University of Virginia Library

The research for this paper was in part supported by travel grants from the General Research Fund, Regents of South Dakota, and a generous grant from the National Endowment for the Humanities, Division of College Teachers, for which I also am indebted to Fredson Bowers, Peter Blayney and Donald H. Reiman for their supporting recommendations. This paper would not have been possible without this support. I also thank Fredson Bowers for suggesting the approach of the paper and encouraging and guiding its progress, and Peter Blayney for commenting and sharing his vast knowledge of the subject.

"Edward Allde as a Typical Trade Printer", The Library, 4th ser., 10 (1929), 121-162. A. E. M. Kirkwood, "Richard Field, Printer, 1589-1624", The Library, 4th ser., 12 (1931), 1-39. Harry R. Hoppe, "John Wolfe, Printer and Publisher, 1579-1601", The Library, 4th ser., 14 (1933), 241-288. These studies concentrate on printer's devices almost entirely with only passing reference to type fonts (see note 5 below). Edwin Eliott Willoughby, A Printer of Shakespeare [William Jaggard] (London, 1934). W. Craig Ferguson, Valentine Simmes, Printer to Shakespeare et al. (University Press of Virginia, 1968) was the first to describe and use a printer's type fonts in identifying his work in sections of shared books. Dennis Woodfield, Surreptitious Printing in England, 1550-1640 (Bibliographical Society of America, 1973). C. L. Oastler included quite unsatisfactory specimens of John Day's type fonts in John Day, the Elizabethan Printer (Oxford Bibliographical Society, 1975) with identifications of punchcutters where possible, pp. 40-43. Akihiro Yamada, Thomas Creede, Printer to Shakespeare and his Contemporaries (Shinsu University, 1981) catalogued appearances of ornamental stock. Clifford Chalmers Huffman, Elizabethan Impressions: John Wolfe and his Press (AMS Press, 1988). See also Peter H. M. Blayney, note 10 below.

"The Prevalence of Shared Printing in the Early Seventeenth Century", Papers of the Bibliographical Society of America, 67 (1973), 437-442.

This limitation does not necessarily apply to longer books, especially prose texts with chapter and section divisions which invite the insertion of ornaments and initials. Several of the quarto and folio prose texts referenced later begin sections with ornamental stock as a matter of course.

This is not a new idea. There are about a half-dozen extant early records of legal proceedings involving the identification of the printers of pirated or seditious books by their type fonts. For some examples, see Talbot Baines Reed, A History of the Old English Letter Founders (London, 1887), pp. 123-133.

McKerrow overlooked the potential of font analysis and defined what seems to have become a prevalent attitude: "Of Allde's types [type fonts] there is little to say. . . . So far at least I have observed, there is nothing in the types used to distinguish Allde's work from that of a dozen other printers" (p. 147) (Allde actually used four easily distinguished pica roman fonts between 1590-1605). Some years later, Fredson Bowers noted the potential value of the description of fonts in a book, Principles of Bibliographical Description (Princeton University Press, 1949), p. 127, and later commented (p. 305): "The font of the type used in the text, whether roman, italic, or gothic, is given; when possible, this type should be identified by reference to books on printing types, but for books of this period precise identification is usually difficult." See my comment on this problem, note 36 below. One purpose of this paper is to lay a foundation for future work leading to the "precise identification" of fonts by reference to Elizabethan books and printers' fonts.

This comment applies to every study of a book printed in two or more fonts that I have seen. The consequent analyses of compositors and presswork inevitably are incorrect with respect to the assignment of compositors, running titles, skeletons, number of presses, sequence of setting and printing and other aspects of printing. For example, W. W. Greg distinguished three sections in The Honest Whore on the basis of type fonts, but missed the EF and G-K division, correctly noting it in the second edition by reference to running titles, "'The Honest Whore' or 'The Converted Courtesan'", The Library, 4th ser., 15 (1935), 54-60, repeated by Fredson Bowers (see note 40 below); and G. Blakemore Evans, "Textual Introduction", All Fools, ed. G. Blakemore Evans, in The Plays of George Chapman: The Comedies, gen. ed. Allan Holiday (University of Chicago Press, 1970), 227-232. Evans attributed three compositors to Eld although two belong to the sharing printers. Even though W. W. Greg had specifically noted that a different font appears in one section, Evans begged off: "I cannot see any significant difference, but Greg's eye was sharp and trained and I hesitate to disagree with his years of experience on such a matter" (p. 229). Likewise Akihiro Yamada ignored the issue and attributed four compositors to Eld in "A Bibliographical Study of George Chapman's All Fools (1605)", The Shakespeare Society of Japan, 3 (1964), 73-99. Both concluded that All Fools was printed on one press,

although they disagreed as to whether casting-off of copy was required. See also: G. K. Hunter, "The Text", The Revels Plays: 'The Malcontent', ed. G. K. Hunter (London: Methuen, 1975), xxxi-xl, and comment below, note 44; Akihiro Yamada, "Q1-3 of The Malcontent, 1604, and the Compositors", Poetry and Drama in the English Renaissance—in Honor of Professor Jiro Ozu (Tokyo: Kinokuniya:shoten, 1980), 107-132, "Press Variants in John Marston's The Malcontent", Studies in the Humanities, 17 (Shinsu University, 1983), 93-113, "Simmes's Compositor A in The Malcontent, Q1-2", Studies in the Humanities, 14 (1980), 121-125, and comment below, note 64; David A. Blostein, "The Text", The Revels Plays: 'Parasitaster' or 'The Fawne', ed. David A. Blostein (1978), 45-52; C. G. Petter, "Note on the Text", The New Mermaids: 'Eastward Hoe!', ed. C. G. Petter (1973), xxxix-xlvi; Allan Holaday, "Textual Introduction", Monsieur d'Olive, in The Plays of George Chapman: The Comedies, 397-402; Akihrio Yamada, "Bibliographical Studies of George Chapman's Monsieur d'Olive (1606), Printed by Thomas Creede", Studies in English Literature (The English Literary Society of Japan: Shinsu University, 1963); Robert F. Walsh, "Textual Introduction", May Day, in The Plays of George Chapman: The Comedies, 311-314. See my discussion of several of the above in "Reproductions of Early Dramatic Texts as a Source of Bibliographical Evidence", TEXT, 4 (1988), 237-268.

