University of Virginia Library

Bibliographical Methods for Identifying Unknown Printers in Elizabethan/Jacobean Books
by
Adrian Weiss [*]

The functions of the descriptive bibliographer and textual editor vary somewhat in purpose and scope but both depend upon knowledge of essential bibliographical facts about a book. Of primary importance is the relation between the identity of the printer of a book and the circumstances of its production. Early books divide into three classes, each of which presents a special problem in this context. First, the need for printer identification is obvious in books which lack a signed imprint and mention of the printer in in the Stationers' Register entry. The second class consists of books whose printer is given in the imprint and/or a Stationers' Register entry but which actually were shared with one or more other printers. Although it has been generally assumed that such printer identifications are trustworthy, the many instances of previously undetected shared printing in this class of books is demonstrable cause for suspicion. As a result, it is necessary to verify these

identifications in order to distinguish a third class consisting of books printed solely by the specified printer. Tentative printer identifications supplied in the revised STC upon the basis of ornamental evidence also must be subjected to this verification process. The significance of the identity of the printer(s) of a book and the distinction between shared and unshared books is self-evident. Descriptive bibliographers and textual editors should, as a consequence, settle these issues as completely as circumstances permit before proceeding with other aspects of description and analysis. However, they should approach the task with realistic expectations, given the unpredictability of the process of searching for an unidentified sharing printer.

The purpose of descriptive bibliography is to record two kinds of information about a book: its physical characteristics and corresponding inferences about its printing history; and other details which may contribute to the general understanding of early printing.[1] Hence, a bibliographer who is working on a specific project involving a group of books or is on the staff of a rare books collection should, as a matter of course, take a long-term perspective by examining every book for evidence of shared printing. Even if the sharing printer(s) cannot be identified, the divisions of labor should be recorded along with any data that may provide clues for the future identification of the

sharing printer(s). Similarly, evidence that confirms the identity of the assigned printer in unshared books is an essential component of a bibliographical description. Typographical evidence is of primary importance in printer identification. The bibliographical descripton of the typefaces found in a book has been discussed in some detail by Fredson Bowers.[2] Although limited to generic typographical characteristics (style of face such as roman, italic, or black letter, and 20-line height), such information can save valuable time otherwise wasted in surveying irrelevant books: a note that type of a specified style and size appears in a book as the text font or that it appears only as the emphasis font, or no mention of a particular type in the description (indicating that it does not appear at all), permits an immediate decision as to whether to examine the book for typographical evidence or simply disregard it. To this may be added the description of more specific information about the fonts found in a book consisting of "gross features" (easily observed font differentiae such as mixed capitals and/or the foul-case cluster). This category of typographical evidence provides the initial clues to the identity of a font and is of utmost importance in the printer-search process. A match-up of the foul-case letters found in a target font in a shared book with those listed in a bibliographical description can lead directly to the unknown sharing printer.

On the other hand, the textual editor is absolutely bound to resolve the issue of shared printing and to define the sections of a shared book before proceeding with the analysis of setting and presswork upon which inferences about the evolution of a text depend. A complete execution of editorial responsibility extends to the identification of the sharing printer(s), the key fact which links the evolution of a text to the production methods and personnel of particular shops and the known manner in which they can affect the textual transmission process. At minimum, a textual editor should attempt to identify the section(s) of a shared book printed by the primary printer, provided that he is identified by the book's imprint (or STC assignment) or can be inferred tentatively from ornamental stock appearing in the book.[3]

The process of settling the shared printing issue, distinguishing sections, and searching for the printers of a book can be a time-consuming, frustrating, and ultimately a sometimes futile effort. However, such an outcome need not reflect upon the calibre of the investigator, given the vagaries and erraticness of early printers' practices. While the search process itself should be guided by logic, the phenomena that it seeks to unravel are sometimes illogical, unpredictable, and off-the-cuff responses to a vacillating business situation. Early printers apparently were reluctant to turn a potential printing job back into the street; rather, part of the job went out the back door and down the alley to another printer who could take it on at a moment's notice. The search for unknown printers is facilitated by a systematic approach grounded in an awareness of the various irregularities that occur in the business and printing practices of early printers. This paper is intended to provide bibliographical scholars with a knowledge of the kinds of evidence which suggest shared printing, methods of avoiding pitfalls in interpreting that evidence, and methods of searching for and identifying sharing printers or printers of books lacking an imprint. For the most part, the principles and methods described in the following are derived from an analysis of the factors which contributed to demonstrable errors in printer assignments and oversights of shared printing as found in the bibliographical literature. As is typical of research limited to a quite small sample of the books printed before 1640, it can be assumed that the following discussion by no means exhausts the probable variations on the problem of printer identification and shared printing. Much work remains to be done. I hope that this paper will create an awareness

of the intricacies of the problem so that bibliographical scholars, armed with the potent tool of font analysis, will eventually complete this sketch of what actually happened in early printing.

I

Several kinds of facts normally noted in the routine examination of a book suggest that the font(s) in which it is printed be analyzed in order to settle the issue of shared printing. Strictly speaking, such facts are not "evidence" of shared printing but rather should be viewed as "clues" to that possibility. In general, these clues represent disruptions in the regular patterns established earlier in a book in one or more aspects of printing style or method of imposition and almost always occur at the boundaries between sections of a shared book. Despite this coincidence, the analysis of the font(s) in the book provides the only entirely reliable evidence that the disruption is attributable to sharing rather than to some irregularity in normal shop procedure during the machining of the book.

(1)

A disruption in the sequence of signature alphabets usually occurs only in long shared books but is not a necessary adjunct of sharing. Essays STC 18041, for example, runs through continuous alphabets although Eld printed Book 3 at 2S-3C and Simmes resumed with Book 4 at 3D. A sharing strategy based upon the division of a text into units such as books or parts or some other form of sub-heading may invite a shift in alphabets that was not necessarily implemented even in shared books. Stafford printed the title, preliminaries and part of the text (L-O2) of Part 1 of Essays STC5775 (1600), while Read(?) did Part 2 about a year later. Despite the gap in production, Parts 1 and 2 are continuously signed with the split at O2/O3. However, the two books of Treatise of Antichrist STC7120 (1603) are separately signed. Eld printed the title and preliminaries and most of Book 2 while Braddock printed Book 1. The division into three sections in An Apology STC19295 is accompanied by shifts to new alphabets which reliably indicate the general pattern of sharing. The title (with Eld's imprint), preliminaries, and Book 1 are signed ¶-E, whereas Book 2 commences with a new alphabet at [F1] running 2A-O; then the third shared section begins with a third alphabet running 3A-N4. In this instance, Eld's sharing strategy was obviously based upon the natural divisions by books and the flexibility in production schedule offered by non-sequential signing. Jaggard's machining of Book 2, in effect, was temporally independent of Eld's work on Book 1. At some point, Eld took over Book 2 after Jaggard's long stint 2A-O and began the third alphabet.

Perfectly logical sharing strategies in Elizabethan printing, however, must be held suspect until confirmed by font analysis. In each of the foregoing examples, font analysis reveals further sharing within sections: the font Creede-3 appears in B-K of STC5775 Pt. 1; Creede-4 appears in STC7120 2H-M; and Jaggard's three fonts alternate in 3H-K of STC19295. In general, no irregularity in signing occurs in short texts requiring only one alphabet,

although a few isolated instances can be found such as Vertumnus STC12555 where Okes printed A-C2v, and Eld began his section 2B-H with a repeated alphabet. It should be noted that both the preliminaries and first sheet of text are frequently signed A as a matter of course. Although fairly uncommon in quartos, a shift in pagination can provide a clue to a sharing division. For example, Islip paginates his first section of De Missa STC23456 1-325, and the sharing printer not only repeats the alphabet but also begins with the second page 1.

Signature-alphabet shifts occur sometimes in a particular class of books consisting of a primary title/text and an appended secondary title/text. It is not always clear whether the two (or more) texts were intended as sections of one work or as complementary works, or were simply printed together as a matter of expediency. Eld printed the first title (with his imprint) and preliminaries of Remaines STC4521 (1604), Simmes did the first text B-2H, and Eld the second title "Certaine Poems, or Poesies, Epigrammes . . ." and text accompanied by a shift to lower-case alphabet a-h. A simultaneous shift occurs from Simmes-S1 to Eld-Y1. However, shifts also occur in other instances involving only one printer. Jaggard appended Genealogie of Vertue with a repeated alphabet 2B-K to Anatomie of Sinne STC12465.5 [old 565] (1603) which ended at I2v; Jaggard-Y1a prints both texts. Finally, the theoretical possibility should be borne in mind that a shift between quarto printing in fours and eights could occur at boundaries of shared sections. For example, this occurs in A Reformation STC3906 after sheet P where the alphabet and pagination repeat along with shifts in the pica roman and italic emphasis fonts.

(2)

The sets of running-titles automatically change between shared sections of a book. Hence, running-title analysis is a fairly reliable method of detecting the possibility of shared printing, especially when a relatively large number of skeletons (e.g., four, five, and six) are found in a play-quarto, provided, of course, that the sets of titles and associated skeletons are correctly identified and charted. If fewer than four sets obtain, shared printing may nevertheless be suggested by a change in the method of imposition and number of skeletons.

In some instances, shared sections are immediately suggested by obvious differences in sets of running-titles. White's pica italic titles in EF of Isle of Gulls Q1-2 STC6212-13 are sandwiched between the large double-pica italic titles of A-D and GH. Irregularities in the style of capital letters in initial positions can be quickly noted. Braddock set an erect Granjon italic capital 'A' in both initial positions in A-C of An Answer STC12988, while Short (D-G), Snodham (H-L), Field (M-P), and Harrison (Q-V) set a variety of combinations in the second position consisting of the Granjon erect 'A' and swash 'A', and lower-case 'a'; an anomalous oversized swash 'A' occurs in

Short's section at D4, E4, G4; and Harrison set a 96mm roman 'A' in both positions at R1 and S4. In other instances, the text of the running-title changes between sections.[4]

However, the tracker of running-titles in shared books faces a dangerous pitfall when the sharing printers set their running-titles in the same italic type-face. Unless the sequential sets of titles exhibit some abnormalities that permit differentiating them, they are easily confused.[5] In Whore Q1 STC6501, the titles of AB and CD can be thus differentiated. On the other hand, the titles of E-K cannot, inviting the incorrect assumption that the section was printed by one printer and the incorrect correlation of some titles in EF with others in G-K. Similarly, the titles of AB and EF in Fools STC4963 have been confused as well as the titles in other plays. In these examples, the detection of the boundaries of the shared sections depends upon subtle differences in the sets of titles which can easily escape the eyes of even seasoned bibliographers.[6] The fundamental value of font analysis as the initial step in examining a book is obvious here. In each instance, a glance at the type below the titles reveals a night-and-day difference that cannot be overlooked and needs no subtle analysis. Whereas the titles of Whore F4v and G1, for example, are actually indistinguishable at high magnification, the radically different type-faces in F4v (Stafford-EFb) and G1 (Eld-Y1) are brutally obvious both in regard to font composition and general appearance. The same is true in Fawne Q1-2 STC17483-84, D4v (Purfoot-Y2) vs. E1 (Windet-S1) and in many other instances.

