lancaster, in which Dimmock finds in the Q/F variants "nuanced differences in the use of strange tongues" (p. 345) that are lost if you conflate them as some editors do. The trouble is, of course, that as critics we are able to detect what we think are nuanced differences in almost any variants, and the question for editors is whether the evidence for these variants being meaningful differences (rather than mere textual corruption) really is strong enough to warrant treating them as evidence of distinct versioning at this point in the two scripts.

    Dimmock thinks that in the original performances the use of Latin rather divided the audience, since some would understand it and others would just hear gibberish and would be, in scenes such as the Jack Cade rebellion which is about linguistic competence, "uncomfortably laughing at their own ignorance" (p. 346). Dimmock reckons that "Latin has little or none of this frisson in the twenty-first century, and to most is simply strange" (p. 346). I would suggest that at Stratford-upon-Avon at least such moments retain their power to divide an audience, as some but by no means all members will laugh or snort at foreign-language elements, presumably to show that they understand them.

    For the part of Owen Glendŵr in 1 Henry 4, Dimmock wonders if the early editions' spellings and elisions in Glendwr's part reflect "a subtle Welsh inflection" (p. 346), which makes it seem that he has not read John Jowett's contribution to this volume on the unlikelihood of that. Dimmock handles the tricky problem of how to modernize the regional accents of the likes of Fluellen and Jamy in Henry 5, which he thinks an editor must treat differently from the other parts and not "naturalize and normalize them" lest their "distinctive strangeness is defused" (p. 351). Dimmock details how the absence or presence of a single comma in Morroco's speech in The Merchant of Venice (2.1.24-26) changes his military nature. He is a loyal servant of the Turks if his is a sword "That slew the Sophy and a Persian prince | That won three fields of Sultan Solyman", meaning that he slew the Persian who had defeated the Turkish Suleiman the Magnificent. But he is a mercenary if his is a sword "That slew the Sophy and a Persian prince, | That won three fields of Sultan Solyman" because the comma makes the second sentence into a separate boast about defeating Suleiman. Dimmock detects a change in the punctuation of the 1600 edition's version of Morocco's speech about the respective merits of the caskets: his "previously swaggering and expansive speech has become halting in its uncertainty and self-doubt; a cascade of clauses run into and over commas and colons" (p. 355), and this he thinks editors should try to preserve since it show Shakespeare's "creative riposte to the bombast associated with the strange tongues of the 'Turk play'" (p. 356).

    Tiffany Stern's wonderfully clear contribution to the collection asks a pertinent question: "Before the Beginning; After the End: When did Plays Start and Stop?" (pp. 358-374). She finds that the trumpet being used in the Swan theatre drawing seems to have the logo of the theatre hanging from its banner. I am not sure I can actually see a trumpet in this picture: where we would expect the lines of the pipe to flare out to make the bell they, in this picture, narrow down to a point. It might just be a picture of someone hanging out a flag. Also, the legs-apart stance of the figure suggests to me carrying something cumbersome, like a flag-pole, rather than someone holding the 'to attention' poise of a trumpeter. Such a trumpet-with-banner would, according to Stern, be useful for proclaiming the forthcoming play by walking around the streets too.

    Because Ben Jonson and William Percy wrote dialogue to be spoken in between the three soundings of the trumpet before the start of a play, Stern wonders if there was normally something going on in these gaps, something not generally recorded in the surviving scripts. Shakespeare's plays-within-plays have a flourish before they start, showing that it was normal before all plays. Where the play's opening entrance is a royal one, the last flourish of the trumpet can also serve as the announcement of the monarch's arrival, as seems to happen with the beginning of The Contention of York and Lancaster. Stern reckons that trumpets were also used at the indoors Blackfriars theatre, because the residents anticipated their disturbing noise and because the word "sound" is used in one account of the Blackfriars and "the term 'sound' was most often used of the trumpet" (p. 365) and because using a trumpet would regularize Globe/Blackfriars practice. I would like to have seen some firm evidence for Stern's plausible claim that the verb 'to sound' was mostly strongly associated with trumpets.

    A typical example of the wealth of fascinating theatre-historical knowledge that Stern brings to her work is the remark that because of their instruments' military connotations, theatrical trumpeters and drummers normally had to be licensed by the crown, at the cost of a shilling a day, but that the King's men were exempted because they were already of the monarch's household. Because it was used to announce the start of the play, any trumpet used within a play was to some extent always metatheatrical, Stern reckons. Possibly the trumpeter who announces the play of Pyramus and Thisbe in A Midsummer Night's Dream--named as William Tawyer in the Folio--was also the trumpeter who announced the start of A Midsummer Night's Dream itself. Indeed, but Stern tends to overwork her evidence on such things. Because Tawyer is not mentioned in the 1600 or 1619 editions of A Midsummer Night's Dream but is mentioned in the Folio text, Stern concludes that "one of the major decisions to make when (re)mounting a play was which trumpeter to use" (p. 367). Perhaps so, or perhaps there was just some other reason concerning the nature of the copy why he is mentioned in one edition and not the others.

    Looking at the other end of a play, Stern remarks that "Finis" generally records the end of the dialogue but not necessarily the action, since the action of an immediately preceding stage direction--as in "Dance. | FINIS" at the end of Folio Much Ado About Nothing--may take some time to complete and implies its own further stage direction(s) and a mass exit before things are really over. I would have said that "Finis" in this case, and in all the others that Stern adduces, is still the end of the action: we just have to appreciate that the preceding dance, or whatever action, takes a substantial amount of time to complete. Also, I do not see the need that Stern insists upon for a mass exit before the play is over: why cannot the actors remain on stage and take their bows without a prior exit?

    We know that prayers for the monarch were usually said at the ends of plays--at least we do if we extrapolate as Stern does from the few bits of evidence to a general principle--so Stern thinks that modern play editions should include such prayers, since they modify what went before. We should treat likewise those end-of-the-play dances that can be part of the final scene and also a post-theatrical celebration of closure, and gigs and improvisations upon themes, such as Richard Tarlton was famed for performing; all these Stern thinks should be somehow part of modern editions' end-matter. Stern speculates about whether the occasional mention of a post-performance announcement of what was to be played the next day tells us that such announcement were routine. She ends by asserting that we have to think about just which of all these pre- and post-text events are really in fact part of the play that our modern editions should represent.

    In "Framing Shakespeare: Introductions and Commentary in Critical Editions of the Plays" (pp. 377-390), Jill L. Levenson gives a history of how editions have handled introductions and commentary, and how from the eighteenth century onwards editors started to explain their thinking about these problems. Levenson quickly skips to the mid-twentiethth century and some classes she took with Alfred Harbage at Harvard in the 1960s. She recalls her work on her Oxford Shakespeare Romeo and Juliet and then surveys the style and structure of some recent editions, Shakespearian and non-Shakespearian, and extols the virtues of digital mediation.

    Eric Rasmussen observes that the history of complaining about collation notes is long and varied, but that most scholars still think them essential ("Editorial Memory: The Origin and Evolution of Collation Notes", pp. 391-397). Rasmussen considers how best to present such notes' information, with a particular eye on digital technologies. In the mid-eighteenth century, Jacob Tonson printed his own version of Oxford University Press's Shakespeare edition by Thomas Hanmer, marking up every occasion on which he thought Hanmer had silently stolen an emendation from the preceding Tonson edition, and so "correction of error in Shakespearean texts became intellectual property" (p. 392). H. H. Furness recorded that he started the New Variorum series because he wanted to know, and the 1863-66 Cambridge-Macmillan edition did not tell him, not only which was the first edition to adopt each reading but exactly what each of all the major editions gave for each reading, which he clearly considered a kind of proxy indicator of the value of the competing readings. Today this kind of compendious information can, once systematically collected, be presented automatically by various kinds of visualization that make it more digestible than conventional notes do.

    In the book's final essay, "Shakespeare as Network" (pp. 398-414), David Weinberger argues that the change to a digital medium should fundamentally change how we do scholarship on Shakespeare. He uses for illustration the Functional Requirements for Bibliographic Records (FRBR) hierarchical standard invented by libraries, which distinguishes among four categories: (at the top) the Work (in the highest Platonic Ideal sense); below comes the Expression (still a Platonic Ideal, since it is not the manuscript but the form that the Work takes in a particular manuscript); then the Manifestation (say, an edition); and then the Item (say, an exemplar from the edition). Having outlined them, Weinberger then mixes up the distinctions and claims that in the case of Hamlet the author's "manuscript is itself an expression" (p. 399). This is wrong: not the manuscript but the form that the Work takes in the manuscript is the Expression. He might also usefully have noted that the FRBR is a practical standard for libraries and that bibliographical researchers have long argued over these categorizations and that there is by no means any consensus about the distinctions that FRBR draws: extreme materialists are likely to argue that there is no such thing as the unembodied Platonic Ideal of a creative work.

    Having offered the FRBR, Weingberger rightly observes that it is inadequate for his purpose--we need "a much richer model" (p. 400)--and argues that when we consider all the ways that Hamlet has appeared in culture, including popular culture reworkings, the resulting structure "would look like a network: many pieces messily joined" (p. 401). I am unclear where the messiness comes into the argument, since what he has described is merely complex. More strangely still, Weinberger asserts that this Hamlet-network is like the WorldWide Web in lacking a central authority and planning system, in having unauthorized links from one item to another, and in being inclusive because anyone can add a link. (In fact Weingberger gives a longer list of defining features, but it is repetitive and boils down to these three characteristics.) Weinberger thinks that the salient distinction is between the WorldWide Web and the "properties of paper-based objects of knowledge" (p. 401) because "Books and journals are limited vessels: relatively few manuscripts are accepted for publication, and the size of their contents is constrained by economics and engineering. Paper media divide knowledge into topics, although they include footnotes and other forms of broken links" (p. 401). I would say that Weinberger is comparing things at different levels of the two hierarchies here. He should in fact be comparing the WorldWide Web with an entire print library that aims for completeness, such as the Library of Congress or the British Library. Once we take in all publishing, not just academic publishing, his distinctions disappear: it is not hard to get published, there is no constraint on size--the Library of Congress and British Library accept virtually every new item published--and because the library has every work in it the links formed by referencing notes from one to another are not broken since the item pointed to can be called up within a few hours. Indeed, broken links are more common on the WorldWide Web than they are in a good research library.

    Weinberger is right that the costs of publishing filters what gets published--it is cheaper to publish on the web than on paper--but he overlooks the concomitant curatorial payoff: what gets published on paper has a considerably better chance of being around and readable in 100 years time than a blog posting does. For much of what Weinberger asserts he offers no evidential basis. For example, he claims that print publication privileges the lone scholar: "we have traditionally thought of scholars as individual experts, working on the issues that have seized them, almost always in a competitive (yet, we hope, collegial) war of all against all" (p. 402). This seems to simply ignore the body of collaborative scholarship, which even if we just confine ourselves to early modern drama is substantial. The Cambridge-Macmillan Shakespeare of 1863-66, the Oxford English Dictionary, the Short Title Catalogue, the Dictionary of National Biography, and the Oxford Ben Jonson are all monumental works of collaboration not lone scholarship. Is Weinberger perhaps thinking only of monographs? He describes a brave new world of networked scholarship in which discoveries made in one place are rapidly conveyed to those making related discoveries in another, and the surprise here is that Weinberger thinks that any of this is new: it is surely what the print medium started to give us 400 years ago. The Internet just accelerates this longstanding process.

    Weinberger goes on to make a series of practical suggestions for how we can best work as networked scholars, and some of it is quite sensible stuff regarding Open Access and the sharing of data. It is not all sensible, though. Weinberger suggests adding hyperlinks to one's prose to gloss basic facts and concepts, which seems naive to me. Adding, as he proposes, a link to the Wikipedia page about Iambic Pentameter the first time one mentions this form is in fact pretty much guaranteed to introduce link-rot into one's work and it is unnecessary since there is nothing to stop the reader who does not know this term from simply looking it up in the best available online resource. Today that may be Wikipedia but it is almost certain that such a link written today will not work in 100 years' time. Moreover, adding such links presupposes that we can anticipate what it is our readers do not know about, and we are likely to misjudge their ignorance.

    Better advice would be to try to write as plainly as possible and to assume that your readers know--and know much better than you can anticipate--when they have hit a patch of their own ignorance and are well able to go looking for supplementary material elsewhere. And to aid readers finding those supplementary materials, we all should give every person's full name spelt correctly and in a standard form when first mentioned (so, not "Lawrence Olivier", p. 405), and we should attribute all works to their creators (so, not failing to mention that "A Hard Day's Night" is a song by the pop group The Beatles, p. 405). Regarding the harm done to scholarship by the use of proprietary formats, including smartphone and tablet apps, and hardware, including the Kindle reading device, Weinberger is entirely correct: they inhibit sharing and linking. He is also entirely right about the positive consequences of making one's work available for others to reuse (or, mashup) in ways that one had not thought of when creating them, and how important it is that projects facilitate this by publishing their Application Programming Interfaces (or, APIs). Likewise Weinberger is right about the importance of Linked Data, the term coined by the inventor of the WorldWide Web, Tim Berners-Lee, to explain his ideas for the Semantic Web.

    There are only two essays of relevance to this review in the collection called Digital Literary Studies: Corpus Approaches to Poetry, Prose and Drama edited by David L. Hoover, Jonathan Culpeper, and Kieran O'Halloran. Both are by Culpeper, and the first is "Keywords and Characterization: An Analysis of Six Characters in Romeo and Juliet" (pp. 9-34). Culpeper explains that by keywords he does not mean words that have "particular social, cultural, or political significance", the sense used by Raymond Williams in his book called Keywords, but rather "statistically based style markers" (pp. 11). I would say that keywords is a much overused and ambiguous term so another should be sought for Culpeper's purposes. Some Americans call a full-text search a keyword-search and many Britons, especially librarians, instead think of keywords as a set of terms that summarize an essay or a book, so that the keywords for Culpeper's essay might be "Literature; Linguistics; Stylistics; Computers; Shakespeare", since none of these appears in his title but they are what the essay is about. The main consideration for Culpeper is that keywords are those that occur in a particular text at a statistically significant higher or lower rate than is usual for texts of that kind, so it means words that the author is, in this text, favouring or avoiding.

    To apply this to Shakespeare, Culpeper came up with his own tagging system to distinguish characters' speeches and he manually applied it to a century-old text of Romeo and Juliet, chosen in order to minimize the problem of variant spellings. I wonder why he did not use a more modern edition--especially one made in the last twently years, when modernization of Shakespeare has become particularly systematic and thorough--and why he thought well established tagging conventions like the Text Encoding Initiative (TEI) were not appropriate. He quotes what he calls a "sample of the tagged text" (p. 13) from Romeo and Juliet, but I cannot see any tags in it and wonder if they were inadvertantly stripped out in the book's production process.

    Finding out the six characters who speak the most in Romeo and Juliet was not hard--they are Romeo, Juliet, Friar Laurence, Nurse, Capulet, and Mercutio in that order--but which text should form the reference text representing the normal frequencies of all the words against which these characters' preferences for and against certain words could be measured? If you choose too broad a reference text--say, all writing from the same period--then you might simply detect not what is special about Romeo and Juliet but rather what is special about all drama. Culpeper settles on comparing the language of the top six speakers in Romeo and Juliet with the language of the rest of the play, defined as everything except the part being considered, and he deals with the problem that proper nouns, such as certain persons' names, are likely to appear more often in the text being examined than in the reference text simply because they are entirely absent from the reference text. Various kinds of filters can be applied to deal with such cases and one can set the threshold of just how unusually frequent or infrequent the words have to be before we consider them to be keywords.

    Culpeper lists the top 10 most frequently used words in spoken and written English today and then the top 10 most frequently used words by his six characters from Romeo and Juliet, and the top 10 most frequently used in the play as a whole; it is not clear if that means the play minus the speeches of these six characters, or including them. Mercutio's and Friar Laurence's most common words are also the top four most-common words in written English today, while the others' most common words include first- and second-person pronouns (markers of interaction with others), which is also true of spoken English today.

    Now Culpeper starts dicing the data in new ways, looking for the words each character uses at an unusually high or low rate. Romeo, it turns out, uses beauty and love at an abnormally high rate, and his own name at a relatively low rate, because the norm here is set by what everybody else says, and since he is the title character everybody else says a lot about him and he does not say his own name very often. Culpeper also looks at where in the play the characters use their words: that is, in which scenes they cluster. Juliet's most abnormally favoured word is if: she is constantly, and anxiously, trying out hypotheticals. Culpeper continues summarizing the other characters, in turn, from their word preferences, and considering whether they are evenly spread across the play or arrive in bursts in responses to particular events, such as the Nurse's six uses of woeful in six lines when she finds Juliet apparently dead.

    Next Culpeper turns to the use of pronouns by each character, revealing for example that Juliet says I and my a lot and Romeo says mine and me a lot, which suggests that he "has an egotistical streak" (p. 27), while Capulet says you a lot because his main function "is to direct other people" (p. 27). In the concluding remarks, Culpeper cautions that the text he used was not lemmatized because the software for doing this work is inaccurate and can obscure interesting results; this covers contractions too: in Culpeper's work I'll does not count as a use of the first person singular pronoun I.

    The other relevant chapter from this book is Culpeper's continuation of the same line of research: "Developing Keyness and Characterization: Annotation" (pp. 35-63). So much (in the previous chapter) for key-words: what else can be considered to be key? Culpeper applies semantic and grammatical tagging to the text to see if he can find keyness in that tagging. After some history of how people, including Roland Barthes in his structuralist phase, have marked up semantic categories in literary texts, Culpeper introduces the practicalities of the eXtensible Markup Language (XML) approach to tagging and the difficulties of getting the machine to do the tagging for you. Culpeper used Paul Rayson's WMatrix software for grammatical and semantic tagging. As is so often the case in publishing, Culpeper gives an URL for this web-service that is already out of date; the solution is just to web-search for "WMatrix", which is sufficiently distinct a name that it is easily found. To handle variant spelling in his modernized Shakespeare text, Culpeper used Rayson's widely admired VARD software. Culpeper explains how WMatrix draws on the Constituent Likelihood Automatic Word-tagging Systems (CLAWS) software to do part-of-speech tagging based on lists of words, their suffixes, and their possible parts of speech, and a table of likelihoods of such things as how often the word after an adjective will be a noun.

    Culpeper presents tables of the parts of speech favoured by Romeo, Nurse, and Mercutio, and the last of these stands out by his frequent use of indefinite articles, definite articles, and possessive pronominal determiners: he "has a tendency to use lists of noun phrases or prepositional phrases" (previous chapter, p. 25), which is what the keyword analysis in the previous chapter found too. Culpeper considers the problem that some of the grammatical categories in his lists, such as Personal Pronouns, are represented by (or at least dominated by) only one or two words from that category that are highly frequent in the character's speeches, while other categories, such as General Adjectives, represent a wider collection of words. As he puts it, ". . . with keywords one is comparing single words with single words, whereas with the grammatical categories one is comparing units that may be composed of varying numbers of types and frequencies of words" (p. 46). Moreover, where a grammatical category has only a few words in it, like Temporal Noun with day in it, a cluster of repetitions at one place in the text--the Nurse's repetitions of day when she finds Juliet seemingly dead--will artificially boost that category's apparent keyness. Culpeper's caveats become so substantial here that I start to wonder if there is any value left in the test of keyness amongst grammatical categories and my impression that there might not be is enhanced when he starts to point out errors caused by the tagger misclassifying certain words.