Two forms of measurement indicate the size of an early typeface: (1) the "20-line" measurement in millimeters is the distance from a point in one line to the corresponding point in the twenty-first line of type above or below set without leading (a thin strip of lead placed between lines in later centuries), the line from which the measurement starts being included in the count; (2) the "bare" measurement in millimeters is the distance from the top of an ascender to the bottom of a descender in a single line (i.e., top of the 'h' to the bottom of the 'y' in "thy") multiplied by 20. The complementary measurements are necessitated by the fact a given typeface could be cast on taller bodies than that for which it was designed.

The technical terminology for the various elements of letter-forms has never been standardized to any degree of unanimity. Philip Gaskell, "A Nomenclature for the Letter-forms of Roman Type", The Library, 5th ser., 29 (1974), pp. 42-51, is the most comprehensive attempt with a glossary of terms and a complete roman alphabet with the term for each element of each letter. Gaskell differs in some respects from the nomenclature traditionally used in historical typography. The physical space occupied by a line of type is defined by four imaginary lines: (1) letters sit on the "baseline"; (2) the "x-line" (Gaskell: "mean line") defines the shoulder height ("x-height") of lower-case letters (m, n, b etc.); (3) the "h-line" ("ascender line") defines the height reached by ascender above the x-line; (4) the "descender line" defines the lowest level reached by letters which protrude below the baseline (g p q y). In regard to other terms, I prefer the traditional "pear" to his "bulbous terminal". "Cross-stroke" (or "cross") seems descriptive of the horizontal strokes in 'A E F T'; the "arm" and "leg" (or "tail") of the 'k' seems preferable to an "upper diagonal" which is topped by an "upper diagonal serif" and a "lower diagonal" terminated by a "lower diagonal serif"; the lower part of the 'g' has always been a "loop", and calling it a "tail" seems inappropriate because it does not look like the "tail" of a 'y' or 'R' or 'Q'; it seems inconsistent to gloss "ascender" as "the extended stem of b, d, f, h, k, l, tall-s" and then use "stem" exclusively to refer to all vertical elements of both short and tall letters, especially when the latter are traditionally "ascenders".

A font composite is compiled by sketching the stylistic features seen in a font's several groups of letters, punctuation and fouled sorts. The value of the font composite arises from the fact that it defines a font in terms of sets of specific distinguishing features. The diagonal letters form one set, the ligatures another and so on. When comparing two fonts, the sets provide an efficient method for quantitatively estimating the chances that the two are the same. See my discussion in "Reproductions as a Source", pp. 251-255.

Reference will be made in the following discussion to typeface specimens of the period. The major sources of reproductions of extant founders' specimens are found in a few ground-breaking facsimiles collections specifically intended as a basis of historical study of early types and punchcutters. (1) J. Dreyfus & others eds., Type Specimen Facsimiles, Vol. I, Nos. 1-15 (London, 1963). The anonymous Folger c.1565 specimen sheet attributed

to Francois Guyot is reproduced as no. 1; the individual specimens of Guyot typefaces are also reproduced in their appropriate section of ref. 4 below. (2) Stanley Morison, John Fell, the University Press and the 'Fell' Types (Oxford, 1967). (3) Hendrik D. L. Vervliet & Harry Carter eds., Type Specimen Facsimiles, Vol. II, Nos. 16-18, "Reproductions of Christopher Plantin's 'Index Sive Characterum 1567' & 'Folio Specimen of c.1585' together with the 'Le Be-Moretus Specimen c.1599'" (University of Toronto Press, 1972). The three extant specimen collections indicated in the title are numbers 16, 17 and 18 in the two volume sequence; each of the three are independently numbered and paginated, a possible source of confusion in references. For clarity's sake, I refer to a specimen by title and specimen number within the section, i.e., "'Le Be' 12" signifies specimen number 12 in the "Le Be" section (no. 18), or by title and page number within the section, i.e., "Le Be" note 13, p. 14. (4) Hendrik D. L. Vervliet, Sixteenth-Century Printing Types of the Low Countries (Menno Hertzberger & Co., Amsterdam, 1968). Specimens are both labeled according to group (i.e., "R17" signifies "roman typeface specimen number 17") and by figure number ("Fig. 186") in this sequentially paginated volume. (5) Joseph Moxon, Mechanick Exercises on the Whole Art of Printing, Herbert Davis and Harry Carter eds. (Oxford, 1962). (6) Fournier on Typefounding. The Text of 'The Manuel de Typographigue', Harry Carter tr. & ed. (The Fleuron Books, London, 1930), esp. Ch. XI "Striking Matrices", pp. 81-88, and Ch. XII "Justifying Matrices", pp. 89-98. The most comprehensive application of historical knowledge about the type production process to a printer's pica roman font is found throughout Peter W. M. Blayney's The Texts of 'King Lear' and Their Origins Vol. I (Cambridge University Press, 1982). The history of typography found in the previous references is complemented in Harry Carter, "The Types of Christopher Plantin", The Library, 5th ser., 11 (1956), 170-179, which reproduces the Varia II folio specimen of the Plantin-Moretus Archives. Also valuable is A. F. Johnson, "The Supply of Types in the Sixteenth Century", The Library, 4th ser., 22 (1944), 47-65; and Paul Beaujon, "The 'Garamond' Types: Sixteenth & Seventeenth Century Sources Considered", The Fleuron: A Journal of Typography, 5 (London, 1926), 131-179, reprinted in Fleuron Anthology, Francis Maynell & Herbert Simon eds. (University of Toronto Press, 1979), 181-214. A frequently referenced paper by Frank Isaac, "Elizabethan Roman and Italic Types", The Library, 4th ser., 14 (1933-34), 85-100 and 212-228, is misleading in its grouping and discussion of pica roman fonts (see further comment, note 36 below). Some sources of potential confusion are worth nothing. The frequently referenced "Specimen Charactervm seu Typorum Probatissimorum" (1592) is variously designated by combinations of "Conrad Berner", "Engelnoff", "Frankfurt", and "1592". Similarly, typeface body-sizes are referenced in the English ("pica"), French ("cicero") and Dutch ("mediaan") vernacular terms that appear in extant specimens. Nomenclature cross-reference tables and measurements are found in The 'Fell' Types (p. 260), Phillip Gaskell, A New Introduction to Bibliography (Oxford, 1972) (p. 15), and Low Countries (p. 16).