In general, it seems that such fortuitous shifts in typefaces occur at boundaries of shared books in the great majority of cases. Dutch Courtesan STC 17475 is an exception that underscores the difficulty of distinguishing same-face fonts both in running-titles and text. Y-fonts appear in both A-E and F-H, so that the boundary at E4v/F1 is not obvious at first glance, although

the first stage of font analysis can produce "gross features" evidence which leaves no doubt that a different Y-font (Jaggard-Y1b) prints F-H.[7] As a general rule, then, the most efficient and trustworthy approach in running-title analysis is first to compare the type fonts on the first and final pages of subsequent gatherings to ascertain whether the typeface itself is different on these facing pages. Overall, a leisurely execution of this procedure requires far less time than the frustrating tedium of sorting out units of, say, seventeen same-face italic letters on about sixty pages and then grouping them into sets of four or eight and charting their movement, including transpositions, through a book. Once a shift in typeface indicates a likely boundary such as at E4v/F1, the task of analysing the running-titles is considerably simplified because the hypothesis of sharing focuses efforts upon just two groups of titles in the two sections of the book.

(3)

In shorter books such as play-quartos, the sections of a shared book almost inevitably exhibit some differences in the various details of layout and setting cumulatively termed "printing style." Compositorial and textual studies of shared play-quartos usually take note of such differences but overlook the possibility of sharing.[8] Caution must be exercised in interpreting such evidence, since it could merely imply a shift between two compositors in one shop. For example, the two Windet compositors who used the Windet-S1, and the Windet-S2 and -F cases differ in some aspects of printing style. Eld's compositors consistently sign all four leaves of play-quartos during 1603-1606 (as do Purfoot's compositors), but after late 1606, the practice changes to the more common signing of only three leaves in most dramatic texts.[9] None of the five printers of An Answer STC12988 (1603) signs the fourth leaf, but Snodham's section (H-L) is the only one signed with roman numerals (H, Hii, iii), an easily recognized clue to the sharing in his section. Such peculiarities of printing style may seem, at first glance, to be potential clues to a printer's work in shared books. However, other factors could be responsible. In printing two of Jonson's dramatic quartos (Sejanus STC14782, 1605; Volpone STC14783, 1607), for example, Eld set the speech prefixes with an initial

pica roman capital followed by small roman capitals, a rare deviation from the standard use of italic in speech prefixes. The same setting style also appears in Braddock's printing of Poetaster STC14781 (1602) and Snodham's Alchemist STC14755 (1610). Thus it seems more probably attributable to Jonson's direct involvement in the printing of these plays than to a shop's practice. The two could be confused and lead to the erroneous suspicion that Eld printed sheets CD of d'Olive STC4983 (1606), given the possibility that Eld-Y1 and Eld-S1, used previously in Eld's books, could have been mixed to produce the font in CD (the fact that Braddock did not use an S-font eliminates him from consideration). Compositorial preferences are a primary consideration. Both Simmes's and Jaggard's compositors periodically switched to a lower-case initial letter in italic speech prefixes possibly because of shortages in the capitals, and / or they punctuated alternately with colons or commas instead of periods. The appearance of the former atypical practice in Isle of Gulls STC6212 (1607; see H) and Dutch Courtesan STC17475 (1605; see H3v-4) provides a preliminary clue to the possibility of Jaggard's involvement; Simmes is easily rejected since his pica roman font is in the wrong face.

II

A different kind of frustration characterizes the reliance upon ornamental stock as a means of identifying printers and detecting shared printing. Despite the fact that ornamental stock has long served as the primary evidence of a printer's work, a comprehensive catalogue has yet to be published and probably never will be. Hence, scanning the standard references to identify the owner of a particular ornament frequently (or usually, in my experience) ends in failure.[10] This seems to have happened to STC researchers as well since, in some instances, a printer has not been supplied for books in which ornamental stock appears. Likewise, shared printing has been overlooked when the stock of a second printer appears along with that of the printer given in the imprint. Furthermore, the reliability of ornamental stock as evidence of printer identity is somewhat lessened by the common practice of borrowing and the existence of duplicate castings and copies. The printer assignments in new STC that are based upon the identification of ornamental stock are thus considered tentative and are cited in brackets; these must eventually be verified by font identification.

Oversights of shared printing and incorrect identification of printers in the new STC seems reason enough to discuss in detail the difficulties associated with the reliance upon ornamental stock. The basic flaw in previous printer-research has been the sole reliance upon ornamental stock.[11] However,

a more accurate and trustworthy conclusion both about the sharing issue and the identity of a sharing printer can result if the appearance of ornamental stock is understood as an initial clue to the possibility of shared printing that must be followed up by font analysis and identification. A discussion of the various patterns of ornamental appearances seen in shared books will contribute to a more effective use of this kind of evidence.

(1)

In general, the usefulness of ornamental evidence increases with the length of a book and the opportunities presented by the kind of text for the insertion of ornamental stock. Unfortunately, no predicatable relationship obtains between the various patterns of sharing and the appearances of ornaments. Many short texts such as play-quartos and theological pamphlets present no opportunity for the insertion of sharing printers' ornaments after the preliminaries and beginning of the text, although this is not an inviolable pattern. The first section of a short shared text could in fact be given to a sharing printer while the printer cited in the imprint (usually accompanied by his ornamental stock in the title and preliminaries) completes the text without the insertion of ornamental stock. Thus Creede's initial Yamada / Creede T6 begins the octavo text of Essays Pt. 1 STC5775 while Stafford prints the final section (L-O); and Eld's rarely used 'W' begins the text of Malcontent Q1-2 STC17479-80 while Simmes prints the second section (F-H). Shorter texts consisting of two sub-titles or a major title and sub-title present the opportunity for the insertion of ornamental stock. In the octavo Of the Calling of the Ministerie Two Treatises STC19733 (1605), Roberts's imprint and ornaments appear in the preliminaries and his two fonts alternate in the first treatise in 1BC. "The Second Treatise" shifts to a repeated alphabet with Yamada/Creede initial I1 in 2B1 heading the text in 2B-K which is printed in Creede-4.[12]

In longer texts, divisions by books and chapters frequently occasion a shift in ornamental stock between shared sections which is almost inevitably accompanied by a shift in fonts except in instances of borrowing. Printers sometimes concluded intermediate shared sections with a favorite ornament to fill space even though the text continued in the next section. In general, it is therefore advisable to check for a shift in fonts before and after the textual appearance of an ornament or initial even when the piece is identified in the references. This approach usually clarifies whether a boundary between shared sections obtains. In some instances, the divisions of shared sections do not coincide with the points at which ornaments appear. For example, it seems clear that a decision to begin each chapter with an initial accompanied the jobbing out of An Answer STC12988 with the result that each of the six printers had the opportunity to insert at least one initial. However, only Braddock's 'H' (Chapter 1, A1) and Harrison's 'Y' (Chapter 8, Q1) appear coincidentally at the beginning of sections and chapters. Other initials are buried within sections (E2, H1v, K4, N4, P1v). The ornaments in the preliminaries and text of Seven Deadly Sinnes STC6522 (1606) present a confusing situation which can only be resolved by identification of the contextual fonts. Allde's imprint and ornaments appear in the title (McKerrow 207; see STC description of four states of the title), the dedication (McKerrow/Allde ornament 15 and an 'I' in 1A3), then two of Stafford's pieces (McKerrow 295b, Lavin/Danter initial I2) head the Induction 2A1-4r which is printed in Stafford-EFb; the short letter "Reader" in an unsigned, apparently extra, leaf begins with an unidentified factotum (similar to Woodfield/Field Fac. 1), is printed in Allde-S2, and concludes overside with an identified Allde piece (McKerrow/Allde 4). Examination of the papers reveals that Stafford's

preliminary gathering was mistakenly bound in between 1A3-4 of Allde's preliminaries. The text (2B1-O4) in black letter begins with an unidentified headpiece and Lavin/Danter I2 followed by a factotum similar to Woodfield/Wolfe Fac. 3 on C2v. The appearance of an identified piece in each of this sequence of sections would seem to suggest that Stafford then continued with the entire text. However, a special problem (see later discussion, note 41) arises in distinguishing the shared sections in the remainder of the book since it prints in black letter with pica roman as the emphasis typeface.

In unshared long books, external identification of ornamental stock is frequently unnecessary since some pieces recur throughout the book. However, a change in the manner of heading chapters and the appearance of new ornamental stock could be misinterpreted. Bearing Jaggard's imprint, Badges STC889 is continuously signed B-2G; each of the four textual units (Books 1-3 and "A Corollary") contains numerous chapters, each headed by an initial. A mixed setting style is established in Book 1 comprised of factotum 1 (C1, E1, F5, F8v, K2), and initial T1 (E4, F3v, F7, G6v, H5v, H7, I4v), with the remaining chapters headed by unornamented, large (7mm) roman capitals. The appearance of previously unseen initial A1 at N4 in Book 2; the shift to setting large roman initials almost exclusively in 20 chapters of Book 3 except for the previously unseen I1 at Q2 and a problematic initial T2 on 2C2v which seems identical to T1 except that single rather than double rules enclose it, making it seem a different initial (it actually is); and factotum 2 in Book 4 (2E1) should cause no difficulty because of the recurrence of previously seen ornamental stock and the mixed setting style. Ornament 1 begins both Book 1 and 2, T1 continues in Book 2 (O7, O8), and factotum 1 appears twice in Book 3 (Q2, R5v). Book 4 raises the possibility of sharing with its new factotum 2 and different printing style, but that can easily be attributed to the change in the kind of text. Moreover, the font which prints the whole book is easily recognized as Jaggard-Y1b.

In some long quartos, sharing may be hinted at by an apparent shift in ornamental stock, although a common problem is that usually too few pieces appear to permit such an insight. Hence, the erratic appearance of ornamental stock somewhat lessens its usefulness as a means of detecting shared printing since pieces may be scattered about in a book without much of a recognizable pattern or with just a few appearances across several gatherings. Furthermore, the intermittent appearance of ornamental stock at likely boundaries between shared sections can create the impression of a sharing pattern less complex than that which actually occurs in a book. Eld's stock and Eld-Y2 appear in Book 1 of An Apology STC19295, followed by Jaggard's stock and fonts in the first half of Book 2. However, a shift to a third alphabet reintroduces Eld-Y2 up to and after 3H-L, where Jaggard's fonts appear.