    Next Culpeper describes the UCREL (= [Lancaster] University Centre for Computer Corpus Research) Semantic Analysis System (USAS) software for content analysis, which takes the part-of-speech tagged output from CLAWS and looks in it for words and phrases that appear in its lists (lexicons) of words and phrases that have already been classified under 21 main headings and numerous subheadings. To make these classifications, it is helpful that the inputted text is already tagged for parts-of-speech, of course. As Culpeper explains, the verb spring is likely to be classified in a different category from the noun spring; the software also takes likely collocations into account when making its determinations. The main headings for the classification are topics such as: "Emotion", "Money and Commerce in Industry", "Movement, Location, Travel and Transport", "Numbers and Measurement", "Education", and "Names and Grammar". My list here is not the one provided by Culpeper but one I got from the USAS website, for which the URL Culpeper gives is out-of-date. The categories' names as given by Culpeper are subtly different from the ones on the USAS website. For example, I would not consider "Language and Communication" (USAS website) to be quite the same as "Linguistic Actions, States, and Processes" (Culpeper), nor "Money and Commerce in Industry" (USAS website) to be the same as  "Money and Commerce" (Culpeper). After all, concerning that last pair, the topic of begging-and-charity belongs squarely within Culpeper's version of the category but not necessarily within the USAS website version because of its qualifier "in Industry".

    Culpeper reflects on the problem of historical change in thee meanings of words, pointing out that cousin in Shakespeare's time did not denote the strict biological relationship it does today, although I do not agree with him that then it "simply denoted a 'friend'" (p. 52). Oddly, Culpeper seems to suggest that the misclassification of terms is not as serious a limitation as it might seem, since modern readers routinely make such mistakes in their understanding of Shakespeare: ". . . reading Shakespeare through the prism of the present-day worldview is the majority experience today" because "Few--if any--of us are sufficiently steeped in Elizabethan social history in order to be able to transcend our own milieu" (p. 52). I should have thought that overcoming, not perpetuating, our misreadings is fundamental to scholarship on early modern English.

    Culpeper next applies the semantic-category classification technique to the speeches of Romeo, Mercutio, and the Nurse. The first three categories (in rank order of frequency) are the same caterogies for all three characters, which is not surprising as they are the categories that function words fall into. Further down the rank order are some interesting differences, such as Mercutio's frequent references to various animals. Culpeper notes, however, that Mercutio's ranking is distorted when the classification software misreads come to mean locomotion when in fact it is most often merely a discourse marker. Likewise, Romeo's activeness is swelled by the misclassification of the verb do as a main verb where it is in fact being used only as a auxiliary, and the Nurse is classified as being concerned with "Time" only because of one burst of repetitions of day in one scene (when she finds Juliet apparently dead). Similarly, Romeo has a bunch of hits in the category "Business-Selling" but they all come from his one short scene of buying poison from the apothecary.

    Culpeper talks the reader through some more successful results from the analysis, but nothing strikes this reviewer as being especially revelatory. For instance, it is unsurprising that Romeo uses the language of liking, loving, and thought. Indeed, this last claim is instanced by Culpeper using the line "This love feel I, that feel no love in this" (1.1.179) for which I would say that feel is here misclassified as being about thought (the category of "Psychological Actions, States, and Processes'")--as we might say "I feel that we should . . ."--when it rightly belongs in the category "Emotion", or even "The Body and the Individual" since in this period perceptions are most often characterized as biological phenomena (via humoral theory) rather than mental ones. More successfully, though, Culpeper is able to quantify Romeo as introspective character who talks about his inner life, which claim I think most readers and playgoers would readily accept. Next Culpeper lists a collection of misclassifications of the Nurse's language, including the interjection faith being misfiled under "Worry" and the discourse marker well under "Good/Bad". Overall, having surveyed the results for Mercutio too, Culpeper concludes that ". . . the semantic analysis also worked rather better for Romeo than the other characters . . ." (p. 59).

    So, what can grammatical and semantic classification tell us that keyword analysis cannot? Culpeper notes that this analysis spotted that Mercutio favours "plural common nouns" and that Romeo uses a wide range of adjectives (p. 60), which keyword analysis would not have picked up because the phenomena are spread across a lot of different keywords. (That is, only the grammatical classification detects this as a pattern, and only because it alone sees what these words have in common.) Culpeper acknowledges that there is a distinct problem, however, of a grammatical category's appearing prominently in this kind of analysis when in fact only a few words from that category are present in the text, but so often that they raise the visibility of the entire category to which they belong. In general, ". . . a straight keyword analysis revealed most of the conclusions arrived at in this chapter" (p.61), so in fact the grammatical and semantic classification made little difference. Just occasionally, however, "key semantic analyses do sometimes reveal well-motivated results not predicted by keyword analysis" (p. 62).

    We turn, then, to the articles relevant to this year's survey. Bryan Crockett claims that Q1 King Lear and Q1 Romeo and Juliet are based on shorthand recordings of the plays' scripts, made during performance ('Shakespeare, Playfere, and the Pirates', SQ 66[2015] 252-85). Crockett gives a history of shorthand systems, and the evidence for their use to take down sermons to produce remarkably accurate piracies, as evidenced by comparison of the unauthorized editions with the authorized. Then he sketches the evidence that shorthand was used to take down plays, and the New Bibliographical resistance to this idea. Plays were not much longer than the longest (two-hour) sermons, and from the printed sermons it seems that preachers spoke about as quickly as actors did, at around 170-185 words a minute. Crockett reckons that this estimate for actors, based on Alfred Hart's counts, might in fact be an overestimate.

    What about the problem that theatres were such noisy places that it would be hard to take shorthand? Crockett reckons that this was not a problem in "an age in which people typically talked of going to hear rather than see a play" (p. 260). Crockett is unaware that this idea about hearing a play has been proved false in an article by the present reviewer. (A web-search for "Hearing or Seeing a Play?: Evidence of Early Modern Theatrical Terminology" will find it.) Crockett also points out that sermons could be pretty raucous events too and that unlike plays there was no opportunity to make a return visit to catch the parts that one missed the first time. What about the objection that anyone taking notes would be ejected from the theatre? There is evidence that audience members were expected to take down bits of what was said and the Induction to John Marston's The Malcontent parodies this when Sly boasts of having most of the play's jests in his table book.

    Crockett gives the main references used to support the Memorial Reconstruction theory, including Thomas Heywood's complaint that his The Rape of Lucrece was copied "only by the ear", which could mean the use of shorthand just as well as it could mean recollection by actors of their parts, and of course Heywood himself complained, in the 1639 edition of it, of stenographic piracy of his play If You Know Not Me, You Know Nobody. Heywood specifically complained that the stenography mangled the text--"scarce one word true"--but in fact the earlier Heywood editions are much more accurate than the Shakespearean suspect texts, which is what Crockett calls the bad quartos. Crockett briefly reports G. I. Duthie's rejection of shorthand as a plausible vector for the transmission of Shakespeare's plays, which he characterizes as special pleading in favour of the alternative explanations. Duthie's rejection of stenography was part of a mid-century expansion in the scope of the Memorial Reconstruction theory, which came to be used for the early editions of The Merry Wives of Windsor, The Contention of York and Lancaster, 3 Henry 6, Richard 3, Hamlet, Pericles, and King Lear.

    As a test case, Crockett takes Thomas Playfere's 1595 sermon The Meane in Mourning for which we have two different shorthand transcriptions and an authorized text. Crockett claims that these texts "have not previously been examined for their relation to shorthand" (pp. 267-8), but in fact Adele Davidson in her book Shakespeare in Shorthand The Textual Mystery of King Lear (reviewed in YWES for 2010) did discuss them (on her pp. 110-11). Like Tiffany Stern's essay arguing that Q1 Hamlet might have been based on a stenographic transcription (reviewed in YWES for 2013), Crockett's essay fails to acknowledge that much of his argument and some of his evidence was first, and recently, presented by Davidson. Playfere's own account of the making of his authorized text indicates that he has included material not present in the spoken sermon--from which the preceding pirated editions were made, as he complains--so of course we cannot easily tell how well the pirated versions represent the spoken version. Crockett offers a table of differences between the two pirated and one authorized versions of Playfere's sermon, and from the pattern of synonyms he concludes that the two pirated ones were made independently of one another using a "synonym-based method like charactery or brachygraphy" (p. 273). He finds other stigmata of stenographic recording too, like getting the first letter of a name right but the rest of it wrong. There are also sound-alike errors, such as "Henry the Second" and "Henry the Seventh"

    Crockett revisits Stern's claim that Q1 Hamlet has some paraphrasing that would be typical of a transcriber trying to catch up when falling behind the actors. Then he turns to Davidson's work on stenography in Q1 King Lear and finally acknowledges it--"my argument here is much indebted to Davidson’s work" (p. 276)--while nonetheless insisting that his conclusions are different because she supposes that a longhand manuscript of the play "authorial or otherwise, was copied out in Willis's stenography, and a longhand transcription of that shorthand version was used as printer's copy" (p. 276), while Crockett's conjecture is that a stenographer simply wrote down the script in shorthand while attending a performance. Also, where Davidson favour's John Willis's stenography system, Crockett favours Timothy Bright's charactery system or maybe Peter Bales's brachygraphy system: ". . . Q Lear has too many mistaken synonyms, for example, to have originated in a phonetically based system like stenography" (p. 276). The key signs of charactery are the use of synonyms beginning with the same letter(s) as the correct word, and mishearings. Crockett lists a bunch of these faults as found in Q1 King Lear compared to the Folio text, including the famous case of "a dogge, so bade in office" (Q1) where the line spoken was "a Dogg’s obey’d in Office" (F). In reaching his conclusion about King Lear, Crockett in fact adds nothing substantial to the case made in Davidson's book, although he differs in the detail of just where the shorthand garbling occurred in the line of transmission.

    Next Crockett turns to Q1 Romeo and Juliet, which used to be routinely claimed as a Memorial Reconstruction but which more recently has been seen as representing a cut-down version of the play for which Q2 provides the authorial original. But abridgement is not all that separates Q1 and Q2 Romeo and Juliet: there are also "verbal substitutions" and paraphrases that seem to Crockett to be symptomatic of shorthand reporting (p. 279). Again he tabulates the examples he thinks most convincing. The paraphrases, of course, could be explained by several other hypotheses, so the crucial aspect is whether we find convincing what Crockett claims are examples of "synonym-based, single-word substitution" (p. 281). They include Q1's enmitie for Q2 mutinie, mariage for nuptial, refuse for denie, happie for blessed, quiet for patient, skies for heaven, and about two dozen more examples. Noticeably, none of these examples has the first-letter-the-same feature that Crockett argued was a marked feature of the shorthand system that he suspects was used. 

    Crockett's next table lists 22 examples that do have the first-letter-the-same feature, starting with engage for endart, breedes for bodes, bowels for bosome, trim for true, and proceeding down to attach for apprehend and barge for barke. Some of these do not at first seem much like synonyms--bowels and bosome for example--but considered in their poetical contexts in the two early editions they are effectively equivalent: "Which wooes euen now the frosē bowels of the north" (Q1) versus "who wooes, | Euen now the frozen bosome of the North" (Q2). Indeed, the problem with this example is that it is if anything too poetical a substitution. If a shorthand-taker were looking for a synonym for bosom that began with b then one would expect breast to come more readily to mind than bowels, which only make sense in the wider context of common bodily analogies for the Earth. The shorthand-taker here seems rather overworked, applying sensitive artistic judgement at the same time as recording the script. Likewise "Cupid hee that shot so trim when young King Cophetua loued the begger wench" (Q1) versus "Cupid he that shot so true, Cophetua lou'd the begger mayd" (Q2). The word trim seems more than merely a synonym for true, and indeed it is arguably the more poetical reading. Certainly some editors--including the 1986-7 Oxford Complete Works, and René Weis's Arden3 edition (reviewed in YWES for 2012) and its predecessor Arden2 edition by Brian Gibbons'--follow Q1 here and explain Q2's reading as a corruption of it, so Crockett's hypothesis that Q1's trim is just a shorthand error requires a remarkable serendipity.

    Crockett lists some other things that shorthand is apt to do, such as corrupt verb tenses and mix up singulars and plurals, but of course other explanations can also account for these garblings. Likewise the lineation differences we see between Q1 and Q2. Crockett has some examples of what he claims are mishearings in Q1, such as Q2's uncharmd becoming Q1's unharm'd but I would argue that this is an error much more easily explained by written transmission, with just one letter missed off, than by aural transmission, since the sound is greatly affected. Crockett also thinks that Q2's limping could be misheard as Q1's lumping and that Q2's houre could be misheard as Q1 honor; in these claims we are a long way from Q1 King Lear's "a dogge, so bade in office" as a mishearing of F's "a Dogg’s obey’d in Office". Crockett offers a short table (just 10 examples) of other claimed mishearings, and several of these seem strained, such as ieere as a mishearing of fleere, sent as a mishearing of lent, and meere as a mishearing of deare. An hypothesis that counts these as mishearings seems capable of explaining almost any textual variation we might encounter, and so loses its claim to uniquely suit the particular circumstances for which it was brought into existence. Crockett thinks that the stage directions being fuller in Q1 than in Q2 ROM is "perhaps a sign that they were composed by someone watching a performance rather than writing a script to be performed later" (p. 284). The obvious study of this phenomenon that Crockett omits to mention is John Jowett's argument--presented in Papers of the Bibliographical Society of America 92 (1998): 53-74--that Q1's long and literary stage directions cluster in the second half of the book printed by Edward Allde in a smaller font than the first half (printed by John Danter) and that they are a space-wasting expedient, probably written by Henry Chettle from memories of performance.

    Also in Shakespeare Quarterly, Daniel Shore coins the term linguistic forms for his statements of Shakespeare's grammatical structures and argues that we need a whole dictionary of them ('Shakespeare's Constructicon', SQ 66[2015] 116-36). Shore laments that (with one exception) all the Shakespeare lexicons in existence "have no purchase on what I will call linguistic forms: abstract, variable, and productive Saussurean signs (i.e., conventionally established pairings of signifier and signified", which are "composed at least partly of categories (blanks, slots, variables) that can be filled in multiple ways" (p. 114). This is a surprising complaint, since almost all linguists no longer think that the Saussurean sign and the structuralist notion of slots-and-fillers is how language really works: most accept some form of the transformational generative grammar that Noam Chomsky introduced  60 years ago. It becomes apparent later that Shore is familiar with Chomskyean transformational generative grammar, which he pejoratively associates with "an earlier generation of critics" (p. 117); but since Saussurean linguistics is even older still it is really not clear what kind of history of linguistics Shore wants his reader to accept.

    Shore's key idea is that we should attend to linguistic forms as verbal constructions that can be expressed in abstract terms such as Subject-Verb-FirstObject-SecondObject, although Shore uses an unexplained abbreviated notation in which that form is expressed as "Subj V Obj₁ Obj₂". Shore wants there to be a reference work that lists Shakespeare's favoured constuctions in that sort of notation, and he illustrates this with some examples culled from a version of Early English Books Online Text Creation Partnership (EEBO-TCP) that has been marked up with parts-of-speech (POS) encoding at Lancaster University to make a database called CQPweb. Shore's writing is full of implicit post-structuralist thinking, as in those references to Saussure and the slots-and-fillers model of language, and also his reassurance that the reference work he envisages would be "not simply a positivist exercise", not merely "an excuse to collect and catalogue more facts about Shakespeare’s language" (which would, in his view, be a bad thing), and also in his approving quotation of Stephen Greenblatt's claim of more than 30 years ago that Shakespeare's writing is "the supreme instance of a collective creation" (p. 116). Shore has not been keeping up with recent empirical studies on the nature of early modern authorship, which challenge the ideas implicit here.

    Shore decomposes Hamlet's "O, what a rogue and peasant slave am I!" into "what [Indefinite Noun Phrase] [Copula] I" and then into a computational regular expression as "what _AT* + * * * * _VB* I", for searching within the POS-tagged EEBO-TCP. Shore explains the elements of his regular expression: "_AT*" returns all articles, "+ ****" matches one to five words, and "_VB" matches all forms of the verb to be. In short, he is looking for "what [a|an|the] [something] [am|was|were] I" in which the italic square brackets enclose alternatives separated by a virgule. It is not clear why Shore looks for all forms of to be when his subject I only allows the forms am, was, or were, unless he hopes to also capture the perfect and progressive forms (have been, am being), which would sound distinctly odd before I and generally come after it.  Shore found from his search that this linguistic form was in print 43 times before Hamlet used it, and he lists all of them. Shore does not make it clear, but in fact his search of CQPweb's POS-tagged EEBO-TCP corpus returned many more than 43 hits--for me on 26 January 2017 it returned 880 hits--so he presumably had to manually whittle away all those for which the [something] in the middle is not an insult, such as "what a case were I", "what the same was I", "what the drift of it is I", "What a great man am I", "what the event may be I", and similar expressions.

    There is nothing in the encoding applied to the corpus that marks up insults as a general category, so Shore's claimed linguistic form is in fact a manual interpretation not something found by the computer. Although it is true that many of the 880 hits seem to have an insult in the [something]  slot, others do not: "what the end is I", "what the cause should be I", "what the name thereof is I", "what the points be I", "what a brave man am I", "what a thing learning is I", "what a yeoman is I", and many more. In fact, Shore should probably have excluded the is form of the verb to be, since is I at the end of the string does not form the same kind of grammatical construction as those ending am I and was I:  the subjects of to be seem, in such cases, to be the [something] in the middle and not the I at all. Shore acknowledges that a good quality rather than a bad may fill the [something] slot but finds that such cases constitute "a deviation from the semantic and pragmatic conventions established by repeated use" (pp. 119-20). I would say that not until he has counted up the good and bad terms used in all 880 cases is he entitled to refer to a pattern established by repeated use: these things are quantifiable, but Shore gives only the 43 that support his claim. Shore finds the same structure of utterance six more times in the 1623 Folio: "what a thrice double Ass Was I", "what a beast am I", "what a fool am I", "what a beast was I", "What a wicked Beast was I", and "What an Ass am I".

    Shore next turns to "Unhouseled, disappointed, unaneled" and constructs a search for three past participles in a row (the search term is "_VVN * _VVN * _VVN") to find similar constructions in EEBO-TCP and claims "11,114 matches in 5,325 texts". I make it 11,144 matches in 5,325 texts. To make sense of this large set, which contains many repetitions of the same phrase, Shore sorts the results by frequency, which puts as the top four most popular hits: "hanged, drawn and quartered", "revealed and made known", "drawn, hanged and quartered", and "published and made known". This result seems to me to vitiate Shore's central claim to be finding distinct linguistic forms, since the second and fourth most-popular results contain two not three concepts. That is, sometimes two past participles in a row (such as made known) are really one compound past participle. Shore simply ignores this result in order to assert his discovery that "With remarkable consistency, three passive past participles were in Shakespeare's day used in legal or liturgical formulae . . ." (p. 123).

