Picture of Gabriel Egan G a b r i e l   E g a n  .  com

"Shakespeare: Editions and Textual Studies in 2018" Not the Year's Work in English Studies

Only one major scholarly edition of Shakespeare appeared this year: King John edited by Jesse M. Lander and J. J. M. Tobin for the Arden Shakespeare, bringing its Third Series (begun in 1995) to completion. King John was the first Shakespeare play to be filmed (in 1899), and Lander and Tobin begin their Introduction (pp. 1-133) by giving a reading of this short film as an epitome of how the politics of historiography operate. The play was first printed in the 1623 Folio and was mentioned in 1598 by Francis Meres in his Palladis Tamia. The earliest possible date of composition is provided by its dependence on the 1587 edition of Raphael Holinshed's Chronicles. There exists also a play called The Troublesome Reign of King John published in 1591, which seems to be a source for Shakespeare's play; if so, 1591 is the earliest date for King John. In 1986-87 the Oxford Complete Works gave a date-range of 1595-96, and the 2016-17 New Oxford Shakespeare gave a best guess of 1596. Lander and Tobin explore some of the play's possible allusions to contemporary events to see if this helps fix the date. They do not.

   King John is Shakespeare's most explicitly political play, considering not only the role of the nation in modern politics but also international religious matters, legitimacy, the power of oaths, and public opinion. In this play ". . . the magic of royal charisma is almost entirely absent" (p. 7). The play depicts the contraction of the English monarch's empire to just the island itself, and then explores whether this really is a viable embodiment of the concept of England. There are no good common Englishmen in this play, unlike other Shakespeare history plays: Hubert apparently comes from the French city of Angers and the Bastard is still a gentlemen even if not royal. The play explores but does not answer the question of what is Englishness.

    The visit of the papal legate Pandulph to England was known to Elizabethans from the account of it in the Homily Against Disobedience and Wilful Rebellion (1570). The play's depiction of Pandulph's excommunication of King John and his declaration that anyone assassinating John will be worshipped as a saint would remind audiences of the papal bull against Queen Elizabeth (Regnans in Excelsis) issued by Pope Pius V in 1570. (True Catholic doctrine does not, of course, allow the sin of murder to be forgiven in advance in the way Pandulph offers.) That the legate so outrageously provides religious cover for political murder would have flattered patriotic English Protestant audiences. The deletion of some lines of the play in the copy of the Folio at the English College at Valladolid in Spain indicates that the play was acceptable reading there once the most anti-Catholic bits, including the claim that a monk was the poisoner of King John, were excised.

    Although the play is anti-Catholic, there is little in King John that is pro-Protestant. The Reformation itself caused an increased interest in casuistry, and in particular the ethics of reconciling one's conscience with one's political allegiance. In the play, the struggle is played out as a tension between conscience and commodity, the latter meaning self-interest, particularly as represented in property. The dispute over the legitimacy of Philip Faulconbridge the Bastard offers a way of thinking about the legitimacy or otherwise of any monarch. In Elizabethan law, a child born in  wedlock, as Philip undoubtedly was, is his mother's husband's legitimate son even if another man actually got her pregnant. (Philip of course renounces his inheritance because following the King and Eleanor offers better chances for advancement.) The complications concerning the legitimacy of King John are equally vexed: it is not clear who is right in the play's arguments about it.

    The debates about whether an impious oath or one that contradicts another oath was truly binding was taken seriously in the period. Pandulph talks King Philip of France into breaking his oath of peace to King John on the grounds that it was religiously invalid when it was made. The topic of divided loyalties was pertinent to the Elizabethan religious present. Peter of Pomfret's prognostication does come true, but not in the way the John understood it. The play suggests that supernatural powers do exist.

    What makes a play a history play? The fact that it relies on historiographical sources is central to this question. For King John these sources are Holinshed's Chronicles and perhaps John Foxe's Acts and Monuments, and in some way (many scholars think) the play The Troublesome Reign of King John. On the other hand E. A. J. Honigmann thought that The Troublesome Reign of King John was derived from Shakespeare's play. The question of the two plays' relationship is not settled by the reliance on such minor sources as Matthew of Paris's Historia Maior (1570) or the Latin manuscript Wakefield Chronicle.

    Although opinion is divided on whether Shakespeare relied on Foxe's Acts and Monuments, Lander and Tobin see clear evidence that he did, especially in the manner of King John's death: Foxe and Shakespeare both use the unusual word burst for what happened to the monk who was John's taster and who committed the poisoning. Lander and Tobin do not think that Shakespeare had a hand in The Troublesome Reign of King John and hence the "W. Sh." on the title page of its second edition, in 1611, was just an attempt to "take advantage of the growing demand for Shakespeare's printed drama" (p. 44). In a footnote (p. 44n1), Lander and Tobin give the references for the major claims that Troublesome Reign was written by George Peele, in the works of Charles Forker, Brian Vickers, and H. Dugdale-Sykes. The likeness between King John and Troublesome Reign resides mostly in "structural similarities" and the "sustained verbal parallels between the two are limited to short phrases scattered throughout the text" (p. 44).

    In Troublesome Reign the audience is told why the Bastard hates the Duke of Austria--the lion-skin he wears once belonged to Coeur-de-Lion, the Bastard's father, and was lost in combat--but in King John the audience is not told. The Bastard in Troublesome Reign is violent, a bully even, and kills the Abbot he suspects of abetting the murder of King John. Shakespeare's Bastard is more thoughtful and has aspects in common with Edmund in King Lear. King John does tone down the militant Protestantism of Troublesome Reign, but rather than being simply pro-Catholicism this might be understood as Shakespeare appreciating the nuance and ambiguity of historiography and the problem that we cannot ultimately know exactly what happened in the past and what people's motivations were. That Shakespeare's play seems to be pointing in multiple directions, that it does not have a simple moral to impart, is not accidental: it is a consequence of Shakespeare wanting to convey the multivalences of historiography.

    In a section on "Rhetorical Patterns" (pp. 51-65), Lander and Tobin offer a study of the schemes and tropes used in the patterned speeches of several characters, especially Constance. Then they turn to the ways that various characters exemplify certain passions, and especially to Constance again, with her grief, and then to King John as wrath personified, and to Hubert as the embodiment of pity. Hubert is tricky because the Folio does not make clear whether he is also the Citizen of Angers who appears on the town walls and editors are divided on this point.

    In "Afterlives" (pp. 65-113), Lander and Tobin note that King John was the English monarch who appeared most often on the early modern stage, in Troublesome Reign and King John and four other dramas; the list is in the Index of Characters in Early Modern English Drama created by Thomas L. Berger, William C. Bradford, and Sidney L. Sondergard. King John itself is not known to have been much performed before the 1730s when it rose in popularity and stayed in regular performance until the nineteenth century. It then fell in popularity and was not much performed in the twentieth century and it remains under-performed and under-read today.

    The expurgation of oaths in the Folio text implies that this edition was printed from a manuscript made after the 1606 Act against stage profanity, and the play's presence in a list from 1669 of King's Men's plays formerly acted at the Blackfriars implies performance there after 1608. The earliest known performance is at Covent Garden in 1737, at a time when religiously inflected Jacobite uprisings were much feared. Colley Cibber adapted King John to make his play Papal Tyranny, which was produced at Covent Garden in 1745, the year in which Bonnie Prince Charlie landed in Scotland to lead the last Jacobite rebellion. Cibber's version puts back the anti-Catholicism found in Troublesome Reign but also nuances the sympathies (there are good Catholics) and it stresses the distinction between spiritual authority (which popes may wield) and political authority (which popes may not).

    Cibber also makes his play allude to the settlement of the Glorious Revolution: his John "rules by authority of law and Parliament as opposed to heredity, and thus exemplifies Whig principles against Jacobite claims in favour of the House of Stuart" (p. 73). In Cibber, the English barons side with the French only to make King John sign the Magna Carta acknowledging their rights, which is an aspect of history that is absent from Shakespeare's play. This is Whiggish: Magna Carta, with its codification of the "fundamental compact between monarch and subject" (p. 73) became an important political landmark only in the seventeenth century. A rival production of King John using Shakespeare's original text was mounted in response to Cibber's version and the effect of this competition and the ensuing public debate about the merits of the original over the revision was to raise the status of the play and build the foundation of its assured place in the repertory for the next 150 years.

    Lander and Tobin survey some eighteenth-century critical responses to the play. In the nineteenth century the play became an opportunity for staging as "a spectacular historical extravaganza" (p. 80), with historical accuracy (especially in costume) being the watchword and selling point. Lander and Tobin do not, as other critics have, dismiss such antiquarianism as a depoliticizing of Shakespeare. An abiding interest in the past may itself be political, and historicizing a play by getting the costumes right is also a way of politicizing it, of saying that this, what we show you, is what we British, collectively, really came from. Lander and Tobin also consider a parodic version of the play that mocked the concerns of Victorian antiquarian theatre. Lander and Tobin spend several pages on the antiquarianism of Charles Macready's 1842 Drury Lane production (pp. 91-95).

    Then they turn to the history of the text, noting that the Folio is so largely correct that little has had to be done. Alexander Pope imported some lines from Troublesome Reign to clarify why the Bastard hates the Duke of Austria. Charles Kean's antiquarian production of 1852 and Herbert Beerbohm Tree's antiquarian production of 1899 at Her Majesty's Theatre followed, and the latter was partly filmed, providing a link to the new medium that would allow antiquarianism (or any other illusion) to be taken even further. Interest in producing the play fell off thereafter. Lander and Tobin turn to a novel that features a production of King John and then Buzz Goodbody's 1970 production for the Royal Shakespeare Company, and finally a couple of other more recent productions.

    Of special interest to this review is the "Texual Note" (pp. 113-33) with which Lander and Tobin end their Introduction. Regarding "The Manuscript Source" (pp. 113-23), they observe that the publishers of the 1623 Folio did not pay for licences to print King John and The Taming of the Shrew although they did for 16 other previously published plays. Overall, the Folio dialogue of King John is not in need of much correction, although the stage directions and speech prefixes are noticeably variable in quality and clarity. Some aspects of the action seem garbled, especially regarding the presence of the characters of the Sheriff and James Gurney in 1.1, the Earl of Salisbury in 2.1, and the prophet Peter of Pomfret in 4.2. Lander and Tobin list the usual New Bibliographical categories of manuscript (with footnoted caveats about disagreements over them), and they find that the copy for Folio King John cannot have been the licensed book because of "the irregularities in its directions text" (p. 117) by which they mean (to infer from list item 2 on page 114) the stage directions and speech prefixes.

    Lander and Tobin conclude that a transcript must have been the Folio copy, and not one made from the licensed book because of those irregularities in the stage directions and speech prefixes. To illustrate this, Lander and Tobin quote the opening stage direction of 2.1 beginning "Enter before Angiers . . .", reckoning that this phrasing is authorial and "not at all characteristic of a bookholder, since there would be nothing on stage that would either illustrate or name the town" (p. 118). But the critiques of New Bibliography that they previously cited repeatedly pointed out that such authorial stage directions did indeed find their way into manuscripts used to run plays in performance. Also, Lander and Tobin object that the full version of this stage direction, which names only the principal characters, omits to mention the army that is needed for this to become a battle scene. Such deficient stage directions lead Lander and Tobin to conclude that the Folio printer's copy was a transcript of the authorial foul papers.

    The transcript must have been made for some theatrical purpose after the 1606 act against profanity for the play has been "haphazardly . . . expurgated" of oaths (p. 119); there would have been no need for expurgation for a non-theatrical purpose. Two scribes seem to have done the copying, one with a preference for the spelling O and the other with a preference for Oh. The state of the text means that editors may be fairly conservative with the dialogue, leaving it largely unemended, but quite interventionist in the stage directions and speech prefixes. Lander and Tobin explain their policies for intervening in stage directions. Where the action is certain from the dialogue but not mentioned in the Folio, they add a direction unless the dialogue also makes it clear exactly what that action is. Where is action is likely but not certain they discuss it in the notes (p. 124). A tricky case is just when King Philip first takes hold of and when he lets go of King John's hand in 3.1.

    Lander and Tobin find that the "Act and Scene Divisions" (pp. 125-33) are particularly faulty in Acts 2 and 3. Things start out alright with the play opening in the Folio as "Actus Primus, Scaena Prima" (on a1r). This first scene is long (276 lines in this edition) and most editions treat it as also the whole of Act 1. The Folio's next marker is "Scaena Secunda" (on page a2r) and what follows, the scene "before Angiers", starts what most editions call Act 2. This first scene of Act 2 is also long (598 lines in this edition) and Lander and Tobin decide that this, the play's second scene, is the whole of Act 2. The Folio's next marker is "Actus Secundus" (on a4v) but Lander and Tobin reckon on the basis of the dramatic continuity that what follows could to be said to constitute a second scene in Act 2, although they instead treat it as 3.1, the start of Act 3.

    The Folio's next marker is "Actus Tertius, Scaena prima" at the top of column 'a' of page a5r yet at the bottom of column 'b' of A4v, the preceding page, there is no scene-ending exit and the catchword is "Actus". This absence of a scene-ending exit leaves Constance on stage (seeming to suggest that she sits on the ground) and yet immediately after ""Actus Tertius, Scaena prima" is an entrance direction that includes her. Lander and Tobin treat this "Actus Tertius, Scaena prima" as erroneous and ignore it, and they remove Constance from the entrance direction since she is already on stage. Thus they simply continue their scene 3.1 with the entrance of King John and King Philip (and their followers) to Constance.

    Next the Folio has "Scaena Secunda" (on a6r, at the start of what Lander and Tobin call scene 3.2) and what follows is a 10-line exchange involving the Bastard, King John, Hubert, and Arthur that ends with "Exit". Then comes "Alarums, excursions, Retreat. Enter Iohn, Eleanor, Arthur Bastard, Huberts, Lords", which Lander and Tobin reckon should indicate the start of a new scene, so they make what follows their scene 3.3. What comes next in the Folio is "Scaena Tertia" (on a6v) which Lander and Tobin emend to make the start of their scene 3.4. Thereafter the divisions make reasonable sense, except for a misprint in the marker for the beginning of the last act, which has "Actus Quartus" where "Actus Quintus" is meant. Thus the problems occur principally in Acts 2 and 3.

    To help make sense of all this, Lander and Tobin summarize Gary Taylor's argument that act intervals were a King's Men's practice that spread to all performances, outdoors as well as indoors, after 1608 when they regularized the practices at their Blackfriars and Globe theatres. The fact that the division into acts is imperfect in King John strongly implies that it was not done for performance after 1608, since for that purpose it would need to be done well. Having removed entirely the marker "Actus Tertius, Scaena prima"--treating Constance's sitting down and the entrance of King John and King Philip and their followers to her as continuous action--Lander and Tobin ponder what it means if there really is an act interval in performance here and they decide that Constance would (like the lovers in A Midsummer Night's Dream) stay on stage across an act interval. Since they believe that the act and scene divisions in Folio King John reflect neither original composition (when continuous performance was the norm) nor post-1608 revival (because they are badly done), Lander and Tobin simply adopt the act and scene division of the Riverside Shakespeare edition, on account of its being the reference text for Marvin Spevack's one-volume concordance to Shakespeare (p. 130).

    Turning to the light the printing of the book might shed on the act and scene divisions, Lander and Tobin give an account of the typesetting of a folio-in-sixes, which is the format the Shakespeare First Folio takes. Their account of how two compositors shared the setting of a Folio gathering gives the unhelpful impression that an exceptional practice used in some gatherings was normal for all. They write that having cast off the copy for the first five pages of a gathering to reach pages 6 and 7, "One compositor would work on this forme [pages 6 and 7]; then when it was ready it would be imposed and printed while he was working on the outer forme (pages 5 and 8); the other compositor (for they usually worked in pairs) would be doing the same thing for pages 4 and 9 (inner forme of the second sheet) and 3 and 10" (p. 131).

    Although each compositor seems to have set a whole forme at the start of the printing of King John, this was not (as Charlton Hinman's investigations revealed) the normal practice across the whole of the Folio. More typically the compositors started by taking a page each of the inner forme of the inner sheet of a gathering (the one Lander and Tobin call pages 6 and 7) and then one man worked backwards through the cast-off first half of the gathering (setting pages 5, 4, 3, 2, and 1 in that order) while the other man worked seriatim setting the second half of the gathering (pages 8, 9, 10, 11, and 12). Paul Werstine's essay "Cases and compositors" in Studies in Bibliography in 1982 showed that casting off the second half of the gathering too would enable a more flexible practice of assignment of compositors to particular pages as the need arose, and this helps explains some out-of-sequence typesetting found in the Folio.

    Lander and Tobin give the order of typesetting by Folio Compositor B as "a3v and a4r, a3r and a4v, b3r and b4v, b2v and b5r; b2r and b5v; b1v and b6r, b1r and b6v" and that of Compositor C as "a2v and a5r; then a2r and a5v; a1v and a6r; ar and a6v; and b3v and b4r" (p. 132). They do not say where they get this order from, but it agrees with the table in Peter W. M. Blayney's introduction to the second edition of Charlton Hinman's Folio facsimile, with one proviso. Within each pairing, Lander and Tobin presumably do not mean to assert which of the two pages were set first. If they do--if their "a3r and a4v" implies that a3r was set before a4v--then they are contradicting Blayney's table.

    According to Blayney, Compositor B's order was in fact a3v, a4r, a4v, a3r, b3r, b4v, b2v, b5r, b2r, b5v, b1v, b6r, b1r, b6v and Compositor C's was a2v, a5r, a2r, a5v, a6r, a1v, a1r, a6v, b3v, b4r. Having laid out all this data, Lander and Tobin attempt to apply it to the faulty act and scene divisions in Acts 2 and 3 of Folio King John. The order of setting of pages matters because in reconstructing the sequence of faulty divisions Lander and Tobin write that ". . . [Compositor] C, on a2r, set 'Scaena Secunda' at the top of the second column, and on a5r, his next page, at the top of the first column, set 'Actus Tertius, Scaena Prima' . . ." (p. 132). But in Blayney's sequence a5r was set by Compositor C before he set a2r, not after it.

    As it turns out, none of this matters since nothing arising from their account of the order of setting pages tells Lander and Tobin anything about why the act and scene divisions are erroneous, other than the fact that both compositors had already set many plays of the Folio (and so were experienced) and that hence the two of them independently making the same kinds of error suggests that the problem was in their manuscript copy. If the manuscript merely had numerals to count acts and scenes then one compositor might easily think that a "II" meant Act 2 while the other took the same numeral to mean Scene 2. This could account for the "Actus Secundus" we see on a4v (set by Compositor B), when what was meant was the second scene of Act 2, and "Scaena Secunda" on a6r (set by Compositor B), which really is the second scene of Act 3. It is hard not to conclude that Lander and Tobin's excursus into the minutiae of the Folio's typesetting was more for show than substance.

    And so to Lander and Tobin's text of King John. As usual, this will be a survey of some interesting textual choices, not a account of every editorial intervention. Their List of Roles has some surprising diction in its descriptions, such as Hubert being an "imperfectly obedient intimate of King John" and Melun being a French lord "with some English blood". Unusually for an Arden3 edition, as well as the edition's own line numbering the right margin provides the Folio signature and column identifiers, here "[a1rb]". No reason is given for this.

    At Lander and Tobin print "Enter a Sheriff [and whispers to Essex]" and credit this stage direction to "this edn". But Edward Capell is the originator of the idea that the Sheriff (who has no publicy spoken lines in the play) whispers the matter of the Faulconbridge dispute to Essex, who raises it with King John. Capell omitted the "and" but he should nonetheless be credited. Capell himself got this solution to the problem from The Troublesome Reign of King John where at the parallel moment the Sheriff whispers to Salisbury who raises the matter with the King.

    Lander and Tobin have the Bastard call Richard the Lionheart "Cordelion" (1.1.53), the Folio's form, rather than modernizing to "Coeur-de-lion" as editors have done since Pope made this change in the early eighteenth century. The cogent reasons for modernizing given in Jürgen Schäfer's essay "The Orthography of Proper Names in Modern-spelling Editions of Shakespeare" in Studies in Bibliography in 1970 seem to apply here, since there are no metrical obstacles to the modernization in "Of Cordelion, knighted in the field". And surely the hyphens are needed to prevent the reader misunderstanding the stress as corDELiON, which is how such a four-syllable word would normally be stressed in English.

    At 1.1.236-237, the Folio has the Bastard, talking of his bastardy, say that "Sir Robert could doe well, marrie to confesse | Could get me sir Robert could not doe it". Lander and Tobin emend this to "Sir Robert could do well--marry, to confess-- | Could he get me? Sir Robert could not do it", crediting this as Henry Halford Vaughan's punctuation that makes "marry, to confess" a parenthetical clause, Pope's insertion of "he" in the second line, and E. A. J. Honigmann's Arden2 question mark in "me?". This solution rather leaves "Sir Robert could do well" out on its own. Do well at what?. Lander and Tobin might have mentioned other solutions that link the lines so that the Bastard asserts that Sir Robert could do well (in modern parlance, would have done well) to get the likes of the Bastard.

    Where the Folio has "Heauen lay not my transgression to my charge" (1.1.256), Lander and Tobin print "God, lay not my transgression to thy charge!". They give the siglum "Oxf", which their list of References identifies as John Jowett's edition in the 1986-87 Oxford Complete Works, as the originator of their emendation of "Heaven" to "God". In fact Jowett's edition did not make this emendation, and neither did "Oxf1", which is A. R. Braunmuller's 1989 single-volume edition in The Oxford Shakespeare series. Curiously, they give no source for their second emendation in this line, of "my charge" to "thy charge", which was first proposed (but not adopted) by Howard Staunton in his edition of 1859. 

    At 2.1.0, Lander and Tobin retain the archaic spelling "Angiers" for the French town whose modern name is Angers. It is not clear why they do this, since at 2.1.152 they, as expected, modernize the Folio's archaic spelling of the province of "Toraine" to its modern spelling Touraine. The Folio has King John say "Arthur of Britaine, yeeld thee to my hand" (2.1.156), in which "Britaine" denotes the French province that we now call Brittany. Using this modern form would wreck the line's meter, but Lander and Tobin's modernization to "Britain" risks misleading the reader into supposing that land on the English side of the Channel is meant. In this case, retaining the Folio's archaic spelling could be defended as a means to stop the reader making that error; Lander and Tobin rely on an explanatory note instead.

    In the Folio, King John describes to the citizens of Angers the French military force and violent intentions directed against the town, which is "All preparation for a bloody siedge | And merciles proceeding, by these French. | Comfort yours Citties eies, your winking gates: | And but for our approch, those sleeping stones" (2.1.213-216 ). The "s" in "yours" is clearly just an easily corrected slip, but the idea of comforting the city's eyes makes little sense, so Lander and Tobin adopt Edward Capell's emendation of the penultimate line so that it begins "Confronts your . . .". This singular form of the verb suggests that its subject is "proceeding" in the previous line. Lander and Tobin perhaps ought to have added punctuation to end the previous sentence after "bloody siege" to make this clear, or else added a note that the more-distant singular "preparation" is the subject and that "proceeding" and "siege" are two things being prepared for.

    The Folio assigns to "Hubert" the line "Heralds, from off our towers we might behold" (2.1.315), which attribution has been much debated. Is Hubert the same "Citizen" of Angers who earlier in the scene appeared on its walls to defy the English and French armies? It is notable that the speech prefix here is spelt out in full rather than being abbreviated as usual. Lander and Tobin write in their note for this line that: "F gives the SP 'Hub.' (Hubert), a denomination which this edition believes is an error that resulted from the doubling of the parts of the Citizen and Hubert by one actor, a doubling that led to a momentary authorial confusion of sequential roles". But the Folio does not give the speech prefix as "Hub." (rather it uses the unabbreviated "Hubert."), and that is part of the mystery. It is difficult to see how doubling the roles of Hubert and a citizen of Angers could cause Shakespeare confusion when writing the play, since presumably such casting decisions came later.

    Proposing the dynastic marriage of the French Dauphin and Blanche of England, the Citizen of Angers says "Such as she is, in beauty, virtue, birth, | Is the young Dauphin every way complete. | If not complete of, say he is not she" (2.1.432-4). The problem is that "complete of" seems to make no sense. Instead of emending, as seems necessary, Lander and Tobin in a note "Freely paraphrase" the wider sentiment (which is in any case clear) and suggest that "of" can mean "thereof" or "therein", but without identifying parallel usages.

    In the Folio, King John begins the description of the dowry he will provide to enable Blanche to marry the Dauphin with "For Angiers, and faire Toraine Maine, Poyctiers", which Lander and Tobin render as "For Anjou, and fair Touraine, Maine, Poitiers" (2.1.487). They credit Pope within emendation of the town of "Angiers" to the province of Anjou, but by the same logic they should emend the town of "Poyctiers" to the province of Poitou. If the point is to correct Shakespeare's shaky French geography, it must surely be done consistently.

    At 2.1.521, Lander and Tobin treat as an emendation (credited to Nicholas Rowe) their rendering of the Folio's "yong-ones"--in "What saie these yong-ones?", meaning let us ask the young couple--as "young ones". This is merely modernization not emendation. Likewise at 3.1.144 the modernization of the Folio's "sea" (in the sense of a job in the church) to "see" is footnoted as if an emendation. And again at 3.1.300 the modernization of the Folio's "whurle" to "whirl" is treated as an emendation. Lander and Tobin elsewhere provide textual notes for corrections that require no editorial judgement because the error is obvious, such as the Folio's "Euvenom" for "Envenom" at 3.1.63. At 4.1.122, Lander and Tobin follow Pope to read "For all the treasure that thine uncle owns" where the Folio spells the last word as "owes". They treat this is an emendation and explain that "owes" provides a "pun suggesting indebtedness rather than possession". This is true, but it is a pun built into the language since owe was simply an early modern form of modern own. Otherwise, Roderigo's remark "What a full fortune does the thick-lips owe" in Othello makes no sense. Likewise Lander and Tobin treat the modernization of "complement" to "compliment" at 5.6.16 as an emendation.

    Sometimes a critical comment from Lander and Tobin seems oddly trite, as in their observation that regarding her marriage "Blanche yields the decision making to uncle John, unlike the more independent Hermia of A Midsummer Night's Dream . . ." (2.1.522). Likewise, it is not clear that anything more than human nature is being described in the comment that "Shakespeare often has a character blame the messenger when the message is painful" (3.1.4), which refers the reader to Cleopatra behaving this way. And again in mid-battle at 3.2.7 when the Bastard reassures King John that his mother is safe with "My lord, I rescued her", to which Lander and Tobin comment: "The Bastard's offstage saving of Eleanor contributes to his expanding image of the heroic in a phrase not quite so memorable as 'Reader, I married him' but nevertheless important". I cannot see what aid is provided by this allusion to Charlotte Bronte's Jane Eyre.

    There are clear signs of the play script underlying the Folio edition having undergone censorship of its religious oaths at some point, as when Constance says "A widdow cries, be husband to me (heauens)" (3.1.108). If the last word were the eminently censorable "God" then the line makes more sense. Lander and Tobin choose instead to make the last word "Lord", which is also reasonable, but they credit the emendation to "Oxf subst.", which in their scheme of abbreviations means that Jowett's 1986-87 edition made substantially the same emendation. But Jowett did not: he emended to "God".

    In the Folio, Pandulph compares King Philip's holding onto King John's hand to holding "A cased Lion by the mortall paw" (3.1.259), and most editors emend cased to caged or chafed or crased (that is, modern crazed), but Lander and Tobin defend "cased" as meaning "with its skin on, i.e. alive". It would be helpful to hear of a parallel for this unusual sense of "cased". The Folio has King Philip dismayed that "A whole Armado of conuicted saile" (3.4.2), meaning his navy heading to England, has been dispersed by a storm. Lander and Tobin note that although "convicted" makes a kind of sense as "vanquished, defeated", much better sense is made by F. G. Fleay's emendation to "convected" meaning sailing together. Unfortunately, Lander and Tobin have no entry in their lists of References and Works Cited for Fleay's edition, but it seems (from other scholars' references to it) to have appeared in the Collins School and College Classics series in 1878.

    At 3.4.48 the Folio has Constance say "I am not mad, I would to heauen I were" and Lander and Tobin change "heaven" to "God", presumably believing it another case of religious censorship. They neglect to provide a textual note recording this emendation. As part of Arthur's pleading with Hubert, the Folio has him say "If heauen be pleas'd that you must vse me ill" (4.1.55). Lander and Tobin again emend "heauen" to "God" and again neglect to provide a textual note. In this case, there is good reason to suppose the Folio correct and uncensored: pleasing heaven is a recurrent Shakespearian idiom and pleasing God is not.

    The Folio has Hubert summon his assistants for the killing of Arthur thus: "Hub. Come forth: Do as I bid you do" ( 4.1.71). Lander and Tobin render this moment as "HUBERT [Stamps his foot.] Come forth! [Enter Executioners with rope and heated irons.] Do as I bid you do". Their textual note garbles the account of this emendation, beginning "71 forth!] Ard2; foot: F . . .". Since the word "foot" is not in the Folio, they cannot mean what this note asserts, which is that they emended the Folio's "foot" to "forth". Presumably they mean only to report that they change the colon after "forth" to an exclamation mark.

    In response to news of the uprising in France, the Folio has King John ask "Where is my Mothers care?" (4.2.117) or perhaps ". . . eare?" since the first piece of type in the last word is damaged and this piece had previously been used sometimes as a c and sometimes as an e. The "eare" reading is not absurd. John is asking how she did not hear of (and react to) the uprising, and the messenger replies by saying what happened: "her ear | Is stopped with dust" in that she died. Curiously, Lander and Tobin make no comment on this crux and treat "care" as if it were undisputedly the Folio's reading. At 4.3.15-17, Lander and Tobin have Salisbury answer Pembroke's question about who brought the letter from the cardinal with the speech "The Count Melun, a noble lord of France, | Whose private with me of the Dauphin's love | Is much more general than these lines import". This is simply the Folio text modernized, and Lander and Tobin make no remark on the peculiar phrase "Whose private with me". One would at least expect a note explaining what the editors think Salisbury means.

    The Folio has the Bastard remark that impatience has privilege "to hurt his master, no mans else" (4.3.33) and Lander and Tobin naturally emend to the Second Folio's ". . . no man else". But in explaining how this simple error happened they make the peculiar claim that "F's reading, 'mans', looks very much like an instance of haplography where the scribe or compositor has let his eye return to manners in 31, the penultimate word in the line as here". The compositor's eye going to the wrong line and finding "manners" two lines earlier is possible, but seeing this word and then setting "mans" would be an unlikely error. And this would not in any case be haplography, which is the accidental setting just once of something that should have been set twice (as in rember for remember), and there is nothing here meant to be set twice.

    At 4.3.41-44, Salisbury says to the Bastard, regarding the dead body of Arthur before them, "Sir Richard, what thinke you? you haue beheld, | Or haue you read, or heard, or could you thinke? | Or do you almost thinke, although you see, | That you do see?". Lander and Tobin follow the Third Folio in emending the end of the first line to "Have you beheld." If the point was to make this the beginning of a compound question ending with "you do see?" then it is incomprehensible why they change the Folio's comma after "beheld" to a full stop. Their "Have you beheld." is neither a question nor an assertion. The Folio's word order is in any case defensible in two ways. The whole of "what thinke you? you haue beheld" may be a single question whose question mark is accidentally displaced from its end to its middle. Or, the first three words are a question and the second three an assertion, and a fresh question starts with "Or haue you . . .". In early modern English it was acceptable to start a question with "Or". The textual note at 4.3.82 seems to be incomplete as it reads "82 God] xxx; heauen F".

    Summarizing what will happen if Pandulph does not act to quell the rebellion, King John says in the Folio "Or ouerthrow incureable ensues" (5.1.16). Lander and Tobin emend to ". . . incurably ensues", which is the Fourth Folio reading. It is not clear why they think the Folio reading incorrect, and since incurably is a word that Shakespeare never uses elsewhere this editorial intervention is heavy handed. As a gloss on the Dauphin's "I did not think to be so sad tonight" (5.5.15), Lander and Tobin offer the trite remark "the Dauphin at least knows the origin of his sadness; contrast Antonio in The Merchant of Venice l.1.1, 'In sooth I know not why I am so sad'". The lines "Why may not I demand of thine affairs | As well as thou of mine?" (5.6.4-5) are assigned to Hubert in the Folio. Lander and Tobin reassign these lines to the Bastard without recording the emendation with a textual note and without mentioning it in the commentary.

    At 5.7.15-16 the Folio has Prince Henry say, of his dying father King John, "Death hauing praide vpon the outward parts | Leaues them inuisible". Lander and Tobin make no emendation of the last word and simply gloss it as "invisibly". It would have been helpful to know what they think this line means, since the notion of Death invisibly leaving something is unclear. Overall, Lander and Tobin seem not to fully understand the Arden series rules about textual notes or have not applied them properly. More than occasionally they omit a necessary textual note regarding a substantive emendation such as changing a word (heaven to God for example) or reassigning lines to different speakers, yet they repeatedly provide unnecessary textual notes for mere modernizations of spelling such as complement to compliment.

    In an appendix Lander and Tobin offer a "Casting Chart" (pp. 331-34) but it does not provide a proposed allocation of all the parts to a given number of men and boys. Instead it is a list of which character is in which scene(s), which is a prerequisite for preparing a casting chart that establishes the doubling. Lander and Tobin provide a few hundred words on some of the doubling options, but the only figure they provide for the size of the company needed to perform the play is T. J. King's rather large determination of 19 men and 5 boys.

    Another scholarly edition published in 2018 is relevant to this round-up, and it would be transformative of the field if its claims were true: 'A Brief Discourse of Rebellion and Rebels': A Newly Discovered Manuscript Source for Shakespeare's Plays edited by Dennis McCarthy and June Schlueter. Unfortunately, the manuscript McCarthy and Schluter present is not what they think it is. They begin by describing the manuscript, A Brief Discourse of Rebellion and Rebels, whose title they abbreviate to "Discourse". It was written by George North in 1576 and supposedly provides a source for the opening soliloquy of Richard 3, the organization of bee society in Henry 5, the likening of human variation to dog variation in Macbeth, the citizens' uprising in Coriolanus, and the fight of Cade with Iden in 2 Henry 6. The manuscript is also supposed to clarify the Fool's Merlin prophecy in King Lear.

    McCarthy and Schlueter provide some background on North's other writings and his biography, including his being at Kirtling Hall (home of Roger, Second Lord North, to whom the manuscript is dedicated) at the same time as Thomas North of Plutarch's Lives fame, but without detailing these Norths' relatedness, which is frustrating. One pertinent contradiction is that North's "Discourse" gives the standard line that all political rebellions are revolts against God and therefore to be condemned, yet he served William the Silent, Prince of Orange, who was in revolt against the rule of Hapsburg Spain in the Netherlands. McCarthy and Schlueter's introduction to the document ends with some history of the manuscript's peregrinations.

    In Chapter Two "Uncovering Connections Between North's 'Discourse' and Shakespeare's Plays" (pp. 11-27), it starts to become apparent how McCarthy and Schlueter have deceived themselves about the significance of their manuscript. The main problem is that they do not understand the mathematics of probability that apply to the discovery of verbal parallels between texts. They start by describing how investigators went hunting for parallel phrases in the days before big digital text datasets. McCarthy and Schlueter think that "NEAR" and "FBY" (which means "Followed By") are "Boolean operators" (p. 13), but of course they are not, they are proximity operators. Boolean operators are terms such as AND, OR, and NOT. McCarthy and Schlueter explain how source identification has in the past used phrases and words that are common to multiple texts.

    Using sections from Plutarch's Lives and Antony and Cleopatra that have the words unarm, condemn, courage, less, woman, and noble mind in common, McCarthy and Schlueter explain the likelihood of this occurring by chance: ". . . matching condemn(ed) would be like randomly picking the right number between 1 and 13,000; repeating unarm(ed) would be like hitting the right number between 1 and 58,000--and then doing it again for noble mind . . . " (p. 16). These large numbers come from dividing the number of times a given word is found in the Shakespeare canon, say 65 for condemn and its cognates, by the size of the Shakespeare canon, 884,400 words, to get for condemn the quotient of around 1/13,000. But the use of these numbers by McCarthy and Schlueter is linguistically and mathematically wrong.

    While it is true that if we pick a word-token at random from the Shakespeare canon the probability that it will be condemn is about 1-in-13,000, that kind of selection is not what writers do when they pick words. Rather, writers pick from the set of different words (known as word-types) that they have in their vocabulary. In his writing, Shakespeare used perhaps 20,000 to 30,000 different words--depending on how you define the word word--and he presumably knew a few thousand more that he happens not to have used in his surviving writing. Thus the better analogy would be that Shakespeare picked condemn 65 times from a set of something around 25,000 words that he could have used, which is 1-in-385 not 1-in-13,000. Worse, with "and then doing it again" McCarthy and Schlueter treat their (overstated ) unlikelihoods as if they ought to be simply multiplied together as we would independent probabilities.

    In fact, words have affinities and they tend to evoke one another. A chivalric context links unarming and nobility so that once we have found one of these two concepts being used the likelihood of finding the other nearby rises. This is not to say that McCarthy and Schlueter are entirely wrong about the evidentiary value of unusual clusters of words, but they clearly are misapplying probability calculations in their attempts to quantify it. McCarthy and Schlueter use the Early English Books Online Text Creation Partnership (EEBO-TCP) dataset (as provided by the commercial digital publisher ProQuest) to show that certain words in the passage from Antony and Cleopatra are clustered in only one other text of the period: North's Plutarch's Lives. The key questions are how one selects the words to search for and how one sets about the parallel hunting, since, as MacDonald P. Jackson has repeatedly shown, if we start with two corpora we think are connected and use EEBO-TCP only to eliminate phrases that occur in other corpora we are bound to find that our two corpora share a number of phrases and collocations that no other corpus has. This is true of any two sufficiently large corpora we start with; we could do it with the corpora of Bob Dylan's song lyrics and the Good News Bible and would find that they share phrases and collocations found nowhere else. This is a raw fact of large language datasets and does not tell us about authorship or sources.