Even when a pattern of recurrence or the appearance of identified pieces is encountered in a shared book, it may not provide a basis for precise differentiation of sections, especially if the two printers print in alternate groups

of gatherings instead of in a simple division into long self-contained sections. De Missa STC23456 (1603) illustrates these difficulties. Bearing Adam Islip's imprint, the text of Books 1-3 is continuously signed B-2T3 but a repeated alphabet is prematurely introduced at [2T5] with the beginning of Book 4. The simultaneous shift from easily distinguished Islip-EFb, which uses Guyot capitals exclusively, to a Y-font suggests, at first glance, a simple division by books. However, the distribution of fonts in Book 4 reveals the actual sharing pattern: the Y-font prints 2A-L, 2P-R, 2T, and 2X; Islip-EFb prints 2M-O, 2S, and 2V. It is illuminating to consider the ornamental evidence against this background. The ornamental stock in Books 1-3 is extensive, including two identified head-pieces and one identified factotum in six appearances, nineteen initials in forty-seven appearances, and three factotums in single appearances. Together with the shift in signature-alphabets, a group of twelve new pieces in 2A-L easily points to a sharing division marked by six recurrent pieces (repeated alphabet assumed in the following): factotum 1 recurs at K2, L7v; 'S' at A1, E1; 'A' at A4v, C5, D4v, G2; 'I' at B1v, H1, K4; 'Q' at C1, D6v, I7v; and 'P1' at C5v, I1v, and R5. In addition, six other initials occur once: 'V' at A8v; 'D' at B5v; 'C' at B3; 'R' at E5v; 'F' at H4; and 'B' at L2v. However, the ornamental evidence is clear only about the 2A-L section of Book 4. Relatively few initials occur after 2A-L, leaving little to go on as to the precise division of sharing. The recurrence of 'P1' at 2R5 (previously seen at 2C5v and 2I1v) links 2R to the sharing printer of 2A-L; otherwise three additional initials appear in his later sections: the 'T' at 2P6; 'E' at 2T8; and 'V' at 2X1. The latter could possibly be confusing since another 'V' appears at 2A8v; it might suggest a third printer rather than one who possesses and uses two different 'V' in the same text. The recurrences of five initials from Books 1-3 permit the assignment of three gatherings of Book 4 to Islip and clearly indicate that the division of labor was not a simple "by books" job; rather, Islip alternated with the sharing printer in Book 4.[13] Overall, internal ornamental evidence clarifies the divisions of labor for most of the book with B-2T, 2N, 2O, and 2S assigned to Islip, and 2A-L and 2R to the sharing printer. "Gross features" font analysis is adequate for settling the assignment of the remaining six gatherings (2M, V to Islip; 2PQ, T, X, to the sharing printer) and confirming the rest.

In general, the appearance of previously identified ornamental stock in alternating gatherings of a shared section such as Book 4 of De Missa STC 23456 provides a short-cut to the assignment of some gatherings. All five recurrent initials noted above belong to Islip, so that a charting of initials through Books 1-3 to establish a recurrence pattern is unnecessary in this instance. However, this is possible only if a previous survey of a printer's work has established that such initials are part of his stock. A rather quick survey of long books (especially folios if available) in the proximate period by

the printer given in the imprint of a shared book will usually locate at least a few pieces which can tentatively establish his sections in the latter. However, in Islip's case, this approach is thwarted by an anomaly in the output from his shop (see later discussion).

Long folios generally contain many pieces so that an internal grouping of ornamental stock exclusive of external identifications frequently can provide an initial clue to the alternation between two printers and sometimes the actual boundaries of sections, provided that the sections are long enough for the recurrence of one or more pieces, or of several initials that belong to a particular alphabet. The latter situation merely hints at ownership by one printer, but is a valuable initial clue to the identity of some printers. For example, an initial from Islip's Blayney/Alphabets 4 and 5 rarely (if ever) appears alone in books which use several initials. Similarly, the Robinson/Braddock stock included at least eight (A B F H I O T S) large (41mm) initials enclosed by two outer rules separated from an inner frame by ringlets (see Plomer T158, F161, H163; Purfoot also used a similar 'T'; see Romish Spider STC5963.5, B1). A second Robinson/Braddock alphabet is characterized by human subject-matter on a filled black background. An appearance of a few initials from these recognizable alphabets calls for a check of the contextual font(s). Moreover, the shift in a long book to previously unseen stock calls for a comparison of the fonts at the previous border between gatherings. Adam Islip and George Eld are identified in the imprint (an extremely rare gratuitous clue) of A General Historie STC12374 (1608) so that the identification of their respective sections should be nothing more than a "mop-up" operation. The new STC assigns to Eld sections O-2D, 3A-3P, 4I-5F, and 5X-6C, with the rest to Islip. The error in the assignment of the first section is puzzling and illustrates the need for consideration of the pattern of ornament appearance as well as font shifts. Islip's ornaments and initials from his Alphabets 4 and 5 appear up to K5. A different set of five initials appears between L1-N. Then Eld's Mermaid head-piece (30mm x 150mm) and a large 'I' (48mm x 45mm) appear on O1. As far as I know, the Horseman-L at L1 and the "Honi Soit"-A at L3v do not appear previously in an Eld book, but the three others do, including the Monkey-D at L5, the Demon-W at M5v, and Scholar-T at N4v. Nonetheless, two recur in later sections assigned to Eld: the "Honi Soit"-A appears later in the first section at V2, in the second section at 3M6 along with the Scholar-T at 3I7, and in the third section at 4X5. These appearances overlap with 15 other of Eld's initials in the correctly assigned sections,[14] a fact that should be easily noted and interpreted as evidence of Eld's work in L-O. Moreover, a shift in fonts occurs at L1 along with the introduction of the new set of initials; why this was overlooked is puzzling, given the fact that other boundaries were detected in the absence of ornamental stock. The shift to Islip's font at 2E1 occurs fairly close to the

appearance of his Mermaid head-piece (34mm x 126mm) and initials P5 at 2E4 and K5 at 2E5v. However, font shifts without ornamental stock occur at 3Q1 [3M6], 5X1, and 6C1 [5P2] at some distance from the previous appearance of ornamental stock (in brackets). In short, it is not necessary to know a printer's entire ornamental stock in many instances if proper attention is given to the pattern of appearances and contextual fonts.

(2)

In general, printer identifications rely upon the recognition of ornamental stock previously identified in the references, or by the appearance of pieces in the context of a printer's other identified ornaments. Peter Blayney's discussion of "the pitfalls awaiting the unwary ornament-hunter" (Texts, pp. 491-497) documents the difficulties created by the common practices of the borrowing and the passing of ornamental stock among printers. The existence of duplicate castings and copies likewise can cause confusion. In one extraordinary instance, eight of Eld's ornaments and initials were specially forged for a pirated edition.[15]

Several strategic factors can impede the discovery of the owner of a given ornament when the piece is not identified in the references. With luck, a survey of the suspected sharing printer's books sometimes turns up other appearances of the ornament. However, in many instances, a given ornament was used very rarely, with appearances separated by a decade or more, so that such a survey can be quite time-consuming and futile, ending in an erroneous rejection of the suspected printer. Moreover, an appearance of the ornament may be found in a shared section whose printer is unidentified. The 'W' that begins Eld's section of Malcontent Q1-2 STC17479-80 (1604), for example, does not appear again until 1608, and not since 1598 when Gabriel Simson owned the shop. These appearances are obviously beyond the reach of a proximate survey of Eld's (and Read's) books so that my search for this initial was a dead-end.[16] An appearance of a borrowed ornament creates a similar confusion. The large "THE" logo frequently shared by Simmes and Eld, for example, was lent to Creede for London Prodigal STC22333 (1605), its only appearance noted by Yamada in Creede's books. The appearance of this unique logo in the title of Prodigal, its total absence in other Creede books, and its frequent appearance in Simmes's and Eld's books could lead one to doubt the veracity of the imprint of Prodigal. The suspicion that either Eld or Simmes ghosted the book, however, is laid to rest by the fact that Creede-4 printed the text. Once another tentative appearance has been located, the two ornaments must then be confirmed as being the same piece rather than duplicate castings or copies. When an ornament or its contextual stock have been verified as identical to those in the references and ownership thus established, the issue of whether the actual owner used the ornament in a given

book can be settled only by the identification of his font(s) as the context of the ornament's appearance. If an ornament appears several times in the context of a given printer's font(s), it is reasonable to assume ongoing possession of the ornament whether or not the printer actually owned it.

For purposes of identification, ornamental stock falls into two categories: (1) duplicate castings which are basically identical and thus must be distinguished by damage; and (2) more or less faithful copies distinguished by major or minor design differences. Differences due to design and damage can usually be detected through direct comparison of simple pieces such as most initials and xylographics. For example, Simmes and Eld frequently used nearly identical title page xylographic ornaments. Small leaves project inward at top-center from the volutes (large scrolls). Eld's right leaf (shaped somewhat like a parrot's beak) is oversized and the tip extends markedly below the left leaf (see Fools STC4963); Simmes's leaves are nearly symmetrical with the tips aligned (see Malcontent Q1-3 STC17479-81). Failure to note this fairly obvious difference probably occasioned the incorrect new STC assignment of Siege of Ostend STC18895 (1604) to Simmes, whereas it is Eld's version that appears in the title, followed by his Goat-head head-piece at 2A1, three of his initials, his pica italic, and Eld-Y1 at 1A2v-3. Jaggard also used the Simmes version of these xylographic ornaments as the outer portions of a head-piece in a folio setting (Foure-footed Beastes STC24123, 1607, ¶6, 2¶5). This particular head-piece (similar to Plomer 52) seems identical to the one used by Eld (see below). Large duplicates and copies like this sometimes exhibit major design differences that can be detected by a brief comparison of the pieces. Three versions of Plomer head-piece 52 were in use. In all versions, the centered cherub's head rising from a cluster of three fishes is flanked by two dolphin-like creatures. These are flanked by feeding rabbits in Simmes's version, but by feeding squirrels in the versions used by Eld, Jaggard, A. Hatfield (Maison Q2 STC10548, 1606), and Melchisadech Bradwood (A Copy STC6164, 1606, C4v; Concerning STC24719, 1607, A2; A Preamble STC18191, 1608, ¶3). The latter can be distinguished in part by the cherub's head detail. The left cheek is straight and forms an angle with the jaw in Hatfield/Bradwood/Plomer 52, whereas the cheek is chubbier and rounded in the Eld/Jaggard version. The tips of the curved leaves flanking the head nearly touch the halo-like protrusions from the head in the Eld/Jaggard version, while they are spaced about a halo's width from the haloes in the Hatfield/Bradwood/Plomer 52 version. Concentration on these two areas of details is adequate for distinguishing the three versions of the head-piece. However, large pieces with no such major design differences present more of a challenge.[17]

In general, ornamental stock is especially prone to a variety of damage because of the complexity of even simple designs. Straight borders and linear elements exhibit bends, gaps created by missing sections, or sheared sections bent away from the original line. In addition, curved elements typically exhibit flattening or wavy damage. Some comment about the differences in the initials of Alphabets 4 and 5 reproduced by Blayney (Texts, pp. 446-447) is useful for illustrating the difficulty of sorting out duplicate castings in a particular style. The best method for comparing two similar initials depends on the nature of the ornamental design. Ornamentation with recognizable subject matter such as human figures, animals, birds, and objects provides clearly defined areas within the initial for comparison (see Blayney W2, E7, H8, N13, C15 etc.). However, abstract, stylized arrays ("arabesques") of vines and leaves such as Blayney/Alphabets 4, 5, 9, and 12 lack recognizable forms for judging symmetry and proportions. It is best to examine a section at a time (e.g., the top third, the lower-right quadrant etc.) for discontinuities, gaps, bends, extra buds, shoots, tendrils, leaves and vines. For example, the diagonal vine passing through the bottom-left section of the upper counter of Blayney B4 shows a short stem or shoot that is lacking in B5. At the upper right corner, the tips of the two leaves touch on the extreme right border in B4, but not in B5. In addition, the letter in B4 is slightly smaller. Note also that the leaf directly beneath the left base-serif of N4 is notably smaller than in N5. Differences in size can be detected by measurements of the rules enclosing a letter (height, width, diagonal) and the spacings between points in the design. If the two initials are identical duplicate castings, the objective of a comparison is to detect damage, usually quite minor, that differentiates the two initials. For example, if B4 and B5 actually were identical, the two could be distinguished by damage at the extreme lower-right, where B4's vine exhibits a sinuous bend and is separated from the leaf. Note also the damage illustrated in States 1 and 2 of T4 (p. 450) and the missing outer section of the leaf at 10mm down and to the right of H5. The differences created by damage in some Alphabet 5 letters owned by Okes (17 letters) and Islip (at least 15 letters) can be noted by reference to Blayney's reproductions.[18] The dates of appearance are important in deciding whether these examples represent

two states of the respective initials (as with Blayney T4). In general, damage is progressive and irreversible. The damage to Islip's initials precedes the appearance of Okes's undamaged initials by several years, a fact which settles the issue.