    Alexander Schmidt's Shakespeare-Lexicon of 1874 is, according to Shore, a predecessor of the kind of lexicon he is advocating, but of course its print form it has limitations that the digital form can partially overcome. If such a lexicon were constructed just for Shakespeare's writing, it could be linked to instances in the works of each linguistic form, which would help overcome the limitation that otherwise the forms can be identified only by the arcane names that linguists give them, such as "Degree_qualifier_realization" (p. 128). As his final example, Shore takes "To be or not to be", generalized to "to X or not to X, where X is an infinitive verb" (p. 130) for which the CQPweb search term is "to _V+I * or not to _V+I" that for Shore produces 190 hits before the year 1600. For me it produces 193 and I cannot explain this difference. Sorted by frequency of occurrence, the most common string that fits this template is "to do, or not to do" and the second is "to be or not to be" with 10 occurrences. As Shore acknowledges, we cannot tell if Shakespeare wrote "to be or not to be" as a filling-in of a mental template of the form "to X or not to X" or was just repeating one of those 10 pre-1600 instances of "to be or not to be" that he perhaps read or heard.

    Shore ends with a rather uncharitable summary (as it seems to a non-specialist like me) of how Noam Chomsky divided lexicon from grammar in his 1957 classic Syntactic Structures and how subsequent research built upon this. According to Shore, Chomsky's theory of language is guilty of "universalism, essentialism, and biologism" (p. 133), but he does not elaborate this accusation or substantiate the charges. Shore seems to dispute a key discovery of Chomsky's early contribution to the debate, which was that contrary to the view of the behaviourists there must be some innate language machinery in the brain because--and this has been empirically demonstrated--children just do not receive frequent-enough positive and negative responses to their early language experiments to account for their prodigiously rapid acquisition of language skills. Shore apparently wants to revive the behaviourist account: ". . . there is no need to posit a universal grammar, as Chomsky does, since the knowledge of these culturally inherited forms is sufficient for the combinatory work of assembling words into the theoretically infinite possible utterances of human language" (p. 134). While it is true that Chomsky's work sometimes divides linguists, it seems most unlikely that many will return to the view that language is an acquired "knowledge" of "cultural inherited forms", since the empirical evidence against that is overwhelming.

    Valerie Wayne tries to make sense of the placing of particular plays within the Folio ('The First Folio's Arrangement and its Finale', SQ 66[2015] 389-408). She decides that the first, The Tempest, is indeed Shakespeare's farewell to the stage and the last, Cymbeline, is "a farewell to his plays, given its highly recapitulatory character" (p. 390). What the Folio especially highlights by its placing of plays at the start and end of each of its three sections is the late, unpublished plays: those that the publisher could easily get the rights to, and those available in good texts. That is, the previously printed plays were tucked away in the middles of their sections to be less visible. Wayne retells a lot of the known narrative of the printing of the Folio and surveys without conclusion the competing theories on such things as the provision of scene and act breaks and the problems of shoe-horning plays into the Folio's three genres. Jonathan Bate suggests that Cymbeline earns its place at the end of the Folio because it shows off Shakespeare experimentation with form, pointing out that it fits Polonius's category of "tragical-comical-historical-pastoral". Wayne lists the ways in which Cymbeline revisits character types and plot actions that Shakespeare had used before, and the debates about whether this is mere recycling of old ideas or an inventive reworking of the major theatrical ideas of his career. She sees it as the latter.

    Finally for this journal, Ross. W. Duffin tracks down a missing song (''Concolinel': Moth's Lost Song Recovered?', SQ 66[2015] 89-94). Asked to sing by Armado at the start of Act 3 in Love's Labour's Lost, Mote says or sings simply "Concolinel" (in Q and F) and no-one has satisfactorily explained this mysterious word. From a suggestion by Claire van Kampen, Duffin reckons that Mote sings the French song "Quand Colinet faisoit l’amour" that appears in a collection printed in Rouen in 1602. Duffin gives the French words and an English translation. The song is about male genitalia and refers to the penis's "belle iaquette" (pretty jacket) which suggests the play's Jaquenetta. From the surrounding dialogue in the play, the tune for the song would seem to be also French, and the words of the refrain appear in the French play La Comédie de Chansons (1640) and are followed by "Est-ce Mars" which was the French name for the tune known in England as Sellenger's Round, which fits the words of "Quand Colinet faisoit l’amour" quite well.

    Michael Ingham and Richard Ingham show that Shakespeare, like other dramatists of the 1590s, used archaic syntactic forms that made their works sound old fashioned, but that this habit fell off sharply in the works off the early Jacobean dramatists ('Syntax and Subtext: Diachronic Variables, Displacement and Proximity in the Verse Dramas of Shakespeare and His Contemporaries', Shakespeare 11[2015] 214-32). They observe that compared to the relatively patriotic and "conservative idealism" (p. 215) of plays of the late Elizabethan period, plays of the early Jacobean period seem to critique the court as a place of corruption and to feature pessimistic, cynical anti-heroes; just how topical these were meant to be is a disputed point. In this context, we can ask whether Shakespeare kept up with his contemporaries regarding two linguistic features: the "use or avoidance of inversion with subject pronouns in declarative clauses, and the use or avoidance of auxiliary do in interrogatives" (p. 218). Subject pronoun inversion ("Thus have I . . ."), now formally called Verb-Second (V2) construction, and avoidance of auxiliary do ("Lives she?" instead of "Does she live?") are archaisms and can be used to distance a play from the time of its own performance; that is, they can be anti-topicality markers. Ingham and Ingham counted these features in history plays and tragedies, by Shakespeare and others.

    Specifically, Ingham and Ingham took 18 late Elizabethan plays (including four by Shakespeare) and 19 early Jacobean plays (including four by Shakespeare) for their tests. Those are my counts from the lists of plays given in Appendix One; strangely Ingham and Ingham elsewhere refer to their "corpus of sixteen late Elizabethan plays (including four by Shakespeare) . . . [and] eighteen early Jacobean plays (including four by Shakespeare)" (p. 218). I cannot account for this discrepancy. Ingham and Ingham's list of their sources suggests that they did their counting by hand on modern printed books. The results show that V2 usage fell markedly from the late Elizabethan to the early Jacobean period, and that although Shakespeare followed this trend he lagged behind, using the form more than twice as often as his early Jacobean contemporaries did. That overall amount ("twice as often") is actually applicable to only one kind of inversion. Ingham and Ingham distinguish multiple kinds, in declaratives and interrogatives, and they do not explain their various abbreviations--XVS, XSV, VS, and V2--well enough for this reader to fully follow their detailed tables and graphs. For example, Figure 1 has a legend that refers to "XVS" and "VS int'rog" but a caption that refers to "Percentage use of VS in declaratives and in interrogatives", and it is not obvious how the abbreviations of the former relate to the one in the latter.

    Ingham and Ingham's main conclusion is that Jacobean dramatists used fewer archaisms than their predecessors because the earlier writers chose to use archaisms to evoke settings that were distant in time and/or place, as part of their conservative idealism, while the later writers wanted to evoke the here-and-now. Notice that Ingham and Ingham consider subject pronoun inversion and avoidance of auxiliary do as conscious archaisms in the 1590s, not as simply the normal speech of the time. That is, they think that the tragedies and histories of the 1590s used these features as stylistic markers chosen for particular artistic effects. To really establish this point one would want to see the data regarding the same linguistic features for earlier decades. In their conclusion Ingham and Ingham claim that "In the late Elizabethan period their [the dramatists'] syntactic choices in these constructions was shown to diverge sharply from the ordinary language of the time . . ." (p. 228) but this was not in fact shown by the present study. Rather this was accepted as the finding of others' studies, referred to briefly at the bottom of page 220. They interpret Shakespeare's slowness to adopt the new language preferences as a result not only of his maturity but also his cautiousness about sounding topical.

    Also in a linguistic vein, MacDonald P. Jackson uses rare-word frequency analysis to broadly confirm the 1986-87 Oxford Complete Works's chronology for Shakespeare's career ('Vocabulary Links between Shakespeare's Plays as a Guide to Chronology: A Reworking of Eliot Slater's Tables', Shakespeare 11[2015] 446-58). In the 1980s, Eliot Slater counted the frequencies of rare words in the plays of Shakespeare, with rareness defined as those words appearing in at least two plays and appearing up to 10 times across the canon. Unsurprisingly, the plays that shared the largest numbers of rare words tended to be ones that we already thought, for other reasons, were written about the same time. That is, Shakespeare's favoured rare words drifted gradually over time. To form his baseline of how often each word should appear in each of Shakespeare plays if chronology were not an influencing factor, Slater used Alfred Hart's counts of how many different word types (not tokens) there are in each play. But Slater made an error in his calculation of expected rates of rare-word connections between any two plays. Say the two plays are called P and Q. Slater divided the number of words (types) in P by the number of words (types) in the canon (excluding P) to get his sense of what proportion of the canon P represents and then multiplied that by the number of link-words in Q to establish how many link-words P should share with Q if chronology were not an influence. So, if we establish that P contains 2% of the Shakespeare canon and Q has 500 link-words to all other plays across the canon then there should be 10 (= 2% of 500) rare words linking P to Q.

    This sounds mathematically reasonable, but it is wrong. The problem is that Slater assumed that P's proportion of the total canon (2%) should be the same as--indeed, can be used as a proxy for--its proportion of the set of all link-words. This is not a valid assumption and in any case we do not need a proxy since we know what all the link-words are and can calculate a play's proportion of that set rather than its proportion of the whole canon. So the correct formula for the expectation--and this revision is Jackson's main contribution here--is the number of link-words in play P divided by the total number of link-words across the canon (excluding the ones in P), to give P's contribution to the set of link-words, which quotient is then multiplied by the number of link-words in play Q to give the expected number of P-Q link-words if chronology were not an influence. Once we know the expected number of links, we can count the actual number of links found and apply standard statistical metrics to decide how significant the difference is. That is, we can calculate how often mere chance variation would produce such a difference and then decide for ourselves if the answer is so small--that is, chance would rarely do it--that we should assume that chance is in fact not the cause of the difference. This final determination is a vexed question in mathematics but does not affect the results here since the same criterion is being applied to all the play pairs and we are interested only in the pattern of relationships amongst them all.

    Jackson revises all of Slater's tables in the light of his improved formula and the result is to weaken the links between certain pairs of plays and strengthen the links between others. Genre naturally effects the results, but chronology is clearly a stronger determinant of which plays share the most rare words. The main conclusion is to confirm the standard chronology of the 1986-87 Oxford Complete Works of Shakespeare. Some caveats are given, perhaps the most important of which is that Jackson did not redo all of Eliot's ground-work in the light of subsequent discoveries about co-authorship, the area to which he has made such extraordinary contributions since the 1970s. To redo everything in the light of co-authorship "would be a mammoth task" (p. 453). Despite this, some conclusions are worth highlighting. Jackson's new tables point to The Troublesome Reign being a source for King John and not derivative of it, but the results cannot tell us if A Shrew preceded or followed The Shrew. Several other plays shuffle a year or two from their usual dates without jumping over a neighbouring play. Regarding Edward 3, the results confirm its co-authorship, since the parts we already believed to be non-Shakespearian have strong links to 1 Henry 6, which is now also believed to be very largely not by Shakespeare. Jackson's revision of Slater's work also confirms the early dates for Venus and Adonis and Lucrece and the late date for A Lover's Complaint, which has rare-word links to All's Well that Ends Well and Cymbeline.

    John Freeman reckons that modern editors of Shakespeare should make a distinction between the words O and Oh based on what discourse analysis tells us about their difference in meaning ('Re-proofing the 'Zero Part of Speech' in Hamlet', CompD 49[2015] 289-312). Freeman starts with the assertion that O and oh have different connotations: the former is "literary" the latter "conversational" (p. 290). I must confess that I have trouble following the logic of Freeman's sentences, and some of them I find baffling. For instance, he writes that ". . . Norman F. Blake demonstrates how the markers 'why' and 'what' function in Shakespeare's plays according to prescriptions laid down by modern discourse analysts" (p. 290). This sounds like the tail wagging the dog; I should have thought it the job of discourse analysis to describe not prescribe what happens in the creative writing. I am not clear why Freeman mentions only O and oh in this essay and not Oh and o, which also appear in the early editions of Shakespeare, if he thinks (as he seems to) that the capitalization makes a difference.

    Freeman asserts that "Genre played a decisive role in compositors' choices of one form over another, histories and tragedies tending toward the O-side of the ledger and comedies toward the oh-side" (p. 291), and his first table in support of this claim counts the occurrences in the 1623 Shakespeare Folio by genre. As well as omitting counts for Oh and o--perhaps Freeman includes these within the counts for oh and O respectively--the table gives only the percentages for each form and not the raw numbers, so one cannot tell if the counts are absolutely large or small. Freeman gives no account of how or whether he adjusts for the Folio compositors making their own decisions about the form to use rather than simply following the form given in their copy, which we now know was in every case a scribal transcript or a preceding edition, not authorial papers.

    Turning to the quartos, Freeman gives raw counts so it should be possible to figure out how he treats Oh and o. I find that the 1597 quarto of Richard 2 has eight occurrences of O, 27 of Oh, none of o, and one of oh, while Freeman's table reports 10 occurrences of O and 28 of oh. Presumably Freeman's 28 for Oh results from his silently including the single oh, in which case our counts agree. It is not clear why Freeman finds 10 occurrences of O where I find only eight: could he perhaps be accidentally including the two within "L O N D O N" on the title page? Freeman does not state how he made his counts; mine were made by copying and pasting into a text editor the transcription of the quarto provided by the Internet Shakespeare Editions and then applying case-sensitive and whole-word-only searching to it. In the 1597 quarto of Love's Labour's Lost I find 73 occurrences of O, two of Oh, six of o, and none of oh, while Freeman reports 73 occurrences of O and two of oh. I cannot account for Freeman not recording the six occurrences of o that a manual check of context shows are present: "But o but o", "o a most daintie man", "fiftie sores o sorell", "o tell me good Dumaine?", and "the other two concludes it o u". Perhaps he applied some additional criteria not stated.

    Freeman's main conclusion is that across the quartos, the preference for O rather than oh goes up over time from 20-30% of occurrences in the early 1590s to 80-100% after 1600. I tested a couple more of Freeman's quarto counts and our differences are small enough to grant that his general point stands: over the period 1595 to 1605 the preponderance of O over oh in Shakespeare editions did indeed rise markedly. The difficulty is in what to make of this fact. It might be that, as Freeman claims, "early modern compositors were indeed aware of the distinction between the more literary O and the baser, conversational oh. When it came time to upgrade the status of dramatic productions, their preference for O in this regard signals that awareness" (p. 294). But Freeman has hardly attempted to exclude other possibilities, including most simply Shakespeare choosing to use O more often.

    When Freeman starts digging into particular cases it is hard to follow the evidence because he silently switches between editions. Having just recorded the 73 occurrences of O in the 1597 quarto of Love's Labour's Lost he subsequently refers to the play's "seventy-six Os" (p. 295) because he is now referring to and quoting the Folio text instead; the only way to figure this out is to compare his quotation with the quarto and Folio. Freeman thinks that modern editors should distinguish (using discourse analysis as their guide) between O and oh rather than flattening everything to an "unrelentingly elevated vocative" O. Freeman believes that this can be conveyed in performance, with O encouraging a elevated style and oh (and presumably Oh too) a colloquial one.

    Arguing explicitly against those who think that O and oh were interchangeable in Shakespeare's time, Freeman offers a table of 16 different uses for oh (or rather, in this table, "Oh"), including "initiates a soliloquy", "a speaker repairs or completes the other speaker's statement", and "the refusal of a compliment". On the basis of his table, he proposes alterations to the forms in O and oh in Folio Hamlet. Again Freeman uses opaque phrasing that makes it fatally unclear just what he thinks are the agencies at work in the early printed editions he is concerned with: ". . . plays that existed only in the Folio format did not require compositors to make choices between the two forms . . . (p. 302). Why should the existence of early editions give the compositors a choice to make, and why would the absence of early editions take that choice away? Certainly the compositors might consult an earlier edition if one existed and they had access to it, but the case for that happening needs to be constructed play-by-play; we cannot just assume that because an earlier edition existed the Folio compositors consulted it and thereby had a choice to make between their copy-text form and the earlier edition. And indeed since their copy text seems to have been in some cases an exemplar of an earlier edition, Freeman's implied distinction between Folio copy and preceding editions breaks down. Moreover, even without an earlier edition to consult the compositors still had a choice to make, since no matter what their copy-text reading they could set O or oh if they wanted to. Freeman's entire argument here is unsatisfactorily laid out. Indeed, his phrasing is baffling at times, as when he refers to "the modern-day compositor who must choose between these competing forms" (p. 302). There is no such person in modern commercial publishing, as the setting of type by hand occurs only in niche craft printing and in education.

    Freeman handles more particular moments from Hamlet using the rules he has constructed. What is most striking is that even if the rules Freeman claims to derive from discourse analysis are indeed the ones that operated in early modern English, he offers no evidence that modern readers can make the distinctions between O and oh that he recommends. Indeed, it has not been shown that modern readers notice this difference at all nor that they make any distinctions in their understanding of the language in response to it. At this point, Freeman introduces prosody as an additional consideration, approvingly quoting John Earle's claim that O is enclitic and oh! is independently monosyllabic. Oddly, Earle seems to think that being enclitic means that O "has no accent of its own, but is pronounced with the word to which it is attached, as if it were its unaccented first syllable" (p. 305). But OED gives that forward-leaning role as its definition of proclitic words, and defines enclitic words as those that "cause a secondary accent to be laid on the last syllable of the word which they follow", so that they are backward-leaning. Thus the whole argument here just gets more murky and more unsatisfactory still.

    On the fringes of our concern in this review is Gary Taylor's argument that Shakespeare had an illegitimate third daughter (possibly called Violante) for whom he wrote his part of Cardenio ('Shakespeare's Illegitimate Daughter', Memoria di Shakespeare: A Journal of Shakespearean Studies 2[2015] 177-94). Now that we know that Shakespeare and Fletcher collaborated on Cardenio, that makes Fletcher the man Shakespeare had the most sustained partnership with: three plays not two. We have now to take Lewis Theobald's preface to Double Falsehood seriously, with its claim about Shakespeare's illegitimate daughter. Rowe mentioned a third daughter besides Susannah and Judith, but said nothing of her and she does not appear in the Stratford records. Theobald wrote that he heard from the "Noble Person" who gave him one of his manuscripts of the play that it "was given by our Author, as a Present of Value, to a Natural Daughter of his, for whose Sake he wrote it". Theobald could have, but does not, connect this natural daughter with the third daughter mentioned by Rowe (whose account Theobald certainly knew well). The illegitimate children of a woman were still the legitimate children of her husband, as Shakespeare dramatizes in King John.