    McCarthy and Schlueter show that in EEBO-TCP the eight words proportion, glass, feature, fair, nature, deformed, world, and shadow collocate within 200 words of one another only in the opening speech of Richard 3 and that they also collocate in the "Discourse". They then show that a passage in the "Discourse" collocates parts, heavens, which, obedience, bees, that by a, nature, creature, rule, order, have, king, some, at home, abroad, burdens, summer, drone, lazy, and actions which also collocate in 31 lines in Henry 5 (1.2.181-211). True, but the spread of the collocations is wide in "Discourse" from folio 4v to folio 8, which is a span of 821 words and the spread in Henry 5 is 218 words. McCarthy and Schlueter consider the possibility that Shakespeare and North independently derived this collocation about bees and social order from Thomas Elyot's The Governor, and they find that Shakespeare is closer to North's version than to Elyot's, which indicates that North is Shakespeare's source (pp. 25-27). The evidence for this claim about closeness is essentially subjective in that it concerns ideas being substantively the same although differently phrased.

    Chapter Three is about "The Final Hours of Jack Cade" (pp. 28-51). There are details in Shakespeare's account of the end of Jack Cade in 2 Henry 6 that do not appear in any of the known sources: a "repentent monologue" (p. 28), an account of his suffering, the weather is hot, he is hiding in fear until starvation makes him enter Alexander Iden's garden, "his limbs are emaciated" (p. 29), he eats grass, he loses his fight with Iden because of his wasted condition, and that Iden plans to drag Cade's body by the heels and leave it for the crows to eat. McCarthy and Schlueter claim that Iden says he intends to drag the body "along the streets" (p. 29) but that detail is not in Shakespeare's play. North's text (in verse at this point) draws on William Baldwin's Mirror for Magistrates.

    McCarthy and Schlueter note that aside from North we have no known source for certain details of Cade's end in The Contention of York and Lancaster: that he was starving and willing to eat grass, "that Iden wished him to hell" (p. 32), that he was forced to wear a halter, and that his corpse was eaten by crows. In fact, the detail about halters comes early in The Contention of York and Lancaster, in scene 4.9, and it applies to Cade's followers not to Cade as North tells it. McCarthy and Schlueter reckon that the death of Cade in The Contention of York and Lancaster is also influenced by North's version of Owen Glendower's end, but the words they highlight are commonplaces of rebels' death, such as rebels, death, axe, and so on. They also hear North's account of Glendower in Richard 3 and list the terms in common, which are all commonplaces: cursed, hap, betide, hate, death, and wretched.

    Listing some words common to North and Cade in The Contention of York and Lancaster 4.4, McCarthy and Schlueter again use EEBO-TCP to show that the words in question (alone, fly, forsake, and rascal) do not collocate anywhere else except in The Contention of York and Lancaster. As before, the problem is methodology: if you start with an assumption that two texts are connected you can always whittle away the non-unique words they have in common to leave a nubbin of words they uniquely have in common, but that is also true of texts that are unrelated. Some of McCarthy and Schlueter's claims are especially carefully worded so as to be strictly true but misleading. For example, they accept that the collocation of like, hare, and hound (found in North) is common in writing of the period but ". . . this [3 Henry 6] is the only Shakespeare play that contains" it (p. 38). True, but the collocation is found in Venus and Adonis too, at lines 678-84, so what is the significance? Likewise the idea of fearing a bush to be an enemy (in North), which they say occurs elsewhere in Shakespeare only in 3 Henry 6, since the famous fear of "a bush supposed a bear" in A Midsummer Night's Dream is "similar but not the same" (p. 38). These sound to me at least as alike as some other supposed parallels that McCarthy and Schlueter accept.

    McCarthy and Schlueter think that the phrases "bereft me of my life" (from North) and "bereft thee of thy life" (from 3 Henry 6) are extremely rare in the period, being found only once elsewhere in EEBO-TCP before the 1623 Folio appeared. In fact this phrasing is common. The words bereft, me, and life collocate frequently in the period and appear as "Hath me bereft of honor, life" in the "Complaint of Henry Stafford, Duke of Buckingham" that was printed multiple times including in the Mirror for Magistrates (STC 1252, sig. U4r) that we know Shakespeare used as a source for his history plays.

    McCarthy and Schlueter think that when North has Cade's ghost say that he was "horst . . . like a hog" and "drawn like a mastie dog" he must mean that "Iden had dragged Cade by the heels" (p. 41) which is what Shakespeare's Iden says he will do. But I do not see why being horsed (meaning to be carried or hoist), or indeed even being dragged, must involve the heels, and in North the line about Cade's being horst comes before the dragging of Cade's body and does not seem to refer to the body being dragged. McCarthy and Schlueter are pressing their evidence more forcefully than it can bear.

    McCarthy and Schlueter report that carrion, crows, and beast do not appear within 20 words of one another anywhere except in North and The Contention of York and Lancaster (outside of the Cade scenes). Again this cut-off seems arbitrary and if we widen the range to 25 words we find an occurrence in Edward Topsell's History of Serpents (STC 24124, sig. M3r) and, widening it a bit further, we find the three words together in a pair of consecutive sentences in Topsell's History of Four-Footed Beasts (STC 24123, sig. V5v). It is hard not to suspect that McCarthy and Schlueter's cutoff of 20 words was chosen because it eliminated these occurrences that run counter to their claim for this collocation's rareness.

    McCarthy and Schlueter claim that the phrase "flesh and fell" is unique "in the seventeenth century" (p. 43) to North and Shakespeare's King Lear " (p. 43), but it is not clear why this "seventeenth century" qualifier suddenly appears, especially as North's "Discourse" was written in the sixteenth century. Perhaps McCarthy and Schlueter do not want their reader to be distracted by the several sixteenth-century occurrences that are easily found. Again McCarthy and Schlueter seem to be cherry-picking their evidence. They adduce further parallels that seem either fairly inevitable--of course Cade in North and Cade in Shakespeare refer to their claim to the throne via descent from Edmund Mortimer--or else banal, as in filth collocating with scum, which can be found in many works. McCarthy and Schlueter insist that filth followed within five words by the bigram scum of is found in no other published work before 1592, and indeed that is what their precise search term ("filth* fby.5 scum of") shows in a search of EEBO-TCP as provided by ProQuest. But a search of the same dataset as provided in the United Kingdom by the state-owned Jisc Historical Texts interface reveals that "scum & filth of" appears in George North's The Philosopher of the Court (STC 19832, sig. B2v), a text that McCarthy and Schlueter discussed earlier. If their parallel hunting were robust one would expect them to mention this hit since it is an even closer match to The Contention of York and Lancaster's "filth and scum of" than the "filthy scum of" from "Discourse" that they make much of here (p. 45).

    Discussing the authorship of 1, 2, and 3 Henry 6, McCarthy and Schlueter make no mention of the ample scholarly evidence that these plays were co-authored. They go on to make a series of further claims for weaker connections between North's "Discourse" and Shakespeare's history plays, and all the above objections apply. Of course one can find unique matches between canons for words and phrases that are in any case likely to occur in writings on these topics, such as traitor, rebel, and death.

    Chapter Four is about "The Fool, Merlin's Prophecy, and the Upside-Down World of King Lear" (p. 52-77). McCarthy and Schlueter find echoes of words from North's Jack Cade in Edgar-as-Tom's speeches in King Lear. For example, Cade lists various kinds of dogs (trundle tail, greyhound, mastiff, and spaniel), as does Edgar-as-Tom. It is not clear just what McCarthy and Schlueter think is the smoking gun that proves that Shakespeare borrowed from North for his list of types of dog. McCarthy and Schlueter then turn to the Fool's Merlin prophecy in King Lear, which appears only in the Folio version of the play, describing at length the critical debates about it and mentioning the attribution of it to Geoffrey Chaucer in George Puttenham's The Art of English Poesy of 1589.

    The claim of Shakespeare's dependence on North here is tortuous and implausible. McCarthy and Schlueter have found that the collocation in North of world followed within 100 words by turned up side down accompanied in the same work by milk within 10 words of honey and 10 words of mouth occurs in only one work in EEBO-TCP, Thomas Pugh's British and Outlandish Prophesies of 1658, and manual examination of Pugh shows that it has many phrases in common with North. In Pugh the prophecy is attributed to Merlin. But what is the connection to Shakespeare? North does not attribute his prophecy to Merlin, only to "our late Poet".

    McCarthy and Schlueter reckon that North must have got his version of the Merlin prophecy from a lost manuscript and that a chance connection to Shakespeare is unlikely since ". . . King Lear is likely the only play in English literature that includes" (p. 65) a version of these supposedly Chaucerian lines. But McCarthy and Schlueter have overlooked Terence Hawkes's demonstration (Notes and Queries 205 (1960): 331-2) that there was in Shakespeare's time a fashion for prophecies attributed to Merlin about states coming to ruin. Because they are unaware of just how many such Merlinesque and Chauceresque prophecies were in existence, McCarthy and Schlueter mistake as unlikely the hypothesis that North and Shakespeare got theirs independently.

    McCarthy and Schlueter think that only in North and Shakespeare does the prophecy invoke a topsy-turvy world: "This intertwining of Merlin with social inversions was peculiar, if not unique, to North" (p. 66). In fact this feature is present in the earliest known version printed by William Caxton in his book of Chaucer, including "And robbery is holden purchas | And lechery is holden solas". When McCarthy and Schlueter tabulate the words and phrases that North and King Lear have in common, they are either commonplaces of prophecy (time, forsee) or are clearly derived from the Caxton-Chaucer original ("come to . . . confusion")

    McCarthy and Schlueter also find bits of North's version of the prophecy in King Lear 1.4, which they claim dramatizes a "young . . . Goneril" interacting with her father. It is not clear why they think that the play indicates Goneril's youth, and performers more frequently contrast a young Cordelia with her elder sisters. McCarthy and Schlueter find analogies that are hard to credit, for example in their remark that "The first line of the Fool's Merlin prophecy--'When priests are more in word than matter'' (3.2.81)--is analogous to the second line in North's Merlin prophecy -- 'God's word is made a waxing nose' (fol. 35v)" (p. 70). According to McCarthy and Schlueter a "waxing nose" is a nose of wax (easily deformed), but it is hard not to hear a Pinocchio-like suggestion of a nose growing in size.

    Regarding another matter of plot, I am not clear why McCarthy and Schlueter believe that by the middle of King Lear ". . . the machinations of Regan and Goneril have depleted Lear of his horses and his large train . . ." (p. 71). Just why Lear ends up almost alone and where his followers have gone are, I think, matters that the play does not explain. McCarthy and Schlueter go on to detect yet more supposed parallels between North and parts of King Lear and but these are just the commonplace words and phrases one would expect in two texts about revolt against a monarch.

    Even when McCarthy and Schlueter find that a collocation they suspect of being rare ("the sun the moon FBY stars") is in fact common, being present in 548 books, they phrase their account to make the evidence sound significant: ". . . this is less than one per cent of the searchable works" (p. 76). From a literary point of view, McCarthy and Schlueter's claim that the presence of Fool's prophecy in King Lear was "obscure until the rediscovery of 'Discourse'" should have rung their critical alarm bells: why would Shakespeare include something that his audiences could not make sense of unless they knew North's manuscript?

    Having used up what they consider their strongest evidence, McCarthy and Schlueter's Chapter Five, "Political Monologues and a Glimpse of Coriolanus" (pp. 78-88), explores some even less convincing parallels between North and Shakespeare's works. Their chapter title promises connections with Coriolanus but McCarthy and Schlueter report that they are saving the full story of that relationship for another book. They end by noting that the real Alice Arden of Arden of Faversham was one of the Norths and lived at Kirtling where she met her lover and co-conspirator Mosby. Roger Second Lord North patronized the theatre and troupes visited Kirtling Hall in the 1570s and 1580s, including the Queen's Men, and hence it is not implausible that Shakespeare had access to the manuscript of the "Discourse".

    In this chapter, McCarthy and Schlueter inadvertantly reveal another flaw in their logic (p. 82n60). They start with phrases as they find them in North and in Shakespeare and search for other occurrences in EEBO-TCP as provided by ProQuest. As they acknowledge, ProQuest's interface to the EEBO-TCP dataset does not treat as a contiguous phrase a set of words split across multiple lines. If we search for "consent to this most" (a phrase from Shakespeare's King John) then EEBO-TCP does not find it in Shakespeare because it appears as "consent | To this most". McCarthy and Schlueter know where the line-breaks fall in the phrases they find in North and Shakespeare and hence know just when they need to compensate for this limitation of EEBO-TCP by searching not for a contiguous phrase but for one or more words in the phrase forming a set that is said to occur near or followed-by one or more other words in the phrase that also form a set. That is, to find the hit from King John they have to enter the search term "'consent' FBY 'to this most'", which will overcome ProQuest EEBO-TCP's limitation regarding line-breaks.

    McCarthy and Schlueter know how to form these sets, taking into account the line-breaks they know about, so that the phrase they enter finds a hit in the Shakespeare canon in EEBO-TCP. But they do not know--they cannot know unless they exhaustively try every possibility--how other divisions of the phrase into sets might have led to other hits in EEBO-TCP that are not being shown because of the ProQuest interface's inability to spot a phrase across a line division. A great many of the search terms that McCarthy and Schlueter use as evidence employ contiguous phrases that might be subject to this problem, which illustrates one of the dangers of arguing from not finding something in EEBO-TCP: the reason you fail to find it might be something others than its absence from the dataset. There is certainly something amiss with McCarthy and Schlueter's ability to find phrases in EEBO-TCP, for they assert that "a spade but a spade" is found in North and has a parallel in Coriolanus ("a nettle but a nettle") but nowhere else since "an EEBO search for a spade but a spade yields no results" (p. 85). In fact "a spade be but a spade" is found in STC 4714 (published in 1575) at sig. Oo4r.

    There is no doubt that McCarthy and Schlueter have found some interesting parallels between their manuscript and Shakespeare's plays. But how much significance should we attach to them? They attempt to answer this question in an afterword called "The Odds that the Parallels are Coincidental" (pp. 89-94). The "Discourse" is only 13,000 words. McCarthy and Schlueter think that the unlikeliness of coincidence is all the greater when not only are rare words and phrases found in common but also "the unique correspondences even occur in the context of the same historical event" (p. 90). In fact the parallels are more, not less, likely to be coincidental if the writers are dealing with the same historical event since the known facts constrain the things they write about. Two accounts of the death of Adolph Hitler in his Berlin bunker are likely to share ideas, words and phrases--such as those concerned with isolation, increasing delusion, despair, and suicide--simply because so many ideas, words and phrases, for example those concerned with delight, exhaltation, triumph and magnanimity, are ruled out by the reality of what actually happened there.

    To put some numbers on the likelihoods of matches, McCarthy and Schlueter take the word insculpt that occurs in Richard Robinson's translation of Gesta Romanorum (1577) and in The Merchant of Venice and they notice that it occurs in EEBO-TCP only 73 times. (Presumably they mean that it occurs in precisely this spelling 73 times since a search for the more common spelling insculped shows another 150 occurrences.) Since EEBO-TCP comprises "more than a billion words", McCarthy and Schlueter calculate that insculpt occurs at a rate of less than "1 out of 10 million words" (p. 91) in the corpus. Since there are only 1,711 words in the casket scenes in The Merchant of Venice, McCarthy and Schlueter divide 10 million by 1,711 to find that the odds of "Shakespeare's randomly inserting insculpt into one of his casket-related dialogues are less than 1 in 5,844" (p. 91).

    This is of course statistical nonsense, since no one supposes that Shakespeare randomly inserted words into his writing. (Writers' selections of highly common function words can be modelled with the mathematics of stochastic systems, which include randomness, but that is not to say that words are chosen at random.) McCarthy and Schlueter's calculation would be correct if the casket scenes were created by randomly choosing one word-token followed by another and then another from the EEBO-TCP corpus, but that is not how plays are written: writers choose word-types from their vocabulary not word-tokens from others' writings. Moreover, even if McCarthy and Schlueter's method were correct, the odds of insculpt or insculped being randomly selected are about four times greater than they calculate since they have ignored the more common of the two spellings.

    Worse still, even though McCarthy and Schlueter's calculation is numerically correct for the (mistaken) picking-words-out-of-the-corpus model of composition, they wrongly express what we should conclude from it. The number 5844 (ten million divided by the length of the caskets scenes in The Merchant of Venice) is how many times one would have to repeat the operation of creating the casket scenes by their random method before one would expect the scenes to contain the word insculpt. This number 1÷5844 is not the likelihood that for any one such scene so created its inclusion of the word insculpt "is a coincidence" (p. 91n65). Indeed, if one adopts the statistical way of approaching such problems that McCarthy and Schlueter favour--the so-called Frequentist approach--then one has to cease using the word coincidence in the way they do.

    In the remainder of their discussion of probability, McCarthy and Schlueter compound their statistical fallacies by asserting that they can simply multiply the unlikelihood of finding one word by the unlikelihood of finding another, as if collocations were what we call independent probabilities when of course they are in fact dependent. For example, in a scene of execution the probabilities of finding words such as sin, regret, and confess or perhaps true, defy, and tyranny are shaped by whether the dramatist depicts the victim accepting or rejecting his punishment. Words collocate according to the tone of a scene and are not independent of one another; it is remarkable that two literary scholars would assume otherwise. These fallacies lead McCarthy and Schlueter to propose probabilities as large as "1 in 40 billion" (p. 93) for some of the matches between North and Shakespeare. The remainder of their book contains a transcription of the manuscript, a photographic facsimile of it, and an index.

    No monographs on our topic were published in 2018, but three book-form collections of essays have some that are relevant. The one with the greatest number of relevant essays is Stage Directions and Shakespearean Theatre edited by Sarah Dustagheer and Gillian Woods. The editors provide an Introduction (pp. 1-16) in which they remark that there was no requirement for the three sources of textual authority for a play--the playbook, the backstage plot, and the actors' parts--to be precisely in agreement about its stage directions. Dustagheer and Woods claim that scholars give lower authority to stage directions than dialogue less because they distrust the stage directions' "problematic provenance" than because they prefer the "'literary' language of speeches" (p. 3).

    Dustagheer and Woods conclude from the evidence about manuscripts that ". . . the provenance of stage directions is not that much more uncertain than other parts of a play-text" (p. 4). But they discuss this provenance only in respect of the manuscripts used in the theatre, and I would say that a good reason to give printed stage directions lower authority than dialogue is the fact that printshop compositors demonstrably did so: they were more likely to alter or omit them than to alter or omit dialogue. In addition, stage directions found in the printed books might not have been performed (the actors could overrule them) and moreover might have been written for readers not players.

    Eric Rasmussen writes that because we number a stage direction by the dialogue line that precedes it and add ".1" to its reference, "Numerically, at least, a stage direction is worth exactly one tenth as much as a line of dialogue" (quoted on p. 8). Dustagheer and Woods develop Rasmussen's idea, claiming that ". . . the value of 'original' stage directions is decimated by the referencing apparatus" (p. 9); but, witty as it is, Rasmussen's idea is wrong, since the system is not arithmetical. In scene 2.4 of  Henry 8, for example, the opening stage direction runs from to, and there are no tenths in that system of labelling. If ".1" meant "one tenth" as Rasmussen claims then after we should see, but in fact we see The clue that the system is not arithmetical is in the presence of multiple full stops in the reference: in arithmetic there can only be one.

    The first essay in the book is "Inventing Stage Directions; Demoting Dumb Shows" (pp. 19-43) by Tiffany Stern, who notes that the term stage direction was coined by Lewis Theobald in 1726 and popularized in his 1733 complete-works edition. Those familiar with Stern's work will not be surprised by her argument that dumb shows were held on separate pieces of paper kept apart from the main dialogue of a play, and hence that they sometimes got inserted into printed plays in the wrong place or bundled together at the end or omitted entirely. She gives examples of these mistakes.

    Were dumb shows written by someone other than the playwright? It is not obvious, Stern observes, why dumb shows should be on separate pieces of paper. She suggests that they were written by "action-experts, not playwrights" (p. 27). She gives the example of Thomas Hughes's Misfortunes of Arthur being printed with a note about the dumb shows' authors. The Master of the Revels George Buc claimed to have written a dumb show for someone else's play, and this turns out to be the only piece of evidence Stern has for a play's dumb show(s) being written by someone other than the playwright.

    Dumb shows needed separate rehearsal because the staging was paramount. Stern details the differences between the Q2 and F versions of the dumb show in Hamlet, and the main one seems to be the use of loud trumpets in the former and quiet hautboys in the latter, suggesting perhaps revision for indoor performance. The loose integration of play dialogue and dumb show is, for Stern, a possible explanation of slippages such as The Murder of Gonzago being said to be about a duke and duchess but the dumb show calling for a king and queen. Stern next tells the story of Theobald inventing the term stage direction in 1726 in relation to the "table of green fields" crux in Henry 5, and then using the term opprobriously in 1733 regarding defective stage directions in Hamlet and Macbeth.

    Stern usefully traces the evolution of the term stage direction in the writings of subsequent editors of Shakespeare, and in particular the use of it for parts of his plays thought to have been written by someone other than Shakespeare and to be reflections on what happened on the stage, and its belated sense of "prescriptions for acting written by playwrights" (p. 37). Stern reckons that ". . . Shakespearean directions are seldom, and perhaps never, for an actor. That means that not only do few to no Shakespearean (or other early modern) paratexts meet the Theobaldian definition of 'stage direction'; they also do not meet the modern one" (p. 37).

    To close, Stern returns to her previous work on the headings such as The letter, The riddle, and The song that are used in play scripts to signal that what follows should be written out on a separate piece of paper to form a property document. These are obviously not stage directions. Stern claims that the instructions to stage hands rather than to actors to put out properties, or get them ready, do not fall under the modern definition of stage direction either since they are not directed to actors. More controversially, she claims that entrances and exits are not directed at actors either but to the scribe making the parts. I would point out that the words Enter and Exit are present in Edward Alleyn's part as Orlando Furioso, and that surely what is in the part is directed to the actor reading the part.

    Echoing Warren Smith, Stern claims that directions for kissing, dying, hand-holding and so on are directed to the prompter warning him that the silence is intentional. I agree that they serve this purpose as well, but surely their main purpose is to tell the actors to kiss, die, or hold hands. All the remaining directions are, Stern claims, fictional not practical: they describe the imagined world not the theatre world. It is clear that Stern is overstating her case when she writes that "Even a simple direction like 'enter Ophelia' is fictional: it asks that Ophelia, the fictional character, enter, rather than that a boy playing Ophelia enter" (p. 40). I would object that the boy playing Ophelia is being theatrical not fictional when he reads this and thinks "I am Ophelia and must enter here". By her argument, Stern arrives at the startling conclusion that there are perhaps no stage directions at all in Shakespeare.

    In Chapter Two of Dustagheer and Wood's collection, "The Boundaries of Stage Directions" (pp. 45-67), Laurie Maguire observes that sometimes the boundary of dialogue and stage direction is not easily drawn. Stage directions are always liminal and serve to mediate between the different worlds of reading and performing. Thus they "police and enable" (p. 48) the boundary between the fictional and the theatrical world. All directions implicity include the phrase as it were, but some (such as two in Coriolanus) actually use this phrase, as if we would otherwise mistake the theatrical for the real. Maguire sketches the claim that as it were might be Ralph Crane's phrasing in Coriolanus, pointing out that Crane's interference in the Folio texts he transcribed remains uncertain.

    Sometimes the language of a stage direction seems to mix theatrical-practical matters with fictional ones, as in "Exeunt into his house" in the anonymous Wit of a Woman (1604) and "exeunt ambo from Wales" in George Peele's Edward I (1593). Maguire explores a series of as if and as from stage directions. Peculiarly, it is with violent directions that the as if bit is often left out and the direction seems to call for actual violence to the actor. Other stage directions flit back and forth between fictional world and theatrical world in adjacent phrases.

    Maguire observes that some stage directions "enter the narrative territory of the novel" (p. 56), for example by repeatedly telling the reader the relationships (his eldest daughter) or the group-memberships (the suitors) of the characters entering. She does not explore the possibility that authors might write such directions as aide memoires for when the author returns to continue the writing after taking a break or passes the manuscript to a co-author to continue. Maguire points out that some directions that seem to contain spurious information about what is going to happen (Enter . . . to drink, Enter . . . to save . . .") can usefully tell an actor how to behave.

    In "'Peter falls into the hole': Nonce Stage Directions and the Idea of the Dictionary" (pp. 69-90), Paul Menzer and Jess Hamlet consider the idea that the professional players of Shakespeare's time may have shared a special language. Maybe there was none, and any dictionary of theatrical vocabulary should acknowledge this. Menzer and Hamlet built a short dictionary of one-off stage directions by pruning from Alan Dessen and Leslie Thomson's Dictionary of Stage Directions all the common phrasings. Thomas Heywood turns out to be the writer most given to unique phrases. Dessen and Thomson's dictionary covers 60 years of practice, across which standards and fashions may have come and gone.

    In a parenthetical remark, Menzer and Hamlet write that "This chapter does not use the words 'problematize' or 'interrogate'; ten years ago it might well have. It might also have been called 'Dictionarying the Stage Direction'. Today's jargon is tomorrow's embarrassment" (p. 78). The editors Dustagheer and Wood allowed this remark to stand although the word problematic comes up in Suzanne Gossett's essay and in Sarah Dustagheer and Philip Bird's essay, and twice in the editors' Introduction, and interrogate is used three times by Andrew Hiscock and twice by Sarah Lewis.

    Discussing the stage direction "Shoots the Dagge" in Thomas Kyd's The Spanish Tragedy, Menzer and Hamlet neglect to explain that dagge (or dag) was a kind of gun, and that this is not an error for dagger; their account of the stage direction is likely to baffle readers unfamiliar with this rare word. Leaning the other way, they make heavy weather of the stage direction "They buffet" on the assumption that the latter was as unusual word a word in 1640 as it is now. (I would have thought that buffet meaning fight is much more widely known today than dag is.) Menzer and Hamlet go through some other rare locutions, including the strange directions of Cyril Tourneur's The Atheist's Tragedy, but they have many more questions than answers and the discussion is inconclusive.

    Regarding methodology, Menzer and Hamlet rightly observe that when counting such things as variations in the phrasing of stage directions the fact that Shakespeare has the largest printed canon tends to skew the results. They have no solution for this. Returning to their discovery that Heywood wrote more unusual stage directions than anyone else, they wonder if this was because he wrote for many different companies. It was not because he wrote about topics that others did not handle. Menzer and Hamlet decide that it was because his unusual Ages cycle drove him to be specially inventive.

    In Chapter Four, "Reading Shakespeare's Stage Directions" (pp. 23-114), Emma Smith considers the inconsistent references to Othello. Although other characters frequently call him by the generic name Moor more often than his personal name Othello, the stage directions and speech prefixes of the 1622 edition of Othello mostly, and those of the 1623 edition consistently, call him by his personal name. The 1622 quarto has three exceptions in its stage directions. When the dialogue draws attention to the impending consummation of his marriage (at the end of 1.3), a quarto stage direction reads "Exit Moore and Desdemona" (sig. D1v), drawing attention to his race as if to feed a fantasy of interracial sex, and again when Othello brings Desdemona back to bed after the brawl started by Cassio: "Exit Moore, Desdemona, and attendants" (sig. F2v). The third example is after Othello has killed Desdemona and found out that he was tricked: "The Moore runnes at Iago. Iago kils his wife" (sig. M4v). From these examples, Smith concludes that the quarto stage direction reserve "Moor" (a term Othello never uses about himself) for sexual or violent moments.

    Smith persuasively argues that theories of editing Othello need to take this discovery into account, especially in relation to the claim that the version underlying Q was subsequently revised to make the play more racially charged and that this revision made the copy underlying F. Of course, the stage directions influence only readers not playgoers. Smith sets out to read stage directions as part of the narrative, at least as far as readers apprehend it, rather than as theatrical information.

    We have tended to read stage directions for what they tell us about i) the textual provenance of printed books, and ii) what could be done on the stage. Editors usually treat stage directions as less authoritative than dialogue, at least in part because someone other than the author may have written them. Smith points out that the typographical distinction between dialogue and stage directions found in printed books is not entirely absent from poetry and prose narratives: Philip Sidney's The Shepheardes Calendar (1579) and Thomas Lodge's Rosalynde (1590) use typography to differentiate direct speech from narration.

    Sometimes the phrasing of a stage direction gives the reader an insight not available to a playgoer, as with "Enter . . . Hermione (like a Statue:)" in The Winter's Tale and "[Falstaff] fals down as if he were dead") in 1 Henry 4, but sometimes not: "Enter Arvirargus with Imogen dead . . ." in Cymbeline. I would say that the first of those is tricky since the word like might refer to a pretence that occurs not in the world of the play but the world of the theatre, calling for the actor to impersonate a statue, since having an actor stand still was a way of representing statues.

    Smith considers some cases where the phrasing of a stage direction seems literary, such as "The Prince killeth Percie" in Folio 1 Henry 4, which is the first time that Hotspur has been called by his family name in a stage direction. The P . . . P alliteration seems to anticipate an ensuing line of dialogue: "Nor can one England brook a double reign | Of Harry Percy and the Prince of Wales". Smith suggests that in such cases a stage direction takes on the characteristics of free indirect discourse. That is, the choice of words is not quite the narrator's voice--the voice of the author of the stage direction--nor the character's voice, but a bit of both.  Likewise, at the start of Antony and Cleopatra, Philo says that Antony is a "fan to cool a gipsy's lust" and then a stage direction immediately follows for Cleopatra to enter "with eunuchs fanning her". The stage direction represents Antony as a kind of eunuch, thus taking up a particular point of view about the narrative. The stage directions of Coriolanus use the same dismissive terms (rabble, mutinous, plebian) about the Roman citizens as Martius does. And likewise in Folio Titus Andronicus, Aaron is called a Moor in group-entry stage directions but given his personal name when entering alone.

    Smith examines the long stage direction for the dumb show in Folio Hamlet and finds that it seems to give Hamlet's point of view, using his preferred word fellow for the poisoner, his favoured adverb anon, and focussing on the unworthy woman's neck: Hamlet objects to Claudius paddling with Gertrude's neck and the Player King puts his head on the Player Queen's neck. The dumb show stage direction also echoes a number of words (loath, crown, ears, decline, seeming virtuous, and sleeping) from the Ghost's account of the murder in 1.4. None of this echoing occurs in the Q1 version of the play.

    E. M. Forster distinguished between just telling a story ("The king died and then the queen died") and conveying a plot ("The king died, and then the queen died of grief"), and Smith suggests that in avoiding causality stage directions are more story than plot. Smith finds a few Shakespearian stage directions that are rather more than just story, for example those that gratuitously mention the relationships between characters, or describe actions as having occurred.

    In Chapter Five, "Shakespeare's Literary Stage Directions" (pp. 115-37), Douglas Bruster explores much the same territory. W. W. Greg pointed out that the need to get a prospective play purchased by playing company would itself drive authors to create stage directions that grab the imagination as well as getting the theatrical business done. Bruster points out that some dialogue lines are effectively stage directions because they are in the imperative and give instructions to an actor. Bruster attempts to dismiss John Jowett's claim that some of the stage directions in The Tempest are Ralph Crane's by showing that the phrasing in them is also found in dialogue by Shakespeare. (Surely that is not the point: what is striking is their having phrasing not found in other stage directions by Shakespeare.) More significantly, Bruster finds three words/phrases in the supposedly Crane-written The Tempest stage directions that are also in Shakespearian stage directions: thunder and lightning, banquet, and vanish. From this he concludes that "The vocabulary of the stage directions in The Tempest is Shakespeare's, not Crane's" (p. 129). This claim overlooks Jowett's ample demonstration that it is not just the diction but the syntax and narrative structure of the long stage directions in The Tempest that seem unlike Shakespeare's stage directions.

    Bruster notices that ambo is used in stage directions for characters whose names end in -o in The Taming of the Shrew and Much Ado About Nothing, and suggests that it is chosen for the rhyme, as does "Enter Protheus solus" in The Two Gentlemen of Verona. He lists notable cases of words and sounds common to stage directions and nearby dialogue, as in "Ile ring it . . . He rings him by the eares" in The Taming of the Shrew, "There's that will sack a city. The PRINCE draws it out, and finds it to be a bottle of sack" in 1 Henry 4, "is arm'd for him . . . Enter COUNT  ROSSILLION, PAROLLES, and the whole army" in All's Well that Ends Well, and "Throws him his purse. PURSUIVANT I thank your honor. Exit PURSUIVANT" in Richard 3. Some of the links that Bruster claims are rather tenuous, such as "I heare the minstrels play. Musicke playes" with its m . . . p . . . m . . . p pattern, and even "Enter Curtis a Seruant. | Cur. In her chamber, making a sermon of continencie" with its c . . . s . . . s . . . c pattern. Bruster reckons that before Eleanor enters in a white sheet in 2 Henry 6 the preceding dialogue has an unusually high number of words with -ee- sounds in them: succeeds, fleet, me, she, streets, feeling, feet, sweet, wheels, streets, see, and miseries.

    In Chapter Six, Suzanne Gossett asks "When is a Missing Stage Direction Missing?" (pp. 141-62). A particularly frequently missing kind of exit is for a character who enters to a scene and later enters again: the moment for the intervening exit is not stated and presumably was left up to the actor. Gossett reckons that the assistance of backstage plots allowed dramatists and scribes to be somewhat casual about the provision of exits. Massed entries of the kind imposed by Ralph Crane in his transcripts give the modern editor the job of deciding when certain characters really enter. We just do not know how complete in its entrances and especially its exits a theatrical manuscript was expected to be, and hence what the early modern theatre professionals would consider a missing direction. The problem for the modern editor, then, is in knowing what to supply, and whether modern actors and readers need more precise instruction than early modern ones. Modern readers generally need more help with non-Shakespearian drama than with Shakespearian drama, simply because the latter is more familiar in its action.

    Gossett discusses some unusual cases, including Slitgut's discovering (in the sense of revealing) Cuckolds Haven by climbing a pole or stage post and affixing horns to the top of it or to the stage balcony in George Chapman, John Marston, and Ben Jonson's Eastward Ho! Just when someone dies is often unclear in printed texts. Q1 Hamlet has fuller stage directions than Q2 and F, but should we import stage directions from a text otherwise considered bad? Gossett explores some non-Shakespearian plays where, because the action is unfamiliar and not obvious at the first reading, editors really do need to add stage directions to clarify what happens. When exits are not provided we can assume that certain characters stay on stage across what we would otherwise consider a scene break, and the minor characters (especially servants) stay on stage to overhear the conversations of more important characters. Editors often disagree about asides. Gossett identifies a few other moments where a case can--but traditionally has not been--made for adding a stage direction, including Hamlet's spotting the hidden Polonius when talking to Ophelia in 3.1.

    In "Editing and Directing: Mise en scène, mise en page" (pp. 163-87), Terri Bourus notes that the narrowness of the 1623 Folio's columns affects verse in particular, requiring a line to be split in two. Marginal stage directions in manuscripts and print tend to get neatly tucked into the main text block in modern editions. In the 1634 quarto of The Two Noble Kinsmen we see genuinely marginal stage directions in the first half of the book, but then they stop and the stage directions are thereafter tucked into the central body of text; presumably it was too much trouble to keep setting them marginally.

    The Folio's narrow two-column boxed layout prohibited the use of marginal stage directions, so if they were included in the manuscript copy (as seems likely for several plays) then the compositors had to decide in each case how to tuck them in. Aside from a few exceptional word preferences that have been indentified, we cannot generally tell the authorship of a play's stage directions. Some stage directions must be Shakespeare's own but others are likely to have been written by various annotators, theatrical and/or scribal. Reprints among the early editions of Shakespeare, as when the Folio The Merchant of Venice reprinted the 1600 quarto, changed and/or added stage directions, and editors have been doing so ever since. The 1986 Oxford Complete Works practice of putting some invented stage directions but not others into broken square brackets is, according to Bourus, thoroughly confusing and it is difficult to use that edition's Textual Companion to work out which directions are merely editorial.

    Where there are two authoritative early editions, as with Titus Andronicus, and they have alternative stagings, as with the 1594 Titus Andronicus quarto treating as one scene what the Folio calls 1.1 and 2.1, a modern edition based on one will typically put in square brackets any material it draws from the other. This rather misleadingly implies that the material is just a guess or an editorial emendation when in fact it is a viable alternative staging that actually was performed on the early modern stage and is present in an early edition. Within certain constraints, editors also move stage directions around (such as from a marginal to a central position in the text block) without square bracketting their interventions. Modern editions cannot entirely solve the problem of how to present stage directions because they are working with competing imperatives regarding authenticity and helping the reader understand the story.

    Bourus gives some illustrations of how marginal paratextual performance notes, as used in the New Oxford Shakespeare of which she is a General Editor (as am I), go some way towards solving all these problems. Folio Antony and Cleopatra contains no act divisions but those invented by Nicholas Rowe in 1709 are routinely foisted on the play even in modern editions. Bourus illustrates her editorial practice regarding performance notes in the New Oxford Shakespeare Antony and Cleopatra, ending with her treatment of the Folio's ghost characters, meaning those given entrances but nothing to say. Instead of deleting them as editors routinely do, Bourus retains them and explains in performance notes the possibly implications of their silent presence.