Since damage is progressive and cumulative, an earlier appearance may lack some damage seen in a later state. A comparison of early and late states of an Islip factotum (Blayney/A3r) illustrates the point. Blayney reproduces a 1607 state (Texts, p. 492) in which two gaps appear in the top border, one large gap at the lower left, and three gaps in the bottom border. Only the gap at the top-left corner is present in the 1600 state (Les Reports STC5493, at *3,A1). Such damage to the rules enclosing a piece is not the firmest kind of evidence, but establishes a strong probability of identity. Hence, this gap, along with most of those seen in Blayney's reproduction, leave little doubt that the factotum at A2 of Six Books STC3193 (1606, Islip) is the same. Islip also possessed what appears to be an identical duplicate casting of this factotum (Blayney/A3r #1) that illustrates one kind of confusion caused by duplicates and copies. The two factotums are virtually indistinguishable except for the fact that A3r #2 is enclosed in double rules and is in much better condition. Both appear in Six Books STC3193 with the inner enclosing rules (the only rules in A3r #1) in obviously different condition: the rules are undamaged in A3r #2 (see Six Books, B1, 2M5v, 2S4v; also De Missa STC 23456 (1603), A2, B1, Z7v, 2F2v; Christian's Sanctuarie STC7113 (1604), B1, K1; and A General Historie STC12374 (1608), C4).[19] Similarly, Jaggard owned the duplicate castings T1 and T2 (noted earlier) which can be differentiated by single vs. double rules; both are characterized by the strange bulge of the whole arabesque array in the upper right quadrant which possibly could be explained in terms of the uneven depth of the mold. Again, simultaneous appearances in one book with differentiating damage leaves no doubt that two pieces are involved (see Foure-footed STC24123, T1 at B3v, T2 at P5, Q6; also Badges STC889 noted earlier).

Large copies and duplicates that exhibit only minor design differences are especially troublesome when the issue is borrowing vs. sharing. An assignment crux arises from the appearance, in the context of identified ornamental stock of one printer, of pieces that are extremely similar to or identical to those known to have been used by another printer. The two samples must be painstakingly compared for differences produced by design or because of the imperfect copying of the original version. In some instances, such differences are so clearly due to design that the issue of damage is irrelevant. However, the absence of distinct progressive damage can leave the question hanging if the ornaments are duplicate castings. These problems are exemplified in

the preliminaries at 1A1-4v of An Answer STC12988 (1603) where the Eliot's Court ornament McKerrow 352 appears on the title page, followed by three ornaments which seem identical to those used by Read and then Eld. This raises the questions: did Eliot's Court lend McKerrow 352 to Eld specifically to be included in the title, with Eld then filling in 1A3,4,4v with his three ornaments and an 'M' initial, or did Eliot's Court own duplicates of all four Eld pieces, or were the pieces borrowed from Eld? The STC assignment of the preliminaries to Eliot's Court is based upon either of the two latter options.

Two of the head-pieces are faithful copies, each with more than one known version. The Eld/Jaggard version of Plomer 52 appears at 1A3; as noted earlier, Bradwood, who joined the Eliot's Court group in 1602, and Hatfield used the Plomer 52 version at least four times. Eliot's Court could have owned both versions, or replaced an earlier Eld/Jaggard version with the Plomer 52 version by 1606.[20] In any event, progressive damage along the upper half of the left edge strongly suggests that the piece in STC12988 is the same as that used by Eld.[21] The initial 'M' seen at 1A3 STC12988 undoubtedly is the same that appears in Eld's books.[22] Damage to two of three areas of the Turk's Head tail-piece in the state seen in 1A4v of STC12988 leaves little doubt that it also is the same piece used by Eld.[23] Finally, the second head-piece, a Goathead flanked by feeding-squirrels and grotesques, exists in two nearly identical versions differentiated by quite minor design

details. Eld's version can be differentiated from that used by Barker and Islip on the basis of at least six such details. On the other hand, design and damage are identical in Eld's version and that which appears in 1A4 of STC12988.[24]

In short, the preliminaries of STC12988 contain one unquestionable Eliot's Court ornament and four pieces which, for all intents and purposes, cannot be distinguished from those used continuously over a long period of time by Eld. So, while ornaments are usually more easily recognized and identified than fonts and frequently lead directly to the correct printer, assignment cruxes do arise that ultimately must be settled by the identification of the contextual font(s). Since A3-3v are set in a Guyot double-pica roman titling font almost never used in texts because few printers owned enough of this type for this purpose, the discovery of samples by either Eliot's Court or

Eld is a project clouded with potential futility. (As luck would have it, Eld used his Guyot double-pica roman [140mm; see Vervliet, Fig. 185, p. 248] to also set the epistle "From the King" in B1-2v of Obedience STC25633. The small size of the two samples is balanced by the small quantity of type in the font so that a dozen identifiable types recur in STC12988 and STC25633, and at least four recur in the two formes of STC25633.) In many instances that involve the more common pica and 96mm roman fonts, it seems more efficient to hunt after a printer's fonts since they are likely to appear with greater regularity than a printer's ornaments.

III

The identification of a printer's font in a sample text provides the only absolute evidence that he printed all or part of a book. The process of verifying a printer's ownership of a font in a shared book involves a survey of books in the proximate period by suspected printers in order to locate candidate fonts, followed by a comparative analysis to settle upon the most likely candidate. The target font in question and the candidate font are then examined for recurrent identifiable types which provide absolute evidence of the identity of the two fonts and thus establish the identity of the unknown printer of an unsigned book or a section of a shared book. The process is encumbered with pitfalls created by the seemingly inexhaustible inconsistencies in early printing. Sharing strategies frequently defy common-sense logic so that anomalous sharing patterns can be expected as the norm. Furthermore, the patterns of font usage varied widely among printers, a factor which frequently confuses the font ownership issue. Finally, the ownership question can be left hanging because of variations in shop output in the particular size of typeface in question during the proximate period.

In general, the process of sorting out the printers in a shared book usually begins with an initial clue as to the identity of one printer: a signed imprint, an identified ornament, or an entry in the Stationers' Register usually implies the assigned printer's ownership of one or more of the fonts found in the text although this assumption cannot be trusted categorically. In some instances, the printer specified in the imprint did only the title or the title and preliminaries. Moreover, some books lack such initial clues so that the search for the printer(s) begins in a vacuum with nothing more to go on than the fonts found in the book. For example, Whore of Babylon STC6532 (1607), tentatively assigned to Eliot's Court in STC, contains no initials or ornamental stock, but prints in two fairly distinct pica fonts in the classic two-section A-B sharing pattern (A4-G, H-L1). Except in longer books which permit insertion of ornamental stock, this situation obtains with shared sections nearly always in shorter books and invariably in play-quartos. Cancel titles and/or preliminaries in reissued or remarketed books present another kind of problem which must be settled by font analysis since the clues are sometimes confusing. Supplemental evidence provided by the distribution of papers (identified by watermarks) in a book (or books) frequently clarifies the

situation. For example, Islip's imprint appears in the cancel title on A2 (A1 blank) of the long Lectures STC7118, but a Dawson initial appears on A3 and the book prints in Dawson-EFc. Half-sheets A1-2 and A3-4 are printed on different papers. Islip clearly farmed out the entire book to Dawson but reserved the title for printing at half-sheet. Seven Sinnes STC6522 is an extreme case since the title page is found in four different states, each exhibiting a different printer's ornament (see STC description). The setting of the preliminaries is identical in at least two states, although changes occur in the title setting. The improper binding of 2A between 1A3-4 in the Huntington copy may provide a clue to the circumstances that produced this anomaly. In general, nothing can be trusted implicitly except an identified font; thus assignments based upon evidence in the title and preliminaries must be verified as a matter of course.

The survey process focuses upon books produced in the proximate period by either an assigned printer or the suspected sharing printer(s) in order to enhance the probability that a candidate font is the one in current use in the shop and in the same state as the target font.[25] Two kinds of books are

encountered in a survey. First, signed books printed in a single font imply ownership of the font in the absence of evidence to the contrary. Signed books printed in two or more fonts must be approached with an awareness of the possibility of shared printing although some printers used two fonts simultaneously. The survey process is affected by fluctuations in shop production. A decrease in overall production may be real, but in some instances is an indication that the shop was involved in printing unidentified sections of shared books. Moreover, the printing of books requiring other sizes of type can depress the use of the candidate font during the proximate period. In either case, the number of samples of a candidate font can drop dramatically. The varied patterns of font ownership and usage that are characteristic of Elizabethan printing thus can complicate a survey aimed at locating a candidate font in a suspected printer's proximate books. Three classes of font ownership and usage occur, each with its own set of subtle variations occasioned by the frequency, methods, and period of usage of the specific fonts that are encountered in a shop's output: (1) the use of a single font for an extended period, or a sequence of single fonts during shorter periods; (2) the use of more than one font either long-term or during transitional replacement periods; and (3) the continuous use of a font which was transformed at some point through replenishment, mixing or fouling.

(1)

The demonstration of ownership is a straightforward matter if the target font belongs to the class of printer who used a single font in a large number of books over an extended period of time in conjunction with his imprint and/or ornamental stock or entries in the Stationers' Register. Simmes-S1, for example, was the primary font used in this manner in Simmes's shop from 1594-1606. Similarly, White-M was White's primary font from 1597 until about 1608 when it was replaced with White-S1. Although Stafford-EFb appeared in a lower proportion of the books printed by Stafford between 1599-1607?, it nonetheless is the only pica roman font that he used. Creede-3 appeared in a large number of signed books 1594-1602. Given an initial clue such as an imprint or identified initial, a check of a few proximate books by such printers usually will produce a candidate font sample. For example, a survey of a few of Simmes's books from 1604 as suggested by his imprint and xylographic (opposed scrolls) in the titles of Malcontent Q1-2 STC17479-80, Q3 STC17481, and Whore STC6501 quickly leads to the identification of Simmes-S1 in his sections of these shared books (sigs. F-H, B-G, and AB respectively).