    If Shakespeare's illegitimate daughter's mother was unmarried, he might be named as the father on official documents, but any documents are lost. But if the mother was married, then her husband would be named as the legal father even though Shakespeare was the biological father. There was no incentive for a man to acknowledge his illegitimate daughters although acknowledging illegitimate sons might be useful if no legitimate sons survived; hence it would not be surprising that Shakespeare's will acknowledges no illegitimate daughter. Shakespeare's biography, including as it does premarital sex and various legends of prowess, is consistent with having an illegitimate child, as is a biographical interpretation of Sonnets.

    There have been attempts to explain away Theobald's reference to Shakespeare's "natural daughter", for example that it refers to William Davenant's mother or that its punctuation is misleading. Davenant's mother was too young, and the punctuation argument takes out the comma to make "a Natural Daughter of his for whose Sake he wrote it", meaning his patron's illegitimate daughter. But Theobald was a careful editor and demonstrably took care in revising this sentence to express himself correctly. But what about Theobald's claim that the play was "a Present of Value" to his illegitimate daughter; is that plausible? If the value is meant financially, this is not implausible: a private transcript of an unpublished play was valuable to publishers and in manuscript circulation and Shakespeare might even have made a deal with the King's men about his share of the take in this case. But the value might also be only emotional, not only in the physicality of the object as a father's handwritten document but also in the fact that he wrote it for his daughter. In the play, Violante's father is entirely absent (as is Violenta's father in Folio All's Well that Ends Well) and perhaps this was the name of Shakespeare's illegitimate daughter that he put into Cardenio. Taylor's article's chief significance for our purposes here is that it reinforces the growing acceptance of Theobald's testimony as reliable evidence about Shakespeare's life and works, which bears on the growing acceptance that his Double Falsehood is an adaptation of Shakespeare and Fletcher's lost original, Cardenio.

    Two items from a special issue of the journal Contemporary Theatre Review on "The Politics, Processes, and Practices of Editing" are relevant to this survey. In the first, Gordon McMullan discusses the various ways in which editing is essentially political, and writes quite a bit about the practical and commercial considerations of doing a new complete works of Shakespeare, the Norton3 for which he is a general editor ('Reflections on the Politics of Editing a Complete Works of Shakespeare', Contemporary Theatre Review 25: Special Issue on the Politics, Processes, and Practices of Editing[2015] 76-79 ). Rather unfairly, to my mind, McMullan questions the democratic politics of the Open Access (OA) movement--calling it only "ostensibly democratic" (p. 77)--and hence questions the politics of the exceptionally fine and entirely free Internet Shakespeare Editions, on the grounds that ". . . university managers in the UK have begun to manipulate Open Access so as to control intellectual property and academic freedom" (p. 77). It is not clear what McMullan is alluding to here--perhaps the difference between Gold OA in which the author pays for publication and Green OA in which she does not--but it seems to have nothing to do with the Internet Shakespeare Editions, which is genuinely free in all senses. The only specific textual question in Shakespeare that McMullan engages with is whether Othello likens himself to a "base Indian" (1622 quarto) or "base Judean" (1623 Folio), for which his conclusion is "we cannot decide" (p. 79). He does not explicitly say so, but the corollary is surely that since the Norton3 treats Q and F Othello as distinct versions of the play the reading at this point in the modernized text should follow the reading in whichever of the two early substantive editions was the copy text for the rest of the Norton edition. That would seem the practical instantiation of the Norton3 edition's (entirely reasonable) faith in the individual integrity of the early editions.

    In the other article from this special issue, Catherine Silverstone gives an account of editing Titus Andronicus for the Norton3 complete works ('Textual Editing', Contemporary Theatre Review 25: Special Issue on the Politics, Processes, and Practices of Editing[2015] 64-7). She credits "Roland Barthes, Jacques Derrida, and Michel Foucault, among others" (p. 65) with puncturing "the fantasy of returning to a 'master text', whereby the printed text 'should ideally give the reader unmediated access to the astonishing force of imaginative power that was the mind of the dramatist'" (p. 65) as Stephen Greenblatt puts it. According to Greenblatt, even the first edition of a Shakespeare work "has been edited" (p. 66). It is clear what Greenblatt is trying to convey, but there is a world of difference between what happened to Shakespeare in the seventheenth-century printshop and what Nicholas Rowe inaugurated in his 1709 edition, the first to be edited in the usual sense of that word. Properly, what Greenblatt and Silverstone mean is that the first editions were mediated, not edited. Silverstone briefly sketches some of the interesting variations between the early editions of Titus Andronicus, but with so little detail that nothing can be said about the difficult editorial choices she had to make for her edition: we are simply promised that she comments on them.

    Shakespeare Bulletin carried two articles of relevance to us. In the first, Clare McManus asks why the Willow Song in Othello 4.3 was omitted from the 1622 quarto and included in the 1623 Folio text of the play (''Sing it Like Poor Barbary': Othello and Early Modern Women's Performance', ShakB 33[2015] 99-120). McManus reckons that the song shows off English boys' ability to perform in the Italian tradition of the abandoned lamenting woman singer. We know that the Willow Song existed as a ballad before the play was written, but I am not clear why McManus thinks that Desdemona's breaking off her singing with "Let no body blame him, his scorne I approue. | (Nay that's not next. Harke, who is't that knocks?" (TLN 3021-22) shows that "her part requires her to adapt the existing ballad" (p. 105). Misremembering the ballad--if that is what Desdemona is doing--is not the same as adapting it; and indeed Desdemona could be mistaken about her misremembering.

    McManus explores editorial responses to the Folio's use of "&c." where we might want to fill in some words from the existing versions of the ballad. She argues that the spoken words to Emilia--"Lay by these", "Prythee high thee: he'le come anon"--show Shakespeare's "interweaving speech and song in a careful, productive tension that brings both vocal modes together" (p. 107), and explores how a modern editor might lay out the text to show that. I am not sure there is much of a problem here: the interweaving is clear enough in the Folio and the editor has only not to obscure it. McManus thinks that at this point the script is particularly hard for a boy actor to perform--the changes of register and addressee being difficult--and gives some credence to the idea that the 1622 quarto, which omits the song, reflects what was done when the company did not have a boy good enough to pull it off. She thinks that the Willow Song is a virtuoso performance intentionally included to show that the boys of the English stage could perform distraught femininity as well as the female actors of the Italian stage, and could do it with a mere English ballad.

    In the second article from Shakespeare Bulletin, Lezlie C. Cross argues that Howard Horace Furness's New Variorum Shakespeare edtions of the nineteenth century were much more interested in performance than has been recognized; indeed they were a century ahead of their time ('Acting in the Paratext: Theatrical Material in Horace Howard Furness's New Variorum Shakespeare', ShakB 33[2015] 191-213). She sketches the prior Variorum tradition of accumulated glosses, which occasionally referred to performance, but did not (as Furness does) use evidence from performance to help establish the text. Actors, Furness believed, could solve textual problems with a bit of business in a way that armchair-bound critics could not. Cross catalogues a series of examples of this from contemporary stage practice employed by Furness. Not unsurprisingly, this sometimes made for absurdly precise accounts of how a character behaves, although Cross insists that Furness was merely describing stage business to give readers a sense of the options and quotes his disavowing a prescriptive intention. An odd note is sounded in all this when the actor Tita Brand thanks Furness (in a letter of 1906) for his use of Folio readings for As You Like It rather than, in Cross's words, "the traditionally used Quarto readings" (p. 206). There was no quarto of As You Like It: the Folio is the only substantive edition.

    Before considering a run of articles on the computational and statistical investigation of authorship, it is worth considering an extended argument that all such research is futile and discovers nothng (''I'll Tell You What Mine Author Says': A Brief History of Stylometrics', ELH 82[2015] 815-44). Jeffrey Kahan starts with the patently untrue assertion that according to the Royal Shakespeare Company (RSC) edition of the Collaborative Plays edited by Jonathan Bate and Eric Rasmussen "William Shakespeare can now be found in Arden of Faversham, Locrine, Edward III, The Spanish Tragedy, Thomas Lord Cromwell, Sir Thomas More, The London Prodigal, A Yorkshire Tragedy, Mucedorus, and . . . Double Falsehood" (p. 815). Kahan thinks that the RSC edition is called Shakespeare and Others but only because he has misread the authorship attribution on its title page as if that were the title. Of course, the RSC edition makes no such claim. As reported in the review in YWES for 2013, Will Sharpe's expert essay on the topic that is included with the edition rejects the Shakespearian attribution of Locrine, Thomas Lord Cromwell, The London Prodigal, and A Yorkshire Tragedy as most unlikely or virtually impossible. Like other experts, including MacDonald P. Jackson, Sharpe thinks that Mucedorus may have some Shakespeare in it, and like a great number of experts he thinks that the cases for Shakespeare's hand in Arden of Faversham, Edward III, The Spanish Tragedy, and Double Falsehood have been proven.

    Having set up this straw man, Kahan offers to knock it down: "Before we embrace what is the largest expansion of the canon since the second impression of the Third Folio (1664), we might . . ." (p. 816). But first he gives a long digression on Isaac Newton, Edmond Halley, and the eighteenth-century Shakespearian William Richardson. Turning to Edmond Malone, Kahan misapplies the notion of entropy as if it merely meant decline, referring to "the gradual entropy of Shakespeare's use of couplets" towards the end of his career (p. 819). The history of ideas is not Kahan's strong suit: he thinks that calculus is "the study of . . . [the] motion of objects" and that the ancient Greeks invented it (p. 819). Calculus is in fact the mathematics of rates of change in any phenomena and it was invented in the seventeenth century. Kahan has some fun pointing out the wilder claims made by F. G. Fleay on the basis of counting metrical features in the plays of Shakespeare and his contemporaries, without mentioning the many things that Fleay got right using his newly invented methods. From this, Kahan bizarrely concludes that science itself is unreliable: "Fleay might have been inaccurate in places, but his technique was purely scientific. Of course, scientists are often wrong" (p. 823). Worse, Kahan seems to reject the idea of all scientific innovation ". . . things that we deem scientifically knowable must be consistent with the rules already in place" (p. 823).

    Kahan objects to what he thinks is the innumeracy shown when A. W. Pollard in Shakespeare's Fight with the Pirates claims that (in Kahan's words) there are "180% more errors in the Q1 of Richard II than in Q2" and that "this difference amounts to Q2 being 'three times as incorrect as Q1'" (p. 823). Kahan reckons that being three times as incorrect "is an increase of 300%" not 180%. In fact Pollard, not Kahan, is arithmetically correct here, since what Pollard actually wrote is that Q2 "has been found to add about 180 per cent. of new errors to those originally made, so that it is nearly three times as incorrect". If the quantity of error in Q1 is x and as Pollard claims Q2 adds 180% more to x then Q2 contains x+1.8x errors which is 2.8x or, as Pollard rightly says, Q2 is "nearly three times as incorrect". Kahan begins his essay with the assertion that "In short, this is a paper about math. This is a paper about Shakespeare" (p. 816) but he turns out to be poor at arithmetic and close reading.

    Kahan's objections to others' mathematical work treat all investigators alike, lumping in Cyrus Hoy and David J. Lake, whose groundbreaking work has since been extensively corroborated, with A. Q. Morton, whose work has not. Kahan's approach is akin to rejecting all cancer research because some of it has been done badly. Having stated earlier that calculus was invented by the ancient Greeks, Kahan now (correctly) attributes it to Isaac Newton and Gottfried Wilhelm Leibniz in the seventeenth century and criticizes W. W. Greg for using the word in the title of his book The Calculus of Variants. In restating just what calculus consists of, Kahan writes that it "has two major branches, differential calculus (concerning rates of change and slopes of curves), and integral calculus (concerning accumulation of quantities and the areas under curves)" (p. 824). These words are lifted verbatim and without attribution from the Wikipedia page on calculus, accessed by this reviewer on 20 December 2016. (I state the date in case the page has since been updated and anyone wants to confirm my claim by looking at its 'history' log.) When undergraduates do this we investigate it as suspected plagiarism; the only defence I can imagine is that Kahan first wrote the Wikipedia page and then repeated himself in this essay.

    Next Kahan criticizes Hugh Craig and Arthur Kinney for dismissing Caroline Spurgeon's goal of "listen[ing] to the sounds of the world" Shakespeare lived in, figuring out what "most affected and charmed him", and "catch[ing] vivid glimpses of the life he saw and the figures in it which specially enchanted him" (Spurgeon qtd. by Kahan pp. 824-5). In fact, at the point in Craig and Kinney's book that Kahan quotes as if it were dismissing all this (reviewed in YWES for 2009) there is no mention of Spurgeon at all. There is just Kinney being characteristically modest about what computational stylistics can hope to achieve in measuring linguistic beauty. Kahan's pile of battered straw men grows higher. MacDonald P. Jackson is next, and Kahan's critique reveals that does not understand the statistical notion of standard deviation, nor even Jackson's exemplarily plain English. Jackson wrote that according to G. E. Bentley's carefully derived approximation "just over 2000 plays were written" in the period 1590-1642, but reporting this claim in order to display his incredulity about it Kahan writes that "Following Gerald E. Bentley, Jackson states that there are roughly 2000 extant plays from the period" (p. 827). Nobody thinks that 2,000 plays survive from this period; we know that around three-quarters are lost.

    Because Marvin Spevack's concordances of Shakespeare list just over 19,000 unique words and the more recent Open Source Shakespeare website counts nearly 29,000, Kahan concludes that ". . . for much of the history of stylometrics, scholars could not even count words properly" (p. 892). And if we cannot agree on the word counts, then "statistics are virtually impossible" according to Kahan. Of course there is no contradiction between Spevack's word count and the Open Source Shakespeare word count, for two reasons. First, they are using different source texts of Shakespeare--and editions disagree on what Shakespeare wrote--and secondly, and more significantly, they are grouping word-forms into words (the process called lemmatization) by different principles, and linguists disagree on how to lemmatize. Made by different principles, both counts are valid.

    A further three pages of Kahan's 'history' of stylometry are wasted on the dead-end of Donald Foster's attribution to Shakespeare of "A Funeral Elegy" and his SHAXICON database. Kahan warns that many Shakespearians do not understand computers and then inadvertantly confirms that he is one of them by asserting the old saws that "computers are machines that do what we tell them to do" and that "we only get the data that we seek" (p. 831). These are patently untrue assertions. Every computer user has experienced the machine not doing what it has been told to do and not getting the data she seeks, and the reason is that modern personal computers work by running millions of lines of program code that no one person fully understands and that interact with one another in subtle and virtually unforeseeable ways. Determinism in computer operations is difficult to achieve, and when it is striven for--in air traffic control systems and military hardware--it is achieved only fitfully and at great cost.

    Kahan next turns to Ward Elliott and Robert J. Valenza's scholarship and complains that one of their techniques--the dividing of plays into equally sized blocks for testing--continues to be used by other investigators, including Craig and Kinney, even though Elliott and Valenza have been shown to get things wrong. To revive our medical analogy, this is like complaining that one's oncologist proposes to use a scalpel despite the long history of failed operations involving scalpels. Kahan is (naturally enough) unable to mount a serious objection to dividing literary texts into equally sized blocks for testing, so when investigators such as Craig and Kinney report that their results are ambiguous he castigates them for not "questioning the premise" (p. 833) that standardized blocks are a good idea. Kahan's criticisms are scientifically illiterate. Kahan complains that Vickers's claimed matches the Additions to Thomas Kyd's The Spanish Tragedy and Shakespeare's writing--reported in an essay reviewed in YWES for 2012--are not perfect linguistic matches but only matches in the underlying poetical ideas. This is true for the evidence quoted here, but Vickers made no other claim for them. In Kahan's judgement, the Additions lack the "rich imagery" that would make the claim persuasive and hence "If we were, for argument's sake, to map this concept as a scientific proposition, it would look something like this: Shakespeare MINUS Shakespeare’s imagery EQUALS Shakespeare PLUS Shakespeare’s imagery OR X-1 = X+1". As Kahan rightly says of what he has just written "this is gibberish" (p. 836).

    Unsurprisingly, Kahan's tour of a field he does not understand leads him to conclude that the field is in disarray and that its proponents' goals are unattainable. Problems of authorship attribution, he decides, "lend themselves to historical, not scientific inquiry" (p. 837). He thinks that after 230 years' work there is "virtually no consensus on what constitutes success or failure" (p. 838) in authorship attribution. If he really means this, Kahan must believe that nothing we have discovered in those 230 years is actually knowledge. He must hold that we just do not know whether John Fletcher co-authored All is True / Henry 8, or whether Thomas Middleton co-authored Timon of Athens, or whether George Peele co-authored Titus Andronicus, or whether George Wilkins co-authored Pericles. He is of course entitled to such a radical scepticism, but it brings no glory to the journal English Literary History that it indulged his deeply ignorant account of how he reached this lonely position.

     And so to some serious scholarship on authorship. Gary Taylor and John V. Nance show that with computational stylistics we can differentiate between the imitation of someone's style and their actual style, in particular when Shakespeare and George Peele are imitating Christopher Marlowe, and can help to solve some attribution problems in 1 Henry 6 ('Imitation or Collaboration? Marlowe and the Early Shakespeare Canon', ShSurv 68: Shakespeare, Origins and Originality[2015] 32-47). Essentially, the imitation of another's style can never be complete because imitators cannot spot all fine detail of someone's writing habits: all they can do is pick out noticeable bits of the style and copy those. Taylor and Nance present a table showing bigrams and trigrams (and the odd 4-gram and 5-gram) for part of a 173-word speech in Titus Andronicus, showing which of the n-grams appear in no other plays from 1576 to mid-1594, which appear more than five times in other plays (and hence are too common to be useful for authorship identification), and which, having 1-5 links, appear in which of the 80 plays from that period. Taylor and Nance point to other studies, by John Burrows and Hugh Craig, that show that imitations--for example, by Romantics of early modern drama, or Henry Fielding of Samuel Richardson--are distinguishable from their models.

    But what of samples as small as 173 words, such as Titus Andronicus 5.1.124-144, Aaron's speech beginning "Ay, that I had not done a thousand more. . . .", which is widely recognized as Shakespeare imitating Marlowe's speech of Barabas listing his crimes in The Jew of Malta? This is the sample used for Taylor and Nance's table, and firstly they counted n-grams occurring in this speech and just one other play. Most often (29% of the time) that one other play was by Shakespeare, which is more than twice as often as the next most-frequently-hit playwright. Then they count which plays (rather than playwrights) had the most unique matches with the speech: Troublesome Reign, Robert Greene's Selimus, Thomas Kyd's The Spanish Tragedy, and Shakespeare's The Two Gentlemen of Verona all had two each, and Marlowe's The Jew of Malta and Shakespeare's Richard III both had three each. Next Taylor and Nance focuss on just Shakespeare, Marlowe, and Peele. (It is not made explicit why Peele enters the frame here, since he was not mentioned before; presumably it is because we know that he co-authored Titus Andronicus.) They broaden their purview to bring in the writers' poetry as well as their drama, and this actually takes some of the hits away because with the larger dataset some terms that previously fell within their range of 1-5 links fall outside it by having more than five links in all to other writings. In this analysis, Shakespeare stands well ahead of Marlowe and Peele because of links to Venus and Adonis and Lucrece.