The height of a typeface, as measured from the top of the ascenders to the bottom of the descenders, defined the shortest body that could carry the face. However, a given typeface could be cast on a taller body; the Claude Garamond "cicero" (i.e., pica) punches in the Plantin collection, for example, produced fonts with heights ranging from 80-84mm (see Type Spec. Facs. Vol. II, "Le Be" note 13, p. 14). Similarly, new punches were cut with shortened ascenders and descenders and the typeface cast on a smaller body than originally intended (see Type Spec. Facs. Vol. II, "Folio c.1585" specimens 38 and 45). In one extraordinary instance, a typeface with a bare measurement of 80mm was cast on a body with a 20-line measurement of 67mm without modification of ascenders and descenders (see Type Spec. Facs. Vol. II, "Folio c.1585", specimen 57, and note 57, p. 10). Vervliet does not explain how the type body was modified to make possible the setting of the specimen and other texts. The body of every sort had to have been kerned to allow for the overlapping of ascender and descender lines.

In the remaining discussion, I will label specific typefaces, variants and subsets by the printer who used them in combination with an initial that was chosen according to an imperfectly defined convention. The "-Y" was selected arbitrarily for Eld's Y-font as first studied in The Malcontent, then the "-Y" was used to designate the same typeface in other

printers' work. The recognition of the second of Eld's typefaces in Sejanus suggested using the first letter of this title for the label in "Eld-S", and so on for other printers using the S-face; similarly, Windet-F was first noted in The Fawne. Lacking the identity of the printer, the short-title of the book and the sheets in which the typeface occurred seemed the logical approach, hence Fools-CDI or Fools-EF. Mixed fonts present a complex problem in that they are combinations of two or more typefaces, and ideally could be described by the printer's name and the label for the faces, as in "White-S+Y", thereby indicating the two faces that were mixed. However, the compromise "-M" for "Mixed" is most expedient until the typefaces in these mixed fonts are isolated and labeled. I rely upon W. Craig Ferguson's labeling of Creede's fonts, although consistency would call for Creede-1S, Creede-2M, Creede-3M, Creede-4S etc. This cumbersome method is necessitated by the fact that we lack seminal specimens of many typefaces that are attributable to specific punchcutters; it was suggested by Paul Beaujon: "by fixing our attention upon the identity of the type [typeface] rather than assigning anything to a definite punch-cutter, we may hope in time to establish a scientific method of approach based on actual evidence rather than rash speculation or (worse) that learned credulity that makes a printed reference a reliable authority. Most of the types [typefaces] to be mentioned cannot be assigned with confidence to any engraver; their labels, for our purposes, may be the titles of the earliest books in which their appearance has been noted," "The 'Garamond' Types", p. 146. I must note that an essential element of the bibliographical description of a font is overlooked in my labeling, namely, its physical size, because of the slant of this paper. Vervliet's discussion of his naming convention should be followed in bibliographical descriptions of the type in a book (see Low Countries, pp. 14-19). I should comment also on my line-numbering system on pages that include titles and stage directions. The line-count begins at the first stage direction, speech prefix or line of text while titling material is ignored; blank lines setting off stage directions etc. are included in the count, although this sometimes presents a problem since the blanks frequently are not equivalent to the height of a line. I use the simple expedient of inscribing a line-scale on self-adhesive labels on one edge of a six-inch ruler and numbering every fifth line.

See reproduction, Texts of 'King Lear' (p. 455), and Blayney's attribution (p. 500) of the font to Francois Guyot by reference to the anonymous c.1565 Folger specimen (in Low Countries, R27, Fig. 203; Type Spec. Facs. Vol. I, No. 1) and the specimen of Henry Bynneman's Guyot font in Isaac, Fig. 6.

Generally, high quality reproductions are useful for discriminating gross features such as the height difference of S-face vs Y-face capitals and stylistic differences (i.e., diagonal group) between two typefaces, provided the images of the letters are distinct and clear. This preliminary stage of analysis is frequently adequate for concluding that two fonts are not the same. However, precise measurements at high magnification such as are referenced in the following discussion should not be attempted in reproductions. The dimensions cited in the present study were taken in originals at The Huntington Library as follows: width is measured from points dead-center in the letter elements as established by the faint trace of the elements in the ink-blot; height is measured at points 0.025mm from the edges of elements. Angular measurements are taken with a specially designed and produced micro-protractor. Consult a graphics design/production shop with a PMT machine for reducing the camera-ready protractor to a radius of about 0.25-inch on celluloid or other suitable material.