   In Chapter Eight, "'By indirections find directions out': Unpicking Early Modern Stage Directions" (pp. 191-211), Martin White observes that some plays have vital parts of the action occur during an act interval, as when Charlois and Beaumelle get married during the interval in Philip Massinger and Nathan Field's The Fatal Dowry. In addition, some other plays might use the interval to set the stage for the following scene. White sees attending to the candles at indoor theatres as the entire reason for their being act intervals, and lighting the candles at the start of the play could be incorporated into the action, as it is in John Marson's What You Will. When there is extensive stage action taking place during an act interval this surely interfered with the necessary adjustments to the candles and hence the need for extra long intervals.

    The book-keeper Edward Knight added Long to two act intervals in Massinger's manuscript of Believe As You List, one of them seemingly written after the rest of the text since it looks crammed in, as if it reflected a need that emerged only in rehearsal. White details a scene in Catiline in which characters find themselves plunged into darkness and argues that while at the Globe this would be an imagined darkness at the Blackfriars it really was practicable to extinguish the candles and/or raise the chandeliers to darken the stage.

    Massinger's play The Guardian has a tricky staging moment in which a husband seems to hear but not see his wife and misconstrues her self-recriminations over her intended infidelity as pious prayers for his safety. The wife has to enter in a chair with a banquet and behind a curtain but no one is carrying the chair, and it is not clear just whose perspective--the audience's? her husband's?--is meant when her position is said to be behind the curtain. There is also the problem of how the husband hears but misconstrues his wife's soliloquy. White reckons that the wife and banquet are revealed in the discovery space and that the husband is under the stage, and enters through the trap door; the character previously referred to the existence of a secret passage into his house.

    Chapter Nine, "'Strikes open a curtain where appears a body': Discovering Death in Stage Directions" (pp. 213-37), is co-authored by Sarah Dustagheer with Philip Bird. When Hieronimo discovers Horatio dead in the discovery space at the end of Kyd's The Spanish Tragedy he is restaging the scene of Horatio's murder in the arbour. We do not know if discoveries were performed behind stage doors or in structures built out from the back wall or in free-standing booths, or in all these ways for different plays. Of course, dead bodies were not the only things discovered. 

    If the discovery space is the arbour where Horatio and later Isabella die, and then is the place used for the hanging of Pedringano, and then is used for the final displaying of Horatio's body, it acquires strong associations with death. Dustagheer sketches some other plays' association of the discovery space with death. Revenge tragedies need a body to concretize the injury done to the revenger and they often reuse a stage space for displaying corpses. Revenge tragedies dramatized the remembrance of, and intercession on behalf of, the dead that the Reformation had removed from religious practice.

    The plays' tableaux of the mourned-for dead mirrored the funeral monuments that became increasingly elaborate because of the Reformation's stifling of other ways to remember the dead. Monuments and pictures were often covered by curtains that were pulled aside for viewing. The bodies in the revenge-tragedy tableaux have not yet been buried, of course--that will come after the revenge inspired by looking on them--so they are in a kind of liminal state that Dustagheer likens to Purgatory.

    The King's Men flipped the trope of discovering the dead with their trick of revealing the surprisingly alive Hermione in The Winter's Tale and Ferdinand in The Tempest. In The Duchess of Malfi the dead bodies apparently discovered in 4.1 are really dummies, which readers will know because the stage direction calls them "artificial" but playgoers may well not realize until Ferdinand tells Bosola that they are only wax figures. In late medieval masses and in religious art the pulling back of a curtain to reveal Christ as the Eucharist or some other sacred figure was a common device. The revealing of a discovery space somewhat breaks the boundary between the private backstage world and the public onstage world: we are peeking into the private side. The discovery space is likely to be less well lit than the rest of the stage and, in conjunction with the fact that no everyone can fully see into it, this dimness adds to the unsettling atmosphere of discoveries, especially when corpses are what is discovered.

    Chapter Ten is "'Enter Macduffe, with Macbeths Head': Shakespeare's Macbeth and the Staging of Trauma" (pp. 238-61) by Andrew Hiscock. He has the verbal tic of starting with "If . . ." sentences that are not in fact conditional, as in "If my sequence of enquiries remains engaged throughout upon a consideration of the staging of trauma in the Folio Macbeth, in general terms, Dessen is surely right to . . ." (p. 243), "However, if the term 'stage-direction' was being used as early as 1726 by Lewis Theobald in . . ." (p. 244), "If Leslie Thomson argued, for example, that . . ." (p. 245), and "If, in Hamlet, we learn that the ghostly presence has descended to a purgatorial netherworld . . ." (p. 256). The editors could have saved Hiscock from the false suggestion that these are all hypotheticals, and also from the misspelling of "Edmund" Malone (p. 245).

    Hiscock alludes to studies that should be fully referenced: ". . . with the resources of stylometric testing amongst others, the balance of critical opinion often returns to viewing the dramatist as a frequent point of origin for such directions" (p. 246). I cannot think of a stylometric study that has established the authorship of any stage directions, beyond Jowett's qualitative one referred to above in relation to Douglas Bruster's essay, and Hiscock's verb "testing" implies something quantitative. It would be pleasantly surprising if anyone had a method that could reliably establish the authorship of such short bodies of writing.

    In "'(From the Dutchesse Grave)': Echoic Liminalities in The Duchess of Malfi" (pp. 263-85), Sarah Lewis argues that the massed stage directions of the 1623 quarto of the play are like the pre-echoes that figure in the action. She considers the appearance of Echo in the Homeric Hymn to Pan and Ovid's Metamorphoses, and also the echo scene in Jonson's Cynthia's Revels. Lewis recounts Ovid's version of the punishment of Echo by Juno (for helping conceal Jove's infidelities), in which she can use her tongue only to repeat someone else's last words, and then her unhappy pursuit of Narcissus and, rebuffed by him, her unhappy retreat into caves. In all accounts, Echo is a sexualized, liminal figure. Lewis traces Echo in the early modern plays that feature her, essentially as either a virgin or whore.

    John Russell Brown argued that the stage direction "Eccho (from the Dutchesse Grave)" requires that the Duchess's voice comes from the onstage grave and that at the end of the scene she enters from, or in, the open tomb. Others differ, and Lewis sketches the competing views. In the Pan version of the myth, Echo is forever on the edge, there and not there, and hence tantalizing, and Lewis suggests that this is true of the Duchess too. In the Echo scene, her agency is confined to choosing what to echo and to bending the sense in echoing it. This liminality of Echo resonates with the more general uncertainty of "the Duchess's subjective identity" (p. 284).

    The last chapter in the book is "Understanding Dumb Shows and Interpreting The White Devil" (pp. 287-310) by Gillian Woods. Hamlet says that only the groundlings enjoy inexplicable dumb shows, but then he commissions a play with one in it. Dumb shows convey plot complexities without words, but in most plays they have to be explained afterwards, so what are they for? Woods reckon they are meant to be contradictory: to aid understanding and also present problems of interpretation. Dumb shows confound the audience's sense of time and space. They have more than the usual number of departures from the one-door-on-one-door-off rule, using forms such as "Enter x at one door and y at another" and "Exit severally". They also seem to involve exaggerated expressions and movements, making up for the absence of dialogue.

    Dumb shows change the rules of representation in a drama, and the characters in them are not quite as real as those in the outer play. Woods surveys some contemporary allusions to dumb shows and metaphorical uses of the term, and no clear picture emerges other than that they "mystify meaning" (p. 302). The show of the eight kings in Macbeth, for example, needs interpretation: Macbeth might think he is being shown a future he can avoid but in fact his being shown this procession makes him take the steps that cause it to come true. Being a respite from the spoken word, such hard-to-interpret dumb shows remind us how much we rely on language.

    The White Devil has notoriously ambiguous characters, each being neither entirely good nor bad though mainly one or the other. It is unclear where an audience's sympathies are meant to lie. The first dumb show is an illusion created by a conjuror and it shows Isabella dying from kissing the poisoned picture of her husband. The ambivalence is that Isabella is a victim of murder but also idolatrous in her veneration of the picture. The second dumb shows the murder of Camillo under the cover of an accident in a vaulting competition. Woods traces some likenesses between the dumb show and the wider action, especially regarding the trial of Vittorio and the subsequent appearances of ghosts.

    Our second collection of essays in this round-up is Shakespeare's Language in Digital Media: Old Words, New Tools edited by Janelle Jenstad, Mark Kaethler, and Jennifer Roberts-Smith. The editors' Introduction (pp. 1-8) gives a history of certain digital projects and is rather impenetrable to anyone who does not already know about them, not least because the writers use abbreviations such as EMEDD, LEME, QME, and DRE before pointing the reader to the section on "Titles and abbreviations", which itself points the reader to the book's index where the expanded forms are given.

    Chapter One is "Beyond the OED Loop: Digital Resources and the Arden 3 Cymbeline" (pp. 13-26) by Valerie Wayne. Her edition of Cymbeline referred to here was reviewed in NYWES for 2017. Wayne describes how the OED overstates Shakespeare tendency to invent new words and how the fossilizing of an erroneous definition can be self-fulfilling: once a word is thought to have a certain meaning, writers use it in that sense. The Text Creation Partnership (TCP) dataset offers a way to break such a loop by consulting usages otherwise unrecorded. Wayne found that the word solicity in Folio Cymbeline, long thought an error and emended to solicits, was in fact a word at the time. In context, the extra syllable in solicity makes the line "To orderly solicity, and be friended" less metrically regular, although it can be defended as a feminine iambic hexameter with a syllable omitted at the caesura; Wayne makes no mention of the metrical consideration here or in her edition.

    Next Wayne discusses imperseuerant and wing-led, both of which were also covered in the our NYWES review. Then she turns to some other words/phrases from Cymbeline and the light that the Lexicons of Early Modern English (LEME) and EEBO-TCP throw on them. She discusses the debate over correction of the Folio's name Imogen to Innogen and tabulates the forms found across EEBO-TCP, in which the latter is predominant. Wayne ends with the mollis aer crux, discussed in our NYWES review of her edition.

    In Chapter Two, "Shakespeare's Hard Words, and our Hard Senses" (pp. 27-46), Ian Lancashire and Elisa Tersigni present evidence about those words of Shakespeare's that were known to need explanation at the time and that we may therefore assume this audience probably had trouble understanding. Lancashire and Tersigni took the two orations to their troops by Richard III and Richmond in Richard III and looked up their words in early modern hard-word dictionaries. Only 12 of them seem to have been difficult for first audiences: attempt, beauteous, bobbed, desperate, exploit, homicide, inferred, ravish, record, tyrant, vomit, and welkin. Of these, only four tend to get glossed by modern editors. Richmond always glosses his own hard words by adding a synonym directly after each one, but Richard never does and he mixes them with low-tone slang words such as scum and lackey.

    Lancashire and Tersigni notice that Richmond uses "two imposing Latinate imports--'tyrant' and 'homicide'--but amply glosses both words for his troops" (p. 28). They do not mention that both words were used twice before in the play: Lady Anne calls Richard a homicide in the second scene, as does Oxford in 5.2, and in the attack on Margaret in 1.3 the killing of Rutland is said to have made tyrants weep, and Margaret herself calls Richard a tyrant in 4.4. Lancashire and Tersigni's claim that these two words had a special effect as hard words in Richmond's mouth near the end of the play overlooks the fact that the play had already made them somewhat familiar.

    Next comes a description of LEME and its automated Hard Word Annotator that takes a block of text and finds all the words in it that appear in early modern hard-word dictionaries. Lancashire and Tersigni illustrate its use to count which characters in The Tempest use the most hard words, and in descending order of percentage (after adjustment for the different lengths of their parts) it is Iris, Ceres, Adrian, Ariel, then Prospero as the most frequent users. The low users of hard words are Alonso, Trinculo, and Miranda. Lancashire and Tersigni think this shows that "The percentage of hard words used by characters reflects their personality, occupation, race, gender, and social status" (p. 32). I would say that apart from showing that supernatural characters tend to use hard words, the pattern is unclear.

    To help solve the textual crux of who speaks the 94-word speech beginning "Abhorred Slaue, | Which any print of goodnesse wilt not take", whether it is Miranda (as in the Folio) or Prospero (as in many editorial emendations), Lancashire and Tersigni look at the speech's hard words. They find that the proportion of hard words in the speech is 5.3%, which is consistent with Prospero's habit (5.1%) not Miranda's (2.1%). The speech claims that the speaker taught Caliban to talk, and since Caliban uses more hard words than Miranda it would be odd, the authors remark, for the teacher to be Miranda. (Surely, once a person learns to speak they are not limited to the vocabulary of the person who first taught them and can pick up words from anyone else, in this case Prospero.)

    Lancashire and Tersigni pursue the claim that Prospero's being out-of-touch with the everyday world is reflected in his frequent use of hard words. Early modern hard word dictionaries tended not to attempt an explanation of a word but to offer simpler, better-known, synonyms. Lancashire and Tersigni look at editors' glossing of distaine in Richard's oration's "You hauing lands and blest with beauteous wifes, | They [Richmond's army] would restraine the one, distaine the other". They object to glosses suggesting that "distaine" means dishonour since "Richard does not believe that a future defeat will bring 'dishonour' to the 'beauteous wifes'; beauty, a matter of appearance, is stainable rather than capable of being dishonoured. Honour is far from Richard’s mind" (p. 35). This seems to me to miss the point: it is not the beauty of the English soldiers' wives that will be dishonoured, but the wives themselves (by rape) because they are beautiful. Lancashire and Tersigni end their essay rather abruptly with the idea that if we read Shakespeare using the assumption that we just need to find an easy synonym for each hard word--which seems to be the principle on which hard-word dictionaries were made--then "The experience might resemble attending a performance and having to understand stage dialogue on the fly, textless, in real time" (p. 36).

    In Chapter Three, "Terms of Art in Law and Herbals" (pp. 47-65), Daniel Aureliano Newman uses plant metaphors and images in Hamlet to argue that books giving terms of art (that is, the professional jargon of a trade or field of study) offer a "powerful tool for exploring his [Shakespeare's] thematic interests as a dramatist, as well as his linguistic methods as a poet" (p. 48). Newman focusses on early sixteenth-century books of terms of art for law and botany and then turns to the light that John Gerard's Herbal (1597) throws on Shakespeare's writing. Curiously, Newman thinks that ". . .  Arthur, who died in 1203, is murdered in Act 4, Scene 3" of King John (p. 58). In fact he either accidentally slips from a high wall or underestimates the danger and risks jumping, but he is alone and it is not murder.

    Newman explains some contemporary points of law that pertain to the bastardy dispute at the start of King John. Next he explains the "nature's bastards" exchange in The Winter's Tale 4.4 by reference to contemporary botanical guides. Newman tries to figure out what flower exactly Perdita means by her term "Gilly' vors" and decides that they are "stock-gillyflowers", which become less fecund when hybridized to produce piedness. Since Perdita herself is accused of hybridity (indeed bastardy) and since by mixing with Florizel she is risking another kind of hybridity--or so she thinks, not knowing herself to be royal--she has reason to dislike hybrid plants.

    In Chapter Four, "'Strangers enfranchised': Shakespeare's Hamlet and the Mother Tongue" (pp. 65-80), Elizabeth Bernath observes that many of the words being glossed in the earliest hard-word dictionaries, those published in the sixteenth century, are used in the definitions for the later hard-word dictionaries, in the seventeenth century. That is, ". . . the mother tongue might expand to accommodate hard words" (p. 67). (Throughout the books, the term for words moving from the hard-word category to the mother-tongue category is enfranchised.) Bernath thinks that Hamlet was first performed in 1604 (p. 69) without saying why. When unfamiliar loan words are used in Hamlet they tend to be joined with familiar native words that act as a gloss on them, as with "tender and delicate prince" in which delicate, a loan word from French, is clarified by being yoked with the familiar tender. Likewise in "rogue and peasant slave", where the first two unfamiliar nouns are clarified by being yoked with the familiar slave. Bernath gives more examples of Shakespeare's tendency to gloss his own hard words. Where he does not do this, the character is intentionally using refined and opaque language that not all the audience will understand.

    In Chapter Five, "Text, Performance, and Multidisciplinarity: On a Digital Edition of King Leir" (pp. 84-104), Andrew Griffin argues that the Queen's Men Editions (QME) do something unique by asking "What does an edition represent?" and "by insisting on an interdisciplinary editorial practice and by resisting some fundamental assumptions of modern textual scholarship" (p. 85). Being online only, these editions can admit longer than usual performance notes and more images, and also incorporate moving images. Griffin gives a sketch of twentieth-century debates about editing and observes that a QME editor seldom has to decide between early editions of competing authority because there usually is only one early edition, so they simply provide a diplomatic transcript of it as their original spelling version.

    Griffin claims that ". . . popular web browsers are unable to display, for instance, ligatured type . . ." (p. 92), which is untrue at the time of this review (October  2021) and unless the essay was written more than ten years ago was also untrue at its time of writing too. Griffin argues that modernizing the opening speech of King Leir tends to destroy the "run-on sentences and ambiguous punctuation of the quarto [which] might suggest something about Leir's character" (p. 93). But there is nothing to stop a modernizing editor conveying in modern punctuation Leir's mental disturbance: it is not the case that only early modern punctuation can do that.

    Griffin dates the emergence of performance studies in English departments to the 1980s at the earliest--I would push that date back to the 1950s in relation to Shakespeare--and finds that recent work on Shakespeare as a literary author, for example by Lukas Erne, assumes a long tradition of performance studies when in fact, Griffin things, there is only a short one. When editors have brought matters of performance into their editions, they have tended to assume a single set of early modern staging practices as if these applied everywhere, and thus they overlooked the fact that performance is "radically contextual and necessarily ephemeral" (p. 97).

    Griffin finds that the same idealist-versus-materialist binarism found in textual studies--deployed in discussions of whether the play exists as a kind of Platonic form distinct from its material embodiment in particular documents--is being reproduced in performance editions. That is, the study of early modern theatrical conditions gives us a singular and idealized originating performance envisaged by Shakespeare, of which modern performances are only debased derivatives. Even disregarding the different productions, no one night's performance of a single production is identical to another's, so we necessarily are generalizing about productions even if we have visual evidence about them. Griffin describes how the QME mitigates these problems, mainly by being aware of them and using copious visual materials to keep the reader aware of all the options and the choices made in particular productions.

    In Chapter Six, "A Digital Parallel-Text Approach to Performance Historiography" (pp. 105-23), Toby Malone is concerned with printed editions and theatrical manuscripts that represent a particular play's performance at a designated theatre. Malone describes a pilot project to create a parallel-text edition of Q and F Richard III combined with various performance texts from the eighteenth to the twentieth centuries. He presents the multiple witnesses in his edition in the form of a spreadsheet, with Folio Richard III as his base text. To judge from his picture of an eight-column, eight-text display (Figure 6.6), readers of his digital edition will need an especially wide monitor to take it all in; Malone acknowledges that this point has been raised before.

    In Chapter Seven, "Storing and Accessing Knowledge: Digital Tools for the Study of Early Modern Drama" (pp. 131-43), Laura Estill and Andie Silva survey of the online datasets called the English Short Title Catalogue (ESTC), the Database of Early English Playbooks (DEEP), EEBO, British Literary Manuscripts Online (BLMO), Early Modern London Theatres (EMLoT), Patrons and Performances, and the World Shakespeare Bibliography. Estill and Silva look in particular at the clarity with which these projects make plain their "data-curation practices" (p. 131) because these practices embody "critical and editorial choices" (p. 131) that utterly shape our access to the materials.

    Estill and Silva complain that ESTC's origins in a printed book are still evident in its structure and inconsistencies. They largely approve of DEEP, but complain that the user cannot see facsimiles of title pages, that its focus on drama distorts the picture for historians of the book, and that it privileges print over performance in giving full re-publication information about books but only first performance dates for plays. They praise DEEP's transparency about editorial decisions. Estill and Silva complain that the metadata of EEBO and BLMO are far from perfect, and that the images in these datasets were digitized from microfilm, which has especially high contrast, rather than from fresh photographs.

    All these complaints are valid, although I would expect that the creators of these projects would in each case explain themselves by invoking the specific constraints under which they were created. For example, re-photographing all the books in EEBO alone would take a multi-million pound project that would be unlikely to get funded because we already have the microfilm images. The EMLoT and Performances and Patons datasets, both from the Records of Early English Drama (REED) project, get generally positive reviews and although Estill and Silva dislike the fact that the World Shakespeare Bibliography focusses only on Shakespeare, they appreciate that it puts him in a global context.

    Chapter Eight is "Past Texts, Present Tools, and Future Critics: Toward Rhetorical Schematics" (pp. 144-56) by Michael Ullyot and Adam James Bradley. Literary criticism seems necessarily exemplary, taking individual instances as examples of wider patterns. But does that have to be the case--do we have to ignore most of the evidence--when we have all the texts digitized, or can we aim to be more definitive? Ullyot and Bradley's case studies use rhetorical devices in writing as the objects of interest, specifically anadiplosis and gradatio because they hope that computers can detect these.

    Anadiplosis is a pattern taking the form A . . . B, B . . . C as in "Featured like him, like him with friends possessed" (Shakespeare Sonnet 29) and "Pleasure might cause her reade, reading might make her know, | Knowledge might pitie win, and pitie grace obtaine" (Philip Sidney Astrophil and Stella 1.3-4), and putting together a chain of instances of this device creates gradatio, as in "For your brother and my sister no sooner met but they looked; no sooner looked but they loved; no sooner loved but they sighed; no sooner sighed but they asked one another the reason; no sooner knew the reason but they sought the remedy; and in these degrees have they made a pair of stairs to marriage" (As You Like It 5.2.31-37).

    To detect anadiplosis computationally we must first lemmatize the tokens in a text and then use the computer technology called regular expressions to look for lemmas that are repeated within an arbitrary window. Ullyot and Bradley settled on looking within 4 tokens of a clause break signalled by punctuation. This might well miss some instances that do not quite fit the rules, and would doubtless include false positives that have to be sifted out by hand. Ulyott and Bradley describe the limits of their method and the patterns that fooled it and list some attractive examples of anadiplosis and gradatio that they found. The aberrant cases are useful in that they help us nail down just what we mean by these rhetorical terms.

    Ullyott and Bradley imagine how their ideal digital tool for analysing rhetorical structures might work, including the possibility of embedding it in a digital edition. They sketch what an Application Programming Interface can do to make digital systems interoperable. I suspect that differences between different systems' implementations of Extensible Markup Language XML encoding, even among implementations conforming to the Text Encoding Initiative (TEI) form of XML, will make the interoperability they anticipate rather harder to achieve in practice than it sounds. If, as some critics have argued, Shakespeare actually thought in rhetorical systems, such investigations might give insight into his cognitive processes as well as the cognitive processes that his characters betray in their language.

    Chapter Nine is "Internet Shakespeare Editions and the Infinite (Editorial) Others: Supporting Critical Tagsets for Linked Editions" (pp. 156-71) by Diane K. Jakacki. The Internet Shakespeare Editions (ISE) edition of Henry 8 is being made by Jakacki using TEI-XML as well as the older format called Standard Generalized Markup Language (SGML) format that the TEI project began with. In this essay, Jakacki explores the features that using TEI might make possible.

    Jakacki extols the virtues of XML over SGML and overstates the case in claiming that ". . . tagging systems such as SGML . . . emphasize representation alone" (p. 159), and that XML allows for mark up of semantic features. Perhaps the ISE tag set defined in SGML emphasized representation alone, but that is not a feature of SGML itself, which is merely a language for defining tag sets and is perfectly capable of defining a tag set that captures semantic features. This is not her only technical slip: Jakacki also claims that when editors submit editions marked up with the ISE tag set ". . . programmers must convert these files to XML in order for them to be machine-readable" (p. 159). The files must already be machine-readable--as opposed to being, say, handwritten on pieces of papers--if the programmers can write programs to convert them from ISE-SGML format to XML.

    Jakacki illustrates the transition from SGML to TEI-XML at ISE with some sample code from her edition of Shakespeare's Henry 8 and shows that the TEI code allows for some more extensive tagging that enables new features, such as making the pronoun her be the content of a <persName> (that is, Person Name) element with a @who attribute that points to a uniform name used across the edition for the person that this her refers to. The description is somewhat clouded by a mistake: the her in question, meaning Katherine of Aragon, is in line 1223 not line 1224 as her explanation has it, and "ANNE  No, not for all the riches under heaven" is line 1244 not 1224 as the explanation has it (p. 161). The explanation is also muddied by the fact that Jakacki's modern-spelling edition uses the spelling Katharine 50 times and the spelling Katherine nine times for the same character, so that the stage direction "Enter Katherine . . ." is followed by the speech "KATHARINE  O Griffith . . .". No reason is given for leaving the Queen's name unregularized.

    Jakacki illustrates how the marking up of pronouns in dialogue allows for a better understanding of how much each character refers to others. Jakacki says that to capture this information she ran "a simple XSLT script" (p. 162), meaning an Extensible Stylesheet Language Transformation, on the XML-encoded text of the play, and it would have been helpful if she had shown the reader this script to prove that it really is simple. (The present author finds XSLT rather a difficult language and recommends the use of the somewhat simpler XPATH and XQuery languages for the same kind of task.) The wider point that Jakacki makes only implicitly is that because her text is in TEI standard encoding, so that for example it uses the standard <persName> element rather than some custom-made element, it is much easier to apply off-the-shelf tools to it in order to perform complex interrogations than it would be if the tag set and structure were custom-made for the ISE project.

    In principle an editor could do to names of places what Jakacki illustrates for the names of people, and thereby enable interoperation with another TEI-conformant project, the Map of Early Modern London. Or at least this will be possible once the two projects figure out some common tagging practices: that both are TEI-compliant does not itself enforce the necessary agreements. Jakacki gives sample code for what she hopes they eventually will be able to do. Jakacki asks the important question of whether an editor would really want to link her edition to every possible TEI-compliant project that it could in principle be linked to, and to mark up every possible referent that any one of those projects might be interested in. One problem is that doing so commits the editor and indeed the whole project to a lot more work, especially in technical maintenance, than would be involved in an ordinary scholarly edition. Another problem is that such inter-project work generates materials that do not obviously belong within either project, and this raises the question of who is going to look after these materials when the funding ends.

    The last book-form collection of essays to be considered this year is A Handbook of Editing Early Modern Texts edited by Claire Loffman and Harriet Phillips. The essays here are unusually short, typically no more than five pages, and most are irrelevant to this review. The first that is relevant is Valerie Rumbold's "Textual Apparatus and Reader Engagement" (pp. 73-77). She favours the abandonment of cryptic abbreviations in the textual apparatus and asks  ". . . could we at least consider the possibility of minimising symbols and abbreviations?" and "Might it sometimes be better to present a longer, more accessible account than a miracle of compaction from which the untrained eye recoils?" (p. 75). I quite agree. Rumbold also suggests the inclusion of "a concise discursive headnote to the textual account of each work" (p. 75) to encourage those readers who may be put off by the incomprehensibility of a traditional textual apparatus to engage in textual matters. Rumbold suggests that for many writers what we really need are monographs that study the writer's entire career from the angle of book history and textual transmission.

    In "The Problems with Old-Spelling Editions" (pp. 97-102), Gavin Alexander shows some cases where excessive fidelity to the readings of the copy text is harmful. He objects to the Greg-Bowers principle of using the earliest text as the copy for accidentals and importing into it the substantive readings of a later text if the latter better represents authorial intentions. The problem is that when such a policy is enacted in original spelling the reader cannot tell that the text is eclectic: the original spelling gives the impression that a document from the period is being represented. I would have thought that this false impression could be dispelled by indicating all the emended words, for example by putting them in square brackets.

    The next essay responds with "In Defence of Old-Spelling Editions" (pp. 102-5) by Roger Kuin. He couches the debate about old versus modern spelling as an either/or, but surely an edition can offer both by presenting the work twice. Kuin thinks that spelling variants introduced into printed books by compositors solely for the purpose of justifying a line of type should be preserved in a modern old-spelling edition since "To reduce such variety to a flat modern uniformity is to mask its historicity" (p. 104). He also thinks that in early printed editions of Edmund Spenser's work the spellings can mark half puns, such as droome (for modern drum) carrying an overtone of doom. But the best reason for original spelling, Kuin concludes, is that it is charming and pleasurable to read.

    Sensibly countering this whimsy is "Modernisation Versus Old-Spelling for Early Modern Printed Prose" (pp. 105-9) in which Joseph L Black argues that the prime reason to favour modern spelling is that the appearance of authenticity in original spellings is misleading: early editions' spellings are largely scribal or compositorial. Black's conclusion is that for different authors it may be best to vary editorial practice regarding the modernization of spelling and punctuation.

    Andrew Zurcher starts his essay "Parting 'with much wee know': Digital Editing and the Early Modern Text" (pp. 171-75) hoping that the digital scholarly edition will take us away from the singular authoritative text that the printed scholarly edition has mandated, and away from lone-scholar practices to greater collaboration, especially across disciplines. Zurcher praises Jerome McGann's Rossetti Archive which "blazed the trail for experimentation in online literary archival hubs" (p. 173). For this review I explored the Rossetti Archive and the parts I happened to browse had a number of resources that were unusable because they rely on proprietary Adobe Flash file formats that modern web-browsers prohibit on account of their cyber-security risks.

    Zurcher extols the virtues of the McGannian digital "editorial machine" that can give us the text served up in multiple different forms with "commentary, annotation, cross-reference, and external linking" (p. 173). Again, my experience differs: in the Rossetti Archive there was so much cross-linking that I quickly got lost and hit dead-ends. Having found from the section "exhibits and objects" the section "Books" I selected a book, Ballads and Sonnets (1881), and tried to see what was in it. All I could find were pictures and transcriptions of the bindings and pictures of the interior pages: every time I tried to see the transcript of a page I got "This plug-in isn't supported" errors because of the reliance on Adobe Flash.

    Zurcher thinks that when texts refer to their own physical conditions as material objects (paper, ink) that necessarily have errors in them, they "ask uncomfortable questions about the virtues of the digital edition" (p. 174), but he does not identify the questions or name the source of the discomfort. He considers some textual tricks in printed books that present difficulties for the editor precisely because they are meant to frustrate the reader, to withhold understanding. But these are surely no more difficult for the digital edition than the print edition to deal with.

    The next relevant chapter is Angus Vine's "Scriptorium: When to Build a Digital Archive Rather than a Digital Edition" (pp. 188-92). Discussing what we set out to do as editors, Vine asserts that ". . . the distinction that needs to be made here is between choosing to build a digital archive or to create a digital edition: where the former focuses on images and the material, the latter privileges text and the verbal" (p. 188). This distinction is mistaken, since the difference lies in what we do to correct the materials. The Text Creation Partnership that keyboards the books whose images are in EEBO in order to create a textual/verbal dataset is as much an archive as EEBO itself; neither is an edition, because when errors are found they are not corrected. Conversely, Charlton Hinman's Norton facsimile of the 1623 Shakespeare First Folio is an edition of images because he chose not one exemplar to reproduce but rather he selected different pages from different exemplars according to their state of stop-press correction and how clearly they appeared when photographed.

    Vine's project called Scriptorium set out to make an archive of manuscript images from various sources. He used XML and TEI but stopped short of providing transcriptions of the manuscripts and he explains this (again, wrongly, I think) with the distinction that his project was "explicitly a digital archive, not a digital edition" (p. 192). The homepage for the project is cited here by a URL that does not seem to work at the time of writing of this review (October 2021). The server structure at Cambridge University seems to have been reorganized since the essay was written and the Scriptorium website can be found at its new home by web-searching.

    The essay most relevant to this review is "On Error" (pp. 209-17) by Cathy Shrank. She considers the lines "Therefore my Mistersse eyes are Rauen blacke, | Her eyes so suted, and they mourners seeme" (Sonnet 127) in the 1609 edition of Shakespeare's Sonnets (Q) and she rejects the emendation of either occurrence of eyes to brows, since without the emendation the lines make sense. Shrank thinks that to emend would be "swapping a certainty for an uncertainty", done only "because we want Shakespeare to be as good as we think he should be" (p. 210). I would say that editors who make this emendation do it because they are pretty sure that Q is erroneous in its reading and that Shakespeare would want it corrected if he had known about it.

    Shrank considers and adjudicates on more marginal cases. In Sonnet 25, Q reads "The painefull warrier famosed for worth, | After a thousand victories once foild, | Is from the booke of honour rased quite, | And all the rest forgot for which he toild" and the problem is the lack of a rhyme in worth/quite. Shrank rejects the common emendations of worth to might or fight or else of quite to forth to because "it is hard to see how 'worth' could be misread as 'might' or 'fight', or 'quite' as 'forth'" (p. 212). She must mean the reverse--how might or fight could be misread as worth or forth as quite--since worth and quite are the problematic readings in Q that we want to explain. Shrank reckons that perhaps the copy for Q was a manuscript in which this sonnet was in the process of revision that was incomplete, and that we should represent its intermediate state by not emending here. I would say that compositors could make substitutions as large as setting might as worth because they are holding words in their heads and synonyms sometimes obtrude.

    What emerges from her examples is that for Shrank error has to be quite certainly present before she will emend. Shrank does not think that some of the question marks in Q should be changed into exclamation marks "Since there were exclamation marks in the compositor's boxes in George Eld's print-shop, where Q was set . . ." (p. 214). But the fact that there were some available does not mean that enough were available, and using what we now call a question mark for what we consider exclamations was not unusual. Shrank has more examples where she disagrees with repunctuation in modern editions of Q. Sonnet 29 in Q reads "Yet in these thoughts my selfe almost despising, | Haplye I thinke on thee, and then my state, | (Like to the Larke at breake of day arising) | From sullen earth sings himns at Heauens gate". Some editors move the closing bracket from after "arising" to after "earth" and Shrank reckons that this "rewrites the meaning of those lines" (p. 215). I dispute that it does, since in likening himself to the lark the narrator is saying that he and the lark are doing the same thing in rising from sullen earth to sing at heaven's gate, no matter where the bracket is.

    Shrank turns to lineation and argues that the Second Citizen in the opening scene of Coriolanus might be "starting to reach for . . . blank verse" in his lines "Well, | Ile hear it Sir: yet you must not thinke | To fobbe off our disgrace with a tale: | But and't please you deliuer". Shrank argues that "Rendering the citizen's speech in prose could be seen as a means of keeping the citizen, and those of his class, in their place, defining them as 'plebeian' . . ." (p. 216), which rather implies that an editor could choose to make verse out of this speech. I would say that these lines could be made into acceptable Shakespearian verse only by emendation of the words, and Shrank requires there to be clear verbal error before she will allow that. Shrank thinks that editors are more keen to correct the faulty Folio lineation when the speakers are socially elevated, but the example she gives of an exchange between Sicinius, Brutus, and Menenius in 2.2 is clearly verse that has been set as prose, while that is not true of the Second Citizen's speech in 1.1.

    We turn next to journal articles. In 2018, Shakespeare Quarterly published two that are relevant to this review. The first is by Ed Pechter ('Against Attribution', SQ 69[2018] 228-55) and I must disclose that he was one of the journal's peer-review readers for this article and recommended publication despite disagreeing with its argument and conclusions. Its merit is in being, in this reviewer's opinion, much the best articulation yet of the important arguments that it seeks to make. Pechter observes that once the publication of complete-works editions of Shakespeare began to appear in the early eighteenth century, there had to be a policing of the border of what was understood to have been written by Shakespeare. But not until the end of the nineteenth century, the era of the New Shakspere Society, and once again at the end of the twentieth century, with the 1986-87 Oxford Complete Works, did it seem to matter much that Shakespearian specialists worked on this problem.

    Pechter makes many approving comments about the current field of authorship attribution, but wants to interrogate just why this topic and its computational methods matter so much to Shakespearians just now. Two reasons given by the authorship scholars themselves are i) that their work throws light on the important practice of collaborative writing, and ii) that they are doing science and hence bringing facts to a domain hitherto dominated by opinions. Pechter attempts to show that neither answer will do and that the authorship-attribution practitioners distort the practice of literary criticism in their account of it. Pechter reckons that ". . . the importance of Shakespearean collaboration is regularly exaggerated in current work" (p. 232). It can hardly be true, he points out, that the entire history of Shakespearean literary criticism is overturned by discovery that our author collaborated with others, and in any case the figures for how much he collaborated can be misleading. Gary Taylor claims that 38% of the plays are collaborative, but if you count the phenomenon by the number of words involved it is under 10%, and a lot of what Taylor counts are Shakespeare's trivial supplements to others' plays, as with Sir Thomas More and The Spanish Tragedy.

    Shakespeare's collaborative work is clustered at the ends of his career, before 1596 and after 1605; in the middle of his career (his most productive years) collaboration is almost entirely absent. Pechter is quite right: in that period there were just the minor contributions to Sir Thomas More and The Spanish Tragedy and perhaps his now-lost contributions to Jonson's Sejanus. This biographical fact does rather counter the claim that collaborative writing was central to Shakespeare's art. If we overstate the importance of collaborative authorship we make the Shakespeare of the middle period seem by contrast to be like Jonson in displaying protectiveness of his individual literary achievement.