    As a check on this method, Taylor and Nance do the same testing for the first 176 words of 2.3.176-179 from Marlowe's The Jew of Malta, the speech Shakespeare was imitating. In their first test, looking for author canons that have the most hits, they find Shakespeare (with three) and John Lyly (with three) and Marlowe (with four), and since Marlowe's canon is the smallest of these three canons those four hits count for more than they otherwise would. (In that case, it is not clear why the differences in canon size was not discussed in the previous set of tests regarding Titus Andronicus 5.1.124-144; if there is some allowance to be made for canon-size differences this allowance needs to apply to all the tests.) In the second test, looking for the plays with the most unique hits, the winner was Marlowe's The Jew of Malta. In the third test, bringing in the poetry of Marlowe, Peele, and Shakespeare as well as the drama, Marlowe's works had the most links to Barabas's speech from The Jew of Malta. It appears that this kind of testing is not fooled by authorial imitation: Marlowe tests like Marlowe and Shakespeare imitating Marlowe tests like Shakespeare.

    Next Taylor and Nance look at Aaron's speech "Now climbeth Tamora Olympus' top" (Titus Andronicus 2.1.1-25), also widely said to be Shakespeare imitating Marlowe, but more recently claimed by a number of analysts as the writing of George Peele. So, does Peele imitating Marlowe test like Peele, Marlowe, or Shakespeare? Numerically, by the three tests described here, it tests like Peele (and not at all like Shakespeare) whether we count by author-canon hits or by play hits. Then they turn to the first 173 words of The Massacre at Paris 2.31-52, which is what critics have long said is the bit of Marlowe that Titus Andronicus 2.1.1-25 is imitating. It comes out on all three tests as Marlowe's work, and all the more strongly if we accept the recent attributions of parts of 2 Henry 6 to Marlowe.

    Convinced by these results that their three tests are reliable, Taylor and Nance now turn to some unsolved attribution questions, starting with the first 173 words of 1 Henry 6 5.3.1-24, Joan's speech beginning "The Regent conquers, and the Frenchmen fly". Taylor and Nance consider the six candidate authors that have been proposed for this speech: Greene, Kyd, Marlowe, Nashe, Peele, and Shakespeare. By canons, the only persons with any hits are Greene (one, or two if you think Greene wrote George a Green) and Shakespeare (two), and Marlowe (nine). By plays, three have more than one hit: John Lyly's Gallathea (two), Samuel Daniel's Cleopatra (two), and Marlowe's Doctor Faustus (four). As before, Taylor and Nance point out that the differences in canon size make these results not directly comparable one with another. At least with these six candidates the final test can be expanded to include all their non-dramatic canons, but we have to be careful because Greene and Nashe have huge non-dramatic canons, making theirs twice to four times the size of the others'. But even allowing for this difference in canon sizes--and Taylor and Nance sound unsure just how to do this and try several ways of thinking about it--Marlowe dominates the third test and Kyd (Brian Vickers's candidate) is far behind the pack.

    Taylor and Nance go on to another attribution problem in 1 Henry 6, the first 173 words of the French General's speech at 4.2.15-38 beginning "Thou ominous and fearful owl of death", and Shakespeare comes out far ahead of Marlowe and then everyone else. Next it is 1 Henry 6 5.3.24 to 5.4.15, which is the continuation of Joan's speech tested above (5.3.1-24, found to be Marlowe's) and the ensuing scene (5.4) of Joan's capture, totalling 166 words and so comparable to the previous samples. Marlowe comes out at the likeliest candidate on all three tests.

    At this point, Taylor and Nance ask the fundamental question "could chance produce these results?" Using Fisher's Exact Test on their results for Titus Andronicus 2.1 (attributed to Peele) and 5.1 (attributed to Shakespeare) they produce a 2-by-2 contingency table for just their Peele and Shakespeare results. The outcome of the calculation is that such a table would be as skewed as it is--with so few hits for Shakespeare and so many for Peele in 2.1 and so many hits for Shakespeare and so few for Peele in 5.1--by the operation chance alone, if authorship were not the deciding factor, just one time in 100,000. Since the counting of these hits was originally for Shakespeare, Peele, and Marlowe as the candidate authors I am not sure that it is valid to make such a contingency table only for Shakespeare and Peele, leaving out Marlowe. The same criticism applies to the following three contingency tables in which Taylor and Nance add in some further results they have already achieved. Since they do not add in all their further results, I am not sure that these are really contingency tables at all, and hence I am not sure that Fisher's Exact Test is appropriate here.

    Whether or not I am right to worry about that, I am sure that the following sentence is not statistically valid: "Fisher's Test gives less than three chances in one thousand that these two sets belong to the same population or were created by the same agent" (p. 47). Much rides on how we interpret the words here, but Fisher's Test is certainly not capable of telling us the likelihood-of-being-true of propositions such as "two sets belong to the same population" or their being "created by the same agent". All Fisher's Test can tell us is how often chance will produce the unusual distribution of numbers we find, if (and this is called the 'null hypothesis') chance alone is all that is causing that unusual distribution. That statistic can help us consider the possibility that an alternative to the null hypothesis is in fact the true explanation of the numbers, but Fisher's Test itself has nothing to say on that.

    Taylor and Nance's method, which they call "microattribution", works on small samples of text. The question of just how large our samples should be for reliable authorship attribution results is comprehensively investigated by Maciej Eder ('Does Size Matter? Authorship Attribution, Small Samples, Big Problem', DSH 30[2015] 167-82). Depending on genre, Ede reckons that the minimum sample size is generally 2,500 to 5,000 words. Eder used the same dataset for all his tests and tried out various well-established attribution techniques but with the variation of "extract[ing] shorter and longer virtual samples from the original corpus, [and] using intensive resampling in a large number of iterations" (p. 169). He does not explain what "intensive resampling" involved, but it presumably means that having derived a set of observations from the data (a sample) that set is then itself sampled in the sense that observations are chosen from it at random to produce a second set that has just as many observations in it. This second set will most likely not be a replica of the first set because the random choosing does not prohibit reusing the same observation more than once. So, if the initial set were observation#1, observation#2, observation#3, observation#4, observation#5, and observation#6, the first resample might contain observation#4, observation#2, observation#6, observation#4, observation#2, and observation#1. This is done many times and the same statistic--say, the mean or the median--is calculated from the resample each time. This gives a distribution showing how that statistic varies with multiple resampling of the same initial set, and the width of that distribution is a comment on how much random variation in the initial sample could affect the results that this initial sample gave us.

     The experiments described here use texts in multiple languages including English, Polish, German, Hungarian, Latin, and Ancient Greek and across the genres of novel, poetry, and non-fiction. The main authorship attribution method was John Burrows's Delta using the 200 most-frequent-words, as described in "'Delta': A Measure of Stylistic Difference and a Guide to Likely Authorship" (L&LC 17[2002]: 267-87), which has been widely used in Shakespeare studies, most notably in Hugh Craig and Arthur F. Kinney's book Shakespeare, Computers, and the Mystery of Authorship (review in YWES for 2009). Additional tests were done using other attribution methods such as Support Vector Machines, but Eder goes into little detail about these. The procedures were standard machine-learning ones: the texts were divided into two groups, one being a training set of one-text-per-author and the other a test set to which the remaining texts were assigned. Samples were extracted from the test set and the machine-learning technique was used to classify each sample against the training set to see how often the technique assigned the extracted sample to the correct author in the training set. Then the tests were rerun with different sizes of sample to see what that did to the accuracy of attribution. For cross-validation to ameliorate the problem of the texts in the training set not being typical of their authors, texts were randomly swapped between the training set and the test set and the whole thing done again.

    How to minimize the problem that different parts of the same text--say, narration and dialogue in a novel--might test differently? Instead of taking words, sentences, and paragraphs sequentially from a novel, the bag-of-words approach picks words one by one at random from different places in the novel. Such a bag might have as few as 500 words or as many as 20,000, and what we want to know is how the bag size affects attribution reliability. Eder graphs the result and of course it is a rising curve of improved attribution sucess as the bag gets larger, leading to a plateau after which attribution success does not get much better as the bag size increases. Below 5,000 words the attribution success falls off sharply (under 80% correct) and below 3,000 words it is "simply disastrous" (p. 170). The effect is different for different languages: English novels gave the best attribution success for a given bag size, then Hungarian novels, then German novels, then Polish novels. Latin and Greek prose get above 80% accuracy once the bag is at least 3,000 words. The saturation points on the curves--after which adding more words to the bag does not make much difference--were somewhat different for different languages. The curves were less well defined (more scattering around the trend line) for poetry, in Latin or in English, than for other genres. Weirdly, with Greek poetry and prose the attribution success actually started to fall off again for really large bag sizes, which Eder finds "difficult to explain" (p. 173) but he notes that a similar effect has been found by another of his studies. An important conclusion for all languages and genres was that after 15,000 words in the bag the attribution success does not get any better.

    What if we do not sample the text to produce a bag of randomly chosen words from it, but instead use blocks of consecutive words? All the tests were rerun using blocks rather than randomly sampled bags of varying sizes and the results are surprising. For any given sample size, the attribution success is worse for this block method than for the bag method, and the scattering around the trend line (that is, the noise) is worse too. The difference in the performance between the block method and the bag method varied by language. Specifically, the less inflected the language, the more it mattered whether we use the block method or the bag method, so it mattered most of all with English texts. What about authors whose individual works are too short to test by either of these methods, so that we have to concatenate various bits of their works to get a big-enough sample to test? Using the results of the previous experiments which seem to show that 8,000 words is a big enough sample to test, Eder set about building samples by different forms of concatenation--"4,096 chunks of 2 words in length (2-grams), 2,048 word 4-grams, 1,024 chunks of 8 words, 512 chunks of 16 words, and so on, up to 2 chunks of 4,096 words"--so that all the samples were 8,192 words in length. The result was that the longer the chunks the worse the authorship attribution success, which rather agrees with the previous experiment since single words are the ultimate short chunks and 4,096-word chunks are like the biggest blocks. But around the middle of the range, it does not make much difference: ". . . there is no substantial difference in attribution accuracy between, say, a few chunks of 500 words combined in one sample and a dozen concatenated chunks of 100 words" (p. 175).

    Next Eder describes the technical details of the cross-validation. Although cross-validated results were less good overall--that is, authorship attribution results were less accurate when the machine rather than the investigator randomly chose the composition of the training and test sets--the important discovery is that the saturation point was about the same: after about 5,000 words the results just do not get much better no matter how many more words you are testing. Switching to other authorship attribution tests instead of Delta did not make much difference to that result: you need 5,000 words or more. Interestingly, of all the methods tried, the counting of most-frequent-words turned out to be the best strategy, beating the counting of character 3-grams and character 4-grams and part-for-speech-tag 3-grams.

    Eder lists the assumptions of the experiments and their associated limitations, including the number of most-frequent-words used (200), which might not be optimal for all languages or genres. In fact, Eder ran tests varying the number of words used in his Delta experiments and it did not make a significant difference to the saturation point. He worries too that the choice of texts to test might be constrained in unfortunate ways that matter, such as the availability of only certain texts in digital form and the existence of stylistic features that are particular to one national literature. Eder's last experiment involves Principal Component Analysis of the frequencies of the 100 most-frequent-words from bags of 1,000 randomly chosen words from 28 English novels, for which he offers two graphs. Unfortunately, his body text does not explain what was done differently to produce the two graphs and the captions to them (Figures 13 and 14) are identical so it is impossible to tell what point he is making.

    Eder remarks that in general the problems of accuracy in authorship attribution are like the problems of accuracy in part-of-speech tagging and syntax parsing--and personally I would add the problems of Optical Character Recognition of early modern books--in that one can "easily achieve a few dozen percent of accuracy, but the actual problem is to gain every next following percent (and the difficulty increases exponentially rather than proportionally)" (p. 179). Importantly, though, the difficulty of authorship attribution and the length of sample you need to get it right varies from author to author: ". . . a fingerprint of both Charlotte and Anne Brontës was considerably well recognizable even for short samples, followed by Richardson, Dickens (David Copperfield, Pickwick Papers), Eliot, Galsworthy, Austen, and again Dickens (Hard Times, Oliver Twist, Great Expectations). It was very hard to attribute Hardy, but particularly long samples were needed to distinguish novels written by James" (pp. 179-80).

    So much for the sample size, what about the number of words we should count when attributing authorship by word frequencies? Jacques Savoy finds that 300-500 is about right ('Comparative Evaluation of Term Selection Functions for Authorship Attribution', DSH 30[2015] 246-261). Savoy went about testing this using a newspaper in English and another in Italian and attribution techniques based on a measure of relative entropy called the Kullback-Leibler divergence, and on John Burrows's Delta technique, and on chi-squared tests. A well-established principle is that although we are looking for discriminators between authors, we should not use ones that one author often uses and a second author never does--such as "while" used a lot by Andrew Hamilton and never by James Madison, in the The Federalist Papers case--not least because these are such noticeable differences that they are open to mere imitation. Savoy applied that rule, but also, to reduce the data to be processed, he threw away terms appearing fewer than 10 times in his corpus. His two newspaper corpora comprised articles on Sports and articles on Politics, and after applying these reduction rules Savoy was left with 6,616 word types for Sports and 5,128 for Politics for his English-language newspaper and for the Italian newspaper it was 6,780 words for Sports and 10,644 words for Politics. The question is, can these datasets be reduced still further to increase the discriminating power of authorship-attribution tests based on them? The mathematics for this takes each word and looks to see if it is disproportionately present in one author's work and absent in the others; in short, can we associate particular words with particular authors? Savoy outlines several mathematical approaches to this question, and finds 10 different ways of calculating the answer; all 10 are used in the analysis.

    Savoy sketches the three authorship attribution method he used. To get a baseline performance for the authorship attribution tests, Savoy used his own set of predetermined words (in English and Spanish) as the one for counting frequencies, to see what kind of attribution performance this would give; the answer was between 75% and 90% accuracy across the different parts of his test set. Then he repeated the experiment using the collections of words from the tests sets themselves, whittled down as described above, working with sets of 150, 300, 500, 800, 1000, 1500, 2000, 3000, 4000, 5000, and 'All' words--the last meaning all 6,616 for English-Sports, all 5,128 for English-Politics, all 6,780 for Italian-Sports and all 10,644 for Italian-Politics--to see which size of set gave the best attribution accuracy rates. The answer was that generally the mathematical approaches for finding the most discriminating words did much better than just using the preselected set of words or using them 'All', typically raising the accuracy by 5-10% from the 80s to the 90s. Savoy applies some complex testing to see which improvements in performance are statistically significant and to see how much the variations in type of text (Sports versus Politics, English versus Italian) affects the performance.

    The results show that 300 most-frequent words is a particularly good setting for English language text attributions, and 500 is good for Italian. Language matters because different languages use different numbers of function words, so for example in Italian "the determiner 'the' (definite article) could be 'il, lo, l, i, gli, la, or le' because the variations in gender and number must be specified in the Italian language" (p. 257.) The results also show that Burrows's Delta's rule for selecting which words to use is a good one.

    Taking a non-technical approach, Warren Chernaik surveys the theatrical and print history of 1 Henry 6 and the arguments for its co-authorship, with special attention to Brian Vickers's methods ('Shakespeare as Co-author: The Case of 1 Henry VI', Medieval and Renaissance Drama in England 27[2014] 192-220). Chernaik clearly knows, because he cites (p. 200n23), MacDonald P. Jackson's article ("New Research on the Dramatic Canon of Thomas Kyd", reviewed in YWES for 2009), in which it is shown that Vickers's 'two-horse race' method (as Jackson calls it) is entirely useless for authorship attribution, but Chernaik seems not to have noticed that this is what Jackson's article is about and that this revelation makes any further critique of Vickers nugatory. Chernaik works though some of the phrases that Vickers finds compellingly common to Kyd's work and 1 Henry 6 and remarks that they are not especially unusual, which is true; as we know, Vickers's means for determining rareness are faulty, as is shown in detail in the New Oxford Shakespeare's Authorship Companion (2017). But this is not quite the point, since even when Vickers finds genuinely rare phrases in common between two works this does not prove common authorship: any two substantial bodies of writing will have in common some phrases that are used almost nowhere else.

    Chernaik then turns (briefly) to Marina Tarlinskaja's unpublished work on the Kyd canon, and Vickers's crediting of Kyd with authorship of Arden of Faversham based on parallels of phrasing. The latter argument is vitiated, Chernaik notes, by the great number of verbal parallels between Arden of Faversham and Christopher Marlowe's Edward 2. Throughout all this Chernaik's approach is impressionistic rather than quantitative, as for example in his remarks that "A wider database . . . might well contain all of these conventional three-word combinations" (p. 199) and "Other parallels . . . could be considered commonplaces" (p. 205) and "more likely to be conscious imitation" (p. 209). Such matters are entirely amenable to empirical investigation and there is not much point relying on anyone's hunches about them.

    In the final authorship-attribution article to be considered this year, Ryan L. Boyd and James W. Pennebaker conclude that Double Falsehood is derived from the Shakespeare and Fletcher collaborative play Cardenio, and is not a forgery by Lewis Theobald ('Did Shakespeare Write Double Falsehood?: Identifying Individuals By Creating Psychological Signatures with Text Analysis', DOI:10.1177/0956797614566658, Psychological Science Online First (Advance Access)[2015] n. pag.). As an alternative to classifying units of language by formal features as unigrams and bigrams, it is possible to classify them "along hundreds of psychological dimensions, including cohesion, time orientation, and sentiment" (p. 2). Function words, in particular, seem to "reveal much about psychological and social processes, including emotional state, cognitive complexity, and sociability" (p. 2). Content words are also revealing of these things, and more transparently so: angry people use words expressing anger and "securely attached people tend to use more words related to inclusion" (p. 2). (The obvious concern that arises from the use of content words rather than function words in relation to drama is that it is the character, not the author, who is being angry or securely attached: one must worry that genre in general and the action of the scene in particular might colour such analyses.) Importantly, Boyd and Pennebaker claim that it has been shown that these characteristics of a person's language use are "consistent within person across time and context" (p. 2).

    Boyd and Pennnebaker took 33 plays by Shakespeare, 9 by John Fletcher, 12 by Lewis Theobald, and Double Falshehood, all believed to be sole-authored except the last one. They stripped out all the paratexts but left in the stage directions; they do not say if they left in the speech prefixes and proper nouns. Then they used software of Boyd's own devising called "Ye Olde Token Converter" to "convert idiosyncratic and outmoded spellings to their U.S. equivalents" (p. 3). Unfortunately the source code of this software does not seem to be available at the URL that Boyd and Pennebaker provide, but the compiled code is available and freely downloadable for testing. It looks to this reviewer as if all the software does is a series of 707 search-and-replace operations so that, for example, wish'd becomes wished and suddain becomes sudden. This of course is necessary for modernization, but it is by no means sufficient on its own. For example, given the first seven lines of Q2 Hamlet, "Ye Olde Token Converter" leaves unconverted the words centinels, answere, vnfolde, selfe, liue, hee, vpon, houre, strucke, and twelfe and alters only nay > no and tis > it is.