However, this ideal situation is fairly uncommon since most pica fonts were replaced at intervals of about 2-4 years, a factor which can present a problem. The survey of the suspected printer's books can yield samples of the wrong font in the sequence used by the correct printer if a font was replaced during the proximate period of the target book. If the sequence involves fonts in different faces, an extension of the survey beyond the proximate period is usually adequate to establish that a replacement occurred and when. The shifts from mixed Creede-3 to S-face Creede-4 (1603) to hybrid Creede-5 (1609) or from Allde-M to Allde-C2 (1591) to Allde-Y1 (1597) are easily established by an extended survey. In contrast, a sequence of two same-face fonts requires closer scrutiny since it is possible to confuse them as a single font in continuous use. For instance, Braddock printed regularly with his sequence of Braddock-Y1,2a from 1598-1601, 1602-1605, with a slight decrease in pica roman output 1601-1602. Y-font samples before and after this transitional period are common enough. Although Braddock-Y1 is easily distinguished by its foul-case cluster, failure to perform this stage of analysis could lead to the erroneous assumption that any Braddock Y-font book 1598-1605 uses the same font. Hence, a comparison of a target font (Braddock-Y1) predating the replacement point of 1602 with a later book (Braddock-Y2a) would lead to the rejection of Braddock as a potential sharing printer. (Further discussion of same-face sequences follows below.)

(2)

The appearance of two or more fonts in an assigned proximate book(s) by a candidate printer raises the possibility of shared printing so that the intermediate issue of his ownership of more than one font must be settled before work proceeds on the target text. In such a case, the pattern of font usage in

assigned books in the proximate period usually provides evidence of ownership. Ideally each font appears separately in a single-font book which can be unquestionably assigned to the printer. Otherwise, proximate books usually provide evidence of the simultaneous use of two or more fonts, especially when the fonts appear in an alternating pattern within gatherings or lesser units. A survey of proximate books should yield books to establish ownership in uncommon instances of a printer's two fonts alternating in sections consisting of gatherings in the manner typical of shared printing. Finally, an extended survey is sometimes necessary to establish the ownership of a replacement sequence of two fonts used simultaneously for a short period before the old depleted font was discarded.

In general, two situations occur: the long-term use of two fonts or the simultaneous use of two sequential fonts during a transitional period. In either instance, the simultaneous use can lead to transformations in font composition that affect the font identification process. Such transformations are usually progressive and irreversible. Reverse transformations through decontamination seem not to have happened except through purging of foul-case italic letters.[26] Hence, a commonsense assessment of the changes necessary to transform a font from suspected early to late states usually suffices to settle the issue. The possibility of the total replacement of individual sorts should be considered in the context of the number of affected sorts. It is conceivable that a printer decided to jettison entire 'a k y ?' sorts and replace them with newly-cast type. However, at this point, there seem to be no examples of this having occurred. When the balance or proportions are observed to shift dramatically to new replenished types in a sort, remnants of the original types are inevitably present. The transformation issue is sometimes critical in establishing the sequence of fonts in a shop during one printer's tenure or in deciding whether a font passed to a new printer along with a shop. In the matter of identifying a sharing printer, a target font in a shared section can be compared to either state of a transformed candidate font. If it can be established that the font was transformed at some point, text samples in the earlier state will contain recurrent-types that can establish the identity of the font in the later state in a target text (or vice versa).

In ideal situations, the ownership of two fonts is easily demonstrated by the interwoven use of two fonts in bibliographical units that eliminate shared printing. New Simmes-S2 prints most of Eliosto STC13509 (1606), but old Simmes-S1 appears at D3-E4v. Such a pattern argues against sharing because of the mutual appearance of the two fonts in the respective halves of a gathering. Ballard took over the two fonts and used them in a similar fashion in Merry Devil STC7493 (1608). Shop lineage, in this instance, provides added evidence of ownership of the two fonts. The alternation of Kingston-Y1 and -EFc, both used simultaneously for a few years, in T. Workes STC12316 (1601)

was probably occasioned by the demands of the large folio setting, although alternating compositorial stints are plausible, and leaves no doubt about ownership. For example, EFc sets 3D1-1v, Y1 sets 3D2:1-49, followed by EFc from 3D2:50-3D2v:38 with Y1 completing the page (:39-54). The shift to EFc on 3D3:1-26 is accompanied by mixing in of some Y1 sorts until Y1 takes over from 3D3:27-47, then EFc sets a long section 3D3:48-3D4:17, Y1 from 3D4:22-D5:34, and EFc concludes the gathering 3D5:35-D6v. The intricate interweaving of two fonts within page-length units in this manner is infrequent but valuable typographical evidence of ownership. The observation of the minor transformation of Kingston-EFc by the fouling with Y1 sorts in 3D3:1-26 provides additional internal evidence of the identity of EFc.

In some instances, the transformation is more extensive, consisting either of a partial or a complete mixing of two fonts in different faces. This kind of phenomenon is usually accompanied by the transitional use of the two in an alternating pattern which leaves no doubt as to ownership and transformation. The mixing of a replacement font into the previous font is illustrated by the creation of Creede-3. Creede-1 (S-face) is used to D in Selimus STC 12310a (1594), at which point Creede-2 (C2-hybrid) takes over; the two alternate, then are promiscuously distributed and fully mixed by the end of the book. The new mixed font Creede-3 immediately begins its appearances in a long sequence of signed books to 1602. The transformation of one case of Windet-S1 into Windet-S2 from 1605 by fouling with Windet-F letters (A D G H M O T a b g k w x y z) has been noted previously.[27] The process can be observed in Fawne STC17483 (1606) where Windet-F and -S2 alternate

by formes in Q1, but in the resetting of pages for Q2, they alternate within pages accompanied by some cross-contamination.[28] With the further contamination of Windet-S1 in Sophonisba STC17488, the clear distinction between Windet-S1 and -S2 becomes somewhat obscured by the low-density mixing of these and additional Windet-F capitals (E F P S) into Windet-S1, an increase in the proportions of these capitals in Windet-S2, and the mixing of Windet-F 'g w' into both fonts. However, Windet-S1 remains relatively free of Windet-F lower-case sorts. Though difficult, it would be possible to sort out the alternating stints of the two fonts in many pages of a later text such as 1 Henry 4 Q5 STC22283 (1608) and Romeo & Juliet Q3 STC22324 (1609) if the need arose in a compositorial analysis.

The transformations in the sequence of three fonts used by Eld 1603-1609 occur in the context of a simultaneous transitional use of Eld-Y1 to Eld-S1 and Eld-S1 to Eld-Y2 in interwoven settings during both transitions. The resulting cross-contaminations yield foul-case clusters in Eld-S1 and Eld-Y2 that aid in potential recognition of the two fonts if encountered in shared sections. Eld clearly planned ahead in acquiring new fonts, but for some reason pushed the old fonts to the brink of depletion, thereby occasioning the transformations of the new fonts. Although his compositors were quite careful, it seems that interweaving a depleted font with a new one presented an impossible distribution situation. Each of the fonts appears separately in signed books; nonetheless, Eld's ownership is clear from the use of the respective pairs of fonts in alternating patterns which eliminate the possibility of shared printing. Put into use in 1603?, Eld-Y1 was the only pica roman that appeared in books from the shop until early 1605, when Eld-S1 appeared in Survey (STC6201, 22 January 1605) with Eld's imprint and thereafter in several others (An Answer STC26002 etc.) to 1608?. The low-density S-capitals that appear in Eld-Y1 early on are possibly remnants from Simson/Read-S1; this suggests that part of Simson/Read-S1 was in the shop until replaced by Eld-Y1 in 1603? If so, it either was not used at all, or only in shared sections, although this seems unlikely because of the overall production of the shop during Read's tenure, which seems to have included but a few short ephemera and no books of noteworthy length after 1601. However, the fouling of Eld-Y1 with additional S-capitals and then with lower-case sorts begins

an upswing in the proximate period of the acquisition of Eld-S1.[29] Unfortunately, the actual transitional period is obscured by several factors, including discrepancies between imprint and entry dates, a common problem in sorting out font sequences.[30] During the nebulous transitional period, Eld-Y1 and Eld-S1 alternate in an anomalous pattern in Eastward Hoe! Q1 STC 4971.[31] Moreover, the sequence of Eastward Q1-2 provides a unique additional kind of evidence of ownership. Standing type in Eld-Y1 from Q1 migrates across page and gathering boundaries into Q2. In the process, lineation is disrupted as the measure is expanded to 100mm in HI to accommodate the transferred type.[32] Tied-up standing type could conceivably be transported between shops, but that unlikelihood is completely dismissed by the dovetailing of standing vs. reset type in Eastward Q2. Eld-Y1 makes a final appearance in two pages (E1v, F3v) of Q3 along with Eld-S1. The period of transition from Eld-S1 to Eld-Y2 likewise is somewhat nebulous due to uncertainty about the date at which Eld-Y2 began use, but Eld's ownership of the two fonts is clear from alternating patterns within gatherings. Eld-S1 appears heavily fouled with Eld-Y1,2 capitals at F2, G1v-2, H4v, I1-2, I4v, K3v-4v, L1v-2, M4-4v of Lingua STC24104 (23 February 1607) and at I(i), K2, M1, M3v-4v, N2v-3 of Volpone STC14783 (imprint: 1607; entry: 3 October 1610). The progressive fouling suggests that Eld-S1 was considerably depleted in the upper-case at the end of its lifetime; the lower-case was contaminated with Y-sorts throughout its lifetime. The process left Eld-Y2 contaminated with S-face sorts (A B C D2 E1 G2 H I L M O P R S T V Y b1 f1,4 g1 k1 n3,4 p1 x2; ligatures: ct fl2 ſi ſl ſt; and '?') in varying low-density proportions. There seems to be a concentration of fouled sorts in sections of later texts such as Troilus STC22331 (1609) which may be useful in compositorial analysis.

Transformations in fonts usually occur offstage. Unless the process is observed in progress, it must be inferred from the demonstrable ownership of the two fonts either in long-term or transitional simultaneous use. This is the case with the two states of Jaggard-Y1. Jaggard-Y1a appeared in at least three books before transformation: T. Anatomie STC12465.5 (1603), T. Triumphs STC18279 (1605), and T. Lamentation STC7606 (1605). At some point in 1605, Y1a was transformed into Y1b by the addition of C2-hybrid sorts. Two factors obscure the time and source of contamination. Jaggard-Y1b's first appearances in signed books occurs in 1606: Essays STC1139, and T. Badges STC889. However, Jaggard used the font earlier in his section (F-H) of Dutch Courtesan STC17475, entered 26 June 1605. The most likely source of the contaminating sorts was Jaggard-C2. However, the date of Jaggard's acquisition and first use of this font is unclear, but it seems to have been later than 1605. The date of its appearance in Westward Hoe STC6540 is confused by the contradiction between the crossed-out Stationers' Register entry of 2 March 1606 and the imprint of 1607. An alternative possibility is that, in order to extend the life of the font, Jaggard replenished Y1a with C2 sorts from the same foundry that supplied Jaggard-C2 before acquiring the latter. In any event, ownership of Y1b and Jaggard-C2 is not an issue because of Jaggard's practice of alternating these fonts within gatherings along with a third font, Jaggard-S1. Jaggard's section (A-G; Eld-Y2 in H-I2v) of Westward Hoe, for example, prints in Jaggard-Y1b except for Jaggard-C2 in D4 and F1, and Jaggard-S1 in B2v-3, C1, C2v-4v, E1-3, E4v, F1v-4v and G2v-3. The three fonts alternate in Jaggard's sections of An Apologie STC19295, shared with Eld, and Jaggard-C2 and -S1 alternate in T. Miseries STC25635 (1607).[33] The two were used at least once in alternating fashion as the emphasis fonts (T. Deade Tearme STC6496, 1608). Just why Jaggard acquired and used the three fonts is a mystery since Jaggard-Y1b was still adequate in 1606 for setting the long (B-2G) T. Badges STC889. Even so, the use of the three is confined to 1607, suggesting a transitional waiting period until Jaggard-Y2 was delivered in 1608 (Jesuites STC1824). The fonts may have been used more extensively than we know. The practice of alternating seems to have gone amuck and fouled one case of Jaggard-S1 in Sermons STC15882 (1607) where a mixture of Y1b, C2 and S1 sorts occurs in R6,7,8, S1,2v,3,4v and later. The font is largely uncontaminated in the short text in A2-A3v:15 of Ruine of Rome STC6641 (1607), but this probably is attributable to the use of the uncontaminated case. A transient phenomenon confined to 1607, the three fonts seem to occur always in alternating fashion and thereby provide a unique kind of font identification composite consisting of the pattern and the cross-contamination.