    So much for the distorting effects of overstatement of the importance of Shakespeare's coauthorship. Pechter turns next to the claim that authorship attribution scholarship is a branch of science. He starts with Peter Kirwan's account of the collection of essays put together in 1923 by A. W. Pollard claiming that Hand D of Sir Thomas More is Shakespeare. Kirwan pointed out this scholarship was impressionistic and failed to perform the necessary negative check to ensure that the verbal parallels apparently revealing common authorship are not simply commonplaces in the wider drama. I would have thought that a more relevant study for Kirwan to critique would be MacDonald P. Jackson's systematic "The Date and Authorship of Hand D's contribution to Sir Thomas More: Evidence from 'Literature Online'" published in Shakespeare Survey and reviewed in YWES for 2006. Jackson showed that even if Hand D is not Shakespeare's (which most scholars think it is), the words in this handwriting were composed by Shakespeare.

    Pechter points out that part of the problem of building attribution studies on parallel passages is that good writing is inherently memorable and imitable, and hence some researchers focus on function words which seem not to be consciously chosen by authors. (Any reader who thinks that authors are in conscious control of their function-work usage might like to predict whether she uses the words the and and more or less frequently than the present reviewer's fairly consistent values of one-word-in-14 and one-word-in-35 respectively. The answer is easily determined by anyone who has in digital form a few tens of thousands of her own words, but in this author's experience no one who takes the test can confidently predict the outcome.) Pechter summarizes the introductory remarks in Hugh Craig and Arthur Kinney's book Shakespeare, Computers, and the Mystery of Authorship (reviewed in YWES for 2009) about the neuroscientific understanding of language and Lene Petersen's thoroughgoing connection of authorship attribution to brain science in her book Shakespeare's Errant Texts: Textual Form and Linguistic Style in Shakespearean 'Bad' Quartos and Co-authored Plays (reviewed in YWES for 2010).

    Pechter is indulgent of Craig and Kinney's and Petersen's excursions into the neuroscience of language, but I would say that we are so far from understanding what the brain does to create and process language--so far from knowing why rates of function-word usage are such reliable markers of authorship--that we should leave the matter aside and focus on discovering which other stylometric tests are objectively capable of distinguishing authorship in those cases where we already know the answer, and that when we have what are demonstrably the most discriminating tests we should apply them to the unknown cases.

    Pechter rightly remarks that the real scandal of Donald Foster's misattribution to Shakespeare of the "Elegy for William Peter" in the late 1990s was not Foster's mistake but "the immoderate haste with which so many Shakespeareans accepted it on faith" (p. 241). The recent discovery that Shakespeare did not have an exceptional vocabulary makes it all the more plain that what he did with words, not the words he used, made him exceptional. Pechter finds this a problem for computational approaches since they are "designed to count words, not to tell us what they 'do'" (p. 242). This seems to me to understate what computational procedures are already capable of, since they undoubtedly can do more than merely count words. Automated morphosyntactic tagging has already reached the point where a machine can, with high reliability, identify which words in a sentence are proper and common nouns, which adjectives, which conjunctions, and so on. Thus a machine can search across thousands of novels for the sequence name, adjective, adjective, conjunction, adjective to find parallels to Jane Austen's "Emma Woodhouse, handsome, clever, and rich", to use an example of which Martin Mueller is fond.

    Pechter thinks that computational analysis necessarily decontextualizes the language it searches for, but there is no reason why the machine may not look for the sequence just described only when it appears at the beginning of a novel (which is a context) and so find novels that begin as Emma does. Computational approaches are not as sensitive to context as readers are, but that does not mean they are entirely insensitive to it. That the attributionists necessarily started with decontextualized approaches--simply looking for, say, all the exclamations in the a text--does not mean they are limited to them. According to Pechter, "Attribution methods . . .  [are] based . . . on computation rather than interpretation, on counting Shakespeare's words rather than trying to account for what they do or how they are used . . ." (p. 245). This is wrong about the methods (there is more going on than counting words) and constructs a false dichotomy between computation and interpretation.

    Pechter is right that a human reader is better able than a machine to give an account of why "I had it of Leah when I was a bachelor. I would not have given it for a wilderness of monkeys" feels Shakespearian, but the cost of human interpretation is that we do not agree about it. Pechter finds that this line "thrusts us into acknowledging a human creature capable of affectionate feelings beyond those conferred on material possessions" (p. 245), but I would challenge his reading. Surely, Shylock has not got past his fixation with material possessions, since he is still coldly calculating values. He does not say that, for emotional reasons, the ring is beyond any exchange value, only that even a vast number of monkeys would not reach its exchange value. There is beggary in the love that can be reckoned.

    Pechter acknowledges and dismisses the case I am making that we are only at the beginnings of computational stylistics and that in future computers will give much more sensitive readings, pointing out that stylistics has been making that excuse since the pre-digital days of the 1960s. Pechter agrees with Johanna Drucker that there is just something fundamentally different between processing a text and reading it; "computers will never read better", he writes, "because reading . . . is not what they do" (p. 248). I would say that this depends on the reader. The machine-learning software called GPT-3 is currently capable of giving a precis of a literary text that beats the best efforts of a typical teenage reader.

    For Pechter, the inevitable impressionism (that is, subjectivity) of literary criticism "is not a defect to be eliminated through the exercise of a meticulous empiricism and confirmability--it's a virtue in its own right" (p. 251), so the scientific method does not trump aesthetics. Whereas each branch of science has one dominant paradigm at a time, a new one periodically replacing an old, in the Humanities there are multiple paradigms in place at once, with competing and incompatible foundations, and according to Pechter this is a good thing. Pechter finds me guilty of slipping from a can (we are able to distinguish authors even though postmodern theory says this is impossible) to a should (we ought to give value to authorship, because it is distinguishable).

    Pechter is happy to offer authorship attribution studies the modest place that it earns by being a contribution to the study of the originating circumstances of a text, on the grounds that the identity of the person who wrote the text is quite an important originating circumstance. He thinks that "Attribution scholars rarely make suggestions along these lines" (p. 254), but I would say that he has merely rephrased the attribution scholars' claim that authorship truly matters. Pechter concludes that what should worry us is the tail wagging the dog in the sense that perhaps we do authorship attribution and the demarcating of boundaries of collaborative authorship simply because we are now enabled to do it by digital technology. That is an acutely percipient observation to which this reviewer has at present no effective response.

    In the second relevant article from this year's Shakespeare Quarterly, Mark Dahlquist argues that a longstanding crux needs no emendation, only explication ('Hamlet and the Snare of Scandal', SQ 69[2018] 167-87). The crux is Hamlet's lines that "the dram of eale | Doth all the noble substance of a doubt | To his owne scandle. | Enter Ghost" (Q2, sig. D1v). Oddly, Dahlquist claims to quote this from the Arden3 edition but he spells the last spoken word as "scandale" where the Arden3 (being modernized) has "scandal" and Q2 has "scandle". The Catholic Douay-Rheims translation of the New Testament used scandalize and scandal where previously English translations used offend and stumbling block. So, scandal could mean offence. The origin was in the Greek word skandalon meaning snare or trap, as in a mouse-trap.

    Dahlquist gives extensive detail about the religious debate over this notion of scandal as something that might lead a Christian into sin. At this time there were new and improved mouse-traps becoming available. The other technology much alluded to in the play is the cannon, associated especially with Claudius. Both devices exemplify "understanding mental error through the metaphor of a mechanical device" (p. 174). Claudius uses cannons not as weapons of war but as part of his attempt to appear kingly and divine by impressively noisy ceremony. The "dram of ale" crux is not, according to Dahlquist, a continuation of the preceding lines' reflection on how one little bad characteristic spoils the whole of something or someone.

    Edward Dowden proposed in his 1899 Arden1 Hamlet that "Doth . . . scandal" is a verb with the meaning of does . . . slander. Dahlquist argues that the slandering (in the sense of ensnaring) is done by evil (perhaps the devil) by making us doubt, hence "of a doubt", and hence we fall into the sin of error. Thus the supposed crux needs only to be explained as saying that "the dram of evil doth the soul, through doubt, ensnare" (p. 179). Dahlquist quotes The Two Gentlemen of Verona where substance means the soul and Shakespeare's contribution to Edward 3 where scandal is used in its doctrinal sense.

    The journal Shakespeare Survey contains two articles relevant to this review. In one, Amy Lidster argues that the success of Andrew Wise's Shakespeare editions of the late 1590s established a particular part of St Paul's Churchyard as the centre of Shakespeare publishing and connected the dramatist with his playing company and their patron ('At the Sign of the Angel: the Influence of Andrew Wise on Shakespeare in Print', ShSurv 71[2018] 242-54). These associations then shaped how others published Shakespeare, particularly regarding the use of his name on title pages and linking him with wider literary and poetical traditions .

    Andrew Wise was the first publisher to focus on Shakespeare, releasing a series of his editions, 11 in all between 1597 and 1602. The plays were Richard 2, Richard 3, 1 Henry 4, 2 Henry 4, and Much Ado About Nothing, and the number of reprints shows that they were successful in print. At this stage, Wise published plays only by Shakespeare and with the exception of Much Ado About Nothing only histories; other publishers tended to be more eclectic in source and subject matter. It may be that Wise's choice was connected to the location of his bookshop in Paul's Cross at the sign of the angel, for this part of St Paul's Churchyard had a number of publishers specializing in medieval English history. Lidster links the publication of Samuel Daniel's The Civil Wars by Simon Waterson in this part of the bookselling district with the publication of Shakespeare's plays on the same topic, suggesting that the success of the Wise quartos might even have led to the expansion of Daniel's The Civil Wars (p. 245).

    That Wise's chief authors, Shakespeare, Thomas Nashe, and Thomas Playere, were all three patronized by George Carey, second Baron Hunsdon, may explain the Shakespeare/Wise link. The success of the Wise history-play quartos seems to have initiated a boom in the publication of works about late-medieval English history, and this perhaps was the reason for the reprintings of The Contention of York and Lancaster and Richard Duke of York in 1600. But on the stages, to judge from the known repertories of the Chamberlain's Men and the Admiral's Men, the range of topics and genres was much broader than just histories, so we have to see publication not merely as a reflection of what was popular in the theatre but as expressing a distinct pattern of selection by publishers to suit what they thought the book-buying public wanted to read.

    Wise made a point of putting Shakespeare's name on his title pages from 1598, and when Shakespeare's name first appears in the Stationers' Register in 1600 it is in a joint entry by Wise and William Aspley. Wise's quartos were the first explicitly to link Shakespeare's name with the Chamberlain's Men playing company, and thereby to indicate the patronage connection. With the reprints of Venus and Adonis and The Rape of Lucrece in the late 1590s by John Harrison and William Leake, the association of Wise's part of St Paul's Churchyard with Shakespeare publishing was strengthened since ". . . all of these editions (with the exception of Harrison's 1598 and 1600 reprints of Lucrece) were offered for wholesale at the Sign of the White Greyhound, just three doors (or about 20 feet) away from Wise's shop in Paul’s Cross" (p. 250). Possibly this is why the second edition of The Passionate Pilgrim in 1599, containing just five Shakespeare poems, names Shakespeare on its title page where his previous poetical publications had not. Title pages were used as advertisements and by the late 1590s Wise's area of St Paul's Churchyard "regularly featured a considerable number of title pages advertising works by 'Shakespeare'" (p. 251).

    Lidster sees the Wise quartos as essential to the establishment of Shakespeare as a literary figure, with other publishers drawing upon them (literally, for books of literary quotations) and upon their practices (such as putting his name on title pages). So how do we explain the sharp falling off of Shakespearian publication after 1603? Perhaps part of the cause was Wise ceasing to publish that year and George Carey, the Chamberlain's Men's patron, dying, and thus the association of one part of St Paul's Churchyard with the Shakespeare publishing industry thereby ended.

    Our second article from Shakespeare Survey is a defence of the unemended "a Table of greene fields" reading in the Folio edition of Henry 5 (Cyrus Mulready 'From Table Books to Tumblr: Recollecting the Microgenres of the Early Modern Stage in Social Media', ShSurv 71[2018] 194-208). According to Mulready, plays were perceived as being made of parts and were inherently fragmentary, and were consumed in a fragmentary way by having lines copied into commonplace books. This suits our modern digital media with their animated GIFs, listicles, and short tweets. Marking particular words in a tweet with a hashtag is like the marginal marking (with manicules, asterisks, and so on) that readers put in early modern books. R. W. Dent's Shakespeare's Proverbial Language identifies 187 proverbs in 1 Henry 4, 169 in 2 Henry 4, and 136 in MWW, compared with 185 in Hamlet and 197 in King Lear. Nearly a third of the proverbs in the first three of those plays come from Falstaff.

    The Hostess's account of Falstaff's death in Folio Henry 5 is rich in proverbs, and contains the crux "for his Nose was as sharpe as a Pen, and a Table of greene fields". Mulready surveys the attempts to retain  table--perhaps some kind of table book, hence the pen--instead of accepting Lewis Theobald's emendation to babbled. Mulready finds the table reading more apt than Theobald's emendation because we can link it to the practice of commonplacing and hence it points to "the copious record of Falstaff's wit and sententiae" (p. 204). If the image is jumbled, that is because the Hostess is prone to malapropism. Mulready explores the generally low opinion of dramatists about some audience members' practice of writing down the best lines (especially jests) as they watched a play. Recurring in them are images of food and swollen body parts--the jest-hoarder being like a glutton--and that suits Falstaff's end too. The mid-seventeenth-century extraction of bits of plays into drolls, including the Gads Hill robbery from 1 Henry 4, again shows that plays were conceived of and consumed as fragmentary.

    Elsewhere, Darren Freebury-Jones sets himself the Herculean task of showing that all the new attributions given in the New Oxford Shakespeare are wrong ('Augean Stables; Or, The State of Modern Authorship Attribution Studies', Archiv für das Studium der neueren Sprachen und Literaturen 255[2018] 60-81). Freebury-Jones starts by praising the twentieth-century authorship attribution methods that have been offering "new insights" for "many decades" (p. 60), picking out the scholarship of Philip Timberlake, Ants Oras, and Muriel St Clare Byrne by name. These traditional methods do not simply use mathematics and Freebury-Jones shares Vickers's view that any new numerical approaches must agree with the results of the older non-numerical methods before they can be accepted. Surely if this were true--if the litmus test is agreement with existing methods--there would be no point developing new methods.

    Getting down to details, Freebury-Jones reports that he thinks that Shakespeare alone wrote 2 Henry 6 and 3 Henry 6 as a standalone pair of plays and that 1 Henry 6 "was designed by rival company Lord Strange's Men to capitalize on their success" (p. 62). Subsequently, Shakespeare was "commissioned by the Chamberlain's Men, founded in summer 1594, to adapt the original 'Harey the vj'" (as Henslowe's Diary calls the Strange's Men's play) to give them the three-part Henry 6 series we know from the 1623 Folio. Freebury-Jones reckons that Nashe wrote Act 1 of this original "Harey the vj" play and Kyd wrote Acts 2-5 and when revising it Shakespeare added what we now know as scenes 2.4, 4.2, and 4.5.

    Freebury-Jones rehearses the historical claims that Kyd had a hand in 1 Henry 6 and offers as a new piece of evidence the likeness of the lines "To be enrolled in the brasse leaued booke | Of neuer wasting perpetuitie" in the 1592 quarto of Soliman and Perseda (often attributed to Kyd) and "Coupled in bonds of perpetuity, | Two Talbots wingèd through the lither sky | In thy despite shall scape mortality" in 1 Henry 6. Freebury-Jones quotes the lines from Soliman and Perseda second-hand from an article in Notes and Queries and precedes them with a discussion of Kyd's idiosyncratic habits of rhyming that are supposed to reveal his authorship.

    In particular, Freebury-Jones endorses James E. Routh's identification of the pattern "aca, where c is an unriming line" (p. 64) as characteristic of Kyd, and while the 1 Henry 6 quotation seems to show that pattern, it is not present in the lines quoted from Soliman and Perseda. But this aca pattern is present in Freebury-Jones's quotation from 1 Henry 6 only because he excluded the preceding line "Anon from thy insulting tyranny". Once we include this preceding line the endings are ". . . tyranny, | . . . perpetuity, | . . . sky | . . . mortality", producing regular rhyming couplets, as indeed are the lines preceding and following this in 1 Henry 6. Freebury-Jones claims that "These examples suggest a single author's rhyming habits . . ." (p. 64) but in fact rhyming couplets are found in almost all the dramatists' works from this period.

    Next Freebury-Jones turns to the habit of giving verse lines feminine endings as explored by Timberlake, listing the "overall range" (p. 65) for writings of contested authorship that he considers to be by Kyd: Soliman and Perseda at 0 to 34.4, Cornelia at 2.4 to 13.1, King Leir at 0 to 25.4, Arden of Faversham at 0.9 to 28.5, Fair Em at 0 to 15.9, and Edward 3 at 0 to 11.7. Freebury-Jones does not cite Timberlake's book as the source for these numbers--although barring extraordinary coincidence it must be his source--and does not indicate what these numbers mean and why they are expressed as a range. The numbers are percentages of the lines in a scene that are feminine and they are expressed as a range because, of necessity, a play will have one scene that gives that play's lowest percentage of lines-in-a-scene that have feminine endings and another scene that will have the highest. Comparison of Freebury-Jones's report with Timberlake's book indicates that he is using Timberlake's strict counts that exclude disputable cases such as lines ending with words that can be monosyllabic or disyllabic, such as heaven and power.

    Freebury-Jones moves straight to his conclusion with the baffling sentence that "Kyd's scenes in Henry VI Part One (in my computations) average 4.7% feminine endings, which effectively rules out Marlowe, who, as Alfred Hart noted, 'did not like a line with an unstressed syllable at the end unless it coincided with a natural pause' (p. 453), and whose dramatic output reaches a peak of 3.7 percent feminine endings for Edward II (1592), which has a range of 0.0-11.1" (Freebury Jones's p. 65). Logically, if one decides before counting that some characteristic makes a scene one of "Kyd's scenes" then one has already ruled out Marlowe, yet if Marlowe's range is 0 to 11.1% then 4.7% falls squarely within that range. Moreover, how can Marlowe's plays reach "a peak of 3.7" percent if the top of his range is the 11.1 percent found in his sole-authored play Edward 2? Even on its own terms, Freebury-Jones's account of the evidence makes no sense and one wonders if some words or phrases have been accidentally dropped.

    In an even more compressed fashion Freebury-Jones reports that the parts of 1 Henry 6 that he attributes to Kyd (Acts 2 to 5) have one compound adjective for every 602 words, which he finds almost identical to the rate in Kyd's The Spanish Tragedy, but he neglects to report Marlowe's rate or Shakespeare's beyond mentioning in a footnote that ". . . Marlowe largely eschewed compound adjectives in his dramatic works" (p. 65n29). Next Freebury-Jones turns to Mueller's work on shared n-grams, claiming that it supports Kyd's presence in 1 Henry 6, and as has been complained of before in reviews of Freebury-Jones's writing his supporting reference is an URL that at the time of writing (October 2021) is broken.

    Freebury-Jones seems not to understand Mueller's method as he glosses his use of the notion of interquartile range as "a measure of variability, based on dividing data into quartiles" (p. 66) when of course the key fact is that this is the midspread of a range of data samples, between the 75th and the 25th percentiles. Freebury-Jones does not mention how Mueller gets from the discovery of n-grams shared between two plays to the quantitative expression of these discoveries and blithely reports that the counts are "weighted according to the lengths of repetitions" (p. 66), as if there were general agreement on how to do such weighting. The reported fact that 1 Henry 6 and The Spanish Tragedy "are placed above the median with a percentage of 96" (p. 66) is simply meaningless without an explanation of what this 0.96 value represents. Freebury-Jones's account of Mueller's work is the verbal equivalent of an x/y graph on which the investigator has neglected to label the axes.

    Turning to the essay by Gary Taylor and John V. Nance "Imitation or Collaboration?" (reviewed in NYWES for 2015), showing rare parallels between Marlowe's work and 1 Henry 6, Freebury-Jones complains that many of them "are in fact meaningless combinations of words" (p. 67). He would know, if he understood it, that Mueller's investigations are open to the same objection, since both methods look for strings of letters and spaces in common without regard to their meaning. Freebury-Jones suggests that Taylor and Nance overlooked relevant evidence in not reporting that 1 Henry 6 and The Spanish Tragedy share the phrase and give me, but of course Taylor and Nance ignored it because they were looking for rare phrases shared by plays and this phrase is common. Somewhat desperately, Freebury-Jones argues that the occurrence in the preceding two lines in both plays of the three words and, that, and me makes the parallel more distinctive, but to substantiate that claim he would need to investigate how often we should expect those three words to appear together in two lines.

    Freebury-Jones finds a suggestive parallel between "feed you with my blood" in 1 Henry 6 and "feed your selves with mine enflamed blood" in Cornelia, but again he makes no attempt to quantify the rarity (and hence the evidentiary value) of this likeness. Thus when Freebury-Jones reports that "By my count, Taylor and Nance miss half a dozen unique phrases and collocations with Kyd's accepted canon . . ." (p. 67) there is nothing to support his claim about uniqueness.

    Freebury-Jones's example of the run of words and, that, and me from 1 Henry 6, in the lines "Now help, ye charming spells and periapts, | And ye choice spirits that admonish me", can be specified as the word and followed by up to five words of any kind followed by the word that followed by one word of any kind followed by the word me. In the commonly used Corpus Query Language this is coded as the following search: [reg="and"] [word=".*"]{1,5} [reg="that"] [word=".*"]{1} [reg="me"]. The TCP dataset of 25,000 early printed books is available for interrogation using Corpus Query Language at the online Early Print Lab hosted by Washington University in St Louis, and for this search it turns up over 7,000 hits. Why Freebury-Jones should think that he is reporting rare verbal parallels is baffling unless perhaps he is unable to perform the necessary corpus searches to quantify rareness and is simply relying on his own instincts.

    Accepting that Taylor and Nance have nonetheless found lots of rare verbal parallels between 1 Henry 6 and the Marlowe canon, Freebury-Jones argues that this would be unsurprising even if Kyd wrote 1 Henry 6 since they "shared a room in 1591, and it is therefore probable that they also shared a reading knowledge of each other's works" (p. 68). Freebury-Jones seems not to notice that this argument runs both ways, and that by his logic it would be easy to mistakenly attribute to Kyd works that are by Marlowe. Thus when he cites the parallel of "this head, this heart, this hand and sword" from Marlowe's The Massacre at Paris with "this sword, this buckler, this head, this heart, these hands" from King Leir, Freebury-Jones does not allow it to daunt his confidence that the latter is by Kyd who simply picked up this phrasing from knowing his Marlowe.

    In order to discredit Taylor and Nance's work, Freebury-Jones attempts what he claims is a kind of replication of their investigation but with the twist that he limited his searches "to matches consisting of at least two contiguous words" (p. 68). He finds that Kyd's The Spanish Tragedy has four unique matches with Marlowe plays: fortune of the war[s], little loss, my loving brother of, and met|meet our army|armies. These are supposed to be unique in the sense that they appear in no other plays in the period 1576-1594, and indeed three of them are unique, but my loving brother of occurs twice in the anonymous play The Famous Victories of Henry 5 first performed in the 1580s: "My louing brother of England" and "I must be King: but my louing brother of France". It is in any case unclear why Freebury-Jones thinks that finding parallels between The Spanish Tragedy and Marlowe's works undermines the case made by Taylor and Nance, since they did not apply his rule of rejecting all but "matches consisting of at least two contiguous words". Freebury-Jones manages to show that his adaptation of their approach does not work but not that Taylor and Nance's study, using different criteria, is faulty.

    Trying a different approach, Freebury-Jones considers a 173-word passage from Soliman and Perseda, which he thinks was written by Kyd, and looks for unique links to other plays from the period 1576-1594. He finds three with Kyd's The Spanish Tragedy but seven with Marlowe plays, and since there are more than twice as many links to Marlowe than Kyd he concludes that the method does not work. Of course, the attribution of Soliman and Perseda to Kyd is not universally accepted, but more importantly Freebury-Jones is wrong that his claimed seven parallels are unique. His first, call[s] on Christ, is also found in the anonymous The Troublesome Reign of King John first performed in 1591: "And call on Christ, who is your latest friend". His second, entreat a pardon, is also found in the anonymous The Taming of a Shrew first performed in the early 1590s: "Intreat a pardon in your lordly brest".

    Freebury-Jones's third claimed unique parallel, a pardon for, is also found twice in Robert Greene's George a Greene first performed in the late 1580s or early 1590s: "And good my Lord, a pardon for poore Robin" and "I pray you, a pardon for the Shoomakers". His fourth, thou|you . . . me and have, does seem to be unique to The Spanish Tragedy and one Marlowe play, but only if we confine the number of words between thou|you and me and have to fewer than 23, otherwise Robert Wilson's The Three Lords and Three Ladies of London first performed in the late 1580s has a match. Freebury-Jones's fifth link, grant|give it thee then, is unique to The Spanish Tragedy and one Marlowe play, although Shakespeare's Romeo and Juliet (which might have been first performed in 1593 or 1594) has the close match "then give it you". His sixth link, is past and I, is unique to The Spanish Tragedy and one Marlowe play, and so is his seventh link, crown[s] and empery|emperies, although Freebury-Jones identifies that play as Part Two of Tamburlaine when it is in fact Part One.

    To sum up, about half of the unique links claimed by Freebury-Jones are simply not unique and it is hard to account for his overlooking additional plays that contain the phrases in question. His point in all this is to show that when Taylor and Nance find evidence of Marlowe's authorship of 1 Henry 6 they may be simply mistaking Kyd's writing for Marlowe's (p. 70). But since Freebury-Jones is demonstrably unable to replicate Taylor and Nance's method--because he repeatedly includes links that do not pass their test for rareness--this claim has no evidentiary basis.

    Having cleansed the Augean stables of others' work to his own satisfaction, Freebury-Jones proceeds to an account of his view of Kyd's contributions to 1 Henry 6 based on close reading rather than item-counting. Then he returns to counting feminine endings, citing Timberlake's tables as showing that 2 Henry 6 has 10.4% feminine endings and 3 Henry 6 has 10.7%, which numbers are above Marlowe's peak of 8% and hence that they are unlikely to be by Marlowe (p. 73). But those values of 10.4% and 10.7% are averages across a whole play and Timberlake's tables show that the range for particular scenes in 2 Henry 6 is 0 to 25% with seven of the play's 22 verse scenes falling below 8%, and that the range for particular scenes in 3 Henry 6 is 3.3 to 28.5% with 11 of the play's 28 scene falling below 8%.

    If we look at the matter using ranges, as Freebury-Jones does earlier in his essay, Marlowe's maximum of 8% feminine endings is compatible with his making substantial contributions to 2 Henry 6 and 3 Henry 6. And this 8% figure for Marlowe is itself only an average: Timberlake's tables show a scene in Doctor Faustus with 12.5% feminine endings, a scene in The Jew of Malta with 9.6%, a scene in Edward II with 11.1%, and a scene in The Massacre at Paris with 12.5%. Nothing in the evidence adduced by Freebury-Jones from Timberlake's tables is incompatible with Marlowe's authorship of parts of the Henry VI plays.

    Freebury-Jones turns to Oras's work on pause patterns in iambic pentameter lines and finds that the parts of 2 Henry 6 that are supposed to be non-Shakespearian are much like the parts supposed to be Shakespearian on this measure. According to Freebury-Jones, multiple authorship ought to reveal itself in different parts of a play showing a difference on this measure. But such a difference between authors is not what Oras's work would lead us to expect, since he showed that the biggest change was across time not across authors. That is, for all writers in the 1590s and 1600s the dominant pause drifted from the first half of the line to the second over this period. Next Freebury-Jones returns to reporting Mueller's results, this time in applying Discriminant Analysis to "lemma trigrams" (p. 74). For this work Freebury-Jones again cites the URL mentioned earlier that is broken and giving no detail of what Mueller counted Freebury-Jones simply quotes the study's conclusion that 2 Henry 6 and 3 Henry 6 "are solely Shakespeare's" (p. 74).

    Then Freebury-Jones considers MacDonald P. Jackson's work on the Shakespearian authorship of part of the anonymously published play Arden of Faversham, citing his own essay "Kyd and Shakespeare: Authorship versus Influence" reviewed in NYWES for 2017. In a single paragraph Freebury-Jones attempts to undermine the essay by Jackson in the Authorship Companion to the New Oxford Shakespeare that provided additional evidence (beyond that adduced in Jackson's book on the topic) for Shakespeare's authorship of part of Arden of Faversham. Giving the matter just 237 words, Freebury-Jones has not even the space to tell the reader what Jackson claimed to have done, let alone critique it. The most Freebury-Jones can squeeze into this space is an insistence that counting things will not do and attributionists "must read the plays closely" (p. 75).

    As with his analysis of 1 Henry 6, Freebury-Jones is sure that if a play is co-authored by two writers then we would be bound to see that fact reflected in a difference between the values on any particular measurement we might make of their putative shares. Tabulating pause patterns of the type measured by Oras, Freebury-Jones finds no significant difference: "What becomes immediately clear looking at this table is the fact that the percentages for 'Shakespeare' scenes are hardly different to those found for the rest of the play, as we might expect were the domestic tragedy sole-authored" (p. 77). This is of course a logical fallacy: there is no reason to suppose that two or more writers will differ on every linguistic feature we can measure.

    Digging into the detail for pause patterns at particular places, Freebury-Jones hits the snag that across Arden of Faversham ". . . the high proportion of pauses after syllable six is anomalous in Kyd's canon . . .", but he has a solution: ". . . this could suggest that the dramatist was experimenting with his verse style to create a more colloquial tone, as indicated in the play's epilogue" (p. 77). Thus Freebury-Jones relies on numbers when they support his case and subjective speculation when they do not. Next Freebury-Jones considers the use of compound adjectives as surveyed in Jackson's 2014 book Determining the Shakespeare Canon on the authorship of Arden of Faversham (reviewed in YWES for 2014) and objects that Jackson paid too little attention to the possibility of Kyd's being the play's author. But this part of Jackson's book, comprising two out of 272 pages, was not meant to be a survey of all the candidates' uses of this feature but rather merely a passing observation that ". . . it is the disparity between scenes 4-9 and the rest of the play that is crucial . . ." (Jackson p. 77). Freebury-Jones ought to be sympathetic to his argument, since earlier in his essay he argued that such an internal disparity is what we should expect to find in a co-authored work.

    At this point, Freebury-Jones's essay gets rather bogged down in minutiae concerning Jackson's claim that Freebury-Jones misrepresented Jackson's book when reviewing it. Freebury-Jones does not bring new evidence to the debate and confines himself to vaguely worded assertions that "Jackson has made an error here" and "Jackson . . . has ignored or dismissed numerous studies demonstrating Kyd's authorship" of Arden of Faversham (p. 79). Freebury-Jones acknowledges that Jackson drew attention to one expression from Kyd's The Spanish Tragedy that appears in Arden of Faversham and complains that Jackson assumed it was just Shakespeare echoing Kyd rather than a sign of Kyd's authorship, and indeed this goes to the heart of one of the problems of using verbal parallels. When a line is famous, how can we distinguish authorial self-repetition from the conscious echoing of the line by someone else, as when Pistol's "hollow pampered jades of Asia" in 2 Henry 4 echoes "Holla, ye pamper'd jades of Asia" from Marlowe's 2 Tamburlaine? Freebury-Jones has no new solution to this perennial problem.

    Freebury-Jones ends his essay with a few words in passing about some other authorial attributions in the New Oxford Shakespeare, finding "unconvincing" (p. 80) the claim that Thomas Middleton inserted the Fly Scene into Titus Andronicus, but without revealing why. As an alternative to the New Oxford Shakespeare attributions, Freebury-Jones recommends Martin Wiggins's essay called "Who Wrote Shakespeare?" that appeared on the BBC website, citing an URL that challenges even the most careful typist since its middle part reads "211LBPTmBYp2rbh4bSQlSTS" and each of those ones and ells must be carefully distinguished if the reader is to avoid receiving a "404 Page Not Found" error.

    This reviewer received Freebury-Jones's essay in the form of a PDF from the publisher in which, surprisingly, the URL in Freebury-Jones's footnote has not been made into a hyperlink connecting directly to Wiggins's article, but instead drops the reader off at the landing page for the section "Programmes" on the BBC website. This landing page contains a text box for searching the BBC's entire collection of programmes, but a search there for "Who Wrote Shakespeare?" does not throw up Wiggins's essay. It does throw up first Ben Elton's splendid television comedy The Upstart Crow about Shakespeare and next an account of the Toronto domestic terrorism vehicle-ramming attack of 23 April 2018. Presumably that significant date confused the BBC's search engine.

    Another of Freebury-Jones's articles this year showed that there are many shared rare verbal parallels between Arden of Faversham and John Lyly's play Endymion ('Exploring Verbal Relations between Arden of Faversham and John Lyly's Endymion', Renaissance and Reformation / Renaissance et Réforme 41.iv[2018] 93-108). He starts by surveying some past scholarship on the authorship of Arden of Faversham, including Frederick S. Boas's in 1901 pointing out that beginning a line with Ay but is a distinct habit of Kyd's found six times in The Spanish Tragedy (universally attributed to Kyd), eight times in Soliman and Perseda (often attributed to Kyd), nine times in Arden of Faversham, and seven times in King Leir (which last two Freebury-Jones would attribute to Kyd). This is all true, but Freebury-Jones does not mention that there are also lines starting Ay but in The Two Gentlemen of Verona, Titus Andronicus, The Comedy of Errors, 1 Henry 6, 2 Henry 6, 3 Henry 6, The Taming of the Shrew, The Merchant of Venice, 1 Henry 4, As You Like It, The Merry Wives of Windsor, Henry 5, Twelfth Night, Troilus and Cressida, Measure for Measure, Macbeth, Othello, Coriolanus, Timon of Athens, Pericles, and The Tempest. In all, these Shakespeare plays have 46 lines starting Ay but and either Freebury-Jones is unaware of them (which seems incompetent) or he is aware but decided not to mention them (which seems dishonest).

    Freebury-Jones turns to literary criticism of Arden of Faversham and he finds parallels with Lyly's and Kyd's writings before returning to the quantitative approach with an account of Mueller's investigations with his dataset Shakespeare His Contemporaries. Mueller is reported to have found verbal parallels between Arden of Faversham and Lyly's plays and also the anonymous plays King Leir and Fair Em, which Freebury-Jones reckons are the unacknowledged works of Kyd. If Freebury-Jones is right that King Leir and Fair Em are by Kyd then this is indeed evidence for Kyd's authorship of Arden of Faversham, but Freebury-Jones offers no reason to think that these two anonymously published plays are by Kyd, not even in the form of citations of other studies in which the case is made.

    According to Freebury-Jones, Mueller sent him the dataset on which these claims are made and Freebury-Jones gives a brief account of the logic at work, which is that plays by the same author tend to have disproportionately more (about "twice as many", p. 99) four-word phrases in common than plays by different authors, when we confine ourselves to phrases found only in two plays. Finding that there are a lot of phrases from Lyly's Endymion in Arden of Faversham, Freebury-Jones presents this as self-evidently evidence against Shakespeare's hand in the latter play, but I cannot see why. He is happy to acknowledge the well-known influence of Lyly on Shakespeare, but because ". . . Mueller's spreadsheet provides no evidence that Shakespeare frequently borrowed large phrasal units from Lyly's Endymion at the very beginning of his career" (p. 100) he rejects Shakespeare's hand in Arden of Shakespeare.

   Instead, Freebury-Jones looks for someone who demonstrably did use Lyly's phrases in his writing and finds him in the person of Kyd. Freebury-Jones starts to list the supposedly shared phrases, and not all are correct. He claims that "Sweet Mosby is the man that hath my heart; | And he usurps it" in Arden of Faversham uniquely finds a match with "sweet Endymion, is he that hath my heart" in Endymion, the link being the run of sweet . . . is . . . that hath my heart with he nearby or within this run. But this is not unique, since Marlowe's A Massacre at Paris has "Sweet Mugeroune, tis he that hath my heart". Freebury-Jones's next example, say[s] true and therefore . . . here, does indeed seem to be unique to Arden of Faversham and Endymion, as is and thou shalt find with begun before or after it, and as is for what . . . life but love and, and as is qualm . . . came|commeth over my heart. Freebury-Jones's claim of rare verbal parallels between Arden of Faversham and Lyly's Endymion is correct: he has found four of them.

    Freebury-Jones next turns to Pervez Rizvi's online dataset called Collocations and N-grams, which includes a list of every run of four or more consecutive words common to Endymion and other writers' plays. Freebury-Jones decides to confine his attention to plays from the period 1580-1600--he does not indicate why he did not do that with Mueller's dataset--and finds that there are 20 such runs common to Arden of Faversham and Endymion. I do not understand what mathematical operation Freebury-Jones is alluding to when he writes that "If we adjust these figures according to composite word counts, we find that both plays average 0.05 rare matches with Lyly's text" (p. 105). There is no such thing as 0.05 (one-twentieth) of a rare match, as matches are counted in whole numbers. This value of 0.05 must be the result of dividing some dividend by some divisor, but Freebury-Jones neglects to say what his dividend and divisor are. Freebury-Jones claims that what he has found supports his view "that Arden of Faversham is a uniform play belonging to a single authorial mind" and not co-authored (p. 105).

    Freebury-Jones lists a few examples of the unique matches from Rizvi's data but it is unclear why since his claim is not that Lyly wrote Arden of Faversham but that ". . . the author of the domestic tragedy [Arden of Faversham] was conscious of Lyly's dramatic language" (p. 107) and that ". . . the evidence that the dramatist responsible for Arden of Faversham had either seen or read Lyly's comedy seems solid" (p. 108). This is uncontroversial but does not advance the authorship discussion: many potential authors of Arden of Faversham are likely to have seen or read Lyly.