    Boyd and Pennebaker briefly discuss the inherent problem that word-counting techniques such as theirs do not take words' contexts into account, and reasonably enough they point out that over large stretches of texts the distortions that this weakness generates cancel each other out: ". . . statistically speaking, the majority of times when people use the word depressed, they are referring to the psychological condition" (p. 3) and not the word's other sense. Boyd and Pennebaker put the function words they are counting into eight broad categories: "personal pronouns, impersonal pronouns (e.g., it, any, thing), articles, prepositions, auxiliary verbs, conjunctions, negations, and high-frequency adverbs lacking direct referents (e.g., very, really, so)" (p. 3) that account for more than half of all writing.

    Pennebaker in his previous work established a measure he calls the Categorical-Dynamic Index (CDI), which is high when a piece of writing uses lots of nouns, articles, and prepositions, as happens in writing showing a tendency to "classify objects, people, and events in hierarchical ways" (p. 4); psychologically this kind of writing comes from people who are practically minded and "emotionally distant" (p. 4). CDI is low in writing that uses lots of auxiliary verbs, pronouns, adverbs and the other categories that Boyd and Pennebaker count, which is what gets written by people "who are dynamic thinkers [and] tend to live more in the here-and-now and like to tell stories, and are often more focused on social matters relative to categorical, complex thinkers" (p. 4). Boyd and Pennebaker's CDI spectrum sounds something like the spectrum from highly systemizing (unempathetic) minds at one end to highly empathetic (unsystemizing) minds at the other that Simon Baron-Cohen has shown correlates well with the medical conditions of autism and Asperger's syndrome. It would be useful to learn more of Boyd and Pennebaker's research in this area, for which the article repeatedly refers readers to "Supplementary Material" provided online by the article's publisher Sage. Unfortunately, when this review was written in late December 2016 the URL given by Boyd and Pennebaker led to a website that read "Sage site is currently undergoing maintenance work" and gave no additional materials.

    As well as their CDI measure, Boyd and Pennebaker measured for each text the average sentence length and the use of long words, and they counted frequencies of content (that is, lexical) words using "the default content categories of the computer program LIWC [Linguistic Inquiry and Word Count" (p. 4). The LIWC software is proprietary rather than open source and not free and although at the time of this review its website had what looked like links leading to a paper on the "science behinds its development" and the "Operator's Manual" these links were inoperable. In particular, this reviewer was unable to determine "the default content categories" used by the software on which the present article's research depended. This kind of obscurity is not at all how publicly funded research projects should engage with their academic readers. The authors of the paper are both employed by the University of Texas and the "Funding" section of the article reports grants from the Army Research Institute and the National Science Foundation, the last of which certainly has a published policy of requiring investigators to share data and encouraging them to share software.

    Yet another kind of analysis undertaken by Boyd and Pennebaker was the use of "meaning extraction" software that detects "themes" in writing; again this seems to be merely a word-counting method that relies on assigning collections of words to themes. Thus, a text with a lot of occurrences of father, mother, brother, and sister would test highly on the "family structure" theme. Again, the detail on the words and their associated themes is supposed to appear in the Supplementary Materials website that, at the time of this review, was unavailable. Boyd and Pennebaker refer to work on the detection of relatively rare words and phrases as authorship markers, referring (without detail) to the reliable scholarship in this field done by Hugh Craig and Arthur Kinney (reviewed in YWES for 2009) and (with detail) to the unreliable work done by Donald Foster; this lack of discrimination suggests unfamiliarity with the materials. Just how Boyd and Pennebaker applied rare-word and rare-phrase tests is said to be detailed in the (unavailable-to-me) Supplementary Materials. Remarkably, Boyd and Pennebaker are still not done sketching their method: they also used Linear Discriminant Analysis, Decision Trees, and Support Vector Machines, of which they do not even attempt to describe the principles let alone the precise workings in this application.

    After these sketches of methodology, the article moves straight to "Results" so the reader never hears just what the experiments consisted of, nor what validation runs (if any) were undertaken. This is most unsatisfactory. Apparently the function word tests, CDI, and measures of complexity all converged on identifying Shakespeare as the author of Double Falsehood when the play is considered as a whole. In support of this, Boyd and Pennebaker provide two scatterplots that they do not talk the reader through, To this reviewer these pictures, regarding "Pronouns, Negations, and Auxiliary Verbs" and "CDI and Large Words", seem to show some kind of clustering by author (Shakespeare, Fletcher, Theobald) but with considerable overlap that should have been explained. Also looking at the play as a whole, the tests using LIWC content categories were not unanimous: different methods of processing the data gave the play to Shakespeare or Theobald. The meaning-extraction, and rare-word tests all gave Double Falsehood to Shakespeare. Boyd and Pennebaker offer a second pair of scatterplots showing "Low Emotionality" and "Properness"--both of which terms are not previously mentioned in the article, let alone described--and these have much clearer authorial clustering, but in both cases Double Falsehood falls between authorial clusters rather than within any one of them.

    Next Boyd and Pennebaker break Double Falsehood into five acts and break each of the comparison plays into five equally sized pieces and repeat their tests. The results, across all the tests, are that Act 1 of Double Falsehood is Shakespeare according to all the tests, Acts 2 and 3 are Shakespeare according to nearly all of the tests, and Acts 4 and 5 are Fletcher according to about two-thirds of the tests. As Boyd and Pennebaker observe, this is broadly consistent with other scholars' findings of Shakespeare at the beginning and Fletcher at the end, and the almost complete absence of Theobald's habits offers "consistent evidence against the notion that Double Falsehood is Theobald’s whole-cloth forgery" (p. 10). Boyd and Pennebaker conclude with some speculations on what their methods might tell us about "Shakespeare's, Fletcher's, and Theobald's unique psychological signatures" (p. 10). Theobald turns out to be the aloof, systematic scholar, Fletcher is the kind of man with lots of close personal and professional relationships, and Shakespeare is a bit of both.

    Robert N. Watson argues that the reattribution to Montague of a line that Q2 and Folio Romeo and Juliet give to Capulet is mistaken ('Lord Capulet's Lost Compromise: A Tragic Emendation and the Binary Dynamics of Romeo and Juliet', Renaissance Drama 43[2015] 53-84). In almost all editions of Romeo and Juliet, it is Romeo's father who answers the Prince's question "Who now the price of his [Tybalt's] dear blood doth owe?" with "Not Romeo, Prince. He was Mercutio's friend. | His fault concludes but what the law should end, | The life of Tybalt" (3.1.182-5). Q2 in fact has Juliet's father say this line, but almost every editor thinks that is an error. Q1 omits the speech.Watson reckons that in fact Shakespeare wanted us to consider this line by Capulet as an early gesture at ending the feud and that editors should leave it alone. Among the considerable objections this theory has to overcome is the fact that both Capulets and the Nurse go on about how disastrous it is that Tybalt has been killed and Capulet's Wife calls Romeo "villain . . . villain . . . traitor murderer" and vows to have him poisoned (3.5.79-92), so if her husband had spoken to the Prince in defence of Romeo she surely would have contradicted him or at least scolded him for it afterwards. As Watson points out, Capulet spoke favourably of Romeo at the feast and reprimanded Tybalt for refusing to "endure" him. But Watson surely takes this too far in suggesting that Romeo has only done what Capulet wanted to do--silence Tybalt--and that Capulet may be "glad Tybalt is out of his household" (p. 56), and "might actually have approved of Juliet's choice as a way of ending the feud" (p. 57).

    Watson refers to investigating Shakespeare's use of words and phrases (up to five words) that are immediately repeated (but just once) by using a computer program in the language Python that worked on the texts as represented in the Project Gutenberg editions. It is not quite clear whether the phrases were allowed to span line breaks, since Watson writes "Prose lines were treated by their natural margin breaks, on the grounds that this would still show a proportion of doublings and would avoid skewing the results away from plays heavy on prose" (p. 69n32). I cannot tell what "were treated by" means in this description. Since a subsequent refinement of the procedure is said by Watson to allow the phrases to appear "across any two lines" (p. 70) I guess that this first test did not allow phrases to span two lines. Not allowing phrases to span line breaks would be an entirely arbitrary prohibition in the case of prose lines, so far from "avoid[ing] skewing the results away from plays heavy on prose" I would have thought that not allowing multi-line phrases would guarantee such a skewing.

    The result of the test was that Romeo and Juliet has indeed the highest proportion of such repeated words and phrases in the canon. Most interestingly, the end of a play shows a marked falling off of this phenomenon, which Watson suggests is a manifestation of twoness-into-oneness at the close. Next, and using the standard Python Natural Language Toolkit with a text in which antonyms are marked, Watson looks at "A doubled word or phrase with its antonym in the same line" (caption for Figure 1) and "A word-and-antonym line following a line with a doubled word or phrase" (caption for Figure 2) and "A word and its antonym in the same line" (caption for Figure 3), and again Romeo and Juliet is ahead of the rest of the canon in each case. The all-pervasive doubleness of the language of Romeo and Juliet makes Watson even suggest that the repetition of one speech ("The grey-eyed morn . . ."), spoken first by Romeo and then by the Friar is intentional, not an error, and likewise the repetitions in Romeo's final speech.

    Also on the theme of repetition is David A. Harper's claim that Sonnet 154 is Shakespeare's first stab at what became the much better Sonnet 153, with which the sequence was meant to end ('Revising Obsession in Shakespeare's Sonnets 153 and 154', SP 112[2015] 114-38). Harper reviews famous duplications in Shakespeare: Romeo and Friar Laurence both welcoming the "grey-eyed morn" in Q2 and F Romeo and Juliet, the two tellings of Portia's death in Julius Caesar, and Berowne's repetition of his speech about women being the inspiration for study in Love's Labour's Lost. Harper credits Grace Ioppolo with the view that Shakespeare's own papers contained his original writings and his revisions of them and that he failed to mark sufficiently clearly the things he had deleted. Ioppolo did claim this in the 1990s, but the credit surely goes to E. A. J. Honigmann who proposed it first in the 1960s.

    Harper reckons the same thing is going on in Sonnets of 1609: number 154 duplicates number 153. He admits that Edmond Malone proposed this in 1780 and that others have revived the claim since then. For bibliographical evidence he turns to Cole Hutchinson's essay "Breaking the Book Known as Q" (reviewed in YWES for 2006) that notes that some of the Sonnets are broken across page-breaks in the 1609 quarto and that only Sonnet 154 appears on a page of its own. The obvious explanation is that this happens because Sonnet 154 is the last one: it appears in the quarto just as we would expect if the page-breaks were allowed appear naturally wherever they fell. Like Sonnets 10, 22, 34, 46, 58, 70, 82, 94, 106, 118, 130, and 142, Sonnet 154 happens to follow immediately after a page break so it sits at the top of a page, and since it is the last sonnet there are no other sonnets after it sharing its page.

    Harper notes that Sonnet 154 begins a new gathering (sheet K) but also completes an opening (I4v-K1r), which Harper thinks is convenient because the running headers for Sonnets reads "SHAKE-SPEARES | SONNETS." and without a sonnet on K1r the running headers for this opening would have had to be altered. This Harper takes to be evidence that "the printer made a conscious decision to ignore Shakespeare's cancellations" (p. 124) of Sonnet 154 to suit himself and keep his running headers the same for this opening as all the previous ones. Harper reckons that no other book printed by George Eld for Thomas Thorpe between 1608 and 1610 "employs a head dividing the author's name and the title across the fold" (p. 125) as Sonnets does, and he thinks this gave the printer a unique problem to solve if he wanted to divide the opening I4v-K1r between the end of Sonnets and the start of A Lover's Complaint. But the question is surely whether or not these books use different running headers on the verso and recto pages, not whether they put the author's name in the running header. The problem to be solved would be the same no matter what was in the running headers, so long as they were different. There are only five other surviving books printed by Eld for Thorpe in 1608-10 and one of them does in fact use different running headers on the left and right side of each opening, just as Sonnets does, and indeed contrary to Harper's unnecessarily specific claim it too "employs a head dividing the author's name and the title across the fold": Epictetus's Manual (STC 10424) has the divided running header "Epictetus | his Manuall". One this example is taken into account, Harper's claimed specialness for the problem in Sonnets disappears.

    Confusing matters still further, Harper thinks that "Eld's 1609 edition of Shakespeare's Troilus and Cressida is the only other Eld/Thorpe edition in those years to use a running head splitting a grammatical unit across the fold" (p. 125). But Thorpe had nothing to do with Eld's Troilus and Cressida, which was printed for Richard Bonian and Henry Walley. More fundamentally, it is not clear why Harper thinks that the need to set up one extra headline to deal with an opening shared between two works would influence any printer's choices about what should appear in his book. There is room enough above the text of I4v for "SHAKES-SPEARES SONNETS" to be spelt out in full there and indeed the headline "SONNETS" would, by Harper's own hypothesis, be lying around unused and available to take this role if K1r contained the head-title and start of A Lover's Complaint. Harper makes this very point himself--that fixing the running headers was easy--when he observes that because A Lover's Complaint ends on the verso page of an opening opposite a blank recto, the running header for that opening alone is altered from "A LOVERS | COMPLAINT." used in previous gatherings to "THE LOVERS | [BLANK]". That is exactly the solution--tweak the verso header and leave the recto blank--that Eld could have employed for I4v-K1r and that Harper thinks that Eld found too tricky and so preferred instead to print Sonnet 154 even though it was marked for deletion in his copy.

    Finally, Harper seems not to have noticed that by printing Sonnet 154 on K1r rather than omitting it and starting A Lover's Complaint there Eld actually made things harder for himself than if he had finished the sonnet sequence at Sonnet 153. Because of the inclusion of Sonnet 154, the forme of type for K(outer) comprised one page of Sonnets (page K1r) and three pages of A Lover's Complaint (K2v, K3r, and K4v). Eld thereby gave himself the task of keeping track of four running headers with three different wordings: one of "SONNETS", two of "A LOVERS", and one of "COMPLAINT". This was by no means the easy option that Harper supposes. In short, Harper's claim that Sonnet 154 was included only to make life easier for the printer makes no sense even on its own terms.

    Next, Harper argues on literary grounds that Sonnet 153 is the proper end to the sequence, not least because its final position provides a thematic link to A Lover's Complaint via its recurrent 'eye' images and sounds (in "Dyans", "liuely", "triall", "desired", "hied", and "lies"; the 'eye' is a vagina and A Lover's Complaint begins with a "concaue wombe"). Sonnet 154, he argues, provides a artistically weaker ending to Sonnets than Sonnet 153 would have. Harper ends with what he admits are speculations about the ordering of the subsequences within Sonnets and the process of revisions that gave shape to the edition, including the revision that led to a version of Sonnet 138 appearing in The Passionate Pilgrim, which transformation Harper takes as a model for how Sonnet 154 got improved by Shakespeare to make Sonnet 153.

    The present reviewer published an article in Studies in Bibliography that considers the clustering of stop-press corrections in Q2 Hamlet ('Press Variants in Q2 Hamlet: An Accident on N(outer)', SB 59[2015] 115-29). Of the 26 known press variants, 10 are clustered on the forme N(outer), which anomaly is explained as a consequence of a specific kind of printing accident. For the most part, the edition's two compositors worked on different sections of the book and used their own sets of headlines, but towards the end they swapped headlines. This would ordinarily make no difference to their work, but they were inadvertantly setting type to slightly different measures, so some type pages ended up with overlong headlines above them and could not be fully tightened. From a fresh examination of all seven extant exemplars of Q2 Hamlet, the article concludes that compositor Y's stick was about three-quarters of a millimeter wider than compositor X's.

    Also on a bibliographic theme, Chiaki Hanabusa gives a superbly complete description of the printing of Q1 Titus Andronicus ('The Text of the First Quarto of Titus Andronicus', Shakespeare Studies (Japan) 53[2015] 1-25). Its title page, printed and published by John Danter, refers to the play's having been performed by three playing companies. If these were the successive owners of the play then it was printed before June 1594 when, according to Henslowe's Diary, the Chamberlain's Men and Admiral's Men performed it at the Rose, since Danter would hardly leave this fact off the title page while mentioning the prior performances by Derby's Men, Pembroke's Men, and Sussex's Men. The Stationers' Register entry on 6 February 1594 puts printing of Q1 Titus Andronicus between 6 February and early June 1594. The presence in Q1 Titus Andronicus of an undamaged printer's device that was damaged during Danter's printing of The Cobbler's Prophecy, entered into the Stationers' Register on 8 June 1594 and printed later that year, also reinforces the idea that printing of Titus Andronicus was completed by early June 1594. Actually, I do not quite follow Hanabusa in this line of argument. The Cobbler's Prophecy could have been printed any time between 8 June 1594 and the end of that year; I cannot see a reason why it must have been printed shortly after being entered in the Register. Its date of entry in the Register tells us nothing we did not already know, which is that printing of The Cobbler's Prophecy was completed after printing of Titus Andronicus. 

    The reuse of two sets of running titles--one of six titles, one of four--suggests printing on two presses, one set on each. Examination of recurring pieces of type "has confirmed that A(i) was not distributed until Sheet E became ready for composition, for none of A(i) types reappears in Sheets B-D but recurs to E(o), the precedent composed forme of Sheet E" (p. 7). I would say that this is not quite confirmation: the evidence is consistent with the conjecture but inadequate to prove it. The reason is that when Hanabusa says that no piece of type from A(inner) appears on either side of sheets B, C, and D he really means that no distinguishable piece of type does so. Most pieces of type are indistinguishable from others in the same sort box--they are made to look exactly alike--with the exception of the few that have become distinctively damaged. Hanabusa found none of the distinctively damaged pieces from A(inner) in sheets B-D but there could still be other undamaged pieces from A(inner) recurring in B-D that by definition he could not detect. The argument is one of probabilities not certainties, and to assess the likelihood of Hanabusa's conjecture one would need to know how many distinctly damaged pieces of type he found in all, so that the likelihood of their non-appearance in B-D by mere chance could be assessed.

    To explain the delay in distribution of forme A(inner) Hanabusa conjectures that it took "time for Danter to gain the two publishers' permission for the typographical design and/or wording of the advertisement on the title-page A2r" (p. 7). Thus Hanabusa is treating Danter as merely the book's printer and the booksellers Thomas Millington and Edward White as the publishers who put up the capital for the publication and made the important decisions. This may be so: the imprint does not make it clear who did what. But earlier Hanabusa had suggested that the printer chose the title-page wording: "It is inconceivable that Danter . . . would have designed the title-page without updating the name of the company" (pp. 3-4).

    Hanabusa tracks irregularities in the spacing of the running heads due to looseness in locking up, and as genuine press variants occurring when pieces of type were replaced. He also completes his deduction of the order in which the formes were printed, based on the evidence of type recurrence. Evidence of ligature shortage suggests that sheet A was set seriatim, since the us ligature is used five times on A3r and four times at the top of A3v before a switch to separate u and s pieces for the rest of A3v and all of A4r and A4v. For the rest of the book (sheets B-K), the irregularity of spacing around stage directions, which he quantifies, suggests to Hanabusa that the copy was cast off and that misjudgements about the amount of text that would fit onto each page were corrected by altering the space around stage directions.