The specimens in this monumental work are an extremely valuable resource for bibliographers desiring to become familiar with actual type appearance at high magnification (20X-30X). The edges of the letters are quite distinct; the only drawback is that modern printing technology has eliminated most of the distorting effects produced by early methods of hand-inking with the balls: originals have to be examined to appreciate the impact of that variable. Such crisp inking is rarely encountered in early books, a fact which renders them quite a bit more interesting in typographical study. Since frequent reference will be made to variations in size as a means of distinguishing variants, a tabulation of the Fell pica typeface (see Appendix II) will provide a useful overview of the accuracy that can be expected of a master punchcutter like Garamond while filing steel

punches by hand over a period of one to two years, the time necessary to complete the entire set.

Bibliographers tend to think of type in terms of the two-dimensional image on the printed page. Several references provide discussions, drawings and photos of the various tools and stages of the type production process which are helpful background for the following discussion. Absolutely outstanding photos of punches and matrices appear in Stanley Morison's The 'Fell' Types, particularly plates 14, 15, 17 and 18, and decorated initials in plate 19, with a detailed glossary of technical terms, pp. 256-259. See also Low Countries for a drawing of the type-casting mold and discussion (pp. 10-11); and Harry Carter, A View of Early Typography (Oxford, 1969), Figs. 2, 3 and 5 for drawing and photographs of a mold, newly cast types, and type-caster at work.

A perceptive former graduate student, Mary S. Radigan-Hunter, responded to this possibility: "Of course: the punchcutter had a bad day, missed a few strikes and crushed his left thumb, so naturally he wouldn't be able to hold the punch perfectly perpendicular, would he?" Fournier explains: "It sometimes happens that the matrix has been struck more deeply on one side than the other, the punch not having been continuously held quite upright in relation to it. . . . Striking the matrix needs a steady and practised hand; which has given rise to a saying that 'A matrix well-struck is half-justified'." (Manuel de Typographique, p. 90).

See Fournier's explanation of correcting height-to-paper, Manuel de Typographique, p. 90. The depth of impression obtained during the striking of the matrix was a "judgment call" as Fournier notes that the punch "is beaten in with a hammer as perpendicularly as possible until it has gone in as much as a twelfth of an inch or thereabouts [my emphasis]" (p. 83). See Blayney's explanation of matrix-wobble during casting, Texts, pp. 179-180.

Morison defines the justification of a matrix: "reducing it by filing to a true rectangular shape such that, when fitted to a mould, it will produce type with a level face and of standard height for a character that is optically satisfactory for uprightness, alignment, and distance from adjoining characters," The 'Fell' Types, p. 257. When the metal of the matrix was displaced during impact by the punch, the shape of the rough matrix blank was distorted by the vectorial dispersion of the impact force. In simple terms, the sides of the matrix bulged around and beneath the impression of the letter. This can be seen in the clear photo of an unjustified matrix in A View of Early Typography, Fig. 6 (between pp. 8-9). Fournier's explanation of the arduous process of striking a matrix reveals the latitude for errors in judgment and provides a possible insight into why certain letters exhibit sharp-edged elements while others seem rounded or blunted: ". . . with a gentle hammer-blow a shallow impression is made [in the blank]. If the letter is observed to slope either to right or to left, the punch is turned round a little and a second impression is made on the first, and after this a third, the punch being carefully got into the right position. So the place for the strike is gradually found. When at last the punch is shown by the preliminary trials to be in its proper position, it is beaten in with a hammer . . . ." (p. 83). Justification involved correcting the kinds of misalignments referred to in my discussion of miscast letters: "The matrix has to be reduced at the sides until it is of the right width . . . in the case of a letter which ought to be upright but leans a little to right or left through having been struck awry, the matrix must be corrected by being filed down on the side away from which the letter leans. In this example . . . where the h leans to the right, the letter must be corrected by having some of the left-hand side of the matrix towards the top filed away" (p. 93). Low-riding letters are corrected: "the distance between the end of the matrix and the head of the strike [i.e., top of the letter] is too great, and some of it [i.e., end of matrix] must be taken off"; high-riding letters are corrected: "the face of the strike is too short. This is cured by a few hammer-blows upon that end of the matrix to lengthen the copper," thereby lowering the letter on the matrix (p. 93).

For example, the fine S2 that seems to lean right (as if the rectangle enclosing the letter were rotated slightly clockwise) and to sit low on the baseline appears in significant proportions in White-M (LLL, A3:16,21, A3v:2,21,36), and in Creede-3 (R&J, A4v:26, B2v:29, B3v:19, C1:32, C3:13,31) amidst normal erect-appearing letters, most of which exhibit some form of distorting damage. A very low portion of Eld's 'S' sort seems to lean although

they are erect, and all sit correctly on the baseline. It may be that Creede's and White's are actually miscast, but there is no point in pursuing the issue, since they are noticeable just the same. I should note that the preceding discussion has focused upon S-fonts because, in general, Y-fonts seem free of horizontally misaligned capitals with the exception of the low-riding 'C L' in Braddock-Y2 in Book I of Antichrist STC7120. The absence of variant Y-face capitals indicates a single set of seminal punches, with the possibility of duplicate or replacement punches only in the 'S T'. The fact that the capitals are uniformly justified strongly suggests that, if several founders supplied Y-fonts, they purchased their sets of matrices already justified by the same workman.