    This year, Freebury-Jones also published an article that he co-wrote with Marcus Dahl ('The Limitations of Microattribution', Texas Studies in Literature and Language 60[2018] 467-95). This looks again at Gary Taylor and John V. Nance's article "Imitation or Collaboration?: Marlowe and the Early Shakespeare Canon" (reviewed in NYWES for 2015) that Freebury-Jones alone critiqued in his article 'Augean Stables', reviewed above. Freebury-Jones and Dahl conclude that Taylor and Nance cannot reliably attribute authorship using the evidence they find, nor tell whether patterns of verbal similarity are due to imitation or shared authorship.

    Freebury-Jones and Dahl begin by suggesting that conscious imitation of one dramatist by another may make them "consciously reproduce function words as well as lexical words" (p. 468) but they provide no evidence for this counter-intuitive claim. Instead of quantitative methods, Freebury-Jones and Dahl suggest that we should go back to "reading-based approaches" (p. 469). They propose that in his searches within Literature Online ". . . it is possible that MacDonald P. Jackson had only Shakespeare’s word associations in mind . . ." (p. 469) and that this biased his searches. This is certainly possible, but Freebury-Jones and Dahl offer no evidence that it did in fact occur. Regarding Taylor and Nance's decision, in "Imitation or Collaboration", to exclude from their evidence any phrases common to the speech being studied and the speech it is widely supposed to be imitating, Freebury-Jones and Dahl write that "A skeptic might consider this to be self-defeating" (p. 470) but they do not say why or what logical flaw they are implying.

    Taylor and Nance decided to consider only the first 173 words of the speech at 2.3.176-199 of The Jew of Malta, so they could compare the results with those from the 173-words of a speech from Titus Andronicus. Freebury-Jones and Dahl write that "they condense" (p. 471) the longer speech, but this is a misuse of the word condense since nothing got denser: they just stopped at 173 words. Taylor and Nance observed (p. 40) that The Massacre at Paris has been claimed as a bad quarto made by memorial reconstruction, based on the evidence of its echoes of other plays. Their previous examples had already shown that we will always find many unique parallels to other plays when testing every n-gram in a sizeable speech, so those echoes of other plays are not of themselves evidence of memorial reconstruction.

    Missing Taylor and Nance's point entirely, which concerns evidence for memorial reconstruction (and whether thereby The Massacre at Paris is invalidated as evidence), Freebury-Jones and Dahl write "This statement, at the very least, seems to raise the question: If all such texts generate large numbers of parallels, how are we to conduct a purely quantitative analysis that ensures that only authorial matches are counted?" (p. 471) The answer, of course, is by the method Taylor and Nance were describing, which is to see whether the parallels accumulate in particular authorial canons and plays.

    Freebury-Jones and Dahl find that "It is curious that Taylor and Nance retain this play [The Massacre at Paris] but omit other 'deeply disputed' plays considered 'bad' quartos, which could have potentially altered their results" (p. 471). This is casting aspersions without actually saying what they think Taylor and Nance did wrong. If their idea is that it was inconsistent of Taylor and Nance to admit The Massacre at Paris as evidence while omitting other bad quartos, then they are simply misreading the footnote they cite (Taylor and Nance's p. 35n14), which does not report that bad quartos were routinely omitted but that where there are multiple early editions of early Shakespeare plays the good texts were preferred over the bad.

    Taylor and Nance reported that the tested speech from The Massacre at Paris has four matches to Greene's canon and five to Marlowe's canon if we confine ourselves to the uncontested canons of these two writers, and that Marlowe is still further ahead if we also accept some recent attributions to Marlowe. This claim Freebury-Jones and Dahl misrepresent as "Marlowe scrapes victory again on the assumption that these previously considered Shakespearean passages are in fact attributable to Marlowe" (p. 472). No, he wins even without the additional attributions: he is already ahead (five matches to Greene's four) when we confine ourselves to attributions that Freebury-Jones and Dahl do not dispute, because nobody does.

    Freebury-Jones and Dahl spot what they think is another problem in relation to Taylor and Nance's testing of the 173-word speech from The Massacre at Paris, when they switch to counting which play (rather than which authorial canon) has the most matches to the speech and find that the anonymously published Locrine has two, Robert Greene's James 4 has two, but Marlowe's Edward II has three. Freebury-Jones and Dahl object: ". . . did not Greene just get four or maybe five matches? Why is he not now at least an equal candidate to Marlowe for the authorship of this 173-word sample?" (p. 472). Well, those four matches (maybe five if you admit a contested play as Greene's) arose when comparing matches to whole canons (and these five represent matches to more than one Greene play) and cannot be directly compared to the smaller numbers we get when looking for which single play has the most matches. Freebury-Jones and Dahl seem not to understand the tests they are trying to critique.

    Likewise, Freebury-Jones and Dahl object that earlier Taylor and Nance had written that "two or three parallels are . . . no more significant than one" and now they, Taylor and Nance, seem to accept the evidence of three parallels to Marlowe's Edward 2. Again, Freebury-Jones and Dahl have misunderstood or are wilfully misrepresenting Taylor and Nance, who actually wrote in respect of Titus Andronicus 5.1.124-144 that "Four playwrights--Greene, Kyd, Lyly and Marlowe--can boast three unique dramatic parallels; obviously, all four men did not write this one speech, so we can deduce that two or three parallels are, here at least, no more significant than one" (Taylor and Nance, p. 36). Clearly, this is a remark about counting canon-wise not counting play-wise. Freebury-Jones and Dahl's quotation uses ellipsis in place of the crucial qualifying clause "here at least", which does seem to suggest wilful misrepresentation.

    Freebury-Jones and Dahl claim to find a parallel, overlooked by Taylor and Nance, between Joan's speech "Now help, ye charming spells and periapts, | And ye choice spirits that admonish me | And give me signs of future accidents" in 1 Henry 6 and Lorenzo's speech "Let this be all that thou shalt doe for me: | Be watchfull when and where these lovers meete. | And give me notice in some secret sort" in Kyd's The Spanish Tragedy. If we represent the intervening words in square brackets, the link is "and [5-words] that [1-word] me | And give me" in 1 Henry VI matching with "that [4-words] me [3-words] and [4-words] And give me" in The Spanish Tragedy. The problem here of course is that collocations at these distances between the function words and, that, and me are common. We can find hundreds of unique function-word collocations in any substantial textual dataset and by mere counting they tell us nothing about authorship.

    The only even remotely distinctive feature of the alleged parallel is the trigram and give me, which Literature Online finds in 17 plays first performed between 1576 and 1594:

John Lyly Endymion (× 3)
John Lyly Mother Bombie
George Whetstone The Second Part of Promos and Cassandra
Christopher Marlowe Doctor Faustus (× 2)
Thomas Heywood Second Part of Edward 4 (× 2)
Robert Yarington Two Lamentable Tragedies (× 2)
George Wapull The Tide Tarrieth No Man
George Peele David and Bathsheba
George Peele Edward 1
Thomas Kyd Soliman and Perseda 
Anonymous The Taming of a Shrew
Anonymous The Wars of Cyrus
Anonymous A Knack to Know a Knave
Anonymous The True Tragedy of Richard 3
Robert Wilson Three Ladies of London
Shakespeare 2 Henry 6
Shakespeare The Taming of the Shrew

This phrase is far too common in the drama to tell us anything. Freebury-Jones and Dahl do not mention its occurrence in Kyd's Soliman and Perseda, presumably because they failed to notice it, although it is mentioned in Freebury-Jones's 2016 PhD thesis.

    Although Freebury-Jones and Dahl consider this collation indicative of Kyd's authorship, they make no attempt to quantify its occurrence: they just notice that Joan's speech in 1 Henry 6 has it and that The Spanish Tragedy has it, and remark that "these parallel passages are at least worthy of qualitative analysis" (p. 473). As this sentence reveals, their approach is non-quantitative and thus forfeits any basis from which to critique a quantitative approach such as Taylor and Nance's.

    At this point Freebury-Jones and Dahl admit that they are not going to try to rebut Taylor and Nance on the numbers and they prefer to do literary-critical analysis of the few matches they are able to find, attending to "parallelism of thought" (p. 473). Freebury-Jones and Dahl claim to have found a match that Taylor and Nance overlooked in "I'll lop a member off" (1 Henry 6 5.3.15) matching "They lopt a collop of my tendrest member" (Soliman and Perseda 4.2.23). To include this parallel requires our window-of-interest to encompass the seven words from lopt and member (inclusive). They claim that such wide windows are unreasonable (p. 486), but even if this example were included as the sole parallel for Kyd it does not affect the result: Marlowe's nine unique parallels outweigh those of the other candidates.

    Freebury-Jones and Dahl find it significant that this last parallel also includes the word collop in Soliman and Perseda since elsewhere in 1 Henry 6 Joan's father calls her a collop "in another scene that Vickers assigns to Kyd" (p. 473). True, if like Vickers you start from the assumption that Kyd had a hand in 1 Henry 6 then these things start to pile up as unusual coincidences, but if you remain neutral and systematically count the parallels they do not. (By contrast, in the earlier example where Taylor and Nance mentioned some matches that occur if we accept recent additions to the Marlowe canon, Marlowe was already in the lead without these additional matches.) The objection about word distance affects Freebury-Jones and Dahl's next claim that Taylor and Nance overlooked relevant evidence: "future accidents" in Joan's speech in 1 Henry VI matches "future happiness. | But what disastrous or hard accident" (Cornelia 3.1.18-19). Here one needs a window-of-interest of eight words to admit the alleged Kyd parallel, and having stretched a point to get one more Kyd match there are still nine Marlowe matches outweighing it.

    We can now consider Freebury-Jones and Dahl's claims that, since the Marlowe canon is twice the size of Kyd's canon, the Kyd matches should count as more significant than Marlowe matches. Let us suppose that we admit the new widely spaced Kyd matches and let each Kyd match count for two Marlowe matches to compensate for Kyd's smaller canon: the result is still four matches for Kyd (up from two) and nine for Marlowe. That is, with all the extra allowances that Freebury-Jones and Dahl want to make for Kyd, there are still more than twice as many Marlowe matches as Kyd matches. It is unclear why Freebury-Jones and Dahl pursue this line of reasoning since it seems to do more harm than good to their claim that Kyd wrote Joan's speech at the start of 1 Henry 6 5.3. At this point, with the numbers so clearly against them, Freebury-Jones and Dahl again turn back to the non-quantitative methods and suggest that Marlowe's words are frequent in the writing that they want to attribute to Kyd because the two men shared a room and read each other's works.

    Freebury-Jones and Dahl describe their work as an "effort to reproduce Taylor and Nance's study" (p. 474), but admit that they had trouble understanding it: ". . . if Dido, Queen of Carthage is omitted from analysis because we 'do not know which parts (if any) are Nashe's', then why not omit Doctor Faustus as a whole . . ." since its co-authored parts are uncertain (p. 474). In fact, Taylor and Nance did not omit Dido from analysis and their remark quoted by Freebury-Jones and Dahl is about how we calculate canon sizes in the light of such uncertainties; indeed, Taylor and Nance's counts of Marlowe's canon and Nashe's canon are each presented "with Dido" and "w/o Dido" (Taylor and Nance, p. 43) to represent this area of uncertainty. It is apparent that Freebury-Jones and Dahl just did not understand what they were reading in Taylor and Nance's article. Mistakenly convinced that Taylor and Nance omitted Dido from their analysis, Freebury-Jones and Dahl ask a series of pointless rhetorical questions about why other plays of a similar condition were not omitted too.

    Freebury-Jones and Dahl try to discredit Taylor and Nance's work by showing that the first 173 words of 1.2 of Kyd's The Spanish Tragedy contain five unique parallels to Marlowe's work and just one to Kyd: "fortune of the war" (The Spanish Tragedy 1.2.3) matches "fortune of the wars" (2 Tamburlaine 2.3.31); "little loss" (The Spanish Tragedy 1.2.7) matches "little loss" (The Jew of Malta 1.2.22); "my loving brother" (The Spanish Tragedy 1.2.15) matches "my loving brother" (The Massacre at Paris 13.8); "met our armies" (The Spanish Tragedy 1.2.24) matches "meet | Our army" (2 Tamburlaine 2.3.42-43); and "full of hope and fear, | Both menacing" (The Spanish Tragedy 1.2.25) matches "full of brags, | And menace" (1 Tamburlaine 3.3.3-4). Freebury-Jones and Dahl's point is that using Taylor and Nance's method this chunk of The Spanish Tragedy that we all accept as Kyd's would seem to be by Marlowe. But in fact they are not applying Taylor and Nance's method, since the key question is not "can we find unique parallels between this part of Kyd and Marlowe's work?" but "does Marlowe's work provide the greatest number of unique parallels to this part of Kyd?"

    Without identifying the totals of the unique parallels they found, Freebury-Jones and Dahl give the impression that these five matches are the only unique parallels between these 173 words of Kyd and all the other plays of 1576-1594. But confining our attention to trigrams, it is easy to find that the phrase on each others in this part of Kyd uniquely matches "on each other" in Edward 3. If we turn to collocations too--as Freebury-Jones and Dahl do in letting "full of hope and fear, | Both menacing" in this part of Kyd match "full of brags, | And menace" in 1 Tamburlaine--then further matches are likely. Moreover, two of Freebury-Jones and Dahl's list of five unique matches quoted above are not in fact unique: little loss occurs not only in this part of Kyd and The Jew of Malta but also in King John at 2.1.307 and, as we saw above in the review of Freebury-Jones's article 'Augean Stables', my loving brother occurs not only in this part of Kyd and The Massacre at Paris but also twice in The Famous Victories of Henry 5.

    In other words, Freebury-Jones and Dahl's list of unique matches to this part of Kyd is not a list of unique matches, and it is unclear why they think it is. The important point here is that Freebury-Jones and Dahl offer no totals for what they have found--they just give imperfect lists that they present as if these contradict the tables of Taylor and Nance--so their approach is not truly quantitative. By supporting a few matches with "qualitative analysis" they imply that they have undermined Taylor and Nance's work. But the two approaches, the quantitative and the qualitative, are incommensurable.

    The same problems beset Freebury-Jones and Dahl's next list (p. 476). They claim to offer a list of unique matches between Soliman and Perseda 4.1.128-149 and all the plays from 1576-1594, purporting to show that Marlowe's works provide more links than Kyd's. They think call[s] on Christ appears only in that part of Kyd and The Massacre at Paris and Doctor Faustus but, as we saw above in the review of Freebury-Jones's article "Augean Stables", it also appears in 2 Troublesome Reign of King John. They think that a pardon for appears only in that part of Kyd and 2 Tamburlaine but (again, pointed out above) it also occurs in George a Greene, and also 1 The Troublesome Reign of King John, and 1 Edward 4.

    In one case, Freebury-Jones and Dahl seem to abandon the principle that although we allow variation in number and case and verb conjugation the matches have to consist of the same dictionary headword, since they claim that I grant it thee then in this part of Kyd matches "I will give it thee then" in 2 Tamburlaine (p. 475). The words grant and give are alike in many (but not all) of their senses, but they are not the same word. Investigators could in principle allow synonyms to stand for one another when identifying matches, but such a principle would have to be applied consistently across an entire study. Freebury-Jones and Dahl also treat as the same word the noun will (in the sense of conscious determination) and the verb will (in the sense of shall) to claim a match between "shall alter by my will" in this part of Kyd and "shall alter, as I hope they will" in The Spanish Tragedy 3.2.92. Whatever it is Freebury-Jones and Dahl think they are doing in constructing their lists, they are not replicating the microattribution method in Taylor and Nance's article.

    At this point (pp. 476-78), Freebury-Jones and Dahl switch to including "discontinuous matches" in their study, without indicating exactly what they mean by that term or how many words are allowed to appear between those that form the match. They give as a match to this part of The Spanish Tragedy (that is, 1.2.1-28) the words countenance . . . speak . . . thus which they cite as Doctor Faustus 1.2.22-23 but surely the last two words should be speak thus without ellipsis. Continuing to search for "discontinuous matches", Freebury-Jones and Dahl think that Soliman and Perseda 4.1.128-149 includes a unique match of maladies . . . eased with Doctor Faustus 1.1.22, but in fact the phrase also matches "I'll ease thy grief and cure thy malady" in Grim the Collier of Croydon which Literature Online dates to 1593, inside Freebury-Jones and Dahl's date-range.

    This last example raises the question, nowhere addressed by Freebury-Jones and Dahl, of whose authority to accept for such cases. The Database of Early English Playbooks (DEEP) generally incorporates the latest scholarship but is no use for Grim the Collier of Croydon because it was not printed until 1662, which is outside of DEEP's range. Alfred Harbage's Annals reckons that this is probably the same play as The Devil and his Dame of 1600, and Wiggins's Catalogue of British drama treats that link as a certainty. An investigator could justify going either of two ways on this problem of authority for dating: i) accepting all the date and authorship assignments in one's dataset, even those one personally disagrees with, or ii) refining the date and authorship assignments in one's dataset by reference to the latest scholarship where it reflects a consensus.

    A justification for accepting one's dataset's dates and attributions is that one cannot then be accused of rigging the evidence, since the dataset is likely to be randomly wrong rather than systematically wrong about the particular cases one is investigating. The justification for using the latter approach and correcting such errors as one can identify is that the dataset might in fact be systematically wrong and we should in any case work with the latest state of scholarly knowledge on our topic. It is not obvious which of these two lines of reasoning should prevail.

    From their collecting of matches, Freebury-Jones and Dahl conclude that the microattribution method is simply unreliable, but as the above counter-examples show their efforts in gathering data miss examples of just the kind of evidence that they consider crucial in reaching this conclusion. For that reason, nothing here constitutes a valid critique of what Taylor and Nance did. The main problem is that Freebury-Jones and Dahl do not tabulate their results in order to compare like with like.

    They next turn to Mueller's dataset Shakespeare His Contemporaries, for which they give an URL that at the time of review (October 2021) returns a "connection timeout" error. Freebury-Jones and Dahl quote at length Mueller's explanation of how he weighted the different kinds of matches one can find, so that long n-gram matches do not count for disproportionately more than short ones, so that matches to many plays count less than matches to few plays, so that function-word matches count less than lexical-word matches, and so that matches in long works count less than matches in short works. In the description quoted by Freebury-Jones and Dahl (pp. 478-79), Mueller does not try to defend by reference to any studies in Information Theory the precise forms of the various equations that he uses to make these weightings.

    One of the formulas is misquoted by Freebury-Jones and Dahl as "log bins" rather than "log(10) bins" and one of Mueller's typos, a dangling adjective "textual" that qualifies no noun, is reproduced verbatim by Freebury-Jones and Dahl: "I treated the aggregate word count of play pairs as a single textual, counted the links of a certain type and normalized their values to relative frequencies per 10,000 word" (p. 479). The reader is left with the impression that Freebury-Jones and Dahl do not understand what they are quoting from Mueller. They are clearly torn about whether to trust the conclusions that seem to arise from Mueller's tables, since his results suggest that ". . . Peele co-wrote the Tamburlaine plays, Lyly co-wrote Doctor Faustus, Fletcher co-wrote The Jew of Malta, and Munday co-wrote Edward II. We could also argue on this basis that Marlowe had a hand in Shakespeare’s Richard II and Richard III" (p. 479).

    Because the results of Mueller's calculations are so unwelcome to Freebury-Jones and Dahl, they suggest that perhaps Marlowe's influence on Shakespeare explains the matches. They make no attempt to demonstrate this to be the case. Instead, having explored Mueller's method (which is not the microattribution method they purport to be critiquing) and found that it leads to unacceptable conclusions they again entertain the proposition that quantitative methods just cannot distinguish imitation from collaboration. Freebury-Jones and Dahl turn to David Hoover's demonstration that long n-gram matches between texts are less indicative of shared authorship than short ones are, and their retort is merely to echo Vickers's claim that Hoover's analysis using nineteenth-century writing is inapplicable to early modern drama because the constraints of form that apply to the latter made the dramatists use a much smaller verbal range and hence verbal echoes between different writers' works are necessarily more frequent.

    Returning to Mueller's data, Freebury-Jones and Dahl find Vickers's claim borne out: ". . . the majority of Marlowe texts in his database are less like other Marlowe plays than those of different playwrights" (p. 485). That is one way to interpret the results, but another is to wonder if Mueller miscounts the matches or errs in the formulas he uses for weighting matches, since he makes no defence of them. But for Freebury-Jones and Dahl the only reasonable conclusion is that we need to return to qualitative analyses. Freebury-Jones and Dahl think that Taylor and Nance failed to follow Muriel St. Clare Byrne's warnings about how to handle evidence of shared phrasing since they allow a match between hand . . . executes in The Massacre at Paris and hands . . . executes in 3 Henry 6, although Freebury-Jones and Dahl do the same regarding number and verb endings throughout their article.

    Freebury-Jones and Dahl think that Taylor and Nance used too wide a window-of-interest when counting verbal matches, and now (p. 486) they suggest that four should be the maximum number of words allowed between matching terms, even though as we have seen their own evidence used windows wider than that. When the window is too wide, lots of matches are thrown up and we ". . . require some form of qualitative analysis in order to determine whether they are indicative of either a single author's idiolect or happenstance" (p. 486). So, Freebury-Jones and Dahl offer to do the literary-critical work that they claim is able to tell the difference between shared authorship and coincidence. They pick through the evidence previously referred to, using their literary-critical skills to determine whether the verbal parallels are significant for authorship. The whole thing is of course subjective, which is fine on its own terms but cannot stand as a critique of a quantitative approach such as that in Taylor and Nance's article to which they purport to be responding.

    Freebury-Jones and Dahl reflect on Dahl's use of a dataset of text that he compiled and then searched using the software called Pl@giarism, arguing that this method is not significantly different from Taylor and Nance's. They admit that the database Dahl used "did not include every text by some of the disputed authors and as such was limited" (p. 488). That is putting it mildly. Vickers used Dahl's searching of this dataset to deny Middleton's hand in Macbeth because a series of phrases in the supposedly Middletonian parts of Macbeth appear nowhere in the Middleton canon. As I was able to show in the New Oxford Shakespeare Authorship Companion, the phrases in question do in fact appear in the Middleton canon and the failure of Vickers and Dahl to find them was due not to their being absent from Middleton's writing but rather due either to their being absent from Dahl's dataset or their being present but not found by whatever method was used to search it.

    Freebury-Jones and Dahl object to my describing Vickers and Dahl's dataset as private in the sense of not being available to other investigators, claiming that it is the same as the online Korpus of Early Modern Playtexts in English (KEMPE). While it is true that this dataset is available for anyone to search online, one cannot see the full texts which provide the resulting hits (as one can in Literature Online used by Taylor and Nance) and if one wants to know what plays his claims are based on then "the list of all plays examined is provided in Dahl's thesis", which he says he is happy to share. However, Dahl has chosen not to let the British Library digitize his 2004 thesis and make it available in Open Access form via their EThOS database, which currently includes over half a million theses by other investigators.

    Freebury-Jones and Dahl end by describing Rizvi's Collocations and N-Grams online dataset and predicting that it will end the disputes about what the verbal matches between early modern plays consist of. This is a forlorn hope, since there are questionable assumptions in Rizvi's dataset, not the least of which is his assumption that it does not matter that the Shakespeare texts he uses, from the Folger Shakespeare Editions, were modernized by human editors (Paul Werstine and Barbara Mowat) whereas his non-Shakespearian texts came from Mueller's Shakespeare His Contemporaries project and were auto-modernized by computer software. Another questionable assumption for which Rizvi has so far offered no defence is his acceptance, for the purpose of identifying the canons which the matches fall into, of Mueller's attribution of 1, 2, 3 Henry 6, Henry 8, Pericles, The Two Noble Kinsmen, Timon of Athens, and Titus Andronicus to Shakespeare's sole-authorship, in the teeth of evidence to the contrary, much of it collected in Vickers's book Shakespeare, Co-Author (reviewed in YWES for 2002).

   It was a bumper year for authorship attribution studies. A special issue of the journal Authorship was devoted to matters of Shakespearian authorship and contained four articles relevant to this review. The first was an Introduction by Brian Vickers ('Introduction [to Special Issue On] Authorship Attribution and Elizabethan Drama: Qualitative Versus Quantitative Methods', Authorship 7.ii[2018] 1-10), which starts with a brief history of authorship attribution scholarship. Vickers then turns to the use of function words, which he thinks cannot help us attribute authorship because authors make conscious choices about them, varied by character. Specifically, he claims that ". . . the function words used by Hamlet are Shakespeare's choice for him, and . . . he made quite different choices for Ophelia . . ." (p. 3). We cannot test what Shakespeare was conscious of doing, of course, but we can test Vickers's claim that he differentiated these characters by their function-word use.

    In modern editions, Hamlet's part consists of 12,465 words, give or take a few depending on the edition chosen. Of these, 282 are occurrences of the function word I, the personal pronoun, and that is a frequency of 2.262% (one word in every 44). Ophelia's part consists of 1219 words (again, depending on the edition but not varying much) of which 29 are I, and that is a frequency of 2.38% (one word in every 42). If we take the function word of the same pattern of likeness emerges: Hamlet uses it 283 times, which is 2.27% of his part (one word in every 44) and Ophelia uses it 26 times, which is 2.13% of her part (one word in every 46). For these words spoken by these two characters, Shakespeare clearly did not make "quite different choices" as Vickers claims. Of course, authorship attribution tends to use not individual characters' parts but whole plays and at this larger scale the tendency of authors to stick to their idiosyncratically preferred frequencies of function words across many works has been repeatedly noticed and successfully used to identify authorship. (We addressed the idea that authors are conscious of their rates of funtion-word usage above in connection with Edward Pechter's article.)

    Next Vickers turns to an account of David Auerbach's essay in this special issue (reviewed below) and claims that it supports Vickers preference for qualitative over quantitative methods. The problem with the latter, Vickers claims, is that they reduce "the language of a complex literary artefact to a set of 'words in a bag', on the assumption that their frequencies of occurrence will provide a reliable authorship identification" (p. 4). In fact, this assumption is not made in reputable studies. Rather, the investigators will--and the ones being critiqued by Auerbach did--run experiments to measure their method's ability to correctly identify the authorship of sample texts where we happen to already know the authorship. That is, the reliability of quantitative methods is itself quantifiable and in all reputable studies it is calculated and reported. No one would use the quantitative methods that Vickers and Auerbach object to if these had not been shown to have knowable and acceptable rates of accuracy at answering the questions we put to them.

    Vickers offers his own qualitative responses to the findings of Brett Greatley-Hirsch and Jack Elliott that Auerbach critiques, using literary criticism to account for what they treat as quantitatively anomalous. To overcome anomalous counts, Vickers repeatedly insists that the numbers "must be interpreted locally, in terms of the interaction between characters" (p. 6). Next Vickers turns to his own method that uses plagiarism-detection software to find phrases in common between various texts derived from "an electronic corpus of all the plays performed in the London public theatres" (p. 7). Vickers does not mention what corpus he is referring to, and a recurrent failing of his applications of this method has been his reliance on a corpus created for him by Dahl that is not available for public inspection. As noted above, this matters when an investigator draws conclusions, as Vickers frequently does, from the discovery that a certain phrase is not in his corpus.

    A much better way to proceed is the subject of Vickers's closing remarks on Mueller's data Shakespeare His Contemporaries, which indeed is available online for everyone to examine and derives from the TCP's project to type up and make available to all investigators a substantial subset of the 120,000 early modern books represented by digitized microfilm images in the dataset EEBO. Vickers also champions Rizvi's creation of an online dataset, Collocations and N-grams, of 527 early modern plays in modernized form and his analysis of it to detect instances of verbal repetition between plays, which Vickers considers the strongest evidence for shared authorship.

    It is pleasant to be able to agree with Vickers that such online datasets as Mueller's and Rizvi's are the best way forward, because (although Vickers does not make this observation) the ability of other investigators to check the accuracy and completeness of the dataset is our best collective defence against error arising from inaccurate and incomplete data. Unfortunately, like Freebury-Jones, Vickers habitually supports his assertions with references to URLs that tend to be broken by the time they reach their readers. At the time of reviewing (October 2021), even Vickers's citation of his own writings on his own website, given in footnote 16, leads to a "Page not found" error. Likewise his citation of an article by Rizvi given in footnote 22 is broken.

    David Auerbach's contribution to this special issue of Authorship is largely a critique of the chapter "Arden of Faversham and the Print of Many" that Jack Elliott and Brett Greatley-Hirsch contributed to the New Oxford Shakespeare Authorship Companion edited by Gary Taylor and I and published by Oxford University Press in 2017 (''A Cannon's Burst Discharged against a Ruinated Wall': A Critique of Quantitative Methods in Shakespearean Authorial Attribution', Authorship 7.ii[2018] 1-16). The key finding of Elliott and Greatley-Hirsch was that when tested by a variety of computational methods, Arden of Faversham appears to be substantially by Shakespeare.

    Auerbach complains that when he asked Greatley-Hirsch for digital files of the numerical results summarized in the chapter and "the parametric constants used in the analysis" (p. 4) he was told they were not available. Auerbach's is a reasonable complaint, as the avowed position of the editors of the Authorship Companion and its publisher is that such data should be made available to anyone who wants to pursue a topic raised in the book. In this case, Elliott and Greatley-Hirsch ran their tests in 2013 and Auerbach asked for the data in 2018, by which time they had deleted it. As one of the editors of the book in which Elliott and Greatley-Hirsch published their findings, I take responsibility for the resulting inability to give Auerbach the data: it should have been retained.

    Auerbach confines himself to critiquing how Elliott and Greatley-Hirsch use the three methods called Delta, Nearest Shrunken Centroid, and Random Forests, and not their use of the method called Zeta as Rizvi has published a critique of that. Rizvi's critique appeared in the journal Digital Scholarship in the Humanities in 2019 and will be reviewed next year. Auerbach objects that when Elliott and Greatley-Hirsch ran tests using the 500 most-frequently-used words in their corpus, the 500th entry in that list, the word less, is actually quite rarely used, and that the most frequently used word and ". . . has hundreds of uses in each text and thousands across the entire corpus under consideration, two orders of magnitude greater than the count for 'less'" (p. 5).

    This is true, but it is hardly Elliott and Greatley-Hirsch's fault. Whenever we put words into a rank order of their frequency of use in any substantial corpus we find them subject to the power law discovered by George Zipf: the frequency at which a word occurs is inversely proportional to its rank. More colloquially, the frequency falls off really sharply after the first few words. Auerbach does not suggest that perhaps Elliott and Greatley-Hirsch should have used fewer than the top 500 words. He turns instead to some specific remarks about the limitations of the Delta method made by its inventor John Burrows and generalizes them to an implicit suggestion that the method is of no use at all in authorship attribution. A survey of the scholarly literature describing Delta's application in a number of cases would soon have shown Auerbach that this cannot be true: investigators working in unrelated genres and historical periods have repeatedly found Delta to be objectively successful in attributing authorship when it is validated using texts of known authorship in trial runs.

    Regarding the Random Forests tests run by Elliott and Greatley-Hirsch, Auerbach claims that the method is unusual "in being in principle non-replicable, because the test is non-deterministic" (p. 6). Auerbach thinks that the random element in the Random Forests procedure makes it non-deterministic, meaning that if one were to run the procedure twice one would not be guaranteed to get the same results. Strictly speaking this is true, but the way Auerbach formulates the point is misleading. I might run an experiment to see if a six-sided die is either fair or is loaded in the sense of preferentially showing one face more often than the others when rolled. The experiment is that I roll the die 6000 times and record the results. If each of the faces comes up about 1000 times then I would declare the die to be fair.

    If I pass the die to Auerbach and ask him to run the same experiment, he will be most unlikely to find the same result that I did on his first roll, and on his second roll, and so on up to his 6000th roll. Since each roll is random, his sequence of results will be quite unlike mine because the process is non-deterministic. But underlying that non-determinism is utter predictability, since if the die is truly fair he also ought to find that each face comes up about 1000 times in total over 6000 rolls. The fact that there is a random, non-deterministic element to the experiment does not mean that it cannot in principle be replicated, as Auerbach claims of the Random Forests method. Indeed, the randomness used in Random Forests is of exactly the type in my hypothetical die-rolling experiment, and it likewise serves to reveal an underlying pattern that is predictable and replicable.

    Auerbach thinks that he has spotted a suspicious anomaly in Elliott and Greatley-Hirsch's published results of their Random Forests experiment on Arden of Faversham. Since the method's error rate is reported by them as between 11.76% and 18.49%, why did it attribute all their 2000-word segments of Arden of Faversham to Shakespeare in all but one experiment? Even if the play were entirely by Shakespeare, Auerbach observes, an error rate of 10-20% would lead us to expect 3-7 of the 35 segments to be misclassified as being by someone else. Auerbach appears not to appreciate that the error rate was calculated from the results with 238 segments of known authorship and that we ought not to expect error to be evenly distributed across all the authors represented. In experiments of this kind it is common to find that error rates vary markedly by author, since some authorial styles are more objectively distinctive than others. This is not to say that the Random Forests method is right to give so much of Arden of Faversham to Shakespeare, only that the fact that it does is not of itself evidence that something is wrong with the experiment, as Auerbach claims.

    In his account of Elliott and Greatley-Hirsch's use of the Nearest Shrunken Centroid technique, Auerbach acknowledges that he cannot know exactly what they did (because their essay does not go into sufficient detail) so his critique "is based on the description of the method given" (p. 8) in the bio-statistics article in which its inventors first announced it. Much hinges on how Elliott and Greatley-Hirsch set the parameters that cause the method to discard frequency data for words that an author uses at wildly inconsistent rates in different places. Auerbach admits that in the absence of this knowledge ". . . any attempts to identify decisive words are speculative . . ." (p. 9). I would say that this is an understatement and his ensuing critique is so speculative about what Elliott and Greatley-Hirsch did as to be meaningless.

    I would nevertheless wholeheartedly endorse the sentiment of Auebach's somewhat inelegantly phrased assertion that ". . . more than simply producing results, such large-scale quantitative analyses must also explicitly produce their criteria by which a classification was made, and this information should accompany any publication of the results" (p. 12). As editor of the Authorship Companion in which Elliott and Greatley-Hirsch's essay appeared, I ought to have done more to ensure this happened, by putting the necessary data in the book's Datasets section (pp. 603-73), although this section was already rather longer than the publisher wanted, or in an online supplement.

    Auerbach reckons that ". . . we know far more about how words relate to each other than how genes do" (p. 13) and hence the fact that Elliott and Greatly-Hirsch got error rates higher than those found when Nearest Shrunken Centroid is applied to genes ought to have warned them off their approach. I do not accept the premise of this conclusion. I believe our understanding of language to be at least as incomplete as our understanding of genetics, but even if we accept his premise Auerbach's conclusion would not follow. We do not need the application of a technique in a new domain to be as successful as its original application in its home domain in order to extract utility from it. Indeed, Auerbach's argument could easily serve as a general prohibition on inter-disciplinary research in general, since quite often the first attempts at cross-fertilization are accompanied by a drop in effectiveness.

    In his final remarks, Auerbach takes Elliott and Greatley-Hirsch's essay as symptomatic of wider failures across the field authorship of attribution by computational means, and offers his own principles for "Rebuilding the Foundation" of the field. Who could disagree that "a greater degree of tentative humility" (p. 14) ought to be shown by investigators? He is entitled to "wonder whether quantitative lexical analyses can ever gain the level of certainty required" (p. 15), but would find that familiarizing himself with the extensive prior scholarship of quantitative lexical analysis (none of which he mentions in his critique) goes some way towards satisfying his curiosity. He betrays his ignorance of the state of the art most starkly in making the suggestion that investigators might in future wish to consider such things as "Sentence lengths" and "Word adjacencies" (p. 15). If only, he laments, our source texts "were to be consistently and thoroughly annotated with markers for parts of speech, word etymologies, metaphors and metaphrands, imagery, other rhetorical figures, computational analyses could take such metadata into account" (p. 16). These things are, of course, currently being attempted in a number of ongoing projects.

    The purpose of Darren Freebury-Jones's contribution to this special issue of the journal Authorship is to establish that the play Arden of Faversham was written solely by Kyd ('In Defence of Kyd: Evaluating the Claim for Shakespeare's Part Authorship of Arden of Faversham', Authorship 7.ii[2018] 1-14). A reader keeping up with Freebury-Jones's publications will feel déjà vu when he starts by describing what MacDonald P. Jackson did in his book Determining the Shakespeare Canon (reviewed in YWES for 2014). As with his other recent essays, Freebury-Jones's source for evidence of Kyd's authorship of Arden of Faversham is n-grams shared between plays as revealed by Mueller's Shakespeare His Contemporaries dataset.

    In his footnote 18, Freebury-Jones reports that Mueller's dataset has since been "renamed Early Modern Print" and points the reader to the Early Print project at Washington University in St Louis. In fact, the "About" section of the Early Print website at the URL that Freebury-Jones provides puts the matter rather differently: ". . . the Shakespeare His Contemporaries Project [was] a pilot effort now subsumed within EarlyPrint". Mueller's own description of Shakespeare His Contemporaries, at the URL that Freebury-Jones cites in his footnote 19, is that "Pervez Rizvi prepared a much fuller and very careful survey of shared n-grams in Early Modern Drama at . . .", upon which Mueller gives an URL for Rizvi's Collocations and N-Grams dataset.