    Hanabusa copied out the watermarks, determined the mould and felt sides of the sheets, measured the chain lines, and recorded the sewing holes. The watermarks seem to suggest that the paper was French. Hanabusa gives a history of the authorship debates about Titus Andronicus, leading to the inevitable acceptance that it is a collaboration of George Peele and Shakespeare. Regarding the nature of the underlying manuscript for Q1 Titus Andronicus, Hanabusa gives the standard New Bibliographical argument that vague or permissive stage directions (of which this edition has examples) could not have stood in a promptbook used to run the play in the theatre and therefore that the copy for Q1 was authorial papers, which could have such directions. Variant speech prefixes for one character are supposed to point to the author, who was thinking of the various relationships a character may have, rather than a promptbook in which such needless variation would necessarily be ironed out. Hanabusa then gives the New Textualist critiques of these arguments, and in particular those in Paul Werstine's recent book (reviewed in YWES for 2013). Hanabusa points out that the kinds of readying instructions and duplicated stage directions that Werstine says are typical of theatrical documents are absent from Q1 Titus Andronicus. Also, Q1 (alone among the early editions) has a speech in which Marcus says that "the noblest prisoner of the Goths" has already been sacrificed, before the action of Alarbus's selection for sacrifice. There are other, lesser signs of authorial untidiness--loose ends not tucked in--so that taken all together the evidence leads Hanabusa to assert that ". . . we can confidently conclude that the copy-text used for the printing of Q 1 was the author's final draft" (p. 16).

    Hanabusa surveys the Stationers' Register evidence for a lost ballad on the play and what might be a prose history version of the same story, of which an eighteenth-century prose version may be the descendent, and concludes that from an analysis of the literary contents we cannot determine the chronological order of these versions. But book history can throw some light on this problem. Hanabusa makes essentially the same argument as Richard Levin did in an article (reviewed in YWES for 2000) that Danter made a habit of entering in the Stationers' Register ballads related to plays, and that these seem to be 'spin offs' commissioned by Danter to exploit the popularity of the play in performance, as witnessed in Henslowe's Diary. The evidence that the ballad did not come first is that these are always mentioned secondly in the Stationers' Register, pointing back to the play of which they are the ballad "thereof". Hanabusa reports that he made this discovery in his PhD thesis before Levin's article appeared. The 'history' of Titus Andronicus that was entered by Danter in the Stationers' Register in 1594 immediately before his entry of the ballad was, then, probably not a prose history separate from the play but the play itself. Hanabusa's appendices detail the distribution of the running titles, the blank lines around stage directions, the watermarks and chain-line spacing. In a short piece on a related topic, Nancy Peters Maude shows that as well as collaborating on printing Q1 Romeo and Juliet in 1597, Edward Allde and John Danter had previously collaborated on printing four editions of two other books in 1596: Ulysses Upon Ajax (STC 12782 and 12783) and A New Discourse of a Stale Subject, Called the Metamorphosis of Ajax (STC 12780 and 12781) ('The Extended Collaboration of John Danter and Edward Allde', The Library 16[2015] 329-42).

    In another fine bibliographical essay this year, Aaron T. Pratt establishes that stab-stitching was the norm for early printed books and thus did not connote cheapness or ephemerality ('Stab-stitching and the Status of Early English Playbooks as Literature', The Library 16[2015] 304-28). That is, most quartos, including those of play scripts, left the bookshop stab-stitched rather than fully bound. Scholars have tended to argue that the sale of plays as unbound quartos shows that these books were not considered important or literary. Pratt has surveyed nearly 1,500 quartos to see if they were once stab-stitched, with single holes punched right through the whole thing and some distance in from the fold, even if they are now bound, with separate holes in each gathering's fold. The discovery is that even high status works were sold stab-stitched and that overall 85% of quartos that are no longer than the average length of a playbook, 40 leaves, were stab-stitched. The length matters because longer books were harder to stab-stitch and might in any case be prestigious simply because they were big books. Excluding the stab-stitched playbooks themselves drops the percentage to 83%, so it is not the preponderance of stab-stitched playbooks that is making this proportion so high. Pratt also considers a kind of hybrid binding that used stab-stitching and thicker thongs to hold the leaves together and to attach some kind of cover. A price list for bindings survives from 1619, including a 10-pence calfskin binding for quartos of all lengths; using sheepskin instead was about half that price. It seems to Pratt that buyers would probably wait until they had a few playbooks they wanted bound and then got them bound as one, rather than paying the binding cost for each one as they bought it; this appears to be what the young Edward Dering did. But there really was no need to have a book bound: many copies survived in stab-stitched form for centuries; this was not a temporary binding.

    In offering to solve a crux, Rodney Edgecombe Stenning relies on the reference ("2.1.110-13") to tell the reader what bit of The Comedy of Errors he is writing about, unwisely so because these numbers do not point to the same lines in all editions ('A Possible Solution to the 'Jewel Best Enamelled' Crux in The Comedy of Errors 2.1.110-13', BJJ 22[2015] 266-8). It seems from the context to be Adriana's "yet the gold bides still | That others touch, and often touching will, | Where gold and no man that hath a name, | By falshood and corruption doth it shame". (These lines, which are all that Stenning emends, follow the line about the "Iewell best enamaled" that gives him his title.) Stenning emends the first two lines to "yet gold bides still | That others touch, and, often touching, swill". Stenning describes this as the emendation will > swill, and seems not to notice that he has also omitted the second word, "the". By "swill" he seems to mean something to do with abrasion from swirling fingers, but this is not explained. He then emends the third and fourth lines to "Whet gold, and no man that hath a name | By falsehood and corruption doth it shame", claiming that "To whet something is to speed up its attrition before our eyes"; the idea seems to be that of enamel being rubbed away to reveal the gold underneath. In this Stenning is ignoring whet's dominant sense of sharpening, which seems out of place here. Stenning paraphrases the lines as they appear in his emended form: "Whet gold--which is to say, chip away at a person's good reputation ('name')--and it will follow that if the reputation is solidly based (as gold is gold through and through), it will not be compromised ('shamed') by falsehood and corruption. And that is because gold (unlike superficial veneers such as gilding or enamel) will survive that chipping, and also because it is incorruptible" (p. 267). I suppose this makes sense as a paraphrase, but I cannot see audience members understanding swill and whet in the ways necessary for them to derive this meaning from Adriana's lines.

    Geoff Ridden asks why Costard begins his account of his being caught with Jaquenetta using the phrase "In manner and form following . . ." (Love's Labour's Lost 1.1.202) ('Costard's Use of a Legal Term in Love's Labour's Lost', ShN 65[2015] 45). Editors rightly point out that this is a stock legal phrase, but Ridden thinks that it more specifically (and incongruously) would put readers and audiences in mind of the phrasing of wills, which often use this language. Ridden notes that Shakespeare's will begins with this phrasing, and cites a few other wills that use it. In fact, a search for this phrase in E. A. J. Honigmann and Susan Brock's Playhouse Wills 1558-1642 gives 33 occurrences, so he seems right to claim that it was particularly common in wills.

    Morteza Lak reports that for his lavishly illustrated Shakespeare of 1791-1802, John Boydell not only had new pictures made but new elegant type cut by William Martin ('Type and Typography as Paratexts of Hamlet in Boydell's Shakespeare Edition (1802)', ANQ 28.ii[2015] 87-93). Contemporary accounts of the type indicate that it was considered particularly stylish, with "refined serifs" (p. 90). Lak takes the line of Gérard Genette and Jerome McGann that such apparently peripheral things matter to how an edition communicates to its readers, but it would be helpful to have concrete examples of this to consider. Instead Lak simply asserts that typography matters in this way and offers as the only concrete example of meaning attaching itself to writing's appearance the phenomenon of "the engraved caption" which although "not a typographic element" is nonetheless "a paratext that is as representative and expressive as both the printed text and the image itself" (p. 90). In fact, Lak does not even press home this point but simply quotes other critics who make this claim. Repeatedly, the article promises arguments that do not appear. The phrases "I will demonstrate . . ." and "I shall demonstrate . . ." occur just 880 words before the end of an article comprising only four and a half pages of writing. As Lak's brief account shows, the technical limitations of printing at the time--getting letterpress words and a picture on the same page was difficult--the choice to have captions engraved rather than typeset was not artistic but practical.

    And so to the round-up from Notes and Queries. John Klause published two relevant notes this year. In the first, he claims that Shakespeare remembered some phrases from his school Latin 'primer'--a devotional manual of prayers, extracts from scripture, and liturgical calendar--and used them across his career ('Shakespeare and the Primer', N&Q 260[2015] 84-9). The parallels are all with the 1531 "Salisbury" primer. None is certain and most seem tenuous; they would take as long to list as the article is itself. Klause reckons that a search in EEBO-TCP and LION for "dominical AND Oes AND machination* AND cataract* AND pristine" will hit only the 1531 Primer and the 1623 Shakespeare Folio. In fact, for me that search term finds only the 1623 Folio (not the Primer) in EEBO-TCP and returns nothing in LION, although in the case of the latter that may well be because ProQuest accidentally broke their search engine for LION when making a software upgrade in 2014 and have not yet (as of 7 February 2017) fixed it. Moreover, even if Klause's claim were true it would not tell us very much since he started by seeking out word/phrases common to Shakespare and the 1531 primer, so of necessity a search term that concatenates with a Boolean 'AND' the terms he found in both texts will match only those two sources. If you look for words and phrases common to any reasonably large texts you will find them, and if you then construct a search concatenating those words and phrases into a single search term that demands that they all be present, you may well get hits only in the two large texts you started with. This does not prove anything except that any reasonably large texts share many phrases, including a few that are not found elsewhere.

    Klause's second note concerns the imprese in Pericles scene 2.2 that have long been known to echo those in Claude Paradin's Heroical Devices of 1591 ('A Shakespearian Scene in Pericles, II', N&Q 260[2015] 578-83). Geoffrey Whitney's A Choice of Emblems (1586) has some of the images and mottoes, but Paradin's language is much closer to the play's. Klause reckons that Pericles's impresa itself comes from Paradin, which has not been noticed before, and it comes right after a mention of a Macedonian nobleman called "Lisimachus", thereby linking a part of Act II with the second half of the play, which everyone agrees is by Shakespeare. Klause finds echoes of Paradin's book elsewhere in Shakespeare's plays too. Also, another of the Knights in Pericles 2.2 has a motto, "Pue Per doleera kee per forsa" in the quarto, that seems to be a garbling of part of John Florio's First Fruits (1578).

    Klause finds nearby in Florio's book a series of sayings that pop up across the works of Shakespeare. Clearly, Shakespeare knew Florio's book and we have no evidence that Wilkins did. But in Wilkins's prose novella spin-off from the play, the correct English translation of this motto appears as "more by lenity than by force" although it is absent from the play quarto. Why? Because Wilkins's prose novella does indeed record what was said on stage, while the quarto garbles the script and omits part of it. Overall, Klause thinks that this all adds up to Shakespeare's being the author of Pericles 2.2, but in fact much of the detail seems not to be directed to that end and the claim rests on just two observations. The first is the likelihood, shown by his multiple uses of it across his canon, that Shakespeare knew Florio's First Fruits (the source for a motto in 2.2) and the absence of evidence that Wilkins did, and the second is the fact that Paradin's book connects, via the name Lysimachus, a moment in 2.2 to the governor in the second, Shakespearian, half of the play.

   C. W. Marshall finds the inspiration for Greene's Groatsworth of Wit calling Shakespeare an "upstart crow" in Virgil's "tum cornix plena pluuiam uocat improba uoce et sola in sicca secum spatiatur harena" (Georgics 1.388-9), which A. S. Kline translated as "Then the cruel raven's deep cry calls up the rain, and, alone with himself, he walks the dry sands" ('A Virgilian Source for the 'Upstart Crow'', N&Q 260[2015] 89-90). But Marshall reckons that improba (= wicked, cruel) has connotations of shamelessness, so cornix . . . improba is "upstart crow". Marshall takes it as uncontroversial that Greene wrote Greene's Groatsworth of Wit, and discusses his education, while of course others think that Henry Chettle wrote it. According to Azar Hussain, one reason to doubt that the "luces" jibe in The Merry Wives of Windsor is Shakespeare's revenge for his treatment by Sir Thomas Lucy is that he seems to have waited 10 years or more to make it ('Dear Deers: Sir Thomas Lucy and Charlecote Park Revisited', N&Q 260[2015] 90-2). But Hussain thinks that The Merry Wives of Windsor was not Shakespeare's first allusion to the Charlecote affair: 4.2 of 1 Henry 6 has Talbot give a speech about deer hunting and then Sir William Lucy (the historical ancestor of Sir Thomas Lucy) enters to bring news of Talbot's being surrounded; this is not in the sources. Also, there are two keepers with crossbows hunting deer (also not in the sources) in 3 Henry 6, and other references to deer-hunting. So, Hussain reckons, The Merry Wives of Windsor's glance at Lucy was not an isolated late occurrence but part of a recurrent pattern in Shakespeare's early works. The objection of course is that these are rather tenuous connections to the Charlecote story and there is nothing in them like the lousy/Lucy jibe in The Merry Wives of Windsor.

    Rosalind Barber has two notes of relevance this year. In the first she disputes David Kathman's claim that the word honey-stalks (Titus Andronicus 4.4.91) is a West Midlands dialect word for clover, a gloss originating with Samuel Johnson ('Shakespeare's 'Honey-stalks'', N&Q 260[2015] 92-3). Tamora says that this plant is dangerous to sheep, but in fact clover is good for sheep and what really hurts them is honeydew, the secretion of aphids while they feed on plants. Since honey-stalks is unknown to EEBO-TCP, Barber reckons that it is unlikely to be a West Midlands dialect word and more likely to be a Shakespearian coinage to denote any plant stalks covered in honeydew. Barber's second note traces the origins of a pair of names (Rosalind Barber 'Bardolph and Poins', N&Q 260[2015] 104-7). R. A. Newhall pointed out in 1933 that a Johan Bardolf was man-at-arms to Sir John Fastolf, captain of Honfleur in 1428, and wondered if that is where Shakespeare got the names Falstaff and Bardolph. Barber sketches the historical Lord Bardolph and his mention in 2 Henry 4, and the context of the Cobham family objection to Shakespeare's character's name Oldcastle in the late 1590s.

    The historical Oldcastle also had connections to a Bardolf and a line of the family Poyntz, all owning property in the same small part of Kent now known as the Hoo Peninsula. The fourth Lord Poyntz was in the 1350s arrested for the kind of wild behaviour (including horse stealing) that Falstaff's companions enjoy in the Henry IV plays. When Oldcastle was renamed Falstaff, the name Russell became Bardolph, which to those who knew about the Hoo Peninsula families would make it clear that Falstaff was really Oldcastle despite the name-change. If Poins became Poins at the same time (having previously been called something different) then this would, according to Barber, "cement that connection" (p. 107).

    Lukas Erne asks why Juliet in the Balcony Scene says that she has forgotten why she called Romeo back when it seems clear that she did so in order to ask what time to send for him the next morning (''I Have Forgot Why I Did Call Thee Back': Editing Romeo and Juliet's Leave-taking in the Balcony Scene', N&Q 260[2015] 93-5) The problem is only present in Q2, as Q1 has them agree the time and then Juliet says "Romeo I haue forgot why I did call thee backe". That extra "Romeo" (absent in Q2) is Juliet calling him back a second time, and then she says she does not know why she did this. That is, Q1 makes clear that Juliet calls Romeo back the first time in order to agree when to send for him the next morning, then they part again and she calls him back again. Erne exhorts editors to add to modern editions the necessary stage direction making this action clear.

    It is often said that Romeo and Juliet was performed by a troupe led by Robert Browne in Nördlingen, Germany, in January 1604 but George Oppitz-Trotman points out that local historians know this to be untrue: the company asked to play and were denied and compensated ('Romeo and Juliet in German, 1603-1604', N&Q 260[2015] 96-8). The nineteenth-century German scholar Karl Trautmann who announced the discovery of the application to play subsequently spotted the rejection of the application and retracted his claim, but Shakespearians seldom notice this retraction. The application to perform does show that the troupe claimed to have been performing plays including Romeo and Juliet and A Midsummer Night's Dream elsewhere in Germany. An application to play at Rothenburg ob der Tauber is similar to the Nördlingen one and seems to come from a non-English company, so that some might doubt that the players arriving at Nördlingen were English. Oppitz-Trotman reckons that the presence of Shakespeare's plays in the Nördlingen application alongside plays by Heinrich Julius of Brunswick, with whom Browne's company was closely associated, makes it likely that the players arriving at Nördlingen were indeed English even though a German-sounding man ("W. Echelin"), presumably their local agent, formally presented their application.

    It is conventional wisdom that Bottom's speech "The eye of man hath not heard, nor the ear of man hath not seen . . ." is a garbling of St Paul's words in 1 Corinthians 2:9 as represented in the Geneva bible which in the next verse calls this "the bottom of God's secrets" ('Bottom's Dream Revisited', N&Q 260[2015] 98-103). Kenji Go points out that the Geneva Bible of 1560 does not say this: it calls these "the deepe things of God". It was William Tyndale's 1526 translation of the New Testament that called them "the bottome of Goddes secretes" and other editions followed, including the New Testament published in octavo in Geneva in 1557. The problem is just which edition we identify as the Geneva Bible: the 1557 octavo of the New Testament or the full Genevan Bible of 1560. Go asserts that they are distinct works, the former merely the "groundwork" (p. 99) for the latter, and the two are highly divergent in readings. Most importantly, the 1557 octavo was not reprinted until the mid-nineteenth century and there is little chance that Shakespeare ever saw it. The Homily Against the Fear of Death quotes 1 Corinthians 2:9 in a way that is verbally closer to Bottom's speech than is any Bible Shakespeare might have seen, and follows it with the phrase "from the bottom of their hearts", so the Homily, Go reckons, is Shakespeare's source for Bottom's speech.

    Tom Reedy has two relevant notes this year. The first focusses on an interlineation above the line "Lin how say yo^u prentisses symple downe w^th him" in the Hand D Addition II segment of Sir Thomas More, which Alexander Dyce transcribed as "now, prentisses" and John Jowett read as "now prenty" so that Lincoln says (in modern English) "How say you now, prentices? Prentices 'simple'? Down with him!" ('A Misread Hand D Interlineation: 'Prenty' Vs 'Prentz'', N&Q 260[2015] 103-4). That is, Jowett understood "prenty" as an abbreviation of "prentices". Reedy reckons that in fact the manuscript reads not "prenty" but "prentz", which makes the case for it being an abbreviation of "prentizes" even stronger. Reedy reproduces the relevant words from a scan of a picture of the manuscript from a 1916 facsimile, which is so blurred that I cannot make anything out in it. 