Some comment by Paul Beaujon shocks the industrial-age mind into an awareness of the accomplishment of these craftsmen: "As Garamont himself wrote that he had cut punches and cast type since the end of his earliest childhood, it is possible to credit the unsupported statement of Lottin that he was practising in 1510, though our earliest firsthand news of him comes some thirty years later. It takes fifteen years, say the punch-cutters, to learn to cut a twelve-point g [i.e., large pica]: the smaller sizes are cut by men who have had more practice," "The 'Garamond' Types", pp. 132-133.

Normal ſh1: Eld-S, Sej, B1v:21; Simmes-S, Mal Q3, B4:13; Creede-3, R&J Q2, A4v:1, 23; White-M, LLL, A4:13, I1v:21. ſh2: Eld-S, Sej, C3:5; Simmes-S, Mal Q3, A4:13; White-M, LLL, A2:10,15; Field-S, Arte, D4:34; Creede-4, d'Olive, A2v:33; Windet-S, Fawne Q1, E2:18. ſh3: Creede-4, d'Olive, A4:30; Windet-S, Fawne Q1-2, I2:5.

In contrast, the cross of the 'e' in early pre-Garamond romans, which the B-face subset in Danter-M resembles (see below), exhibited a upward slant of as much as 24 degrees or so, obviously an intentional stylistic feature. Similarly, the cross of the Le Be 'e' consistently slants upward, while Henry Denham's (An Apology STC6077 [1566]) slants downward; these variant 'e' definitely derive from two punches, while the S-face 'e' from a single punch.

Commas are difficult to work with, given their physical shape and size. Another problem is created by medial vs terminal justification which affects the spacing between a comma and the preceeding letter. See Blayney's discussion, Texts, pp. 179-180.

Generally, "obvious to the naked eye" means that the presence of a possible variant can be sensed; however, some form of magnification is always useful in confirming the variant, such as a 3-inch high-quality magnifying glass. Bear in mind that improper orientation of the glass or the viewing angle distorts the viewed image. The AGFA LUPE 8X viewer, a quite useful magnification device designed for examining slides and contact prints, is pocket-portable, inexpensive and can be purchased at most photo shops. The precise resolution that is required for the kind of measurements indicated by the term "high magnification" simply cannot be achieved with "a powerful magnifying glass and a fine scale [ruler]" as one typographer suggested. In the typical steel or plastic ruler calibrated in millimeters, the tick marks are 0.2mm or wider; what's worse, the scale becomes more inaccurate as the lenght increases. Most simple high magnification devices, such as the Radio Shack or the shirt-pocket varieties, invert the image, decrease light about 75%, and distort the image. I use a Wolfe 8x30 Field Microscope with lens extender, an internal scale graduated in 0.05mm, and negligible-light-loss non-inverting optics. Since it took about two years to locate a source for this instrument, I pass along the information: Catalogue #591305, Carolina Biological Supply Company, Burlington, NC 27215 (about $125 or higher, depending upon the dollar-yen exchange rate).

A comment by Harry Carter, one of the great scholars of early typography, is worth noting in this context: "Some of the types [typefaces] . . . can be reduced to three or four very diverse elements, more of them than I thought at first. These conjunctions and confusions are material for another, more technical discourse, of interest to people who underrate the difficulty of identifying types [typefaces] by looking at them . . .", "Types of Christopher Plantin", p. 171.

I have discussed the inherent deficiencies of reproductions for typographical study in "Reproductions as a Source", see esp. pp. 238-239 regarding some of the factors that distort the image of a type; other problems of textual transmission that undermine the