    Because Wiggins in British Drama 1533-1642 gives a date of 1590 for Arden of Faversham, Freebury-Jones concludes that the play "antedated the whole of Shakespeare’s corpus", hence "that the Shakespeare matches with Arden of Faversham are indicative of Kyd's influence on him, rather than Kyd's authorship" (p. 3). Unless Freebury-Jones means those last three words to be "than Shakespeare's authorship" I cannot follow the point he is making. Wiggins's is not the only chronology available, of course. Alan Farmer and Zachary Lesser's DEEP assigns the first performance of Arden of Faversham to the date-range 1588-1592 and the New Oxford Shakespeare gives the range as 1587-1592. But even if 1590 is correct for Arden of Faversham, it is entirely possible that Shakespeare had already written The Two Gentlemen of Verona and Titus Andronicus by then, whether we use the DEEP or New Oxford Shakespeare chronology.

    Next Freebury-Jones turns to the rare-phrase matches between Kyd's plays and the Quarrel Scene in Arden of Faversham, as reported in his essay "A Raven for a Dove" (reviewed in NYWES for 2016). He tries to explain the matches between Arden of Faversham and Robert Yarington's play Two Lamentable Tragedies by the fact that Yarington was apprenticed to Thomas Kyd's father, the scrivener Francis Kyd. Thus ". . . Yarington may have had access to Kyd's manuscripts and could have been involved in making fair copies of his plays" (p. 4). Indeed he may, though why Kyd would give his father his manuscripts is not explained.

    Freebury-Jones repeatedly asserts that Jackson overlooked matches between Arden of Faversham and Kyd's works but does not say what they are, so presumably he is still recycling the claims of "A Raven for a Dove". Then Freebury-Jones turns back to a blog posting in which Mueller reported finding a disproportionate number of links between Arden of Faversham and Kyd's work. As is usual with Freebury-Jones's writing, the URL for this blog posting given by Freebury-Jones (in his footnote 32) produces a "Page Not Found" error at the time of writing this review in October 2021. The same thing happens with the URL given in Freebury-Jones's footnote 33, and since footnotes 34 to 39 simply point back to these earlier two footnotes the entire scholarly apparatus in support of this section of Freebury-Jones's essay is missing, making it impossible to follow up his claims. Such weakness of reference would be penalized in the marking of an undergraduate essay.

    Freebury-Jones is on more solid ground when he turns to Rizvi's online dataset called Collocations and N-grams, which is at least at the website that Freebury-Jones gives in his footnote 43. But when Freebury-Jones reports that Rizvi's texts are "modernized and lemmatized texts--drawn from Mueller's corpus and the Folger Shakespeare Editions website" (p. 5) he omits to mention that Rizvi drew on the former for all the non-Shakespearian plays and the latter for all the Shakespearian ones. The different provenance is not necessarily trivial since, as mentioned earlier, in Rizvi's corpus the modernizing of the non-Shakesperian plays was done by machine while the Shakespeare plays were modernized by human editors. This introduces into Rizvi's raw data a systematic difference running along authorial lines. It would be reassuring to see a demonstration that this matter of provenance does not affect the results when Rizvi's data are used to distinguish authorship.

    Freebury-Jones describes experiments that Rizvi did with his data for which the supporting footnote cites an URL containing more than 60 pseudo-random characters (of the kind "79964A7BF09"), and the reader who has carefully and accurately typed them will be disappointed to receive another "Page Not Found" error. The file that is meant to be provided at this location appears to reside on the Microsoft OneDrive online service, which is notorious for the way its frequent reorganizations change the URL that must be cited. Freebury-Jones claims that the evidence in this file "counters Jackson's claim" (p. 6) that the long-accepted Kyd plays (The Spanish Tragedy, Soliman and Perseda, and Cornelia) are stylistically distinct from the ones that Vickers and others including Freebury-Jones have recently attributed to Kyd: Fair Em, 1 Henry VI, Arden of Faversham, and King Leir. It is impossible to evaluate this claim since the evidence is not to be found at the URL that Freebury-Jones gives.

    Next Freebury-Jones reports the findings of his essay "Exploring Co-Authorship in 2 Henry VI" that was reviewed in NYWES for 2016, of which the central argument was that finding phrases in common between different parts of a play that other investigators think were written by different authors means these investigators are wrong and the play was sole-authored. For Freebury-Jones, sole-authorship is proved by such phrases that appear across a play, even if (as was shown in the NYWES review) the phrases he finds are commonplace..

    Relying on plagiarism-detection software to find phrases common to different texts, Freebury-Jones misses relevant evidence. For example, he claims that the phrase to let thee know "cannot be found in Shakespeare's entire dramatic corpus" (p. 7). Since you and thee are different forms of the same word, the second person singular pronoun, and since scribes and compositors are known to have altered one form to the other when transmitting a text, we should count as occurrences of this phrase its appearance as "To let you know of it" (Hamlet) and "To let you know my thoughts" (Othello), and perhaps its close relatives "I will let you know" (Julius Caesar) and "If you will let your lady know" (Twelfth Night). Arguments can be made for discounting such matches, of course, but Freebury-Jones does not make them and seems unaware of this Shakespearian evidence.

    Freebury-Jones objects to particular claims in Jackson's book Determining the Shakespeare Canon, mainly using arguments of the form "Jackson does not mention . . ." (p. 8) and "Jackson does not acknowledge . . ." (p. 9), referring to studies such as those by Marina Tarlinskaja and Timberlake that Freebury-Jones prefers to Jackson's. As with his other studies, Freebury-Jones finds that consistencies across Arden of Faversham, such as the rate at which verse lines have feminine endings, point to sole-authorship rather than co-authorship. In Freebury-Jones's data on page 10, the rate at which this prosodic feature is used in a scene of as few as 11 lines is treated as equally significant as the rates derived from scenes of as many as 607 lines. A short scene is of course unlikely to have a rate representative of a writer's general habit, so averaging across scenes (as Freebury-Jones does) is meaningless when the scenes show rates varying from 0.9% of lines to 28.5%.

    Regarding pause patterns as counted by Oras, Freebury-Jones reprints the tabular data from his essay "Augean Stables" reviewed above, showing that those of Arden of Faversham are like those found in Kyd's plays. The objections to this claim given above apply again here. Freebury-Jones's next table is also a reprint of one in "Augean Stables", and the editors of the journals Authorhip and Archiv für das Studium der neueren Sprachen und Literaturen and Authorship in which these essays appeared have a right to be irritated about this duplication. Presumably Freebury-Jones does not have to worry about the United Kingdom's national research audit, the Research Excellence Framework (REF), which takes a dim view of such duplication between assessed publications. The duplication runs to more than just the tables, for an entire passage of 153 words from "the percentages for . . ."  to ". . . respect, except Cornelia" (pp. 11-12) is identical to one in the essay "Augean Stables" (pp. 77).

    Freebury-Jones's last witness in support of Kyd's authorship of Arden of Faversham is Lene Petersen's book Shakespeare's Errant Texts, reviewed in YWES for 2010. Without mentioning what Petersen counted, Freebury-Jones reports that when the counts were processed by Discriminant Analysis and Principal Component Analysis she found that ". . . Arden of Faversham cross-validates as Kyd . . ." (p. 12). Freebury-Jones evidently does not understand the concept of cross-validation that Petersen is describing at this point in her book, or he would not quote this sentence of hers, since it ends ". . . while [The Two Noble] Kinsmen registers as Marlowe" (Petersen, p. 214). These are not attributions by Petersen, but part of a discussion of the limitations of her methodology. Freebury-Jones ends his essay by reporting on an essay by Rizvi, "The Interpretation of Zeta Test Results", which will be noticed in this review next year.

    The last essay to be considered from this special issue of Authorship is by Brian Vickers who claims that Thomas Dekker, rather than Shakespeare, wrote "To the Queen", also called "As the dial hand tells o'er" ('The 'Dial Hand' Epilogue: By Shakespeare, or Dekker?', Authorship 7.ii[2018] 1-18). The New Oxford Shakespeare's attribution to Shakespeare of this short text, which might be a prayer or an epilogue to a play, was based on an essay by John V. Nance reviewed in NYWES for 2016. Vickers relates the known facts about the prayer/epilogue and then focusses on his two key pieces of evidence: "the verse form (trochaic tetrameter couplets), and the optative verb 'may', which is not distinctive in itself but gives an important clue to the poem's genre" (p. 3). Vickers compares the prayer/epilogue's "little rhythmic emphasis on the first and last syllables of each line" (p. 4) with Jonson's use of trochaic couplets in a masque "with its emphases on the first, third, fifth and seventh syllables" (p. 4). Then Vickers compares the prayer/epilogue to some trochaic verse by Dekker (of which there is not much), which keeps slipping back into iambic pentameter as does the prayer/epilogue.

    Next Vickers turns to the problem that the prayer/epilogue does not really have any epilogue-like features: it does not refer to the play just shown and it does not ask for the audience's indulgent acceptance of it. Prayers to the monarch typically make use of the word may to express a wish, as indeed does the British National Anthem: "Long may she reign". So does "To the Queen" and a similar end-of-play prayer by Dekker, the "Epilogue at Court" to his Old Fortunatus, which Vickers quotes. Vickers traces some thematic connections between the use of circularity in relation to human mortality in "To the Queen" and in Dekker's writing, and finds a previously unobserved occurrence of "dials-hand" in Dekker's play The Wonder of the Kingdom. He finds other verbal parallels, such as empress and grave between "To the Queen" and Dekker's writing.

    Vickers turns to Nance's quantitative method for assigning "To the Queen" to Shakespeare and points to "several critiques" that "have shown its limitations" (p. 16). There are three: the one by Freebury-Jones and Dahl ("The Limitations") reviewed above, and two by Rivzi that will be reviewed here next year. Vickers confines himself to objecting that the verbal parallels found by Nance are not valid evidence because when found in "To the Queen" they have "a quite different sense" (p. 17) from their use by Shakespeare, and "the total meaning is completely different" (p. 18). Even more subjectively, Vickers finds that a parallel with Othello "relates to a completely alien world" (p. 18). For Vickers, these objections make the fact that Shakespeare's writings supply the most verbal parallels with "To the Queen" altogether nugatory because "they turn out to have completely different meanings or referents" (p. 18). Once these parallels are dismissed, everything points "not to Shakespeare, but to Thomas Dekker" (p. 18).

    The journal Digital Scholarship in the Humanities published four articles on our topic in 2018. One is by a team including the present reviewer and for that reason it cannot be reviewed here in detail (Mark Eisen, Alejandro Ribeiro, Santiago Segarra and Gabriel Egan 'Stylometric Analysis of Early Modern English Plays', DSH 33[2018] 500-28). The article applies the Word Adjacency Network method of authorship to a series of plays by Jonson, Middleton, Chapman, Marlowe, Shakespeare, John Fletcher, Francis Beaumont, and Massinger. The most surprising of the articles in this journal are two by Hartmut Ilsemann, one claiming that Samuel Rowley is the main author of the play Sir Thomas More and the other that Marlowe did not write most of the plays we atttribute to him.

    In the first article, Ilsemann attempts to overturn the conventional wisdom that Sir Thomas More was written by Anthony Munday and Henry Chettle, with later revisions by Heywood, Dekker, and Shakespeare ('More News on Sir Thomas More', DSH 33[2018] 46-58). Ilsemann assumes that the reader understands the Rolling Delta method for authorship attribution and points those who do not to an explanation in the form of an article by Maciej Eder, Jan Rybicki, and Mike Kestemon that he says appeared in 2016 in the journal of the R software package and is available online. At the time of writing (October 2021), the URL supplied by Ilsemann resolves to that journal's list of accepted articles, but Eder et al.'s (with the title 'Stylometry with R: A Package for Computational Text Analysis') is not among them. The correct online location for this article can be found by putting its title into a web-searching engine.

    Ilsemann goes into the detail of the culling feature in the R Stylo software package that he uses, which is where one or more of the features being looked for in a text (say, the presence of a certain word) is excluded from the final results if that feature is not found in at least a certain proportion (a user-defined threshold) of the samples of text being studied. Ilsemann gives a list of the texts that he used as samples of each candidate's sole-authored and securely-attributed works and for authors with substantial canons the samples are needlessly small. In Shakespeare's case, for instance, we have at least a couple of dozen sole-authored and securely attributed plays but in Ilsemann's experiments Shakespeare's writing is represented by just Richard 3 and 1 Henry 4. Since we know that history plays are a particularly distinct genre--tending to be less authorially individuated than comedies or tragedies--this is an unusual decision and Ilsemann offers no explanation for it.

    The article's Table 1 has the caption "Sequential attributions with mf1w, mf2c, and mf3c", and those three abbreviations are explained in a footnote: they decode as most frequent single word, most frequent pair of successive characters, and most frequent three successive characters, respectively. But there is no narrative to help the reader make sense of the content of the table. For instance, the column headings ("window", "sizes", "words") are cryptic and the body of the table uses typographical styling of italicization and boldface the significance of which is nowhere explained. Where Ilsemann attempts explanations of his experiments, I am unable to grasp what is meant by, for instance, the following two sentences: "It may be seen as a constraint that the returned delta values depend also on the number of reference texts involved, and when authors can be discarded, tests with a small number of reference texts should be chosen. In this case, however, the various hands in the additions could not be left out, and heuristic initial tests indicated that 1 Henry IV bores on the assessment of additions 1-6" (p. 49). I suspect that bores in the second sentence should be bears, but making that substitution does not clarify how 1 Henry 4 bears on the assessment of the Additions to Sir Thomas More.

    I have similar trouble making sense of Figure 2: although the x-axis is labelled "window size in words" with values running from 2250 on the left to 19,500 on the right, the narrative of the experiment reports that "The following two charts return, for example, attributions for window sizes of 4,000 words. Whereas in Fig. 1 word frequencies were investigated, Fig. 2 recalls mf3c attributions" (p. 49). Ilsemann concludes from these inscrutable visualizations that ". . . Chettle, Dekker, and Heywood are most unlikely as contributors to the original text of the play" (p. 49), which of course runs counter to the widely accepted authorial attributions. If I am understanding correctly the significance of the preponderance of the letters "R" and "D" in Table 1, Ilsemann's finding is that Samuel Rowley and Shakespeare wrote the original version of Sir Thomas More, and the preponderance of "D" and "W" in that table shows that the Additions were mostly likely by Shakespeare and Webster. (Actually, Ilsemann later decides that the results for the Additions are unreliable since supplementary tests prove contradictory.)

    Whatever the actual merits of Ilsemann's experiments--on which I withhold judgement because I cannot fathom what he did--I doubt that his conclusions will attract many adherents so long as his method remains as poorly explained as it is here. Ilsemann describes at length what happens to the results when certain variables (such as window size) in R Stylo are tweaked, and then the effect of switching from Rolling Delta to Support Vector Machines and Nearest Shrunken Centroid as the classifiers, neither of which is explained. He remarks that "One might well conjecture at this point that the various mathematical kernels of the classifiers have a mind of their own and will yield different results in the respective contexts of the variables" (p. 53), which indeed anticipates a thought that this reviewer was about to record. But for Ilsemann the right approach is to consider the "overall tendency of attributions . . . [that] can be deduced from the majority decision" (p. 53). Ilsemann ends with visualizations in the form a  "Bootstrap Consensus Tree" and a "Cluster Analysis" that he claims show support for Samuel Rowley's authorship of Sir Thomas More but he does not explain how these visualizations were made. Ilsemann mentions in passing that his other experiments have shown that the first four acts of Thomas of Woodstock were written by Samuel Rowley and the last by Shakespeare (pp. 56-7). Rowley's authorship of the whole of Thomas Woodstock was convincingly demonstrated in an article by MacDonald P. Jackson reviewed in YWES for 2009.

    In the second article, Ilsemann applies the Rolling Delta method to problems of authorial attribution with special attention to the works of Christopher Marlowe ('Christopher Marlowe: Hype and Hoax', DSH 33[2018] 788-820). He starts with a caution: when he used to believe that George Peele wrote the play Edward the First, Ilsemann's results showed that Peele wrote many more plays as well, but now that he excludes Edward 1 from the set of securely attributed plays of Peele, the other plays cannot be attributed to Peele either. (The reason most people think that Edward 1 is by Peele is that the words "Yours. By George Peele Maister of Artes in Oxenforde. Finis" appear at the end of the first edition.)

    Ilsemann thinks that the problem he found with Peele besets studies of Marlowe's authorship and that the reason the New Oxford Shakespeare project found so much evidence of Marlowe's hand in the Henry 6 plays of Shakespeare is that we have all been treating non-Marlovian works as if they were by Marlowe. Using essentially the same methods as his other article this year, Ilsemann examines first 2 Tamburlaine and his Table One shows that it all looks like Marlowe's work. As previously, Ilsemann's set of the sole-authored, well-attributed works of the author under investigation is much smaller than that used by other investigators. The legend to Table One indicates that for this first experiment he treated 1 Tamburlaine as the only representative of Marlowe's works. The obvious objection is that there are at least four other plays--Edward 2, Doctor Faustus, The Jew of Malta, and The Massacre at Paris--that virtually everyone agrees are also by Marlowe and could have been used as a more representative sample of his style. Ilsemann's aim, however, is to overturn the consensus that these plays are Marlowe's.

    It is not surprising to find that 2 Tamburlaine tests much like 1 Tamburlaine and if one is being radically sceptical, as Ilsemann seems to want to be, then there is as much reason (namely little) to doubt Marlowe's authorship of 1 Tamburlaine as to doubt his authorship of 2 Tamburlaine. Having established that 2 Tamburlaine is similar to 1 Tamburlaine, Ilsemann remarks that "The same applies vice versa . . ." (p. 791), and I cannot see any logical possibility of finding otherwise, given what we mean by the notion of similarity. Next Ilsemann turns to the play Locrine, published in 1595 as "Newly set foorth, ouerseene and corrected, By W. S.", which must be an attempt to imply that it is by Shakespeare without actually stating so. In his Table Two, Ilsemann finds it entirely by Marlowe.

    The play Dido Queen of Carthage was published in an edition of 1594 that describes it as "Written by Christopher Marlowe, and Thomas Nash. Gent.", but Ilsemann finds Marlowe present only at the beginning and only with some of the variables at his disposal--the length of the n-grams being sought and the size of the window used for rolling--set to certain values. Overwhelmingly, in  Ilsemann's Table Three Dido Queen of Carthage comes out as the work of Kyd, with Nashe substantially present only when we look for most-frequent character bigrams. The Massacre at Paris was published in an undated edition (probably of 1594) that has "Written by Christopher Marlow" on its title page, but Ilsemann's Table Four finds it entirely by Samuel Rowley, unless we count just the most frequent single words, in which case it becomes a collaboration of Marlowe and Rowley. Ilsemann remarks that "It is obvious that these findings do not correspond to conjectures by traditional scholarship which attributes the play erroneously to Marlowe" (p. 795).

    Table Five shows as the results for The Jew of Malta--whose first edition's title page reads "Written by CHRISTOPHER MARLO."--that the predominant authors are Kyd, Nashe, and Shakespeare. As Ilsemann rightly remarks, The Jew of Malta has long been a problematic case as various tests by different investigators have shown its dissimilarity to known Marlowe plays. Possibly part of the problem lies in its first edition being published in 1633, more than 40 years after its likely premiere. Next it is Doctor Faustus--the first edition in 1604 has "Written by Ch. Marl." on the title page--and Ilsemann's Table Six gives the first version of the play mainly to Nashe and Shakespeare and Kyd if we look at most-frequent character bigrams or Shakespeare and Kyd if we look at most-frequent character trigrams, or mainly to Kyd and Henry Chettle if we look at most-frequent single words. The revised version of the play, first printed in 1616 ("Written by Ch. Marklin"), is attributed in Table Seven mostly to Nashe, Samuel Rowley, and Shakespeare (by most-frequent bigrams), or Kyd, Nashe, Samuel Rowley, and Shakespeare (by most-frequent trigrams), or Kyd, Greene, and Samuel Rowley (by most-frequent single words analysis).

    Table Eight attributes Edward 2 (first published in 1594 as "Written by Chri. Marlow Gent.") either mainly to Samuel Rowley (according to most-frequent bigrams) or mainly Shakespeare and Samuel Rowley (according to most-frequent trigrams) or about equally Shakespeare and Marlowe (according to most-frequent words). Ilsemann reflects that ". . . one might be tempted to think that Rolling Delta mirrors all sorts of things, but bears no relation to acknowledged facts and to the authorship attributions of secondary literature" (p. 803). I should think that the greatest obstacle to the acceptance of Ilsemann's new attributions will be that they require a rather large number of early printed play books to have false authorship attributions. Title pages do sometimes lie, of course, but the scale of the deception about Marlowe that Ilsemann requires us to swallow is dwarved only by the scale of deception that the anti-Stratfordians posit about Shakespeare.

    Four more plays complete the analysis. Ilsemann gives the anonymously published Edward 3 mainly to Kyd and Marlowe (Table 9), with maybe some Greene and Nashe. Ilsemann revives a theory previously expounded by Thomas Merriam (reviewed in YWES for 2011) that Shakespeare's Henry 5 has some Marlowe in it because Shakespeare took over an existing Marlowe play on this topic. Ilsemann finds that this earlier play was not The Famous Victories of Henry 5 and that Peele and Nashe contributed to Henry 5 (Tables Ten and Eleven).

    That Marlowe's writing is present in some plays that were long thought to be solely Shakespeare's has recently been claimed by the present reviewer and others working on the New Oxford Shakespeare, which credits Marlowe as co-author on the three Henry 6 plays. Ilsemann thinks that the New Oxford Shakespeare investigators are wrong about this, having been misled by their use of the traditional Marlowe canon as their baseline for his style. According to Ilsemann, 3 Henry 6 was actually written by Shakespeare and Kyd (Table Twelve). Ilsemann's last attribution is King John as it appears in the 1623 Shakespeare Folio (Table Thirteen), which is "unambiguously a Shakespeare play" (p. 812) with small signs of Marlowe, Peele, and Greene too.

    Our last article from Digital Scholarship in the Humanities is by the late John Burrows and introduces a new way of processing word-frequency data using Spearman Correlation, which appears to enable authorship attribution using sample sizes considerably smaller than was previously thought the lowest threshold, down to 500 words samples from larger works ('Rho-grams and Rho-sets: Significant Links in the Web of Words', DSH 33[2018] 724-47). At the core of this claim is how we count things. We can count all uses of the words from or ball as essentially one thing because although they vary in sense in different contexts they do not vary much. But some common words such as to and that have so many varied meanings that we ought not to count them as all essentially one thing: we need to distinguish the sense being used in each occurrence.

    As an example, Burrows counts the occurrences of hath in 28 sole-authored well-attributed Shakespeare plays and 16 sole-authored well-attributed Middleton plays and shows that it occurs much more often in the former than the latter. He then sketches how an investigator might combine the counts for several words using Principal Component Analysis, noting the drawback that for short texts (fewer than 2,000 words) the sharp divisions start to disappear quite rapidly. The problem with n-grams is that words can be grammatically linked while sitting far apart in a sentence, as with the I am n-gram in "I, who offer myself as . . . X . . . and who have served you faithfully as Y and Z, am now . . .". Conversely, it is so and so it is and it so is have the same in meaning but form three distinct n-grams.

    Moreover, some words' presences correlate not with others' presences but with others' absences, which we can capture with a correlation matrix. So, for 200 words or phrases we can have a 200 × 200 table and put in each cell the Spearman correlation which ranges from 1 where the word at that row always appears with the word at that column, through 0 where the two words show no positive or negative correlation, down to -1 for words where the presence of one coincides with the complete absence of the other. Such a pair of words, with maximal (towards 1.0 correlation) or minimal (towards -1.0 correlation), Burrows calls a rho-gram. Burrows makes such a collection of rho-grams for how two sets of Shakespeare plays and one of Middleton plays use the word hath, looking for words that correlate either especially positively or especially negatively with it.

    A feature of Burrows's method that I do not understand is his calculation, for each play's rate of using the word hath shown in Table One, of what he calls the "Fraction" for that play. His description is "The set fractions of the last column allow equitable comparisons between the scores for high-frequency and low-frequency words. Each of them represents the proportion that its score occupies as a decimal fraction of the total score for that variable" (p. 728). The total for the "Percentage" column in Table One for the 28 Shakespeare plays is 6.482, so from this description I would expect the "Fraction" for the first play, All's Well that Ends Well, to be its "Percentage" of occurrences of hath, which is 0.228, divided by this total for all the percentages of 6.482, which I calculate to be 0.035 (rounded to three decimal places). But Table One has "0.033" as the "Fraction" for All's Well that Ends Well. For the next play, Antony and Cleopatra, the hath percentage is 0.178 so I would expect the "Fraction" to be 0.178 divided by 6.482 which is 0.027 but Table One has it as "0.026". And so on down the table: my calculations differ from Burrows's by just enough to make me think that I have not understood exactly what he is doing.

    Curiously, the total of the "Fraction" column in Table One is 0.951 ("0.950" according to Burrows) yet from the definition of the calculation of this column--"the proportion that its score occupies as a decimal fraction of the total score for that variable"--I would expect this column to necessarily sum to 1. If in each case I multiply my "Fraction" calculation by 0.951, I get the same "Fraction" figures as Burrows. Either there is an error in Burrows's Table One or, more likely, there is a subtlety to its calculations that I have failed to grasp from the description. These "set fractions" (p. 732) feature in what follows, being used to select Burrows's rho-sets (p. 732), and he explores whether these can help in the analysis of texts shorter than the usual 2000-word threshold below which the accuracy falls off sharply with other attribution methods.

    Burrows took the rho-sets for the 40 most-frequent words and ran a t-test to see which were the best at discriminating Shakespeare from Marlowe, leaving him with hath, doth, and, thus, and did as the favoured-by-Shakespeare word and has, all, now, never, and a as those favoured by Middleton. For each of these 10 words the method gave him two rho-sets: the words that appear most often when each of these is present, and the words that appear least often when each of these is present. The most discriminating of these, again derived using set fractions in a way I do not understand, gave Burrows "an overall frequency profile" (p. 736) of words favoured by Shakespeare and another of words favoured by Middleton.

    Applied to 1000-word segments of Macbeth, The Revenger's Tragedy, and Hengist King of Kent, these two rho-sets gave Burrows a successful Shakespeare/Middleton classifier for such small samples of writing. Doing this all again but retaining the 10 best rho-sets gave Burrows a successful Shakespeare/Middleton classifier able to distinguish 500-word samples, yielding "an unprecedented level of accuracy for such short texts" (p. 738). Burrows acknowledges that attributing 500-words segments from one play in this way might be giving better results than we should expect when attributing standalone 500-word segments such as short poems, since the segments from one play are likely to be similar to one another, whereas an author's standalone poems might well not be. Glancing at the question of whether this corroborates Taylor's claim that parts of Macbeth were written by Middleton, Burrows will not go that far.

    Chopping a play into 1000-word or 500-word segments necessarily cuts across scene boundaries, and Burrows notes that his method attributes the sleep-walking scene in Macbeth to Middleton, when of course no one has doubted its Shakespearian authorship: "These rho-tests are clearly sensitive to stylistic difference, but they do not, of themselves, interpret it" (p. 740).  Burrows summarizes his algorithm in an appendix but it does not clear up my uncertainty mentioned above about the notion and calculation of set fractions.

    Pervez Rizvi published this year a critique of the Word Adjacency Network method of authorship attribution used by the present reviewer and his collaborators ('Authorship Attribution for Early Modern Plays Using Function Word Adjacency Networks: A Critical View', ANQ: A Quarterly Journal of Short Articles, Notes and Reviews 33[2020] 328-31). Their raw frequencies are not the only aspects of function words that turn out to be distinctive of authorship. Also distinctive are the habits of putting certain function words near to other function words, so that a writer may prefer to use and near to in more often than other writers while avoiding putting the near to of. It seems likely that such habits are largely unconscious, but whether they are or not is irrelevant to the use we can make of them for authorship attribution.

    The Word Adjacency Network method captures the data about the proximities of large numbers of function words, one to another, with a comprehensiveness that no previous method has attempted. It succeeds in large part because it uses a data structure called a Markov chain to hold the data representing each author’s preferences and uses methods from Information Theory (principally relative entropy) to compare the preferences of different author with regard to these proximitie. We claim a successful attribution rate of 89.6% to 93.6%, in the sense that when the Word Adjacency Network method is applied in repeated trials to early modern plays for which the authorship is already known, the method points to the writer already believed to be the author in 89.6% to 93.6% percent of the cases. The range arises from the method’s sensitivity to the size of the sample it is given: the more text it has to work with, the more reliable the method.

    Rizvi's first objection is that the Word Adjacency Network method attends only to function words. "The method", he writes, "takes no account of an author's choice of non-function words, still less their meaning. It regards as indistinguishable the phrases 'O, for a muse of fire' (Henry V) and 'O, for a couple of faggots' (Fletcher and Massinger's The Little French Lawyer), since they use the same function words and in the same places" (p. 328). Rizvi is right about this.

    What if it is invalid to consider function words in isolation, because the function words chosen by an author are shaped by the lexical words chosen? This is Rizvi's second objection to the Word Adjacency Network method, and to pursue it he undertakes computational investigations of the same plays we considered:

The simplest search is to look for cases where the same function word always follows a non-function word. For example, upside is listed in the concordance as occurring 23 times, and it is followed every time by down. Or consider the word looker and its plural lookers, which occur 48 times and are followed by on every time. Similarly, devoid occurs 24 times and is followed every time by of. The obvious inference is that the non-function word influenced the choice of the function word that followed it . . . If it is true, as both our intuition and the textual evidence suggest, that non-function words exercise an influence over the function words that follow them, then the premise of the method I am considering [that is, the Word Adjacency Network method] becomes highly suspect. (pp. 328-29)

Rizvi here makes a subtle but fundamental and false logical leap. He assumes that since our method does not take into account non-function words, we must believe that these words do not affect the choice of function words and are irrelevant to studies of authorship. We do not believe that.

    The fact that a given inference method excludes a feature is not an explicit claim that this feature lacks explanatory power or correlation with the included features. To provide an example, imagine that we were trying to estimate the probability of hurricanes occurring in Texas. A successful method might be based on daily values of temperature, pressure, and rainfall. The fact that this method excludes humidity values does not imply that we think that the probability of hurricanes is unaffected by humidity, or that humidity does not affect rainful, which is a value we include. (Humidity clearly does affect rainfall.) Satisfactory performance of the method would imply that there is enough information in the considered features to estimate the probability of hurricanes. Non-function words (excluded from our model) do have information about authorship and do have a relationship with function words (which we include). Our exclusion of non-function words does not mean we assume that authors choose "the non-function word and the function word separately" (p. 329). We neither claim nor assume this at any point, and the validity of our model does not depend on it.

    Rizvi's next objection concerns another category of evidence that he thinks we have ignored: the fact that a given author chooses not to follow one of the function words we are interested with another one of those function words. Again, Rizvi's own searches underpin the objection:

I found 2,670 pairs of function words in which the first word is followed by the second in a Shakespeare play but never in a Marlowe play. For example, Shakespeare follows about by at (e.g., "they may have their throats about them at that time" in Henry V) but Marlowe never does. Conversely, I found 116 pairs of such words in Marlowe but not Shakespeare; for example, Marlowe follows after by any (e.g., "to imagine that after this life there is any pain" in Doctor Faustus), but Shakespeare never does. It is difficult to see how we can have confidence in a method that disregards highly relevant evidence such as this, simply because it does not know what to do with it. (p. 329)

The suggestion that our method "does not know what to do" with this evidence arises because our calculations would assign a value of zero to the occurrence of word x followed by word y if an author never followed x with y, and since at a later stage in our method another number would have to be divided by this zero--an operation not allowed in mathematics--we just (according to Rizvi) ignored this evidence.

    A key point that must be considered here is that we are analyzing texts of finite length. Imagine that in a text of 10,000 words author A uses the transition from the word a to the word of 100 times and the transition from a to in one time, whereas in another text of the same length author B uses the transition from a to of ten times and does not use the transition from a to in at all. Which transition encodes better the difference in style? Rizvi suggests that it is obviously the case that the latter transition--the one that occurs in author A's work and not in author B's--is more distinctive of authorship than the former.

    This assumption is not borne out by the evidence. The finite length of the texts introduces uncertainty in the observed transitions. It might well be that if we were to observe a further 1,000 words from author B, we would see a transition from a to in and what Rizvi assumes to be an utterly distinctive difference in style would vanish. The transition from a to of constitutes a quantifiable habit shared to different degrees by the two writers. That is, we prefer cases where we have positive data to cases where we do not. Our method encodes this preference by discarding the transitions that appear fewer than k times. For the published method, we selected k = 1, so that transitions that do not appear at all are disregarded in the computation. Of course, it is possible to vary this value along the positive integers as another free parameter and implement known procedures (such as cross-validation) to choose a preferred value for parameter k.

    A further consideration regarding what we might call habits of omission (where an author does not follow one word with another) and their relative usefulness for authorship attribution when compared to habits of commission (for which we have positive data) is that Rizvi searches within the Shakespeare and Marlowe canons looking specifically for the latter: word-pairs found in one canon and absent from the other. Of the 10,000 pairs that can be formed from a set of 100 function words, he locates 2,670 occurring in Shakespeare's work and not Marlowe's and 116 occurring in Marlowe's work and not Shakespeare's. We looked for all such pairings and recorded their strengths for all positive transitions, whereas Rizvi confines himself only to those for which the value in one canon, the k parameter, was equal to zero.

    No matter how it does what it does, an authorship attribution method deserves scholarly attention if it turns out to be a good 'predictor' of authorship already known from other independent evidence. Performed properly, the validation of a method such as ours should be able to quantify how often the method correctly identifies, from its words, the author of a play for which other independent evidence already tells us the author. We claim 89.6% to 93.6% accuracy in this regard. Rizvi thinks that we are mistaken about this because we made fundamental errors in our validation. The first error he claims is that:

The 93.6 percent success rate is claimed for "the ninety-four plays whose authorship is not in dispute." We are told that the method attributed all but six of those plays to their known authors. However, these attributions were obtained after training the method to recognize the correct authors for those very plays. As the article admits: "Each play is attributed . . . based on the adjacencies of the one hundred function words that were found in training (based on the full, undisputed sole-authored canon of each dramatist) to be the most discriminating” (Segarra et al. 2016 243). The argument is circular. Since the function words used in the tests were changed until they yielded the claimed success rate, the success rate can hardly be used as evidence of the method's correctness. (p. 330)

Here Rizvi has simply misunderstood our method. At no point in our work does a given play affect its own classification. Rather, the attribution of a given play is affected only by information derived from other plays. This can be mathematically formalized through an argument of statistical independence.

    The procedure we choose is a standard one that goes by the name of leave-one-out cross validation. Imagine we have made a specific choice of function words and want to test the value of this particular choice by measuring its efficacy in authorship attribution. Picture a bag that contains all plays of known authorship and imagine that we select from it a play that we will pretend is of unknown provenance. This play is removed from the bag, we use the remaining plays to train our classifier, as the jargon has it, and then use the classifier to attribute the play of supposedly unknown provenance. We record whether or not the method identifies as the author of this play the person who we know, from other evidence that scholars agree on, really is the author. We then replace this play into the bag and extract a different one.

    We repeat the procedure many times and end up with an accuracy score for this particular choice of function words. The procedure is then repeated for a different choice of function words, and a different score is computed. The choice of function words that performs best is the one that we use to classify plays of unknown authorship. Contrary to Rizvi's assumption (based on misreading our account), the choice of function words is the same for all of the attributions in a test run. Adapting the set of function words to a specific play would indeed artificially increase classification accuracy and would indeed be circular logic. It would also be self-defeating because it would then be impossible to decide what set of words to choose to classify a play that really is of unknown provenance.

    A second error that Rizvi claims to identify in our selection of function words arises because for different experiments we make different selections from the set of 100 function words. As Rizvi notes, our article "mentions at one point that it selected 76 out of its 100 function words for use in some experiments, and 55 in some others" (pp. 329-30). How can we be sure that we picked the right function words? As Rizvi puts it:

Now, the number of ways in which we can select several dozen words from a hundred words is astronomically large. The number of ways of selecting 76 words out of 100 is approximately 8 followed by 22 zeroes; and the number of ways of selecting 55 out of 100 is almost a million times greater than that. Therefore, it is no great surprise that if we pick many different subsets from a set of 100 function words, we will eventually find one that gives us the result we are training our method for. Tellingly, despite having literally trillions of subsets of the 100 function words to choose from, the software used in Eisen et al. failed to find one that works well for all authors. Instead, it was obliged to use different subsets for different authors. This means that the claimed success rates are of no great significance. (p. 330)

The assertion here is that there are so many possible combinations of function words that we are bound to find one that works well by sheer luck. This is false. We follow what is known as a greedy approach where we first rank the words (in terms of frequency) and then choose how many words to include. In this way, we first reduce the complexity of the method to be given by the number of function words instead of the set of function words. Second, the complexity of a classifier offers no guarantee that it will work well. In fact, quite the contrary: a lot of effort in machine learning is devoted to finding classifiers of low complexity.

    Perhaps this is best illustrated with a counterexample. Let us propose a clearly bogus classification method in which the authorship of a play is determined by the number of times each of 100 function words appears on the first page. There is an innumerable number of possible choices for these 100 words, but none of them would yield a reasonable classification accuracy. We should also point out that while the choice of function words does have an effect on classification accuracy, the effect is noticeable but not dramatic. We can still classify reasonably well with alternative choices of function words. These tradeoffs are described and analyzed in our published articles.

    Most importantly, Rizvi misunderstands us when he writes that "Since the function words used in the tests were changed until they yielded the claimed success rate, the success rate can hardly be used as evidence of the method's correctness" (p. 330). No, all of the attributions on which our claimed success rates of 89.6% to 94.6% percent are founded were made using the same set of function words.