    In his second note, Tom Reedy reports that in 1617-18, John Weever made a previously unnoticed transcription of John Combes's epitaph (beginning "Howe'er he livèd, judge not") written by Shakespeare, giving us an earlier version of it than the one known from Nicholas Burghe's manuscript of 1640-60 ('John Weever's Early Transcription of John Combe's Stratford Epitaph', N&Q 260[2015] 127-30). Where Weever's transcription differs from later ones, Weever's reading is to be preferred. Reedy remarks that this new early transcription of one of Shakespeare's alleged epitaphs for Combes does not clear up the problems regarding the other, more famous, one beginning "Ten-in-the-hundred here lieth in-graved" that seems so jokily unsuitable for a monument in the chancel of a church. Another mystery is why, if Weever saw "Howe'er he livèd, judge not" in 1617-18, it was missed by William Dugdale who recorded epitaphs in the church in 1634 and 1649. Reedy speculates that the substantial monument called for by Combes's will took a while to construct. Combes died in 1614 and the verses were placed somewhere around his tomb that enabled Weever to see them in 1617-18 before the monument was constructed and obscured them.

    Shakespeare's use of Cicero's Ad Herennium is widely recognized, and Colin Burrow thinks Prince Hal' speech beginning "I know you all, and while a while uphold | The unyoked humour of your idleness" (1 Henry 4 1.2.192-214) was influenced by Cicero's account of memory taking more note of rare events than everyday ones (''Rare Accidents' in 1 Henry IV and the Ad Herennium', N&Q 260[2015] 107-9). In particular, the phrase "rare accidents" in this speech is rare in the period and comes from Cicero's "raro accidunt" in Ad Herennium (3.22.35). More widely, the passage in Ad Herennium on eclipses being particularly memorable influenced Shakespeare's recurrent 1590s depiction of kings as celestial figures and his recurrent motif of monarchs' public appearances as rare spectacles that inspire wonder. 

    The "galled jade" is said to "wince" in modern editions of Hamlet, using the Q1 spelling over Q2/F's "winch". Steven Doloff accepts that this is just a spelling variation of the same word, but he thinks that "winch" should be retained for its associative richness ('Hamlet's Winching Jade', N&Q 260[2015] 109-11). In sketching these associations, Doloff claims that Hamlet is the only one of Shakespeare's plays that uses galled more than once, but Troilus and Cressida does too: "Their eyes o'er-gallèd with recourse of tears" (5.3.57) and "Some gallèd goose of Winchester would hiss" (Additional Passage B.22). Doloff reckons that Hamlet's "galled jade" glances as Gertrude, whose sexual behaviour he disapproves of, and that retaining the spelling "winch" would make Hamlet's line also allude to the prostitutes operating openly in the Bishop of Winchester's jurisdiction, as mentioned in Troilus and Cressida.

    Andrew Hadfield asks what exactly is a "bodkin" that Hamlet suggests could be used for suicide ('A Bare Bodkin', N&Q 260[2015] 111-2). Hadfield cites a variety of definitions covering such things as fancy hatpins exchanged as intimate gifts and concludes that in this case it "probably means a love token given to a beloved as well as a dagger with Hamlet referring to the imminent end of his prospective marriage to Ophelia" (p. 112). Since Ophelia has yet to return Hamlet's love tokens--that happens after this soliloquy--it is not clear to me why an audience or reader would be thinking of the end of his prospective marriage. The point of Richard F. Hardin's note is to remind editors that Eleanor Lumley pointed out in 1910 that Hamlet's "Assume a virtue if you have it not" (3.4.151) seems inspired by "Saltem, tute si pudoris egeas, sumas mutuom" (= "If you lack shame, you could at least borrow some"), which are Amphitruo's word to Alcumena in Plautus's Amphitryon ('Plautus and 'Assume a Virtue If You Have it Not'', N&Q 260[2015] 112-3).

    The first record of Othello that is usually cited is the court performance on 1 November 1604, but Brandon S. Centerwall has found an allusion to it in the prologue to the poem Saint Mary Magdalene's Conversion published in 1603 with an epistle bearing the date 31 January 1603 ('An Allusion to Othello, 31 January 1603', N&Q 260[2015] 113-6). The poem's first stanza alludes to a series of Shakespeare works--in turn, Julius Caesar, Romeo and Juliet, Othello, Troilus and Cressida, Richard 3, and Lucrece--so although considered alone the allusion to Othello seems tenuous the existence of the entire sequence itself makes it certain. The allusion is "and the rage, | Wherwith that passion is possest withall, | When ielousie with loue doth share a part, | And breedes a ciuill warre within the harte". This dating of the first performances of Othello to no later than January 1603 means that it can easily be understood as a source of Thomas Heywood's A Woman Killed With Kindness rather than the other way round.

    The Mother Goose rhyme "Hark, hark, the dogs do bark, | Beggars are coming to town; | Some in jags, and some in rags | And some in velvet gowns" is not known before 1672, but Lila Marz Harper thinks that the lines "Hark, Hark | The watch-dogs bark | Bow-wow, bow-wow" in The Tempest come from a version of this rhyme ('Hark, Hark!: Nursery Rhymes in The Tempest', N&Q 260[2015] 118). Harper reckons that hearing a snatch of it from Ariel would make Shakespeare's audience anticipate seeing beggars and "some in velvet gowns" and in a way this is what they get as the shipwrecked Ferdinand is led on stage by Ariel. An obvious objection to this claim is that the lines from The Tempest are not correctely quoted by Harper. The Folio has Ariel sing "Harke, harke, bowgh wawgh: the watch-Dogges barke, bowgh-wawgh". Modern editions sometimes rearrange these words so that "Bow-wow" is said by other characters as the "burden" that the preceding stage direction refers to, but Harper would need to claim that "Bow-wow" is somehow said or sung during Ariel's line in order to arrive at his singing this as "Hark, hark, the watchdogs bark", upon which her argument rests. This is not impossible, but it needs to be argued for explicitly.

    Margaret Jones-Davies finds significance in 64 being the number of squares on a chessboard and the number of words in the speeches of Miranda and Ferdinand playing chess (The Tempest 5.1.174-82), from "Sweet lord, you play me false" to "cursed them without cause" ('The Chess Game and Prospero's Epilogue in The Tempest', N&Q 260[2015] 118-20). The first 32 words end with Miranda's reply to Ferdinand "call it fair play" and the second 32 words are the speeches of Alonso, Sebastian, and Ferdinand. Jones-Davies notes that Ferdinand up to this point thinks himself king because his father is dead, so there are two kings on the island and then from this moment just one, as at the end of a chess game. A famous commentary on chess by Evrard de Conty was probably known by Shakespeare and in it the movements of the pieces are likened to a ship's manoeuvres to avoid being wrecked. As in The Tempest, this commentary has verse constructed of 64 units, in this case eight octosyllabic lines. These are somewhat like the octosyllabic lines of Propero's epilogue, referring to favourable winds and taking the form of a kind of prayer.

    Perhaps, suggests Geoffrey Caveney, the "W. H." to whom Thomas Thorpe's 1609 quarto Sonnets is dedicated was the stationer William Holme, who knew Thorpe and Edward Blount and worked closely with Adam Islip (''Mr. W. H.': Stationer William Holme (D. 1607)', N&Q 260[2015] 120-4). Thorpe and Holme's families were both from Chester. Caveney details the professional connections of these men and George Eld and John Wolfe, but amidst the welter of biographical details the point of this note emerges as a mere speculation that after Holme's death in 1607 ". . . Thorpe and Eld may well have come upon the manuscripts of the plays they entered [into the Stationers' Register] in 1607" among his possessions, and perhaps "they also obtained Shakespeare's sonnets in this way" (p. 123). This, according to Caveney, makes sense of calling Holmes the Sonnets' "begetter" and making its title-page look and sound funereal.

    Domenico Lovascio surveys the evidence from Ben Jonson that Shakespeare originally had Julius Caesar say "Know Caesar doth not wrong but with just cause" (Julius Caesar 3.1.47) and then revised away the illogicality before publication in the 1623 Folio where it appears as "Caesar doth not wrong, nor without cause | Will he be satisfied" ('Julius Caesar's 'Just Cause' in John Fletcher and Philip Massinger's The False One', N&Q 260[2015] 245-7). Lovascio has found new evidence for this argument in John Fletcher and Philip Massinger's play The False One written around 1620, in which Caeser (looking at a masque displaying great riches) remarks "I am asham'd I warr’d at home, (my friends) | When such wealth may be got abroad? what honour, | Nay everlasting glory had Rome purchas’d, | Had she a just cause but to visit AEgypt?" (3.3.77-80). Fletcher and Massinger elsewhere in the play refer to Shakespeare's Antony and Cleopatra and Julius Caesar and it seems unlikely that they would have put "just cause" into Caesar's mouth in their play if the phrase had not been in Caesar's mouth in Shakespeare's play. Gray Campbell asks just what Rosalind means when she says that she is as native to the forest "As the cony that you see dwell where she is kindled" (As You Like It 3.2.329-30)? Campbell reckons that cony here connotes the victim of the cony-catcher (an urban con-artist). That is, Rosalind calls herself a victim of deception but, being in disguise herself, is really the deceiver.

    James P. Bednarz reckons that ". . . the current theory that Twelfth Night was first staged late in 1601 or early 1602 is primarily based on the erroneous assumption that in it (at III.i.57-58) Shakespeare applauds Dekker’s mockery of Jonson’s clichéd diction in Satiromastix (I.ii.134-136, I.ii.186-188 and V.ii.324-327)"  ('Suspect Evidence for the Late Dating of Twelfth Night', N&Q 260[2015] 563-7). The evidence from Satiromastix is that Feste says he avoids using the phrase "out of my element" because it is "overworn", and it is mocked three times in Satiromastix (performed 1601). Bednarz's reports that the standard narrative is that Satiromastix was first performed by the Chamberlain's Men in August-November 1601 and hence Twelfth Night must be later than that. The dating of Satiromastix relies on its clear reference to John Weever's Whipping a' the Satyre entered in the Stationers' Register on 14 August 1601, so Satiromastix was written after 14 August 1601 and before its own entry in the Stationers' Register on 11 November 1601. This puts Twelfth Night between 14 August 1601 and 2 February 1602 when John Manningham saw it as Middle Temple Hall. But Bednarz thinks that Ben Jonson was the first to use the phrase "out of my element" mockingly rather than approvingly, upon with Shakespeare picked it up and mocked it in Twelfth Night, and then Dekker got the mockery from Shakespeare and used it in Satiromastix. Bednarz has no solid evidence for supposing that Shakespeare gave Dekker this mock rather than receiving it from him, so he tries to shore his claim up by moving Troilus and Cressida a bit earlier than its usual dating so that it more immediately responds to Jonson's Poetaster in late 1601. Pushing Troilus and Cressida earlier allows Twelfth Night to be earlier too, so that it was written early in 1601 rather than late in 1601. Bednarz may be right, but there other considerations that should go into any comprehensive chronology of Shakespeare's works, including analyses of drifts in Shakespeare's style over time.

    The dating of "The Phoenix and the Turtle" (otherwise known as "Let the Bird of Loudest Lay") relies on Love's Martyr being dedicated to Sir John Salusbury who, it is thought, got his knighthood on 14 June 1601, so the book at least was published between 15 June 1601 and the end of the year. Boris Borukhov announces discovery of a poem to Salusbury confirming that he got his knighthood for his service in putting down the Essex rebellion, and that he received this honour on 14 June 1601 ('A More Precise Date for Shakespeare's 'The Phoenix and the Turtle'', N&Q 260[2015] 567-9). It is not clear why Borukhow thinks that this precise date is news, since he quotes John Buxton reporting it in an essay on "The Phoenix and the Turtle" in 1980, although he implies that scholars had good reason to mistrust Buxton. 

    Thomas Merriam published two notes of interest to this review. In the first he considers the bits Jesuit William Sankey deleted from Henry 8 and King John in a copy of the 1632 Folio at Valladolid in Spain some time around 1641-51, taken as evidence that Shakespeare's plays are not pro-Catholic ('Shakespeare and Christian Doctrine in Retrospect', N&Q 260[2015] 133-6). Merriam points out that these bits were mainly in the Fletcherian end of the play. Sankey cut bits in King John that imitate The Troublesome Reign of King John and that Merriam has previously argued are non-Shakespearian too, probably by George Peele. Thus, Sankey's cuts cannot be used to argue that Shakespeare did not have Catholic tendencies, since we need to factor in the co-authorship of the plays that he cut.

    In the second note, Merriam revisits the claim in the 1987 Textual Companion to the Oxford Shakespeare Complete Works that the variation in the frequencies of use of 10 function words (but, by, for, no, not, so, that, the, to, and with) in Pericles show that it is co-authored ('Function Words as Author Invariants: Pericles as Test Case', N&Q 260[2015] 574-8). Merriam measured these words' frequencies in the two halves of Pericles, the two halves of Timon of Athens, all the other plays in the 1623 Folio, George Wilkins's Miseries of Enforced Marriage, and Thomas Middleton's A Game at Chess, The Ghost of Lucrece, The Revenger's Tragedy, and The Witch. He then performed Principal Component Analysis (PCA) on the resulting 10 (words) by 45 (works) matrix to reduce the 10 words to two PCs, plotted on an x/y scatterplot as his Figure 1. The Middleton works do not, on this plot, form a group apart from the 36 Folio plays: they are right in the thick of that cluster, except for The Ghost of Lucrece, which might be separate from the Middleton cluster because of its narrative rather than dramatic mode. Merriam concludes, rightly I think, that the frequencies of these 10 words alone do not make a good Shakespeare/Middleton discriminator. However, it is worth noting that the 1987 Textual Companion did not claim that they were good for that discrimination, and that cutting-edge analysis of function words as authorship discriminators now tends to use around 100 words not 10 for such work. In Merriam's plot the parts of Pericles are distinguished, but some of the 10 function words are considerably more involved in that discrimination than others, leading Merriam to write "This suggests that distinctions should be made between function words whose relative frequencies are author invariants with respect to Shakespeare and Wilkins, and those whose relative frequencies are not" (p. 575). Again, I agree: but cutting-edge work in this field already makes precisely that distinction when choosing which words to use.

    Merriam rightly asserts that this count alone tells us that Pericles and Timon of Athens are co-authored, but does not identify Shakespeare's co-author. He notices that of the 10 words, the word to is responsible for most of the separation between Shakespeare's plays and the Wilkins part of Pericles, so he reruns the test using only nine words, omitting the word to. Now, the Wilkins part of Pericles does not look so unlike Shakespeare. How, then, to choose the right function words for author discrimination? Merriam is right to assert that we need to find the function words that each author uses consistently rather than the ones they are inconsistent about. But he seems unaware of the extensive work done in the field by other researchers, and in particular John Burrows's Delta test that weights function words in just the way Merriam favours.

    Merriam uses a cumulative-sum (cusum) analysis (a method described in a review of his work in YWES for 2014) of 100-word segments from all the Folio plays, Miseries of an Enforced Marriage, and the two halves of Pericles, in each case counting the frequency of the word to. This shows its highly variable rate of usage across different parts of Miseries, so it is not good for discriminating Wilkins, or at least not if we use Miseries as representative of his work. Then Merriam performs the same analysis for the word the and finds it more consistently used by Shakespeare and avoided by Wilkins, concluding that it is an authorship distinguisher and indeed that it shows that all of Pericles is by Shakespeare (because the Pericles Acts I-II rates are like the Pericles Act III-V rates). None other of the 10 function words turns out to be as consistently used or avoided as the word the. This is all interesting, but the fact that one test is able to tell Merriam which function word to count (the) and tell him that this word's counts indicate Shakespeare's sole-authorship of Pericles should have rung alarm bells: we would want the test for the former to be conducted independently of the latter, using different samples. Most importantly, there are other, now well-established, methods for selecting which function words to count, and Merriam would need to show that his method, using cusum analysis, is better than those.

    Roger Stritmatter continues his series of articles claiming that Ecclesiasticus (or Wisdom of Sirach), not to be confused with Ecclesiastes, is a source for Shakespeare (''My Name be Buried Where My Body Is': The Influence of Ecclesiasticus 41 on Sonnets 71-74', N&Q 260[2015] 583-6). Here he wants to show that Ecclesiasticus 41 strongly influenced Shakespeare's Sonnets. Stritmatter claims that Ecclesiasticus 40:12 has "long been recognized" as the source for Iago's insincere speech about reputation, but his footnote in support of the claim that same verse is the source for Mowbray's similar speech about reputation in Richard 2 reads simply "Copy missing", which I take to be a comment from Notes and Queries's typesetter rather than part of the essay itself. In the remainder of the essay, Stritmatter quotes bits of Ecclesiasticus and asserts that each "calls to mind" (p. 584) a bit of Shakespeare, and in particular Sonnets. There are indeed verbal parallels, but Stritmatter does not establish that they are any more than commonplaces and none is especially close in phrasing.

    In his book Determining the Shakespeare Canon (reviewed in YWES for 2014), MacDonald P. Jackson used searches of the database Literature Online (LION) to show that unusual spellings found in A Lover's Complaint (published at the end of Sonnets, 1609) are disproportionately present in the Shakespeare canon and not in other writers' canon; the obvious conclusion is that this is because (as Thomas Thorpe's edition claims), A Lover's Complaint is by Shakespeare. But, as reviewers have pointed out, the database Early English Books Online Text Creation Partnership (EEBO-TCP) gives a much wider sample of spelling of the period than LION does, so Jackson repeats his searching but using EEBO-TCP instead of LION to look for spellings that the latter told him were rare ('A Lover's Complaint and Early English Books Online', N&Q 260[2015] 586-9). The result is that the spelling a twaine appears in A Lover's Complaint and Folio King Lear and is exceedingly rarely (two times across 53,000 books) in EEBO-TCP. Broadening out from LION to EEBO-TCP adds just four hits for the rare spelling fillial. Likewise the spellings sheelded and beseecht become a little less rare (gaining 14 and 22 more hits, respectively) when we move out to EEBO-TCP, but those numbers make them still rare.

    Jackson acknowledges that some spellings that seemed rare when searching in LION, such as addicions, become markedly less so when we broadened the searching to include EEBO-TCP. But moving to EEBO-TCP also allows Jackson to revise his baseline for rareness and run his tests again against the combined LION+EEBO-TCP set. There are just 12 spellings that appear in works by no more than 10 writers across this set--a twaine, daft (meaning doffed), fillial, gases (meaning gazes), greyned, hewed (meaning hued), laundring, leaueld, palyd, radience, sithed (meaning scythed), and wepingly. Jackson lists the hits for these 12 words and the results are obvious: Shakespeare's works come up far more often than one would expect if this were just a random sampling of the dataset. Switching from LION to EEBO-TCP only strengthens the evidence that A Lover's Complaint is by Shakespeare

Books Reviewed

Hoover, David L., Jonathan Culpeper and Kieran O'Halloran Digital Literary Studies: Corpus Approaches to Poetry, Prose and Drama. Routledge [2014]. 216 pp., £35.99, ISBN 978-1138210547

Kidnie, M. J. and Sonia Massai, eds. Shakespeare and Textual Studies. Cambridge University Press [2015]. 482 pp., £74.99, 978-1107023741

Lesser, Zachary Hamlet After Q1: An Uncanny History of the Shakespearean Text. University of Pennsylvania Press [2015]. 304 pp., £52, ISBN 978-0812246612

Shakespeare, William Macbeth, ed. Sandra Clark and Pamela Mason, The Arden Shakespeare. Bloomsbury [2015]. 408 pp., £8.99, ISBN 978-1904271406