editorial process are detailed in G. Thomas Tanselle, "Reproductions and Scholarship", Studies in Bibliography, 42 (1989), p. 29. A more specific comment about the deficiencies of the reproduction processes upon which scholars rely may be useful at this point, since they vary greatly in their imaging fidelity. In general, regardless of the kind of process, essential information about a typeface is either obscured or lost in reproductions. This is partly due to the fact that the end-product used by the scholar is of at least "twice removed" quality, that is, an original is usually photographed, and this copy is then duplicated for the printing process. Fidelity decreases with each step of duplication. A xerox is a high-contrast image in which lightly inked areas are either cancelled or filled-in, depending upon the density of inking in the original or the adjustment of the machine. Essential information is inevitably lost. The trade-off between the usefulness of a xerox and the damage to the original that results from the duplicating process suggests the ethical principle that direct xeroxing be avoided without exception. If the point to be made is important enough, it will require the examination of an original for final confirmation in any event. When examined in a microfilm viewer, microfilms are virtually useless for typographical study except for sorting out categorical stylistic differences (i.e., roman vs italic typeface). The usefulness of hardcopy produced either with the photoflow (wet) or xerographic (dry) processes varies with the quality of the microfilm itself, of the copying machine, and the reproduction ratio, which is never "actual size". It is frequently impossible to judge the size of a typeface (i.e., 82mm vs 94mm) unless a ruler scale is reproduced along with the book (technicians are extremely erratic in this essential step), but this is useful only for distinguishing between different categories of typeface sizes such as pica (82mm) vs english (96mm), and certainly not for distinguishing fonts within one size range (i.e., an 80mm vs 84mm pica) or measuring 20-line height. Once it has been determined that a font is of a given size such as pica, the usefulness of the hardcopy depends entirely upon the clarity of the images, but it should never be considered an adequate substitute even for high-quality facsimiles. In regard to these, I am not really technically knowledgeable about the actual processes used, but note the following from my examination of the end products. Early W. W. Greg Shakespeare Quarto Facsimiles such as Pericles (No. 5, 1940), and Hamlet Q1 (No. 7, undated), and Shakespeare's 'Titus Andronicus': The First Quarto (Charles Scribner's Sons, New York, 1936) were reproduced by the photographic unscreened collotype method and are of quite high quality; however, at high magnification, the edges of the type-images are fragmented so that details such as nicks and precise locations of bends are lost. The collotype facsimiles vary in quality. Later Greg quarto facsimiles (such as Romeo & Juliet Q2, No. 6, 1949) were screened collotypes with a dot resolution of 0.18mm. Others used considerably finer screened lithographs with a dot resolution of 0.12mm (Henry V, No. 9, 1957, and LLL, No. 9, 1957). Screened type-images, regardless of dot resolution, do not transmit details of the image between the dots (see discussion and specific examples, "Reproductions as a Source", note 13, p. 258). The Scholars's Facsimiles & Reprints (New York, Gainesville) series varies widely but generally are of poorer quality, as are the hardcopy reproductions from University Microfilms. The Scolar Press Limited (Menston, England) offset reproductions are quite good, as is the Shake-speares Sonnets from the Nottingham Court Press (London, 1984) despite the fact its source is a British Library microfilm. The facsimiles in Shakespeare's Plays in Quarto, ed. Michael J. B. Allen and Kenneth Muir (University of California, 1981) were printed from screened line-negatives (as I understand the note, p. 899) with a dot resolution of 0.18mm; a very interesting editorial decision occurred at the final stage: "To avoid a disjointed appearance on the facsimile page, each Quarto page was overprinted with a faint, uniform, tint block. No loss in image accuracy has resulted [my emphasis], although the book has become easier to use." Numerous sheets such as R3 M1-3v definitely suffered from the tinting. Tanselle comments on the philosophical implications of such editorial modifications of the original in the duplication process ("Reproductions and Scholarship", pp. 30-32).

A mandatory exercise for the new typographer consists of recognising the effects of the reproduction process on the transmitted image of a letter by comparing the Fell 'a' with the "CICERO DE GARAMOND" screened facsimile (p. 124; also in Type Spec.

Facs. II, "Le Be" specimen 10). The AGFA LUPE 8X magnifier is marginally adequate for this exercise until a proper instrument is obtained. The features of the 'a' samples are quite uniform in the Fell specimen: the pear-tipped fineline shoulder extends beyond the bowl and rises at an acute oblique angle (28 deg.); a top-side medial bend is formed where the shoulder-stroke reverses direction and broadens into the erect back; the long up-curving base-serif symmetrically balances the stressed top of the back; the bowl's top-stroke at 28 degrees produces a high junction with the back and a crimped small counter between shoulder and bowl; the top-stroke curves down to form the left edge of the bowl. A comparison with the "CICERO DE GARAMOND" facsimile reveals that, although some 'a' samples (line 7 "coagulasti") in the facsimile seem to conform to the Fell 'a' traits, the bowl's top-stroke seems more horizontal and shoulder-pear detail is lost in the others. The direction of ink-flow away from the type face and the displaced paper fibers seen at high magnification in an original usually permit detection of such features, which are irretrievable by any means in a reproduction. Next, a xerox of a xerox of the Fell specimen can be compared to the original to appreciate the "twice removed" quality of reproductions. Note particularly the cancellation and fill-in effects; examination of a xerox of the "GARAMOND" specimen reveals how the screened dot pattern is transformed into a solid image. In general, if variant letters exhibit "major" stylistic differences, an analysis of a moderately-large sample in high-quality reproductions can lead to a tentative conclusion that they derive from different punches. For example, Carter and Vervliet trace most (but not all) of the matrices of the "CICERO DE GARAMOND" specimen ("Le Be" 10) to the set of Garamond punches responsible for three other sets of matrices represented in the "c.1585" specimen 38 and "Le Be" specimens 12,13 ("Le Be", note 13, p. 14). A sufficient number of 'a' occur in the four reproductions to reveal such stylistic differences: the shoulder is formed by a symmetrical, tight curve with no medial bend and terminates short of the bowl; the back leans to the right and is frequently bent in at the bowl so that the letter seems to lean right; the bowl's top-stroke curves closer to the horizontal (junction angle of 15 degrees) and joins the back at a lower point, resulting in a larger counter and a more open letter. Although no single 'a' precisely embodies the combination of traits seen in the group of samples, the consistency of the major stylistic differences suggests a different punch. However, an examination of the Le Be 'a' in originals is necessary to confirm these differences and the existence of a second punch (Jean de Tovrnes [Lyons], De la Vie des XII Cesars 1569 [Hn Shelf No. 404078], A4v; Christopher Plantin [Antwerp], B. Svlpicii Severi . . . quae extant Opera 1574, [Hn Shelf No. 388522]).

See Gaskell, A New Introduction, p. 38.

Blayney's statistics and discussion of the approximate size of Okes's pica font is most informative. See Texts, pp. 145-148.

It is interesting to note the use of Eld-Y1 and Eld-S in the former's final appearance in Eastward Hoe! Q3 F3v. The compositor first drew an 'H' and two 'ſſ' ligatures from Eld-Y1 on F3, then switched to Eld-Y1 on F3v. Eld-Y1 apparently was so depleted that substitutions were required with S-face 'k fi ct ſi ſſ A C E M P S'.

W. Craig Ferguson notes that Simmes acquired a new font in 1606 which prints all of Eliosto Libidinoso STC13509 except for D3-E4v, the final appearance of the long-lived Simmes-S. He printed only four books in the new font before he "was officially replaced as a printer" in March, 1607. See Valentine Simmes, Printer to Drayton, Shakespeare et al., pp. 9, 40-41.