One circumstance is shared by the Islip-Y1a,b and Jaggard-Y1a,b transformations.

The contamination respectively with EFc and C2 sorts occurs in rough proximity with the acquisition of a new font. In both instances, a similarity in the nature of the transformation and the lapse between the introduction of the new font perhaps implies something about supply sources, that is, replenishment from a local source which supplied a particular hybrid set of letters. Islip-Y1a appears uncontaminated in 1598 in A Treatise STC 17291, but by 1599 exhibits the replenished EFc sorts in Prenobilis STC11196. Meanwhile, Islip-C2 makes its final(?) appearance in the same year (see T. Love STC19540). Islip-EFb first appears in 1600 in EMOH STC14767, which establishes its first date of usage and not necessarily its date of acquisition. Its appearance in the incorrectly assigned and dated (A. Jaffes, c1585) Articuli STC4584 (see Pica Roman Type, Plate 76, Jaffes-1) probably predates EMOH (8 April 1600). The fact that two versions of the EF-hybrid are involved raises a question about the connection between the two acquisitions, implying perhaps that two supply sources were involved, one of which supplied the EFbhybrid with Guyot capitals. Perhaps the replenishment of Islip-Y1a was an attempt to extend the life of the font, and once completed, Islip decided to order the new font Islip-EFb. This is plausible in view of the fact that Islip had obtained Islip-S1 by 1602. The font was used along with, but independent of, Islip-EFb until 1604; similarly, Islip-Y1b continued in use until 1602. Overall, a pattern emerges in which a sequence of three fonts was acquired and used separately during two-year overlapping periods, although this degree of precision may be illusory.

In general, the continuous use of a single font in two states can be demonstrated by recurrent-types in the early and late states. In some instances, a font is already contaminated in what seems to be its earliest appearance. Circumstances may indicate the plausible source of contamination. For example, Allde-Y1 appeared in 1597 (R & J Q1 STC22322; and Granados STC 16902, 1598) with a low-density wrong-face mixture in place, including S-face sorts (A1 D1 G H O T; low-riding C F1 G2 H I L M P; a d2 f1 h2 p1 u1; ligatures: ct fi2 fl2; and '?') and C2-hybrid sorts (b2 k1 g2), all of which are probably from Allde-C2 which last appeared in 1596. However, Jaggard-Y2 appears with oversized capitals (2.95-3.15mm) already in place which do not seem to come from Jaggard-S1 (A B C D G H I K M O P R S T V; see T. Pathway STC898, 1609).

The transformation issue can be clouded when the similarity exhibited by a sequence of fonts is such that transformation seems the logical explanation. The general principle that transformations are progressive usually permits settling the issue. Sorting the sequence of fonts in Purfoot's shop 1598-1607? is complicated by the fact that three same-face fonts are involved (see also later discussion of Robinson-Y1a,b). The sparse known output during the transitional periods between Purfoot-Y1, -Y2, and -Y3 forces an extension of the survey across 1601-1606 in order to locate samples of each font. The fiveyear period represents sufficient time for major transformations to have occurred. The early font, Purfoot-Y1 (last seen in 1601?) exhibits mixed Y- and

S-capitals, an obvious characteristic of Purfoot-Y2 (first seen in 1605?), inviting the inference that the two are the same.[34] A closer examination reveals that whereas Purfoot-Y2 exhibits an uncontaminated lower-case, Purfoot-Y1 exhibits a variety of moderate-density oversized wrong-face roman letters: (S-face: b3, p1; C2-hybrid: k1; B-face: a b1,2 c d e h i m n o p q r u; Guyot 'w'; 015). Although the dearth of output eliminates the opportunity to observe the transformation of Y1 into Y2, the hypothesis that Purfoot-Y1 capitals were mixed in with a new Purfoot-Y2 to produce the apparent continuity in the capitals can be rejected on the basis of the new condition of the latter. There is no doubt that the rest of Purfoot-Y1 was discarded since its contaminated state precludes the kind of modification that would be necessary to permit mixing the lower-case into Purfoot-Y2. Furthermore, the acquisition of uncontaminated Purfoot-Y3 during the four-year interval separating the use of Purfoot-Y1 and -Y2 clearly indicates a strategy of total replacement of Purfoot-Y1. On the other hand, the sequence of uncontaminated Purfoot-Y3 in 1603-5, followed in 1605 by Purfoot-Y2 with mixed S-capitals, looks like an internal transformation of one case of Purfoot-Y3 by replenishment with wrong-face capitals. The two-year period could conceivably deplete the uppercase of Purfoot-Y3, but the replenishment of just one case is improbable. The fact that both Purfoot-Y2 and Purfoot-Y3 were adequate for setting prose quarto texts in the same year (1605: Spider STC5963.5, Papisto-Mastix STC 17913, Downefall STC1819 etc.) as well as play-quartos contradicts the depletion hypothesis. The issue can be settled with finality on the basis of the general principle: if a seminal font is split into two through replenishment or mixing, recurrent-types will remain in both states. Recurrent-types first seen in 1603 are still in Purfoot-Y3 in 1605, but absolutely no sharing of recurrent-types with Purfoot-Y2 is detectable.

(3)

A fairly common problem arises if proximate books were printed during a transition in shop ownership, raising the question of whether a font passed to the new owner. On the one hand, a survey may reveal a gap of several years in the use of what seems to be a single font in intermittent use in a shop. In some instances, the interval is such that common sense rejects the notion that a single font is involved. Font analysis usually can confirm this view, especially if a significant transformation in the composition of the suspect font(s) would have been necessary. For example, a seven year gap occurred between Middleton's last use of a Y-font (1581), which was extensively contaminated with wrong-face letters, the passing of the shop to Robinson, and

the introduction of Robinson-Y1 (1588?). The fact that at least thirty lowercase letters and ligatures and most capitals had to be purged to transform the former into the latter leaves no doubt that these are different fonts. The four-year interval separating the last use of Read-S1 (1601) and the appearance of Eld-S1 (1605) is not as great, but only a transformation through a highly improbable decontamination could have produced the latter.[35] Read-S1 is contaminated with Y-face letters: the high-density 'a' (about 30%) and 'k' (about 40%) variants and low-density 'A D E F I M P p', and the low-density crimped 'w', which is not seen elsewhere at this time. Eld-S1 is uncontaminated in its first three appearances in 1605 and does not exhibit Y-face fouling until its use in Eastward Hoe! Q1. Moreover, fundamental differences are seen in the S-face composition of the two fonts. For example, Read-S1 contains worn t1,2 and moderate-density oversized t3 whereas the new t1 is exclusive in Eld-S1; the old worn S-k1 is dominant in Read-S1 except for moderate-density replenished S-k2 and fouled Y-face 'k' variants; Eld-S1 sports new S-k2 and no Y-face 'a k p' variants, capitals or ligatures. Read-S1's capitals are correctly cast and justified; the miscast D2 and G2 are exclusive in Eld-S1; and Eld-S1's punctuation sorts are new and uncontaminated. In addition to these and other compositional differences, Eld-S1 is, in sharp contrast to Read-S1, a crisp clean font in 1605. In short, the differences between the two fonts preclude any possibility of transformation.

Settling the issue of ownership from shop lineage is a relatively simple matter if the production of books in the font size in question was consistent before and after the change in shop ownership. This is true even in instances where overall production was consistently low as long as a few books in the given font can be found. For example, shop output was sparse both during Simson's, Widow Simson's, and Read's tenures. Nonetheless, the lineage of Simson-S1 to Read-S1 is easily demonstrated. The font is used in combination with ornamental stock in proximate books before and after the passing of the shop to Read.[36] The unmistable composition of the font bridges the transition without modification. The passing issue is more difficult to resolve if the fonts used before and after the passing of a shop are in the same-face and lack such significant differentiae. In general, a recurrent-types survey must be employed to resolve the same-face passing issue in this situation.

Two factors present a problem in determining whether Robinson-Y1b

passed to Braddock in 1598. Because of the lack of books from the shop in the preceding year, Braddock-Y1 of 1598 must be compared to the Robinson-Y1b seen in at least four books in 1595-96: in the text of Salomon STC18194, 1596; the preliminaries (¶2-3v) and sheet P of The Discoverie STC20634, 1596; long quotations in A Comparison STC4098, 1595; and as the emphasis font in The Second Time STC18246, 1596. Furthermore, Braddock-Y1 exhibits a more extensive contamination with wrong-face sorts than Robinson-Y1b, giving it a different overall appearance. However, the limited foul-case cluster present in Robinson-Y1b (S-face A P S; italic I O S; turned 'p'; 108mm ',') is consistent with that seen in Braddock-Y1. Progressive contamination through fouling and replenishment during 1597-1598 is a plausible explanation for the expansion of the cluster since the transformation requires no elimination of sorts found in Robinson-Y1b. In this instance, the large font samples from 1596 and 1598 create a favorable situation for a successful recurrent-types survey to either confirm or reject identity. The passing and continuous use of the font is demonstrated by recurrent-types from the 1595-96 Robinson-Y1b in Braddock-Y1 from 1598-1600.[37]

(4)

The failure to locate recurrent-types that positively confirm the identity of the same-face fonts in samples from before and after the suspected passing must be viewed as negative evidence and interpreted with caution. The simple

fact is that two reasonably adequate samples of one font will eventually yield recurrent-types. The length of the samples must increase proportionately with the interval separating them to adjust for the on-going process of purging battered types, a factor which varies among shops. In general, samples consisting of several quarto gatherings at both ends of a long interval (three years or more) are usually adequate for the purpose, provided that the font is reasonably battered to start with. The variables that must be weighed in interpreting the failure of the survey include (1) investigator oversights, (2) a less-than-thorough search for recurrent-types, and (3) the possibility that each sample text was set from a different case thereby precluding recurrence of identifiable types. The first clue to the latter possibility is that recurrent-types appear in both formes of a gathering, a fairly certain indication that only one case was used in setting the text.[38]