    Rizvi's final objection is that we manipulated our results to make them look more significant than they really are. This concerns the supposed misrepresentation that occurs when we subtract a constant from each of a series of experimental results. When making a series of comparisons between a play to be attributed and the canons of each of a set of candidate authors, using the measure called relative entropy, we first calculate for each play its relative entropy with the entire set of all the plays by all the dramatists in order to get a kind of background reading of just how far this particular play differs from the collective norm. Each time that play's relative entropy with the canon of one of the dramatists is calculated, we deduct from that relative entropy the background reading for that play.

    To try to show how this distorts the results, Rizvi asks the reader to consider the difference between a student who gets 101 on a test and a student who gets 105. If we take 100 away from each result before announcing it, the figures become 1 and 5, and the second student seems to do much better than the first. By contrast, in the original data of 101 and 105, the second student seemed to do only marginally better than the first.

    Two points are worth mentioning with regard to this comment. The first one is that investigators often subtract a constant from their experimental results before comparing them. Studies of the health risks arising from the explosion at the Chernobyl nuclear power station necessarily subtract the naturally occurring background radiation that we all receive anyway in order to make sense of the increase caused by the explosion. In such cases, the number we are concerned with is not the total but rather the difference between the readings before and after the phenomenon we are interested in.

    Naturally, the constant being deducted has to be a reasonable one, not one chosen simply to exaggerate the results. Rizvi's own example of student grades may serve to illustrate that such deductions can be reasonable. Consider the scores for a multiple-choice test in which the candidate must, for each question, select the right answer from a menu of five suggestions. Someone choosing answers randomly will get one-fifth (20%) of the answers right by chance, so it would be reasonable to discount the first 20 percentage points before comparing students' grades.

    Consider two students who score 20% and 60%. On these raw numbers the second student seems to have done three times as well as the first. But if we discount the first 20 percentage points because they represent the background reading--that is the score that a person would get anyway even without knowing any of the answers--we arrive at a comparison of 0 with 40. This reveals that the first student demonstrated no knowledge of the subject, doing no better than would a person selecting answers at random. Only after deducting the background reading do we see that the second student's attainment was not three times better than the first student’s attainment, but vastly better.

    The second point worth mentioning is the difference between relative and absolute comparisons. In the test score example that we are discussing, a relative comparison is the ratio 105/100 between the student scores and an absolute comparison is the difference (105 minus 100) between their scores. A relative comparison can be changed, indeed, dramatically changed, by subtracting a baseline. Absolute comparisons cannot be. In our case, the method attributes a text to the author whose profile is closest to the Word Adjacency Network of the text to be attributed. This is unchanged by a constant shift of all the distances, as is apparent and clearly stated in our articles. The difference between 101 versus 102 and 1 versus 2 would not affect our attribution results in any way; both the validation accuracies and attribution decisions being made would be the same whether this subtraction was performed or not. Therefore, it is perfectly reasonable to deduct this constant to improve the readability of the results. At no point do we use the relative difference between the distances as a measure of attribution, and none of the correct or incorrect attributions depends on this.

    Rizvi reached the conclusion that our results should be disregarded. However, all his arguments are based on misunderstandings of our method. Our analyses cannot be used to conclude irrefutable proof of authorship. In our published work, we go to some trouble to explain the limitations of the technique, specifying that it can be used only as evidence and not irrefutable proof. Future work may bolster our conclusions, or it may prove them wrong. But nothing is achieved in misrepresenting how we reached those conclusions.

    One last study in authorship attribution must be noticed before we can leave the topic. In it, Brian Vickers attempts to establish that Thomas Kyd wrote the play King Leir ('Kyd's Authorship of King Leir', SP 115[2018] 433-71). He starts with the known facts: King Leir was anonymously published in 1605; it is plausibly assumed to have been performed by the Queen's men around 1589; and it was a source for Shakespeare's King Lear. Vickers traces some plot and action details that King Leir shares with Richard 3 and other Shakespeare plays. To prepare his case for Kyd's authorship of King Leir, Vickers shows that it shares three plot details with Kyd's The Spanish Tragedy: "an intrigue plot; the use of comedy in a context involving imminent death; and a vengeful woman, ready for violent action in her own interest" (p. 438).

    Vickers devotes 14 pages to tracing the two plays' similarities under these three heads, and while they are no doubt present it is unclear from Vickers's account that they are especially rare parallels within the wider body of drama. The second half of Vickers's essay is devoted to finding phrases common to King Leir and Kyd's The Spanish Tragedy, or reporting ones found by previous investigators. As ever, the value of this evidence is not only in the total number of phrases in common but in the general rareness, since finding that two early modern plays had in common a large number of occurrences of the phrases good life or my heart would of itself tell us nothing since these occur throughout the written language.

    Vickers's first example is "To be inroll’d in Chronicles of fame, | By never dying perpetuity" in King Leir matching "To be enrolled in the brass leaved book | Of never wasting perpetuitie" in Soliman and Perseda, and EEBO-TCP does indeed show this to be unique across all the books in that dataset. But Vickers's next example is not: "First to the heavens, next, thanks to you, my sonne" (King Leir) is certainly like "First, thanks to heaven; and next to Brusors valour" (Soliman and Perseda), but it is also like "first thankes to heauen, next to my foreward country-men" in the anonymous play The True Tragedy of Richard the Third published in 1594 and also "First thanks to heauen, & next to thee my friend" in the 1594 quarto of the play The Contention of York and Lancaster, usually attributed to Shakespeare and identified as an alternative version of his 2 Henry 6 published in the 1623 Folio collection. It is surprising that Vickers is unaware of these matches in well-known works, or at least makes no mention of them.

    Vickers's next example illustrates the same problem. "And add fresh vigor to my willing limmes" (King Leir) is indeed like "And ad fresh courage to my fainting limmes" (Soliman and Perseda), but so is "add fresh strength to these my withered limbs" from the anonymous play Guy of Warwick first performed in the early 1590s, and so is "adde fresh vigour to thy feeble limmes" from Robert Armin's play The Valiant Welshman first performed in the early 1610s. Again, the two parallels that undermine Vickers's claim regarding distinctiveness are not mentioned in his essay.

    The foregoing are Vickers's strongest pieces of evidence. What follows are looser parallels, apparently offered to set the seal on a case for common authorship that has already been established, yet worth explicitly refuting for the sake of completeness. Vickers finds a parallel between "mortall toung can tell" (King Leir) and "tongues can tell, | Or pennes can write, or mortall harts can think" (The Spanish Tragedy), but this loose parallel is bettered by the exact ones of "no mortall tunge can tell" in John Bale's play God's Promises first performed in 1538, and "no mortall tongue can tell" in the anonymous play The Merry Devil of Edmonton first performed around 1600, and "nor tongue of mortal tell" in Shakespeare's Love's Labour's Lost. The last two of these three additional parallels are far from obscure and it is unclear how Vickers missed them. In any case, the phrase from King Leir is evidently a commonplace of the language and EEBO-TCP has many further non-dramatic examples.

    Vickers's next example is the collocation of winter, nipped, and blossoms in "Whose blossomes now are nipt with Winters frost" (King Leir) and "Deaths winter nipt the blossomes of my blisse" (The Spanish Tragedy). This too seems a commonplace of the language, as John Speed in his The History of Great Britain (published 1611) wrote that "winter had nipt all those faire blossomes" and EEBO-TCP has several other examples of these three words collocating as closely as they do in King Leir and The Spanish Tragedy. Vickers finds a collocation of and, pale, death, and my in "And pale grym death doth wayt upon my steps" (King Leir) and "And now pale Death sits on my panting soule" (Soliman and Perseda). Unsurprisingly, this is a common phrase in the language according to EEBO-TCP, found in "and recouery | Pale Death himselfe did | loue my Philomel" by Francis Davison in A Poetical Rhapsody published in 1602 and "And though pale death consume my flesh" by Francis Quarles in Divine Poems published in 1633, and elsewhere.

    Vickers's list of parallels gets increasingly loose as he proceeds, as witness his next one, where flattery is followed by I flatter followed by I flatter not with up to seven words in between. This has no match elsewhere although James Shirley's play The Traitor (first performed in 1631) comes close with "Tis hansome flattery, and becomes a Courtier. | PETRUCHIO I flatter not my Lord". The point of course is not that Shirley is a plausible candidate for authorship of King Leir, but that Vickers nowhere attempts to quantify the rarity of the phrases that he adduces as evidence of authorship, and many are much less rare than he supposes.

    The next link that Vickers adduces is the collocation of wrath and exasperate in either order and with up to four words in between, found in King Leir, The Spanish Tragedy, and Cornelia, and once again there are many other examples unmentioned by Vickers. In the most notable, Queen Isabella in Marlowe's Edward 2 says "But that will more exasperate his wrath". Vickers next explains how he found the parallels that he thinks have added to the case for Kyd's authorship of King Leir, paying special attention to the software he used called WCopyfind. Of course, there are well-established computational algorithms for finding particulars words and phrases in a digital textual corpus, but what matters is not so much the particular software used--so long as it is competently written--but the particular digital textual corpus being searched.

    Vickers writes that for his searching he "compiled a corpus of the fifty-four plays performed in the London public theaters between 1579 and 1596, excluding those privately performed (in schools, universities, and the Inns of Court) and plays of dubious provenance, the so-called 'bad quartos'" (p. 457). Use of such a small hand-compiled corpus goes some way toward explaining Vickers's failure to find the additional matches that make his parallels between King Leir and Kyd's plays much less valuable than he supposes, but it cannot be the whole story. The plays The True Tragedy of Richard the Third, The Contention of York and Lancaster, Guy of Warwick, and Edward 2 meet Vickers's narrow criteria of being first performed in public between 1579 and 1596 and ought to be in his corpus, yet Vickers overlooks matches found in them.

    After this discussion of his method, Vickers returns to listing verbal parallels but confining himself to those in which "the dramatic situation is the same in both plays" (p. 458), which is of course a subjective filter. He needs such a filter, since in every case there are additional matches beyond the ones Vickers finds for the phrases, which are "for fear the", "I know not what", "favour me so much", "bow|bend their knees to", "will stand between me and", "what avails to", "sovereign of my soul", "whence . . . the ground of this", "divorce betwixt my . . . and me", "the heavens are just",  "pieces with my teeth", "should . . . the person of the king", "comfort my life", "I long to see thy face", "never can be . . . but by", "my love was . . . her|hers", "my heart will break", "even to the very", "it is a world to", and "I, I, I" (pp. 458-63).

    Some of these are, unsurprisingly, found many hundreds or even thousands of times across books published in the period since they are common phrases in the language, such as "I know not what" and "even to the very". Just one of Vickers's parallels between King Leir and a Kyd play, Soliman and Perseda, is in fact unique: "have pricked the heart". This is certainly an interesting parallel, but on its own it does not count for much, since if one lists all the phrases that any two substantial texts have in common one will invariably find an occasional example that is unique to those two texts; that is just a fact of how rarity in language works.

    As has repeatedly been demonstrated in this annual review, it takes no great technical skill to find parallels that undermine Vickers's authorship claims regarding Kyd and Shakespeare. The version of the TCP data offered free of charge to all users as the splendid Early Print database hosted at Washington University in St Louis is sufficient.

    The last quarter of Vickers's article concerns the likeness of the verse style of King Leir to Kyd's verse style. King Leir has a fairly high rate of feminine endings, 10.8% of lines, and Kyd is known for having a high rate. Likewise with the pause patterns quantified by Oras: King Leir tests like Kyd's The Spanish Tragedy. In both these matters, Vickers tabulates only King Leir and Kyd's habits, so we do not get to see if King Leir is like Kyd's work and unlike other candidates' plays, which would be more convincing than simply showing that it is like Kyd's work.

    Vickers's Figure One is a photographic reproduction of a pair of graphs from Oras's book showing authorial preferences for the placement of pauses in verse lines and it consists of two shapes that look rather like traces from an electrocardiograph. There are no labels on the horizontal and vertical axes but we can infer from the presence of nine circles on each trace that the horizontal axis represents the nine positions where a pause may fall between any two of the syllables of an iambic pentameter line. What is measured along the vertical axis is anybody's guess, but that the two traces are not to the same scale is implied by the fact that the peak of the first is labeled "848" and the second, though reaching to nearly the same height as the first, is labeled "1339". According to Vickers the traces' "profound similarity can hardly be coincidental" (p. 466).

    Vickers then quotes Tarlinskaja's work on verse habits and her agreement with him that King Leir is by Kyd. Vickers's last verse feature is the frequency and type of line-ending rhyme, and again his only comparison is with Kyd's plays so the reader gets no sense of how distinct from others' styles Kyd's writing was. Vickers ends rather abruptly after the section on verse style with a conclusion comprising a one-sentence paragraph of 29 words asserting that all the evidence he has adduced supports the case that Kyd wrote King Leir.

    According to the New Oxford Shakespeare, the 1603 quarto of Hamlet represents the play as it was first written by Shakespeare in the 1580s, as argued by Terri Bourus in her book Young Shakespeare's Young Hamlet. MacDonald P. Jackson adduces evidence that where Q1 Hamlet is like Q2 Hamlet, the Q1 parts can be dated to the 1600s not the 1580s, which is hard to reconcile with Bourus's account ('Vocabulary, Chronology, and the First Quarto (1603) of Hamlet', Medieval and Renaissance Drama in England 31[2018] 17-42). To explain the origins of Q1 Hamlet scholars have tended to look at its worst bits, but the really hard things to explain are its best bits, to explain why they are "so good and so exemplary of Shakespeare's mature style" (p. 18)

    As an example, Jackson picks out 1.42-131 in the Arden3 edition of Q1, which contains several notable examples of hendiadys, not a conspicuous proficiency in early Shakespeare (just 14 occurrences in his first 10 plays), but which by 1600 he had mastered. Jackson also points to verse bits of Q1 that are virtually identical with their Q2 counterparts, such as 2.110-125, 2.155-164, 2.166-169, 4.15-32, 5.8b-18, 5.41-53, 7.31-51, and 7.340-348, making 102 lines in all. Of these passages, almost half the lines (except speech-ending ones) are run-on, which is high even for early seventeenth-century writing by Shakespeare--typically 23-39%, depending on whose counts you accept--and is more typical of his final period. Shakespeare's early plays have far fewer run-on lines than this.

    Jackson turns to pause patterns as identified by Oras, for which the clearest shift is from pauses in the first half of the line in early Shakespeare to pauses in the second half of the line in late Shakespeare. Jackson reproduced Oras's method using all three kinds of break within a line--(A) breaks marked by any kind of punctuation, (B) breaks marked by punctuation heavier than a comma, and (C) breaks marked by a change of speaker--and applied it to the above 102 verse lines in Q1 that are virtually identical to their counterparts in Q2. In the table of results, it is clear that these parts of Q1 do not resemble early Shakespeare, which tends to have more pauses after syllable 4 than syllable 6 while these results have more pauses after syllable 6 than syllable 4, which is something Shakespeare starts to do beginning with Twelfth Night in 1601.

    Jackson then turns to the evidence of rare-words, using Eliot Slater's counts of dictionary headwords that occur in at least two of Shakespeare's plays but no more than ten times in all across his canon. Slater showed that plays written about the same time tend to share more such rare words than plays written at different times, with words that occur no more than six times in the canon--what Slater calls "two-six-fold words"--being the most useful evidence. Confining himself to just these words used in the Q1 passages and in "not more than five other plays" (p. 22.) gave Jackson this list of 41 words: alley (n.), annual (adj.), apparition (n.), arrest (n.), blazon (n.), canonize (v), cap-a-pe (adj.), competent (adj.), conveniently (adv.), dawning (n.), eruption (n.), exactly (adv.), foreknowing (v.), fretful (adj.), gage (v), garbage (n), glimpse (n.), harrow (v.), hearsed (v.), heraldry/heraldy (n.), illusion (n.), implement (n.), impress (n.), invulnerable (adj.), jelly (n.), juice (n.), jump (adv.), observant (adj.), platform (n.), ponderous (adj.), porch (n.), quicksilver (n.), quill (n.), rugged (adj), russet (adj.), sable (adj.), skirts (n.), smear (v.), stalk (v.), sweaty (adj.), and truncheon (n.).

    These 41 words appear 107 times in all in Q1 Hamlet. Jackson divided the Shakespeare chronology in the New Oxford Shakespeare into six groups: The Two Gentlemen of Verona to Richard 3 (1588-1592), The Comedy of Errors to A Midsummer Night's Dream (1594-1596), King John to Henry 5 (1596-1599), Julius Caesar to Othello (1599-1604), Measure for Measure to Antony and Cleopatra (1604-1607), Pericles Acts 3-5 to All is True (1608-1613). From this six-way chronology Jackson omitted of course Hamlet because it is the text under examination and also Arden of Faversham and Edward 3 on account of their co-authorship. These groups were chosen to put roughly the same number of occurrences (around 5,800) of the "two-six-fold words" in each group. For each group Jackson lists how many times one of the 41 rare words in the Q1 Hamlet passages under consideration appears in plays in that group, and the results are 9, 16, 20, 28, 21, and 18 for the six groups. The result is that the rare-word links on these passages from Q1 Hamlet are weakest of all to earliest Shakespeare (just 9 links) and strongest of all to the plays from 1599-1604 (28 links). The co-authorship of the earliest plays cannot be the explanation for this, since the first group share rare-word links to one another even when they are co-authored.

    Within the1599-1604 group, the largest numbers of links to these Q1 Hamlet passages are found in Troilus and Cressida, Othello, and Measure for Measure, which were written 1602-1604. That is, these passages of Q1 Hamlet look just like Q2/F Hamlet. Also, where Jackson looked only at hendiadys before in 1.42-131 he now looks for it across these Q1 Hamlet passages and finds it. What if Q1 is an early Shakespeare play but these particular passages in it reflect Shakespearian revision of it in the early 1600s? Jackson takes up Gregor Sarrazin's list for each play of the words in it that occur twice or thrice (dislegomena and trislegomena) elsewhere in Shakespeare; for each play these links are most commonly to other plays of the same period. 

    Jackson takes the dislegomena and trislegomena Sarrazin gave for Hamlet and finds that 33 of the 80 occurrences in Q1 Hamlet link to plays up to Julius Caesar and 47 link to the plays from As You Like It to All is True. These results place Q1 (all of it) with canonical Q2/F Hamlet and Othello, well away from the earliest Shakespeare, and the dislegomena and trislegomena are not confined to the passages in Q1 tested above for their prosody: they are right across Q1. Again, co-authorship of early Shakespeare cannot be the explanation since the Sarrazin links are found equally in the sole-authored and the co-authored early Shakespeare plays.

    What if Q1 Hamlet represents a first draft made in the early 1600s that was soon after revised by Shakespeare to make the canonical Q2/F version? Jackson looks at Q1's 2.1-6 and makes the point that it has some particularly odd errors, such as a contradiction about whom Claudius has written to, Fortinbras or Norway, and calling Norway "impudent" (a mishearing of "impotent"?), and with significant metrical roughness. The corresponding bit of Q2 is metrically regular and lacks these problems, and the difference looks like textual corruption perhaps alongside abridgement. 

    Jackson examines Gertrude's account of Ophelia's death as represented in Q1 and Q2, and argues (subjectively) that the former looks like a faulty recollection of the latter, with phrases such as "'twixt heaven and earth" and "sat smiling" brought in from elsewhere in the play. Likewise the rough meter of many lines in Q1's version of this speech is suspicious, as is the uninventive string of present participles "Having . . . Sitting . . . Smiling . . . Chanting". Bourus's explanation for this infelicity in language was that Gertrude is emotionally distraught.  

    In Q1's scene 14, not present in Q2/F, in which Horatio reveals all to Gertrude and she joins their faction against her husband, the phrase "sugar o'er" appears, and Literature Online's only other occurrence is in canonical Q2/F when Polonius and Claudius eavesdrop on Hamlet and Ophelia. Since "sugar o'er" appears nowhere else, it either got into Q2/F from Q1 (or rather their underlying manuscripts) or vice versa. Jackson reckons that Q1's getting it from Q2/F is more likely than the reverse, on account of the general weakness of Q1's scene 14. Jackson details that weakness, including the absurd "He [Hamlet] being set ashore, they [Rosencrantz and Guildenstern] went for England", which of course never happens: why would they go on for England after letting Hamlet disembark in Denmark? Also, by omitting the story of the pirates intercepting the ship, Horatio's account in Q1 makes almost no sense.  Jackson highlights other Q1-Q2/F alignments where Q1's wording is hard to explain other than as a garbling of Q2/F.

    The first part of Jackson's article shows that if Q1 does indeed represent a version of Hamlet then it was one written shortly before the canonical Q2/F version rather than (as Bourus has it) long before. It is hard to see why, if it existed, Shakespeare "would release it for publication in print" (p. 35). Alfred Hart showed that the bad quartos of Hamlet, 2 Henry 6, 3 Henry 6, Romeo and Juliet, Henry 5, and The Merry Wives and Windsor "share from 86.5 to 93.3 per cent of their word-types with their canonical counterparts" (p. 35). This is considerably more than are shared between The Troublesome Reign of King John and King John or between King Leir and King Lear. If Q1 was Shakespeare's first stab at the play, Hart pointed out, he then unaccountably deleted 400 lines of it, fixed up the remainder, and added nearly 2,000 new lines that were not needed for the plot. This is not, Jackson insists, how plays got revised. The key problem with Q1 is its mix of the fine writing that it shares with Q2/F with the really poor stuff. The evidence Jackson adduces bolsters the explanation that this is because of textual corruption in the transmission that led to Q1 rather than, as Bourus argues, Q1's being a much earlier version of the play.

    Matt Vadnais has also published an article on the topic of Shakespearian revision ('The Distribution and Ordering of Speeches and Shakespearean Revision', ShakB 36[2018] 657-86). He argues that where we have a short quarto and a longer quarto or Folio version of a Shakespeare play, as with Romeo and Juliet, Henry 5, The Merry Wives of Windsor, and Hamlet, the longer version was an authorial expansion of the the shorter one, perhaps prepared for a court performance, as Richard Dutton argued in a book reviewed in NYWES for 2016. The evidence for this is that in each case the longer version is just as theatrically economical and sensitive to the needs of performance as the short version, and that hence we do not have to construct the binary of authorial-long versus performative-short versions as we have tended to do.

    The relationship between the 1600 quarto edition (Q1) and the 1623 Folio edition of Henry 5 was understood by Pope and others of his time to be simply that the former reflected the play as Shakespeare first wrote it and the latter reflected his subsequent refinement of it in revision. But by the New Bibliographers of the twentieth century ". . . the Folio text was understood to have been imperfectly derived--via piracy, memorial reconstruction, or some other unauthorized method--from an authorial copy of the text written prior to the opening of the Globe" (p. 657). Here Vadnais must have typed "the Folio text" where he meant "the quarto text", since it is Q1 not F that was thought to have been imperfectly derived this way; there has always been widespread agreement that F is a relatively uncorrupted text set from the author's papers or a transcript of them. This is an unfortunate (and for the beginner, confusing) slip right at the start of an essay on the Q1/F problem.

    Also leading the reader astray is the essay's first footnote in which Vadnais advises readers to "see . . . Egan, 'What is not Collaborative' . . ." for "textual and contextual discussions of the variant texts of Henry V". In fact that article by the present reviewer (published in Shakespeare Survey in 2014) contributed nothing to the debates about the variant texts of Henry 5. Vadnais sketches the claims that Q1 Henry 5 represents the play as first performed in a script cut down from the too-long authorial first draft represented in the Folio, especially as this argument is made by Erne. Rather than just counting the lines in a play, Vadnais thinks it better to track how many speeches there are. He surveys the critical tendency to associate a script's brevity with its performativity, and some recent arguments that in fact plays were not confined to the two hours mentioned in the prologue to Romeo and Juliet, including Michael C. Hirrel's article on this that was reviewed in YWES for 2010.

    If plays were routinely cut because authors wrote more than could be performed, it is odd, Vadnais points out, that we have only a handful of references to this happening, such as Richard Brome's note to the reader at the end of the 1640 quarto of his play The Antipodes. The surviving manuscript playbooks that show cuts do not have anything like the extensive cutting that would be needed to get Q2 Romeo and Juliet down to the length of Q1, or the Folio editions down to the lengths of the first quartos of Hamlet, Henry 5, or The Merry Wives of Windsor. Also, paying playwrights to write more than was needed would be uneconomical, as would the time spent making the cuts, even if the actors managed to do it themselves without engaging a professional writer (p. 663).

    Vadnais attempts to show that in the cases of Henry 5, Hamlet, The Merry Wives of Windsor, and Romeo and Juliet, the quarto and Folio editions share features (specifically, speech patterns) that "were intended to ease the difficulty of performance according to cued parts" (p. 664). In particular, the longer Folio versions do not have proportionally more speeches than the shorter versions: the extra length comes not so much from new speeches as the lengthening of existing speeches. Thus the disruption to the cue-script system is minimal.

    For example, Folio Henry 5 is twice as long as Q1 but has only 40% more speeches. This suggests the reluctance to require players to give and listen for new cues. Vadnais calculates what he calls the "speech burden" (p. 665), which is the average number of speeches and cues a player has to remember between coming on stage and going off (where he can refresh his memory from a written document), as between 6 and 10 speeches-per-entrance. This is true of the long Q2 Hamlet and the short Folio The Comedy of Errors. Scott McMillin pointed out years ago that long scripts provide more doubling opportunities than short ones and so, contrary to prevailing assumptions, are more readily played by a small cast. Vadnais extends this principle by pointing out that long scripts give the actors more time off stage and fewer appearances in consecutive scenes.

    The long Folio versions also have more two-character scenes that are especially easy to play because each actor simply has to wait for the other to stop speaking. Much the same is true of easily played scenes in which one actor speaks every other speech but with multiple interlocutors rather than a single interlocutor taking the other speeches, which Vadnais calls "speech stems" (p. 668); these are no more common in the short, supposedly more theatrical, quarto versions than the Folio versions of the plays he considers. Perhaps at times the actors wanted to play a three-and-a-half-hour version of a play and at other times they wanted to play a two-hour version so they appreciated having both.

    Next Vadnais turns to the surviving  playbook manuscripts: the 18 studied by William B. Long and the 21 studied by Paul Werstine. The big picture is that cuts in the manuscripts tend to be mostly within speeches, thus preserving the cues. Vadnais infers that the cuts were made after the actors had begun to learn the parts, else why bother to preserve the cues? But the manuscripts do not show the kind of single-word variation that characterizes so many Shakespearian Q/F differences: the manuscripts have whole lines present or marked for cutting. And as noted earlier, the cutting marked in manuscripts is much smaller (no more than 7% of the text) than the Shakespearian Q/F differences, and this remains true if we look at the number of speeches rather than the number of lines.

    Vadnais concludes that the Q/F differences represent two versions of each play, both entirely and equally suited to performance. This situation might arise through cutting a long playable text down to a short playable one, but Vadnais thinks it inherently more likely that the reverse happened: the long versions came about by expansion of the short versions. Thus the scripts underlying the quartos really were chronologically prior to the scripts underlying the Folio. The revision was probably not gradual and spread over time: more likely, it was one-off and substantial since the approval of the Master of the Revels was needed. Also, to advertise a play as new and improved, and hence to attract back those who saw it before and might think it old hat, the changes would have to be substantial (p. 675).

    In Shakespeare, Court Dramatist, Dutton argued that performance at court was the occasion for the substantial expansion of a number of Shakespeare plays, and that the expanded plays were subsequently performed in the commercial theatres to recover some of the costs of the expansion. The fact that the longer versions of the Q/F plays are as theatrically refined, in the sense of economical speech distribution discussed above, is evidence that the longer versions are authorial expansions of the shorter ones.

    Vadnais invents a new term, "French scene" (p. 677), to mean the span of a play from any one entrance or exit to the next entrance or exit: whatever adds or subtracts one or more characters from the stage. Although Q2 Romeo and Juliet is much longer than Q1, it has only 12 extra French scenes to add to the 86 French scenes in Q1. Of these 12, several entail just the exitting and re-entrance of single characters or the expansion of a stage direction that has multiple exits and entrances in it. The 120 speeches in Q2 Romeo and Juliet not in Q1 are spread over 46 French scenes, which Vadnais reads as an attempt to spread the additional burden around, "thereby limiting the relative increase in a player’s speech burden" (p. 678). Vadnais suggests that this pattern of spreading the burden evenly across the play is a sign of authorial expansion, but surely it might be interpreted as a sign that the shorter version was cut down from the longer version. After all, would we not expect that an author would add whole new plot elements via whole new scenes rather than just inflate the elements and scenes that already existed? Vadnais finds the same broad patterns in the expansions of Henry 5, Hamlet, and The Merry Wives of Windsor, noting that where new speeches are added they are disproportionately added to two-player or stem-speech scenes that are "likely among the easiest scenes in the play to perform according to cued parts" (p. 679).

    Ericka Boeckeler thinks that the orthography and typography of Q1 Hamlet have meanings of their own that intersect with the play's wider meanings, and that the first readers would have noticed them ('The Hamlet First Quarto (1603) and the Play of Typography', Early Theatre 21[2018] 59-86). She has the annoying habit of introducing old-spelling word forms into her prose when she is not quoting, as in "vndiscouered" (p. 61) and using long 's' in place of modern 's' (p. 62), to no discernible (and certainly no discussed) purpose. Her major point is that in original orthography Q1 Hamlet has some interesting internal resonances that are lost in modernization or performance. She means things such as "a loofe" (= modern aloof) looking like "loose" (= modern lose or loose).

    "I do not insist that intent lies behind the readings I suggest below", Boeckeler writes (p. 63). It it hard not to silently reply to oneself, "then who cares?" The clouds make interesting patterns sometimes too, but as Hamlet points out you can see what you want to see in them. The whole article chases these quite possibly illusory patterns such as do/deaw/doe/doo't/duetie/dutie across Q1 without offering any evidence that these spellings are more than just random compositorial choices. Boeckeler's only meaningful examples have no orthographical element. For example, in "sit still my soule, foule deeds will rise" the soul/foul pairing does indeed suggest that Shakespeare was exploiting a visual likeness (which is all the stronger when soule is spelt with a long 's') but this likeness is unconnected to the original-spelling 'e' on the end of each word.

    Boeckeler finds meaning in italicization too and in spellings such as "noone" (= modern noon) that also implies no one, which suits the context in which this spelling is used. She finds meaning in "To be, or not to be, I there's the point" having 'I' instead of modern ay (since it is a speech about the self) and its spelling of modern judge as "Iudge" with another capital 'I'. Boeckeler reckons that the capital Ts in the first two lines of Q1, "STand: who is that" and "Tis I", serve to "visually evoke the image of the two 'Centinels' standing in front of each other" (p. 72). Boeckeler does not mention that it was simply conventional whenever a dropped capital was used for the first word in the first line of a play, as it is in "STand", for the second letter to be capitalized too. The capital 'T' in the next line's "Tis" is of course due to its being the first word of a sentence.

    Boeckeler has many more examples of what she thinks are significant typographical choices that most experts have understood as ordinary habits of early modern book making, and she offers in defence of her idiosyncratic explanations nothing but the inherent interest of the patterns as she sees them. In her conclusion Boeckeler turns to letter-play, number-play, and verbal patterning in printed poetry, including Shakespeare's Sonnets, and argues that some of this can be found in the plays too. Hers is a plea for critics to take more account of the visual as well as the oral aspects of words.

    In her conclusion, Boeckeler seems to confuse serifs with kerns when she writes, regarding the practical reasons for variable spelling in printing, about "the insertion of letters to protect serifs on type" (p. 79). Serifs, the short strokes or lines at the end of longer lines in letters--especially the ones that run horizontally along the bottom of letters to provide a kind of intermittent base line on which the letters seem to sit--are always contained within the body of a piece of letterpress type, and choices of spelling could not protect them as Boeckeler suggests. Kerns, the long strokes that project beyond the body of the type in some letters in some typefaces (notably italic ones), could clash in adjacent letters and adding an additional, orthographically unnecessary letter between was indeed an accepted way to avoid this.

    As well as helping make sense of dramatic authorship and revision, computational analysis is beginning to make considerable advances in our understanding of genre, as evidenced in a new study arising from work on Shakespeare's texts at the University of Lancaster (Jonathan Culpeper 'Measuring Emotional Temperatures in Shakespeare's Drama', English Text Construction 11.i[2018] 10-37). The investigators took the SentiStrength tool for sentiment analysis of texts, developed by Mike Thelwall at the University of Wolverhampton, and adapted it to suit early modern English and then applied it to the plays in the 1623 Shakespeare Folio. The idea was to see if genre is correlated to a play's overall sentiment, in the sense of emotion.

    The source text was a digitization of the 1916 Oxford Shakespeare edited by W. J. Craig, to which additional tagging was added and then the University of Lancaster's Variant Detector (VARD) software was used to regularize the spelling. SentiStrength is software designed to "classify the strength of positive and negative sentiment in short informal texts, such as in Twitter" (p. 15), giving each text a score from 1 to 5 for increasing positive sentiment and -1 to -5 for increasing negative sentiment. It does this by having a list of 2608 words or word-stems with assigned scores, so love is assigned a +3 and hate is assigned a -4. The software takes into account negations and intensifiers such as very. The authors admit that the software does not take into account sarcasm and words that can be good or bad such as wicked.

    The adaptation to early modern English consisted primarily of tweaking the assigned scores, and the authors give some detail of which words were affected and by how much. They also added some new words with particularly strong emotional connotations in early modern English and took out words that had not yet been coined. To test the resulting Early Modern SentiStrength they compared its scores for sentences of Shakespeare to those of expert human scorers and then calculated the strength of the correlation between the two sets of scores. These correlations were then used to "automatically adjust the weights of words in its sentiment lexicon" (p. 19) to make Early Modern SentiStrength more closely match the human judgements.

    The authors start to give their results, and their first figure shows for each play the mean-average of the sentence-unit positive and negative scores, using the absolute value for the latter (thus a -3 would be treated as 3). Thus this average shows the total strength of emotion (whether it is positive or negative emotion) in each play. At the top of the rank order, with the strongest emotions, are history play and thereafter come tragedies and comedies interspersed in the rank order (so not clustering as genres). Of the histories, Richard 2 is the most emotionally marked and 1 Henry 4 and 2 Henry 4 the least. Unsurprisingly, of the tragedies Titus Andronicus is the most emotionally marked and Antony and Cleopatra the least, and of the comedies A Midsummer Night's Dream is the most and The Merry Wives of Windsor the least. There is least variation among the comedies so they are the most cohesive group.

    The plays from the beginning of Shakespeare's career emerge as the most emotionally intense and thereafter there is a falling off. If we confine our attention to negative emotion, which the authors show by subtracting positive emotion from negative emotion, the histories and tragedies come out on top. In criticism today three comedies are thought to have problematic endings: Love's Labour's Lost because the love affairs are broken off, The Two Gentlemen of Verona because of the assault on Silvia and the trading of her between the men, and The Taming of the Shrew because of the subjection of Katherine. In this analysis these are the only three comedies that have more positive emotion than negative and hence end up with negative numbers when the positive is subtracted from the negative.

    The four comedies with the most negative emotion are The Comedy of Errors, Measure for Measure, Cymbeline, and The Tempest and these results accord well with the critical reception of these plays as having dark elements verging on tragedy. When placed chronologically the plays do not show the periodization that Dowden hypothesized with his descriptors "out of the depths" and "on the heights" and so on, but they do show his hypothesized habit of following an emotionally intense play with a markedly less intense one. The authors end with some speculation of where Early Modern SentiStrength might go next in acquiring more words and their scores, looking beyond Shakespeare, looking at particular characters, and assisting in authorship attribution.

    Also using computers, but much less successfully, is an attempt to make the machine spot a rhetorical device (Adam James Bradley and Michael Ullyot 'Machines and Humans, Schemes and Tropes', EMLS 20.ii[2018] 1-16). This article covers much the same ground and arises from the same experiments as its authors' book chapter reviewed above. The authors explain the figure of speech called gradatio, which goes AB BC CD, as in "they appear to men like angels light. Light is an effect of fire, and fire will burn" (The Comedy of Errors) where A is angels, B is light, C is fire, and D is burn, and the trope called incrementum where B is bigger than A, and C is bigger than B, and so on. Then comes a long and rather tedious quasi-philosophical discussion of the difference between reason and imagination, leading to claims about the ability of computers to find quantifiable schemes such as gradatio, and their inability to find supposedly unquantifiable tropes such as incrementum. I do not agree that incrementum is unquantifiable: it is just harder to quantify than simpler features of language.

    The authors built a software tool for finding anadiplosis, which is the B-to-B and C-to-C linking by repetition in the above example. They started by copying Mueller's Shakespeare His Contemporaries dataset of plays and then "tokenizing" the words (p. 12). They do not reveal what this tokenizing consisted of, and the term has more than one meaning in data science. Then they set about "lemmatizing or stemming those tokens" (p. 12) and again they do not explain what they mean nor whether by or they just mean in other words. The authors discuss the problem of how far apart the words forming a repetition might be allowed to fall and still qualify as anadiplosis and they settled on four words as the upper limit. They give no details of exactly what their software did and simply report that it found "112 instances of gradatio in these 400 plays" (p. 13) from Mueller's collection.

    Or rather, the software reported 112 instances and when they looked they found that it had detected lots of repetitions that were not anadiplosis forming gradatio. The authors seem surprised to discover that "The assumption that proximate and sequential anadiplosis is the universal formula for gradatio proved false, in some instances" (p. 13). They conclude that sometimes "humans alone can recognize a trope" (p. 14) and while this is true of complex tropes they might usefully have tried to refine the algorithm to see if they could better formalize just what it takes for some writing to display gradatio.