See W. Craig Ferguson's discussion, "Thomas Creede's Pica Roman", p. 149.

Stanley Morison indicates that the set of Fell pica punches, for example, includes later replacement punches in the 'C J N R U W g' sorts: "Their presence in a uniform set of matrices suggests that a strike from Garamond's punches with these additions or replacements was justified in Holland in the seventeenth century" (The 'Fell' Types, p. 133); the english (96mm) roman matrices were supplemented by punches cut at Oxford c.1688 to produce duplicate matrices in 25 sorts (pp. 144-145). See also Type Spec. Facs. II, "Index 1567", note 26, p. 4.

The letters are reminiscent of pre-Garamond early romans, especially in the bold

'e' with the up-slanted cross. These letters provide a better size match than normal S-face letters to the sometimes overwhelming S-face ligatures, especially the gross fl2, apparent ſſ2, and ſt2, which match the height of the tallest capital ('G').

Modern typography benefits from the fact that a Caslon, Garamond or Times Roman letter is easily identified by reference to standard specimens. However, the current state of the history of early typography is a serious obstacle to historical font analysis since only a few early pica roman specimens are known that can be definitely attributed to identifiable punchcutters. The problem of extant specimens in font analysis merits further comment. The most frustrating aspect of working with Elizabethan and Jacobean books is that the published specimens do not match with the typefaces in these books, with few exceptions like Henry Bynneman's Francois Guyot pica font. In general, typographers have long resorted to comparatives such as "very similar to" or "a close copy of" in recognition of the fact that similarities in design, but not precise correspondences, can be detected between specimens and fonts in books. A. F. Johnson deals directly with this practice in "The Supply of Types". I believe that the old, frequently invoked, axiom "the same type can look quite different in the hands of different printers" can be rejected out of hand. The bare measurement of a typeface and individual letter sizes are constant physical facts: precise measurements of letters of a typeface simply do not vary significantly from one book to another or one printer to another. This is obvious from thousands of such measurements: a Y-face 'm' is always narrower than the S-face letter, and always by an amount that hardly varies. Similarly, referring to differences as "a later development in" an identified typeface is legitimate only if it can be shown that the seminal set of punches, or an intermediate set of matrices, is identical except for the differences. Ignoring these basic physical facts leads to nebulous references to typefaces which are a hindrance to font analysis and the bibliographical description of the typefaces in a book. Frank Isaac, for example, notes of the "extremely popular" pica roman: "most of the quarto plays are printed in this size. They were modelled on the 'Cicero de Garamond' of the Berner sheet which is also found in Plantin's Index," "Elizabethan Roman and Italic Types", p. 89. Isaac groups these typefaces according to the styles of 'g' and 'w' and makes a few supplemental distinctions in surveying the varieties of the two letters used by some 18 printers (e.g., "A w with a short second stem . . . was used in this group by Bishop in 84[mm], Bynneman in 86, Day in 82, Orwin in 82, Purfoot in 82 . . .", p. 90). Isaac's listing lacks references to sample books and dates, a real problem. For example, Purfoot indeed used this 'w' at least from 1570 (The Hateful Hypocrisy STC10951) in an F-font that by 1579 (The Forest of Fancie STC4271) had become mixed with an S-font. Determining when Purfoot-Y was introduced needs further investigation, but it does not use the Guyot 'w' and appears in Fawne with Windet-F which does, an occasion for confusion. Beyond that, pica roman fonts simply cannot be grouped according to a few variants. The confusion caused by such a simplistic rationale is apparent when the six pica roman specimens are examined. Henry Bynneman's font (Fig. 6) derives exactly from the Francois Guyot punches that produced the Guyot specimen of the anonymous Folger folio sheet (Type Spec. Facs. I, No. 1; Low Countries, Fig. 203, p. 268), except that Bynneman's font is cast on an 82mm rather than a 78mm body; the bare measurements of the two match (72mm). Richard Field's and Valentine Simmes's S-fonts (Figs. 8, 9) are stylistically distinct in all respects from Garamond's pica romans. Similarly, John Windet's pica (Windet-F) (Fig. 12) represents a different design concept and scale (the x-line is closer to the h-line) and is closer to faces cut by Ameet Tavernier (see Low Countries, Fig. 206, p. 267). Thomas Orwin's is a mixed font (Fig. 10) with assorted S-face letters; John [Henry?] Denham's font (Fig. 11) combines letters from "Le Be" specimens 12,13 with others probably cut in the Netherlands, but this font needs further study. [I append this brief reference to W. Craig Ferguson's Pica Roman Type in Elizabethan England (Scolar Press, 1989), which arrived too close to press-time to permit detailed comment on what seem to be fundamental problems in approach and documentation. Too many conflicts between this paper and Pica Roman Type occur to permit specific discussion; hence a general statement must suffice. Although my identifications of fonts and printers have been

presented in the context of fonts evidence because of the focus of this paper, a fundamental principle has been tacitly followed: identified recurrent types provide the final, absolute evidence of the identity of a font. Hence, my identifications, based upon such evidence, can be trusted to either confirm or supersede those presented in Pica Roman Type.]