The lack of adequate font samples during a transitional period can produce an ownership crux that can be resolved tentatively at best. For instance, shop lineage is established by Read's inheritance of Simson-S1 along with the shop's ornamental stock, which then passed to Eld. The situation is complicated by the assumption, based upon the first appearances of Eld's imprint, that he took over the shop in 1604.[39] The appearances in 1601, 1602,

and 1603 of Y-fonts in suspected Read books, and then Eld-Y1 in signed books in 1604, create the appearance of the continuous uninterrupted use of a single Y-font and raise the possibility that it passed to Eld along with the shop. Overall, these font samples exhibit no significant differentiae that would raise suspicion about their identity. Two Y-font appearances in early 1602 contribute to the appearance of the continuous use of a single font, but the sample is less than minimally adequate in both instances. The Y-font that appears in one quarto-page (B2v) of verse in Epitaph STC3415 (entered 24 February 1602) is considerably worn and battered; the Y-font seen in speech headings in A Dialogue STC18892 (entered 25 February 1602), a very short quarto, is in distinctly better condition. This observable difference in condition is a reasonable basis for rejecting the identity of the two. It is plausible that the font of A Dialogue STC18892 is that seen in the other books but the amount of type in A Dialogue is simply too limited to provide the evidence needed to establish identity. On other hand, the consistent appearance of Read/Eld stock in pages set in a Y-font in the octavo text of Essayes Pt. 2 STC5775 (entered 19 October 1601) and Tragedies STC26076 (1601, not entered), and two later dramatic quartos suggests a single font. However, two factors prevent reaching a definite conclusion about the identity of these font samples in this instance. As sometimes happens, the practical reality of limited library holdings can present an obstacle: since The Huntington lacks an original copy of Cromwell STC21532 (entered 11 August 1602), it is impossible to perform a recurrent-types survey which could link Essayes Pt. 2 to the first demonstrable appearance of Eld-Y1 in Nero STC12251 (entered 23 February 1603) via identifiable types in Cromwell. Cromwell could be an important source of evidence given the quite good condition of the type in Essayes, Tragedies, and Nero, a factor which makes it difficult to locate identifiable types in the latter books. In general, some identifiable types remain in a font throughout its lifetime. However, the probability of locating recurrences of specific types is affected by the random recurrence phenomenon and hence is directly proportional to the number of such types in a font. Although the interval between Essayes/Tragedies and Nero is hardly enough in itself to explain the apparent absence of recurrent-types in the two books, the relatively few identifiable types in each compounds the problem presented by the interval. A third set of identifiable types in Cromwell would dramatically increase the chances of finding overlapping recurrences if the fonts are actually the same. It is also possible that repeating the recurrent-types survey of Essayes/Tragedies and Nero would eventually yield evidence of identity. Nonetheless, the ornamental evidence in combination with the composition of the complementary Granjon pica italic font points to a single font in Essayes/Tragedies, Cromwell, and Nero although specific typographical evidence is lacking.[40] In general, the suspicion about ownership raised by the failure to locate recurrent-types in reasonably battered candidate and target fonts from the proximate period is usually a sound indication that a search for a sharing printer is necessary.

Finally, the ownership crux is unresolvable if a font appears in only one assigned book but in the absence of an interwoven setting which could demonstrate ownership. The presence of identifiable ornamental stock in such an instance is inadequate to prove the ownership of the font because of reasons discussed earlier. On the other hand, ownership can be assumed if the single appearance of the font occurs in combination with an identified complementary font (e.g., italic, black letter) in quantities and patterns such as preclude sharing. Very short settings that are inserted into a text which is set before and after in another font style or size present a special problem since transportation from another shop is not unrealistic. The page of short verse noted above in Epitaph STC3415 is such an instance. Page-length prose epistles or dedicatory verses in preliminaries seem to be another class of settings to approach with caution. It seems wise to require a verifiable second appearance of a font in these instances before concluding ownership. The need for such caution can be illustrated by two cases involving books signed by Roberts. His infrequently used 76mm Guyot pica roman appears in three short settings in the context of identifiable fonts and ornamental stock. The passages at B4-4v and I3-3v of Scourge of Villanie STC17485 (1598) are inserted in a text in Roberts-S1. The epistle in A3-3v of A True Discourse STC7293 (1604) shares A3 with a Guyot pica italic such as Roberts used and Roberts-S1 serves as the emphasis font in the black letter text A4-D3. Similarly, the epistle on A2-4 of Euphues STC17075 (1597) is headed by Roberts's Trumpets-T, followed by a Granjon italic in A4v-B1v such as he used elsewhere, and Roberts-S1 on B2-B2v headed by his Angel-G, followed by Roberts-S1 as emphasis font in the black letter text B3-2F2.[41] In these instances, the contextual materials and their pattern of appearances would be sufficient

to allay any doubts about Roberts's ownership of the Guyot pica roman. Fortunately, Roberts used the Guyot pica at least two other times as the text font, leaving no doubt of ownership (See Englands Heroicall STC7196 [4th ed., 1600], STC7197 [5th ed., 1602]; not checked: STC7193 [1st ed., 1597] and STC7195 [3rd ed., 1599]). However, two appearances of Y-fonts in books signed by Roberts are deceptive and lead to an erroneous inference of ownership. The minimal sample of a Y-font seen in the few speech heads and emphasis in the black letter text (A2-C3v) of Clim of Clough STC1808 (1605) lacks distinguishing features (except for the odd use of the 'fl' ligature in setting "Cloudesse" which may indicate a lack of the 'ff' ligature in the font) and would be cause for suspicion of sharing if it was the only appearance of a Y-font in a book signed by Roberts. However, the entire text of Scourge STC 17486.5 (1599) (Folger 17486a; British Library C.39.b.43) also prints in a Y-font which, despite the gap of five years and the minimal sample in STC1808, would seem adequate to demonstrate ownership. Furthermore, Roberts-S1 accumulates an extensive foul-case set of Y-face capitals from about 1600 on, a factor which suggests the use of a Y-font in the shop. However, the Y-font of STC17486.5 exhibits an extensive foul-case cluster that makes it easily identifiable (even a single-page reproduction such as Plate 124, Pica Roman Type, contains highly suggestive evidence of identity). Roberts actually farmed out STC17486.5 to Braddock and the font is Braddock-Y1 in a state similar to that seen in Midsummer Night's Dreame Q1 STC22302.[42] Finally, a short passage printed in a font distinguished by such "gross features" can sometimes suggest the identity of the printer of a section in another font style or size. For example, the appearance of 18 lines of Stafford-EFb at D1 in the 96mm roman text font of sheets CD in T. Whole Magnificient Entertainment STC6513 (1604) suggests the assignment of the section to Stafford.

(5)

The practice of sharing sometimes diminished the known output from a shop and created gaps in the use of a given font because it was used in unknown sections of shared books. The current state of knowledge about this phenomenon is by no means exhaustive. The stages of progress reflected in the expanded list of sharing printer assignments added to Morrison's Index and old STC in new STC, and the assignments yet to be added on the basis of typographical evidence, suggest that a significant portion of the output from printers with known sharing activity may lie buried in shared sections.

Two factors affect the practical implications of this possibility: (1) these sections may actually constitute a large portion of the pica (or other sizes) output from a shop; and (2) since font identification requires recurrent-types identified in original copies, a virtual shortage may occur because the library at which one is working lacks a copy of the key assigned book in which the critical font sample is found. The unknown output in Eld-Y1 provides an informative example. The six books assigned to Eld in Morrison/STC for 1604 include two Eld-Y1 texts: Epigrames STC5672 (four octavo sigs.), and A Loyal Subjects STC25760 (9 quarto sigs.), equivalent to eleven quarto gatherings (plus one minimal use as emphasis font in Palladis STC26014). New STC adds three Eld-Y1 texts (Supplication Q2-3 STC14429.5, 14430.5, ten quarto sigs.; Q4 STC14430, one sig. [only] standing from Q3 [see later discussion]) comprising eleven quarto prose gatherings. In addition, Eld-Y1 printed Malcontent Q1-2 STC17479-80, B-E (eight sigs.); Q3 STC17481, HI (two sigs.); Whore Q1-2 STC6501,6501a, G-K (eight sigs.); and two pages in Siege STC18895, or slightly more than eighteen gatherings. Overall, the fact that eighteen of a total of forty gatherings for 1604 are hidden in unidentified shared sections obviously is a significant factor in a survey for Eld-Y1 samples in 1604. Similarly, one Eld-Y1 appearance in a Read book in 1603 is found in Morrison/STC in the tentative assignment of Antichrist STC7120; new STC assigns 2A-G to Read (seven sigs.) as well as Nero STC12551 (eighteen sigs.) for a total of twenty-five gatherings. In addition, Eld-Y1 appears in Book 3 of Essayes STC18041, an enormous Eld-Y1 sample in a prose folio in sixes which dwarfs the previously known output.

A different kind of problem is encountered in a survey of Eld's 1605 books. Both Morrison/STC and new STC assign Survey Q1-2 STC6200,6201 to Eld. Q2 prints in Eld-S1, while Q1 prints in a Y-font which could be easily misinterpreted as a healthy eight gathering sample useful for identifying Eld-Y1 in a shared section of another book. However, it is not Eld-Y1. Papers used in Q1 are shared with An Apology STC19295 (1607) and do not seem to appear in any other Eld or Simmes book 1603-1608. In private correspondence, Peter Blayney suggests a 1607 piracy. The 1607 dating of Q1 is unquestionable, but the issue of piracy can only be settled by the identification of the Y-font. Given a list of about 100 identified Eld-Y2 types which recur in half-dozen 1607 books by Eld, whether Eld-Y2 prints STC6200 could be easily settled, but The Huntington lacks a copy, a situation which amounts to a virtual gap in shop output and leaves the issue hanging upon another research trip (note also previous discussion of Cromwell STC21532). Nonetheless, the question remains: how is it that papers concurrently used by Eld in 1607 would find their way into a pirated text bearing Eld's imprint and a false date of 1605? Adam Islip's 1598-1605 output represents an extreme case that probably is unique but is worth mentioning since more moderate cases may occur. Pica roman output is lacking in 1599-1601 books bearing the Islip imprint and listed by Morrison/STC. New STC reassignments to Islip of books bearing the Thomas Wight (and/or Bonham Norton) imprint reveal the reason for the apparant lack of pica roman output: Islip-Y1a,b appears

in at least seven such books 1598-1601. Islip-S1 appears at least in one Islip book 1602-1604, but in at least six Wight books.[43] The impact of new STC reassignments such as this will become clear once Volume 3's "Index of Printers" appears.