    We have seen that Brian Vickers and Darren Freebury-Jones insist that quantitative analysis of texts must be accompanied by qualitative interpretation of the results, and Jeffrey R. Wilson essentially agrees ('Shakestats: Writing About Shakespeare between the Humanities and the Social Sciences', EMLS 20.ii[2018] 1-38). After a survey of recent work done in the computational analysis of Shakespeare's writing, Wilson alights on Franco Moretti's analysis of "characters who speak to each other" (p. 10) in Hamlet, finding that it tells us nothing that would not be obvious to anyone who has read the play.

    Wilson does not mention that Moretti's much-emulated analysis was based on an unjustified assumption about drama. Unless a human has marked up the Shakespeare text with decisions (based on expert opinion) about who is actually speaking to whom, the only way a computer can form the kind of character-network drawn by Moretti is for it to assume that characters who are on stage at the same time, to judge from the overlapping of their Entrance and Exit directions, are speaking to one another when they talk, which of course is not necessarily true.

    Wilson then describes the contents of an unpublished essay of his own that examines the preponderance of male characters and their speeches over female ones in Shakespeare's drama, and decides that although not a misogynist ". . . the Shakespeare who wrote Hamlet still held an unconscious bias against women" (p. 12). I would say that the gender imbalance in Shakespeare's writing is at least partly due to women's parts being performed exclusively by apprentice boy actors, of whom there were only a few (typically no more than four) in a playing company, which limited the number and lengths of the female roles he could write. The gender imbalance is also somewhat due to Shakespeare mostly choosing to tell stories about powerful people, which in his time meant mainly men.

    Wilson describes some quantitative studies that did not need computational assistance, including Erne's tallying of the early editions of early modern dramatists, attempts to tally from performance records which of his plays were most popular in the theatre across the ages, and studies of which plays get taught most frequently in universities and which plays get written about most frequently by scholars. Wilson approves of such gatherings of raw data, but sees them as only the start of the process since numbers do not speak for themselves but need interpretation. This is of course quite true. The last third of his article is about Wilson's pedagogic practice and has nothing of relevance to this review.

    Mariko Nagase argues that in preparation for reprinting the 1623 First Folio, the editor of the 1632 Second Folio eliminated stage directions for flourish and sennet because doing so was a recognized way of making a theatrical text more literary, as had been established by Ralph Crane in his work creating manuscript copy for the First Folio ('Traces of Jonsonian Neoclassical Editorial Convention in Shakespeare's Second Folio', Shakespeare Studies (Japan) 56[2018] 35-60). For the most part the Second Folio simply reprints the First Folio, but for some reason certain stage directions were removed in the reprinting. The First Folio has 105 occurrences of flourish of which only 64 make it into the Second Folio, and 18 occurrences of sennet of which only 12 make it into the Second Folio. Nagase tabulates which plays have these terms and how many there are in the First and Second Folios.

    Matthew W. Black and Matthias A. Shaaber thought that the editor of the Second Folio simply did not understand the meaning of the word sennet, which had indeed become archaic by the 1630s. But when changing sennet to sonnet the Second Folio editor might simply have been imposing an alternative spelling, since sonnet was a valid spelling for the sound cue. The plays in the First Folio that were printed from Crane transcripts almost completely lack calls for offstage sound effects. Crane's habit of massing mid-scene entrances at the beginning of a scene necessarily entailed suppression of accompanying mid-scene sound effects, such as a flourish for royal entry. The Second Folio editor may simply have been extending this practice.

    Nagase considers some quarto/Folio differences in 2 Henry 4 in the light of the fact that ". . . the copy-text of F1 2 Henry IV was prepared by Crane, who 'could have removed "theatrical apparatus" in his copy' . . ." (p. 56). Here the words "could have . . . in his copy" are attributed to an article by John Jowett, although Nagase gets the location of these words wrong, citing Jowett's pages 284-285 when the words actually appear on page 283 of Jowett's essay "Cuts and Casting: Author and Book-Keeper in the Folio Text of 2 Henry IV" in AUMLA: Journal of the Australasian Universities Language and Literature Association 72 (1989): 275-295.

    More importantly, Nagase's phrasing implies that in his essay Jowett agrees with her that Crane created the manuscript from which Folio 2 Henry 4 was printed. But Jowett's essay does not identify Crane as the scribe of Folio 2 Henry 4 and although Shaaber tentatively suggested this attribution it is not generally accepted. Nagase does the same thing again when writing that "Crane's copy-text for the F1 play is believed to have been either a theatrical script or another scribal manuscript of the theatrical script staged after 1606" (p. 56) and cites in support of this sentence pages 93-99 of René Weis's edition of 2 Henry 4 for The Oxford Shakespeare series. But Weis's edition does not identify Crane as the scribe of the Folio copy for 2 Henry 4 either.

    The point of Nagase's digression into the quarto/Folio relationship of 2 Henry 4 seems to be that Crane's elimination of some stage directions as he copied a play tells us that scribes often did this when preparing plays for publication, which might explain the reduction in occurrences of flourish and sennet when the First Folio was reprinted as the Second Folio. The problems with this thesis are obvious: there is no agreement that Crane did this to Folio 2 Henry 4, nor any reason to think that those making the Second Folio were following Crane's lead, and in any case a large number of occurrences of flourish and sennet survived the process and ended up in the Second Folio. Nagase does at least address the last of these and decides that the supposed Second Folio editor "recognised he was acting beyond his commission [to merely correct errors] and decided to leave more than half the directions untouched" (p. 59). Thus, according to Nagase, the process begun by Jonson in making his plays look neoclassical--that is, the classicalizing that Ralph Crane was known for--continued into the 1630s.

    Jonathan Lamb offers evidence that the additions to the anonymous play Mucedorus that were first printed in the 1610 edition are Shakespearian in character, although he does not quite commit himself to the assertion that Shakespeare wrote them ('William Shakespeare's Mucedorus and the Market of Forms', Renaissance Drama 46[2018] 57-86). Without these additions, the audience does not know that Mucedorus is a prince in disguise. As well as making this disguise clear, the additions give Mucedorus the opportunity to tell the audience about the power of, and the purpose for, adopting it. They also give irony to various other characters' remarks about how this lowly shepherd has a nobility of character to him. All these effects of the additions are typically Shakespearian practices. Lamb's Figure Two is an impressive photo-reduction of all the pages of the 1610 quarto with the lines that are additions shown with a shaded background; this is most useful for seeing at a glance where they occur in the book. The 1610 quarto of Mucedorus adds the adjective conceited to the title-page description of the play and this word was especially associated with Shakspeare's output: "Throughout the period, Shakespeare was the only playwright whose dramatic writing was described as 'conceited' on multiple title pages and for multiple plays" (p. 82).

    Emma Smith has found a new corrected proof sheet in the Shakespeare First Folio ('A New Corrected Proof Sheet from Shakespeare's First Folio (1623)', The Library 19[2018] 69-72). The handwritten corrections are on page rr2r (containing part of King Lear) in the exemplar once owned by the actor David Garrick and now at Queen's College Oxford. This adds a sixth proof sheet to the five previously discovered. The new sheet adds no new variants to those already known, but it helps settle one debatable variant. Some exemplars have Lear refer to Cordelia's husband as "the hot-blooded France" and others have "the hot-bloodied France". Other corrections on the same forme show that "hot-bloodied" is the state of the type after correction, but it is possible for a compositor to misunderstand a corrector's intentions and make changes that were not called for. This new proof sheet shows that the corrector did intend this correction: there is "an insertion mark in the text and an 'i' just visible in the gutter" (pp. 70-1).

    Emma Depledge has found that although Q6 and Q7 Hamlet are dated "1676" on their title pages, the latter was in fact printed in 1683-4, to judge from the fact that a watermark in Q7 is found in other books from 1683-4 but none from 1676 ('False Dating: the Case of the '1676' Hamlet Quartos', PBSA 112[2018] 183-99). Depledge reckons that Richard Bentley probably financed the edition, in collaboration with Jacob Tonson and the printer Robert Everingham, again on evidence of shared watermarks and also Stationers' Register entries. In the journal Shakespeare Studies, Amy Lidster published a condensed version of the evidence and argument presented in her article "At the Sign of the Angel" reviewed above from its appearence in Shakespeare Survey.

        And finally to the round-up from Notes and Queries. Brian Vickers argues that Thomas Kyd wrote the anonymously published play Arden of Faversham ('Verbal Repetition in Arden of Faversham: Shakespeare or Kyd?', N&Q 263[2018] 498-502). In the play's Quarrel Scene, Scene 8, there is an exchange that repeats the word love: "[MOSBY]  It is not loue, that loues to anger loue. | ALICE  It is not loue, that loues to murther loue. | MOSBY  How meane you that? | ALICE  Thou knowest how dearly Arden loued me". Vickers objects to the way that MacDonald P. Jackson handled this evidence in his article "Shakespeare and the Quarrel Scene in Arden of Faversham" (reviewed in YWES for 2011), claiming that Jackson's account of rhetorical devices was inaccurate.

    More generally, Vickers objects that in counting matches between passages in Arden of Faversham and passages in Shakespeare, Jackson accepted passages that were undoubtedly written after Arden of Faversham "and were therefore likely to be the result of recollection . . . or of imitation" (p. 499) rather than shared authorship. Vickers instead focusses on plays by Kyd that pre-date Arden of Faversham, starting with The Spanish Tragedy's passage "Drew forth the manner of my life and death. | This knight (quoth he) both liu'd and died in loue: | And for his loue tried fortune of the warres, | And by warres fortune lost both loue and life". Vickers seems to think that this passage is self-evidently like the quotation from Arden of Faversham, and moves on to another from The Spanish Tragedy: "Yet what auailes to waile Andreas death, | From whence Horatio proues my second loue? | Had he not loued Andrea as he did, | He could not sit in Bel-imperias thoughts. | But how can loue finde harbour in my brest, | Till I reuenge the death of my beloued. | Yes, second loue shall further my reuenge. | Ile loue Horatio my Andreas freend, | The more to spight the Prince that wrought his end:|  And where Don Balthazar that slew my loue". And then a third: "O saue his life and let me dye for him, | O saue him brother, saue him Balthazar: | I loued Horatio but he loued not me".

    Linking these quotations, Vickers offers literary critical remarks about the action of the play but nothing about the strength of these Kyd's passages as matches with the tightly patterned sequence of love . . . loves . . . love . . . love . . . loves . . . love . . . love in the quotation from Arden of Faversham. Then Vickers moves on to Kyd's Soliman and Perseda and again the parallels with the extract from Arden of Faversham are barely perceptible to this reader: "Did not I change long loue to sudden hate? | And then rechange their hatred into loue: | And then from loue deliuer them to death?", and "Calst thou me loue, and louest another better", "Parting him from his loue, in spight of Loue", "That though Loue winke, Loues not starke blinde", "Shee loued me deerely, and I loued hir", "If so your life depend vpon her loue, | And that her loue depends vpon his life", "And with her life, I likewise loose my loue, | And with her loue my hearts felicitie", and "[LOVE]  . . . Stung them both with neuer failing loue. | DEATH  But I bereft them both of loue and life. | LOVE Of life, but not of loue, for euen in death".

    Some of Vickers remarks are uncharacteristically vague, as when he observes that "In both tragedies Kyd made the terms 'love', 'life', and 'death' central to the action" (p. 500). In case the reader is unconvinced by these lines from Kyd, Vickers offers a second quotation from Arden of Faversham, "Nay he must leaue to liue, that we may loue, | May liue, may loue, for what is lyfe but loue? | And loue shall last as long as lyfe remaines, | And lyfe shall end, before my loue depart", and asserts that "The similarity between these lines and those found in Kyd's two tragedies is unmistakable, and provides a further reason to recognize this domestic tragedy as his unaided work" (p. 500).

    Vickers proceeds through Jackson's quotations from Shakespeare where the word love clusters much more thickly than in the Kyd quotations, and dismisses them each in turn on literary critical grounds: "The difference is due to the relative emotional moods of the dramatic situation" (p. 500), ". . . but we hardly feel concern for this serial deceiver" (p. 501), ". . . word-play here is calculated manipulation, not expressing genuine emotion" (p. 501), and ". . . but when restored to its dramatic context we can see that her words have nothing of the intense feeling expressed in Kyd's plays . . ." (p. 501). Vickers's conclusion is that authorship attribution has to be qualitative: ". . . each item needs to be evaluated in its dramatic context, paying strict attention to the sequence in which plays were performed, so that we can distinguish originality from reminiscence and imitation" (pp. 501-2).

    Thomas Merriam has three articles in Notes and Queries and they may profitably be taken together. In the first, he defends his previously offered minor redistribution of authorship divisions in Shakespeare and John Fletcher's play Henry 8 ('Henry VIII, All Is True?', N&Q 263[2018] 84-8). The New Oxford Shakespeare assigns to Shakespeare the parts of the play it numbers as 1.1, 1.2, 2.3, 2.4, 3.2.1-204, and 5.1 and Fletcher the rest, which latter is more than half the play. Merriam's book on Henry 8 reviewed in YWES for 2005 reassigned parts of some scenes, giving Fletcher 2.3.50-80 and 5.1.86-157 and giving Shakespeare 2.2.1-17, 2.2.116-142, 3.1.1-23, 3.2.228-235, 3.2.255-325, 4.1.37-80, and 4.2.31-99. The present article is a response to MacDonald P. Jackson's article 'All is True or Henry VIII: Authors and Ideologies' (reviewed in YWES for 2013) that resisted Merriam's reassignments.

    Merriam summarizes Jackson's evidence and points the reader to his earlier response to Jackson ('A Reply' reviewed in YWES for 2014) before offering new evidence for his claims. Merriam used the R Stylo software package to count the frequencies of the 79 most-common words in a set of 13 Shakespeare plays and seven Fletcher plays and then plotted the first and second Principal Components (PC1 and PC2) of the 79 data points for these 20 plays to produce his Figure One, in which the separation of the two authors along PC1 (the x axis) is visible. Then he adds in the nine passages from Henry 8 about which he and Jackson disagree on the authorship and reduces the number of words considered from 79 to 64--no explanation is given for this change--and the result is his Figure Two. The phrasing of the methodology is rather hard to follow, but if I understand it correctly then Figure Two seems to show that along the PC1 axis the disputed passages that Merriam gives to Fletcher are like the known Fletcher parts of the play and those that Merriam gives to Shakespeare are like the known Shakespeare parts.

    Merriam's second article argues that the play King John in the 1623 Shakespeare First Folio was co-authored by Shakespeare and George Peele ('A Contrast between Acts I and IV and Acts II and V in King John', N&Q 263[2018] 524-30) Merriam counts the frequencies of the words a, I, is, me, my, not, and you in in each of the 204 100-word chunks that make up King John and found that Act 1 has 10.4%, Act 2 has 6.4%, Act 3 has 9.4%, Act 4 has 11%, and Act 5 has 7.8%. In each case, the percentage is the proportion of all the words in that part of the play that are instances of any of those seven looked-for words. I replicated this exercise using the electronic edition of the 1986-87 Oxford Complete Works, first stripping out the speech prefixes but not the stage directions in order to match Merriam's method, and got roughly the same results.

    Merriam represents his results using a cusum chart, which I do not think tells us very much. Such a visualization highlights the phenomena of a counts rising above the average and falling below the average across parts of a text, and without comparison to other plays we get no sense of how a typical sole-authored play might exhibit such local rises and falls. Acts 1 and 4 of King John are favourable to Prince Arthur while Acts 2 and 5 are favourable to King John, and critics have remarked on the play's unevenness, which Merriam atttributes to the differing views of the two co-authors.

    Using R Stylo's most-frequent-word-n-grams and most-frequent-character-n-grams functions, Merriam counted the most frequent bigrams (both kinds) for six Peele plays (The Battle of Alcazar, The Arraignment of Paris, David and Bethsabe, Edward I, Act  1 of Titus Andronicus, and The Troublesome Reign of King John), 14 Shakespeare plays (1 Henry 4, 2 Henry 4, Much Ado About Nothing, The Comedy of Errors, Hamlet, King Lear, Love's Labour's Lost, A Midsummer Night's Dream, The Merchant of Venice, Othello, Romeo and Juliet, The Taming of the Shrew, The Two Gentlemen of Verona and The Winter's Tale), and the whole of King John and King John split into the putative Peele and Shakespeare sections as Merriam has previously determined, showing the results as a Principal Component Analysis scatterplot. Merriam records that he used "thirteen Shakespeare plays" but he lists 14 of them (p. 526).

    A setting in R Stylo was chosen so that only bigrams found in all the texts were included, so a zero count of any feature for any text would exclude that feature from the experiment. The result is that the part of King John that Merriam thinks is Peele's sits close to the Peele group, the part that he thinks is Shakespeare sits close to the Shakespeare group, and the whole play taken together sits in between the two groups. Merriam thinks that this shows that "The stylometric evidence favours King John as a play co-authored by Shakespeare and Peele" (p. 526). I would say, rather, that the evidence shows that if King John is co-authored by Shakespeare and Peele then the divisions that Merriam has proposed are plausible, since they put the two halves close to their respective authors' clusters. But Merriam does not here consider other candidate authors nor show that these results are not obtained when we arbitrarily divide a sole-authored-well-attributed play and test it the same way.

    Merriam then turns to Pervez Rizvi's Collocations and N-Grams dataset, pulling out the data for 4-grams and more-than-4-grams shared by King John and the two-part play The Troublesome Reign of King John (putatively Peele's). Of the 40 matches, over 32 are between The Troublesome Reign and just Act 2 and Act 5 of King John, the acts that Merriam thinks are Peele's. Merriam uses Rizvi's weighting formula and Mueller's calculation of how often plays by different authors share n-grams to quantify how strongly we should treat this evidence of 32/40 matches., Since no one has solved the problem of how we weight such counts to allow for the fact that different authors have differently sized canons, I find this part of the article less convincing than what precedes it. Merriam ends with the three long n-grams and collocations shared between Troublesome Reign and King John, well known to scholars and debated as either signs that Shakespeare wrote both or that one play copies phrases from the other. They are "[subjects: | subject and] For him, and in his right, we hold [this town | our town]", "them a bastard of the king's deceased", and "a will that bars the".

    Merriam's last article argues that we do not know that Munday wrote the original version of Sir Thomas More, as is commonly assumed from the fact that the manuscript is in his handwriting ('Who Did Not Compose the Original Text of Sir Thomas More?', N&Q 263[2018] 51-3). Jowett in his Arden3 edition of the play (reviewed in YWES for 2011) argued that it is inherently implausible that Munday would copy out a play that he had no part in composing. In two unpublished papers, Ward Elliott and Robert Valenza reported equivocal results from their tests for the authorship of Sir Thomas More and Merriam concludes that the matter is far from settled.

    One more article concerns authorship in the sense of being part of a larger effort to show that William Shakespeare of Stratford-upon-Avon had no part in writing the plays widely attributed to him (Rosalind Barber 'Shakespeare and Warwickshire Dialect Claims', N&Q 263[2018] 549-51). In a public talk for which Barber gives the YouTube URL, Jonathan Bate claimed that four words found in Shakespeare's works are Warwickshire dialect terms: "keech for a cake of wax, cradlecloth for a baby's blanket, dowle for a soft feather, and dey for dairy" (p. 549n3). Barber goes through each turn and shows its use in works not associated with Warwickshire.

    Barber points out that Abraham Veale used keech in his An A, B, C for Children of 1570 where it appears "in one of its word lists" (p. 550). This is not quite right, as Veale puts keech in an exemplary list of "English Terminations" (sig. B2r-B2v) in which he cycles through letter combinations ("Bych pych lych mych" and so on), rather than listing actual words. Barber's second example of keech is from John Florio Florio's First Fruits of 1578 but there, as her quotation shows, keech is used as a verb to mean serving meat or drink, not as a noun for a cake of wax as Bate claimed.

    Barber finds dowle meaning feather in Arthur Golding's 1567 translation of Ovid's Metamorphoses--a printed, non-dialectical source certainly known to Shakespeare--and other printed books with no connection to Warwickshire. Bate's third word cradlecloth is not a single word in Shakespeare but two words, cradle clothes, with no dialectical specificity: it just means the clothes in a cradle. Bate's last word, dey, is found in Old Norse and Old English and the OED has multiple uses of it before Shakespeare used it in Love's Labour's Lost, none dialectical. As Barber points out, in Love's Labour's Lost the word does not in any case necessarily mean dairy, as the Folio line is "shee is alowd for the Day-woman" (substantively the same in the 1598 quarto) meaning that she is approved to serve as one, and OED shows that a day-woman could be a female servant with no specific connection to a dairy. Thus Barber is right: Bate in his public talk should not have claimed that these are Warwickshire dialectical words and hence are evidence for Shakespeare's authorship of the plays. That claim rests on the firm foundation of other evidence.

    John Klause argues that an apparent crux in Shakespeare and George Wilkins's play Pericles does not need emendation, just explanation and repunctuation ('A Crux in Pericles, II.i', N&Q 263[2018] 542-3). In the 1609 quarto of the play, a fisherman says "what a man cannot get, he may lawfully deale for his Wiues soule" (sig. C3r). This is widely read as meaning that a man can always make money by renting his wife out to other men for sex. Klause finds this contradictory, since if he "cannot get" what he wants how come he can get it by renting out his wife? And how could such a renting out of her be lawful?

    To ease the awkward phrasing of "deal for his" editors usually emend to "deal for with his". But even so, it is an odd thing for the fisherman to say in this dramatic context, which is Pericles's expression of desire to compete at Simonides' tournament. Indeed, the meaning assumed by editors ought to offend Pericles but he does not react as if offended. Klause suggests that the sentence has nothing to do with a man prostituting his wife. We should, he argues, understand get to mean winning or acquiring love and deal to mean fighting for, which has a Shakespearian precedent in 1 Henry 4's "I never dealt better since I was a man", and soul to mean heart in the romantic sense. Then all is needed to make sense of the passage is repunctuation as "what a man cannot get, he may lawfully deal for--his wife's soul" (p. 543).

    Edward Paleit corrects what he sees as a misunderstanding about the state of classical knowledge among writers of Shakespeare's time ('Shakespeare's Faulty Learning? Classical References in 2 Henry VI and the Authorship Question', N&Q 263[2018] 506-11). The New Oxford Shakespeare drew on Paul Vincent's work, which in turn drew on the work of J. A. K. Thomson, to argue that errors of classical allusion in 2 Henry 6 point towards Shakespeare's authorship of the parts of the play containing these errors, since of the writing team he was the one without a university education and hence most likely to make these slips. In the first supposed error, the Queen refers in scene 11 to the god of the winds, Aeolus, having "brazen caves". Thomson thought this a mistake, but other early modern writers including Heywood, Nashe, and Marlowe (if he wrote the requisite part of Dido Queen of Carthage) refer to brass caves holding the wind.

    Thomson also thought that Aeneus not Ascanius told Dido of the destruction of Troy and that it was not Ascanius but Cupid disguised as Ascanius who bewitches Dido, so that the Queen's allusion to this story in scene 11 of 2 Henry 6 is wrong. These too are not errors, according to Paleit, but allude to a part of Virgil's Aeneid that Thomson and Vincent did not consider. Suffolk's remarks to the sailors who are about to kill him in scene 13 of 2 Henry 6 were read by Thomson as simple errors of knowledge, but a better understanding of early modern allusions to classical literature, especially those by Marlowe, shows them to just as plausibly be attempts to push at the boundaries of interpretation of classical stories.   

    Something goes awry with the referencing in Florence Hazrat's note on The Merry Wives of Windsor ('Psalms or Anything: New Evidence for Psalm Allusions in The Merry Wives of Windsor', N&Q 263[2018] 71-4). On page 71 she writes "(see below p.7)" but there is no page 7 below, and her footnote 2 seems to date the Modern Critical Edition of the New Oxford Shakespeare to "1777" (it was 2016) and the Critical Reference Edition of that project to "1666" (it was 2017). Also, in the body of her text she refers to the "Modern Critical Edition of the Complete Works of The Oxford Shakespeare of 1986" (p. 71) but those first three words were not used by the 1986 Oxford Complete Works of Shakespeare to describe itself.

    The substance of Hazrat's note concerns the moment in the Folio text of The Merry Wives of Windsor when Mistress Ford says that Falstaff's disposition and the truth of his words no more go together than "the hundred Psalms to the tune of Green-sleeues". The trouble is, no hitherto known collection of the Biblical Psalms has 100 of them; 150 is the usual number. Editors have generally emended, for example to "the hundredth Psalm" or "the hundred and fifty Psalms" (on the assumption that 150 or the Roman numeral cl got misread as 100). Hazrat points out that there was a popular tune called "The Old Hundreth" with lyrics derived from Psalm 100, but Hazrat thinks that whether you could sing it to the tune of Greensleeves is disputable. I do not see how this presents a problem: surely Shakespeare's joke depends on the words and the tune not going together.

    Hazrat points out that a Dutch version of the Psalms containing exactly 100 of them was printed in London in 1561 and suggests that Shakespeare wrote, or his printer set, "the hundred Psalms" in The Merry Wives of Windsor "because of an unconscious memory of the Dutch collection" (p. 74). The Dutch collection was written by Jan Utenhove, who was active in the Dutch Stranger Church in London, which was known for its evangelicalism. Perhaps, Hazrat suggests, connecting Falstaff with Dutch Protestant evangelicalism was Shakespeare's way of alluding to the character's origins in a version of the Protestant martyr Sir John Oldcastle. Thus Hazrat thinks editors should not emend this crux but merely explain it.

    Natalie Roulon reckons that the ballad of King Cophetua and the Beggar Maid contributed more to Love's Labour's Lost than has been acknowledged (''King Cophetua and the Beggar Maid' and Love's Labour's Lost: Echoes of the Ballad in the Play', N&Q 263[2018] 521-4). It has long been accepted that the ballad is referred to in Love's Labour's Lost, but Roulon thinks an additional allusion to it has been overlooked. Mote sings "If she be made of white and red, | Her faults will ne'er be known, | For blushing cheeks by faults are bred | And fears by pale white shown. | Then if she fear or be to blame, | By this you shall not know, | For still her cheeks possess the same | Which native she doth owe". Roulon finds this to be a parody of part of the seventh stanza of the ballad: "The beggar blushed scarlet red | And straight again as pale as lead | But not a word at all she said she was in such a maze".

    Don Armado refers repeatedly to his love Jaquenetta being base, and in stanza 7 of the ballad the beggar maid refers to her own baseness. Also, the ballad mentions King Cophetua's disdain of women, which is like Navarre's decree against women coming within a mile of the court. Cophetua goes to his palace gate to meet the beggar maid, just as Navarre's group meet the visiting women at the gate between the royal park and the palace. Boyet in the play says that Navarre's face shows "amazes" once he is entranced, and it is said of Cophetua that the beggar maid "did 'maze his eyes". Roulon traces a few looser parallels: giving a loved one a chain, asking a name, making a curtsy, and blaming Cupid. The play and the ballad use the fairly rare word salve for being healed of heart-sickness. With all these allusions to the ballad, it would, Roulon suggests, make sense for Mote to give a snatch or two from it when he is asked to sing at the start of Act 3.

    Adrian Streete proposes a link between Measure for Measure and Jonson's play Sejanus's Fall ('Sejanus, Measure for Measure, and Rats Bane', N&Q 263[2018] 75-6). If Shakespeare played the role of Emperor Tiberius in Jonson's play, as Philip J. Ayers proposed, then he would have been learning the part around the time he was writing Measure for Measure. This might explain the parallel phrasing of Jonson's "We that know the evil | Should hunt the palace-rats, or give them bane; | Fright hence these worse than ravens, that devour | The quick, where they but prey upon the dead" and Shakespeare's "Our natures do pursue, | Like rats that ravin down their proper bane, | A thirsty evil; and when we drink, we die".

    It would be typical of Shakespeare, remarks Streete, to turn Jonson's noun raven (one who eats the living and the dead) to the verb ravin (seizing and devouring food like a raven). Streete's argument does not really need Ayers's speculation about which role Shakespeare played in Sejanus's Fall since the 1616 Folio edition tells us that Shakespeare was one of the original performers, so would have known the play no matter what part he performed. Streete might also have mentioned the suggestion in the New Oxford Shakespeare that Shakespeare's was the "second pen" whose contributions to the play were suppressed by Jonson, as he tells us in the 1605 quarto.

    Frederick Kiefer finds a source for Hamlet's "What a piece of work is a man" speech in Nicholas Lesse's An Apology or Defence of the Word of God publised in 1548 ('Hamlet's 'What a Piece of Work is a Man'', N&Q 263[2018] 74-5)). Lesse's book reads "O what a goodly pece of worke, and wel framed building wold this be? How even and well proporcioned a matter, how excellent a frame", which matches the play's "this goodly frame, the earth, seems to me a sterile promontory. This most excellent canopy What a piece of work is a man! How noble in reason". Why might Shakespeare read Lesse? Because his Apology was appended to Lesse's translation of Philip Melanchthon's The Justification of Man by Faith Only. Melanchthon was a friend of Martin Luther and like him taught at the University of Wittenberg.

    Inspired by Julie Maxwell's investigation into the influence of the work of the Swedish brothers Johannes and Olaus Magnus on Hamlet, Rhodri Lewis has concluded that Olaus Magnus's Historia de Gentibus Septentrionalibus [History of the Northern Nations] gave Shakespeare his "monsters of the deep" image for humanity preying upon itself in King Lear ('Shakespeare, Olaus Magnus, and Monsters of the Deep', N&Q 263[2018] 76-81). The idea of eating yourself up recurs in Troilus and Cressida and Lewis distinguishes this special monstrosity from merely preying on your own kind, as mentioned in Sir Thomas More and Coriolanus. For the latter, Lewis cites Martius in 1.1 telling the rebellious citizens that without senatorial restraint they would "feed on one another", but, as Ed Pechter has pointed out to me in private communication, Lewis overlooks Volumnia's "I sup upon myself" in 4.2, which is precisely the special kind of monstrosity he is interested in. In Book 21 of his Historia, entitled "Of the Monstrous Fish", Olaus Magnus describes a fish that eats itself when hungry. The accompanying picture of this fish might have appealed to Shakespeare as a version of the ouroboros, the serpent that eats it tail, which was widely understood as an image of the universe's perpetual self-regeneration.

    As the source for Prospero's speech "the great globe itself, | Yea, all which it inherit, shall dissolve", Naseeb Shaheen proposed the Biblical source "the heavens shall pass away with a great noise, and the elements shall melt with fervent heat . . . Seeing then that all these things shall be dissolved . . . [at] the day of God, wherein the heavens being on fire shall be dissolved, and the elements shall melt with fervent heat" (2 Peter 3.10-12). But Manabu Tsuruta reckons that a closer source is Psalm 75's "The earth and all the inhabitants thereof are dissolved" ('Psalm 75 and Prospero's 'Great Globe' Speech in The Tempest', N&Q 263[2018] 83-4). In Romeo and Juliet, Capulet's Wife and the Nurse agree that Paris is a flower: "Verona's summer hath not such a flower" and "Nay, he's a flower, in faith, a very flower". Herb Paris (often not capitalized) was actually the name of a flower, also known as Truelove, and Lisa Hopkins traces how Shakespeare got his botanical knowledge, including this detail, from his reading and from various people we know he associated with ('Herb Paris, Romeo and Juliet, and Thomas Hesketh', N&Q 263[2018] 530-3).

   The following exchange in Antony and Cleopatra clearly has cosmological significance: "CLEOPATRA  If it be love indeed, tell me how much.| ANTONY  There's beggary in the love that can be reckoned.|  CLEOPATRA  I'll set a bourn how far to be beloved.| ANTONY  Then must thou needs find out new heaven, new earth". Peter D. Usher explores Shakespeare's possible reading about the nature of the universe, especially whether it is finite or infinite, that might have shaped this exchange ('Shakespeare's Antony and Cleopatra and the New Astronomy', N&Q 263[2018] 81-3). He finds a new and previously unexplored source: Thomas Digges's poem A Perfect Description of the Celestial Orb published in 1576. In Digges's model "the stars extend outward 'infinitely' to where angels hold court" (p. 83), which being a new idea provides Antony's "new heaven", while Nicolaus Copernicus's sun-centered model provided Antony's "new earth".

    Joyce King argues that, like the fox of beast fables, Falstaff is cunning, outside of polite society, and always hungry, favouring chickens; they also share the particular trick of the sham death, Falstaff's being near the end of 1 Henry 4 ('Falstaff and Fox Fables: A New Source', N&Q 263[2018] 533-5). The fox fakes death to lure others into a false sense of security and then steal from them, just as Falstaff attempts to steal the honour of killing Hotspur, whom (an original stage direction tells us) he takes up on his back, which is the way a fox carries off his prey in many accounts and depictions.

    It is widely accepted that Shakespeare relied on the "Life of Marcus Antonius" in Thomas North's translation of Plutarch's Lives of the Noble Grecians and Romans for Antony and Cleopatra, Julius Caesar, and Timon of Athens, but Mark Houlahan and Aidan Norrie argue that he also used the "Life of Demetrius" for his representation of Cleopatra's baleful influence on Mark Antony ('Shakespeare's Antony and Cleopatra and Plutarch's Life of Demetrius', N&Q 263[2018] 539-42). The description of the funeral barge of Demetrius is particularly similar to the famous description of Cleopatra's barge ("The poop was beaten gold . . .") given by Enobarbus in Antony and Cleopatra.

    Jeanette Tran reckons that In Twelfth Night, Olivia's reflection on "how much the better | To fall before the lion than the wolf!" alludes to the Old Testament Book of Jeremiah, 5.3-6 ('An Allusion to the Book of Jeremiah in Twelfth Night', N&Q 263[2018] 537-8). The Israelites are warned of their idol worship and told that "The lion will slay those who accept their fate immediately; the wolf will destroy those who attempt to delay facing God’s judgment; [and] the leopard will tear to pieces those who are foolish enough to try to escape God's judgement" (p. 537). Cesario is the idol whom Olivia realizes she has been worshipping.

    According to Douglas Arrell, the property eagle needed to fly Jupiter in Cymbeline was probably the same property eagle used to fly Jupiter in Thomas Heywood's The Golden Age. Heywood claimed in the 1632 preface to his two-part The Iron Age that his plays were "acted by two companies, upon one stage at once" and the Revels Accounts for 12-13 January 1612 show the King's Men and the Queen's Men playing Heywood's The Silver Age together at court. This co-operation makes likely their sharing of an expensive property such as Jupiter's eagle. Perhaps this sharing was itself the prompt for Shakespeare to add the descent of Jupiter into Cymbeline--where it is rather extraneous and odd--some time after original composition of the play.

    Camilla Caporicci finds that an unacknowledged source for Venus and Adonis lines 229-240 is the Biblical Song of Songs, especially with the idea of a woman presenting her body to a man as an enclosed landscape for him to explore and enjoy ('A Reference to the Song of Songs in Shakespeare's Venus and Adonis (229-240)', N&Q 263[2018] 50-1). With this idea, Shakespeare overturns the Petrarchan image of the female as a deer to be hunted, instead making the male a deer to be indulged. When Shakespeare mentions musk-roses (as he does in A Midsummer Night's Dream three times), Mark Griffiths and Edward Wilson think that he meant Rosa mochata, a non-native import, not the native English field rose Rosa arvensis ('Sweet Musk Roses: Botany and Lexis in Shakespeare', N&Q 263[2018] 53-67). In this distinction, Shakespeare was following the period's standard sense of the term musk rose.

    Akihiro Yamada has found that the William Johnstoune who put his ownership inscription in one of the Tokyo exemplars of the First Folio is not the owner of the hand that made the annotations throughout that Folio ('William Johnstoune, Signatory in Shakespeare's First Folio, and its Owners', N&Q 263[2018] 551-6). We can tell because Yamada has found a sample of Johnstoune's handwriting from 1699 and it does not match the annotations. After a fascinating exploration of possible numerological punning in Shakespeare's Sonnets (1609), B. J. Sokol offers an emendation to Sonnet 76 that would fit with this punning ('Significant Numerology in Shakespeare's Sonnets 1609, and a Possible Alternate Emendation to Line Seven of Sonnet 76', N&Q 263[2018] 511-8). Line 7 in the 1609 edition reads "That euery word doth almost fel my name", which is meaningless, and Sokol proposes instead ". . . spell my name".

Books reviewed

Dustagheer, Sarah and Gillian Woods, eds. Stage Directions and Shakespearean Theatre, The Arden Shakespeare. Bloomsbury [2018]. 320 pp., £80, ISBN 978-1474257473

Jenstad, Janelle, Mark Kaethler and Jennifer Roberts-Smith, eds. Shakespeare's Language in Digital Media: Old Words, New Tools. Routledge [2018]. 216 pp., £120, ISBN 978-1472427977

Loffman, Claire and Harriet Phillips, eds. A Handbook of Editing Early Modern Texts. Routledge [2018]. 292 pp., £106.40, ISBN 978-1472474780

McCarthy, Dennis and June Schlueter 'A Brief Discourse of Rebellion and Rebels': A Newly Discovered Manuscript Source for Shakespeare's Plays. D. S. Brewer [2018]. 464 pp., £65.21, ISBN ‎978-1843844884

Shakespeare, William King John, ed. Jesse M. Lander and J. J. M. Tobin, The Arden Shakespeare. Bloomsbury [2018]. 408 pp., £12.99, ISBN 978-1904271390