(6)

The detection of shared printing is often aided by the fortuitous accident that sections of a shared book print in dissimilar fonts. The alternating pattern of dissimilar fonts in units of a gathering or more provides an obvious clue to the possibility of sharing and reveals the possible divisions of labor. The divisions are usually obvious in a book shared in the classical two-section A-B pattern. For example, it is impossible to overlook the fact that Eld-Y1 appears in Malcontent STC17479 Q1 B-E with Simmes-S1 in F-H; or Creede-4 in A-2D and Windet-F in 2EF of Regiment STC1827. The same holds for books shared in three or four sections so long as a sequence of dissimilar fonts emerges, as, for example, in Honest Whore STC6501, where Simmes-S1 prints AB, Creede-4 prints CD, Stafford-EFb prints EF, and Eld-Y1 prints G-K. In such instances, the dissimilar typefaces or obvious "gross features" differentiae are easily detected in the initial seriatim font analysis. However, a crux frequently emerges in books of more than two sections when two of the sections print in same-face fonts separated by one intervening section (or more) in a dissimilar font. This raises the intermediate issue of whether the two separated same-face fonts are the same and therefore indicate one printer as opposed to two printers using same-face fonts. In instances involving separated appearances of a font distinguished by "gross features" or obvious differentiae, the matter can be easily decided. Creede-3's unique mixture, for example, appears in Parnassus STC378 (1600) at B-S and 2G-K, separated by Purfoot-Y1 at T-Z and Simmes-S1(?) at 2A-F. Purfoot-Y1 could be as easily distinguished in a split appearance because of its unique composition. However, instances such as the appearance of Windet-S1 in EF and HI of Fawne Q1 STC17483 present a problem.

Same-face fonts lacking obvious differentiae usually can be identified in separated appearances only by a recurrent-types survey. Perfectly practical and logical assumptions about sharing strategies cannot be trusted in such instances.[44] For example, the appearances of same-face fonts in widely separated

sections of long books invites the erroneous assumption that, given the length of the book and individual sections, a sequence of printers were involved. A second factor frequently contributes. Printing quality and appearance easily change (sometimes due to different job lots of papers) in long books, leading to the suspicion that a second font appears in a later section. The font distribution in An Apology STC19295 (1607) is complicated by the fact that this quarto in eights was the work of two printers who respectively used two and three fonts during the proximate period. The major difficulty involves determining whether a single Y-font appears in widely separated sections in Book 1 (¶, B-E), 3A-G, and 3L-N. The low-density foul-case S-face sorts in this state of Eld-Y2 (A C D M S T Y b1 g1 k2; 'fl' ligature) do not recur in these isolated sections with the frequency and consistency needed to infer a single font so that evidence of recurrent-types is necessary to establish that Eld-Y2 indeed is the font. Sorting the remaining fonts is assisted by the font distribution. Although Eld was using Eld-S1 at the time, the alternating pattern of fonts in Jaggard's section, which includes Jaggard-S1, eliminates the possibility that Eld-S1 was involved. The combined appearances of Jaggard's three fonts in several gatherings clearly indicate a single printer in those sections;[45] individual full-gathering appearances in Book 2 of Jaggard-C2 (2D, 2G, 2K, 2O) and Jaggard-S1 (2LM) can be resolved by an analysis of cross-contamination. The combined appearances by halves of Jaggard-Y1b and -S1 in 3H and 3K separated by Jaggard-C2 in 3I repeat the pattern of Book 2 and furnish the evidence necessary for assigning the section (3H-K) to Jaggard.

Although shared printing is usually implied by a sequence of fonts in units of a gathering or more, the possibility exists that a printer with two fonts used them in this manner, which is typical of shared printing. The alternation of Lownes-S1 and -Y1 in A Modest STC5882 (1604) invites the inference that three printers were involved as suggested by an S-font in R-2C between Y-fonts in B-Q and 2DE.[46] As in STC19295, the widely separated

fonts in B-Q and 2DE seem different because of printing quality. Nonetheless, recurrent-types indicate that Lownes-Y1 printed both sections with Lownes-S1 in R-2C. The rationale underlying this in-shop division of labor between two fonts is unclear, but the similarity to Jaggard's alternation of Jaggard-C2 and -S1 in STC19295 is obvious. Both may be related to compositorial stints.

The sequence of four editions of A Supplicatione culminates in an undesirable kind of same-face crux: the Eld-Y1 and Simmes-S1 sections of Q3 are reset in Q4 in a second Y-font except for one standing gathering in Eld-Y1 and a few pages in Simmes-S1. Q1 STC14432 prints entirely in Simmes-S1 (incorrectly assigned to the English Secret Press in new STC). Both Q2 and Q3 are in the two-section A-B sharing pattern. B-E is in Eld-Y1 and F-O is in Simmes-S1 in Q2 STC14429.5. In the resetting for Q3 STC14430.5, F of Q2 shifts to Eld-Y1 with Simmes-S1 printing G-O. The redistribution of labor and the sections of standing type seen in Q4 STC14430 provide one clue to the sequence of editions.[47] The complex sharing pattern constructed around standing Eld-Y1 in F and standing Simmes-S1 in M4 and O1-2v would be impossible except in a line-for-line resetting of a text previously containing the standing sections. The pattern exemplifies the problems associated with separated appearances of same-face fonts. White-M with its mixture of nearly complete S- and Y-fonts is easily distinguished in D and G of Q4. Creede-4 could possibly be confused at first with the minimal sample of Simmes-S1 in the latter's standing section; moreover, the separated appearances of Creede-4 could also suggest two more sharing printers; however, Creede-4's differentiae permit verification of separated appearances in BC, H, and L. Another major pitfall is created by the appearance of Y-fonts in the remainder of the pica roman sections of text. The previous use of a same-face font in a section(s) of an earlier edition creates a predisposition to assume that it appears in the same section (or part thereof) in the later edition. Given the fact that Eld-Y1 appears in B-E of Q2 and B-F of Q3, it is easy to assume that Eld-Y1 therefore appears in EF, I, N and M3v of Q4. The analysis of "gross features" and composition is frequently inadequate to distinguish same-face fonts in these situations: the few wrong-face capitals in Eld-Y1 in Q3 B-F (B E L P S, Guyot 'S') seem consistent with the 'E P' in Q4 E (E2v:1, 3), the 'P' with the crimped bowl (Q4, E4:9, I:13, N2v:16) seems at first glance consistent with that seen in Eld-Y1 in Q3 (B2v:15, B3:25, C1v:35, D1v:3, E3v:3), and the few fouled italic 'P S' seen in Q4 could easily be attributed to transient fouling. A recurrent-types survey of Q4 is necessary to identify Eld-Y1 in standing F and distinguish it from the second Y-font. The remainder (E, I, M3v, N) was reset (very probably)

in Braddock-Y2a. The correct sequence of 14430.5 and 14430 is implied by the fact that the final sections of Q3 in Eld-Y1 (F) and Simmes-S1 (M4, O1-2v) were undistributed and hence remained standing for Q4. Similarly, resetting patterns in the absence of standing type may provide useful evidence as to a sequence of editions. For instance, the font distributions shift between the two editions of Pericles. Q2 STC22335 is set in the classical A-B pattern, with White-S1 in A-D and Creede-5 in E-I3v. In Q1 STC22334, however, sheet E is printed by White and sheet B by Creede so that the resulting distribution is: White-S1: A, C-E; Creede-5: B, F-I. In general, the simple two-section pattern usually occurs in the first edition while the disrupted pattern occurs in a line-for-line resetting in the second edition.

The most treacherous crux consists of the appearances of two same-face fonts in contiguous sections of a shared book. If other clues to sharing are present, the possibility of two fonts should be tested immediately. The shift of ornamental stock in Book 2 of Antichrist STC7120, for example, calls for a comparative analysis of the Y-fonts in Books 1 and 2 in order to differentiate Braddock-Y2a from Eld-Y1. Although an uncommon case, the Dutch Courtesan STC17475, a book bearing Purfoot's imprint and printed in two Y-fonts, epitomizes the various complications that can confound the search for sharing printers. Some aspects of printing style and the running-titles shift between A-E and F-H. Purfoot's ownership of both Purfoot-Y2 and -Y3 is a complicating factor in detecting shared printing in this instance. The S-capitals and italic foul-case cluster in A-E leave little doubt that Purfoot-Y2 printed the section. However, the Y-font in F-H exhibits a deceptive similarity to Purfoot-Y3. The italic foul-case cluster in F-H is consistent with that seen in Purfoot-Y3 although the two overlap in only a few sorts. This difference could be attributed to progressive fouling and/or purging. The presence of the turned 'p q' and the oversized 'b' and 'ct' ligature in F-H, Purfoot-Y2 and -Y3, in contrast, suggests cross-contamination and renders the inference of identity more difficult to dismiss. In any event, this obvious evidence is ambiguous. The recurrent-types survey of F-H produces another complicating factor: several positively identifiable types appear in both formes of sheet F,[48] indicating one-case setting with distributions after each forme, a sound basis for inferring that only half of the font appears in F-H. In turn, this undermines the conclusiveness of the failure of a recurrent-types survey of F-H and Purfoot-Y3 to yield identified types in both fonts. The identity of the target font in F-H, in short, can be determined only by the discovery of a candidate font which yields recurrent-types. The obvious differentiate exhibited by the F-H font lessens the difficulty of settling upon a suitable candidate such as Jaggard-Y1b. If it is the correct candidate, recurrent-types usually emerge

rather quickly. However, the search process seems destined to failure when a target font lacks obvious differentiae, and worse yet, exhibits minimal to virtually no damage and wear. For example, identified Eld-Y2 types recur throughout Byron STC4968 (B-O, QR) except in sheet P, where a Y-font in virginal condition appears. This font is entirely uncontaminated, almost all types print with crisp clean edges, and only a few types exhibit minor damage that could support positive identification in recurrences. Identification of this font would be a tenuous procedure even if the correct candidate font were located. The Y-font alternating with Islip-EFb in De Missa STC23456 is a similar instance but not quite as extreme.

(7)

Although concluding on such a negative note seems poor rhetorical strategy, it nonetheless is indicative of the psychological experience that sometimes results from a futile search for a sharing printer. The commitment to printer research assumes a willingness to endure such frustration. The discovery of sharing, however, is important in itself and is a source of satisfaction. In fact, shared printing is the only area where new discoveries in relatively large numbers are within easy reach, given the high probability that we have found all the extant manuscripts of plays by important Elizabethan/ Jacobean authors. Beyond that, an enormous amount of work remains to be done by employing font analysis to verify the tentative assignments in new STC that are based upon ornamental stock. A practical problem exists in regard to the recording and dissemination of the new information that will be generated by typographical analysis. Detailed evidence that affects our understanding of the transmission of early texts should find its way into print as a matter of course. However, the overwhelming majority of early books probably do not merit such treatment although typographical information about them can be extremely valuable in the context of printer identification. Publication of such information in printed format is obviously out of the question: the Short Title Catalogue with its abbreviated descriptions exemplifies the practical limits both in terms of economics and dedication (bibliographers cannot help feeling a sense of gratitude to Katharine V. Pantzer every time the revised STC is consulted). The creation of on-line computer databases, however, offers the exciting prospect of instant (or nearly so) dissemination of current bibliographical information that can be expanded and updated as the need arises. The proposal by Henry L. Snyder (University of California, Riverside), Director of the Eighteenth-Century Short Title Catalogue for North America, to include the STC in the on-going project of converting the ESTC to an electronic database has, in fact, been underway for some time.[49] Although the abbreviated entries of the revised STC serve as the basic records for the database, Snyder envisions expanding them "so that they

would be comparable to ESTC records in fullness, content, and format." The possibility of adding to the database once it is completed should provide additional motivation for bibliographers to record routinely the kinds of evidence that are pertinent to shared printing and printer identification.

Notes