The combinations of typefaces noted in regard to mixed fonts suggests many hypothetical explanations for the minor variations in same-face fonts. For example, the fonts were purchased at different times from one founder who had replaced certain matrices in the meantime; or from two founders who were supplying type cast in hybrid sets of matrices that differed in but a few sorts; or one of the two printers was forced by circumstances to replenish the ligatures from a founder who supplied only S-face pica type. The Plantin and Le Be inventories of punches and matrices suggest another possibility. Plantin had at least four sets of pica roman matrices from two sets of punches, one by Claude Garamond; the Le Be inventory listed eight sets of pica roman matrices (see Type Spec. Facs. II, "Le Be", notes on specimens 10, 12, 13, 19, p. 14). Secondly, although it illustrates an 18th-century foundry, a woodcut of William Caslon's casting shop shows casting stalls and casters at work along two walls of the room (see fold-out illustration at the front of Talbot Baines Reed, A History of Old English Letter Foundries [London, 1887]). Taken together, these suggest a fundamental economic rationale that could explain the mixing of typefaces either piecemeal or more extensively. A temporary heavy demand for pica roman types could be met by assigning several casters to the task using available sets of pica roman matrices; or, by assigning two casters to high-frequency sorts (a e p etc.) using duplicate matrices not necessarily in the same typeface. I must admit these explanations are purely speculative although attractive.

The following examples are based upon my own oftentimes regrettable experiences. Trips to rare books collections inevitably place a premium on time that, at least in my case, often invites short-cuts in methodology that cause later problems.

The search process will be discussed in a later paper. Relevant details about this case are as follows. While previously identifying Eld-Y1 in G-K of Whore, I had noted that EF used the large awkward 'W', the Guyot '?' and several other potential discriminants. The completed Fools EF composite demanded a follow-up analysis of EF of Whore and the second edition, The Converted Courtesan. Some Fools EF discriminants emerged in sections of Limbo-Mastix during a survey of other books by Eld, but it is not the same font. Fools EF discriminants were later encountered in books by Simon Stafford and Edward Allde, leading to the former's identification. I had noted some traits of d'Olive EF while analysing Creede-4 in A, B(i), and these suggested the comparison to Fools CD, I.

See Fredson Bowers, "The first quarto of 1604 was set up in three sections comprising sheets A-B, C-D, E-K. . . . This final section of the play, E-K, appears to have been machined on two presses, one of which printed sheets E, F, H; and the other, sheets G, I, K" (p. 4) and other discussion of the compositors, running titles and skeletons in the two editions, "Textual Introduction", The Dramatic Works of Thomas Dekker Vol. II (Cambridge, 1955), pp. 3-14. I might note that identifying running titles is a difficult proposition simply because they are set in same-face letters. As I note in "Reproductions as a Source", most of the letters in typical running titles are unidentifiable because of the lack of specific damage (see pp. 242-244); hence, without specific evidence to the contrary, concluding that the same running titles move through a book is based upon an assumption of regularity in compositorial procedures at one press or in one printing shop. Shared printing is inherently irregular in this sense since a minimum of two compositors, two presses, two skeletons and casting-off of copy is required. The confirmation of printing in one shop by a font analysis must, therefore, precede running-title analysis. In the event that two fonts are found together within a single gathering (as in Eastward, Fawne E-I, and d'Olive B), a survey of the printer's work is an additional prerequisite in order to confirm the possibility of simultaneous printing in a two-press shop. (Aside from d'Olive, I am unaware of any instance of two printers' fonts appearing together within a gathering.) Otherwise a dangerous trap ensues when the running titles in sections of a shared book are set in same-face fonts without distinguishing peculiarities such as a distinctive damaged letter

or stylistically distinct letters (e.g., italic 'swash-curl-M, swash-P' in A-D, Granjon 'M P' in E-I). This is evident in several of the studies cited in note 6, where the same-face running titles of one printer are assigned to another of the printers in a book.

Locating low-density foul-case letters such as these is especially tedious. I usually do a minimum of three searches of a text (or texts). One cardinal rule is to note the location when the letter is first detected, since a repeated search is no guarantee that it will be found again. It is worth noting again that non-appearance does not mean that the letter is not in the font.

The complementary black letter font was a major source of foul-case punctuation, but I know of no resident fouling with letters from this typeface except for punctuation marks. However, infrequent instances of textual fouling with the black letter 'ye' occur (Antichrist STC7120 [1603], [2]H3v:32; The Woeful Cry of Rome STC1833, B2:3, B2v:14; Regiment STC1827 [1606], I1:18; W. Craig Ferguson noted 'ye' fouling in Creede-3 also, "Creede's Pica Roman", p. 152). This 'ye' fouling should be viewed as transient fouling resulting from the compositor's justification habit in prose texts, and not as resident in the roman cases. A rather unique kind of textual fouling with a large pre-Garamond roman font occurs in Isle of Gulls (STC6412, 1606), where the left-third or so of F4v:31-39 are set in this oversized Venetian(?) type. Fouling with 'C' occurs elsewhere throughout this section (EF) (F4v:13 etc.). The discovery of this unique font in a signed book would automatically identify the printer, given its rarity in Elizabethan printing.

The three styles of italic capitals in use during the period provided obvious discriminants. In the most common, stems and ascenders of the capitals are straight and terminate in serifs (as if a roman face were titled to the right); this style appears in the pica italic faces cut by Robert Granjon (see The 'Fell' Types, p. 139) and Ameet Tavernier (Low Countries, Figs. 233-234, p. 305). Granjon also cut a second style of italic capitals characterized by extended "swash" strokes. The left stem of the 'A M' is a long, graceful curve; the bowl-stroke of the 'B D P R' begins with a swash, the 'G' terminates in a descender swash, the tail of the 'K Q R' terminates in a swash below the baseline, the diagonal stroke of the 'N' incorporates both the shoulder and base swash, and the 'E' is like a reversed numeral "3". Both styles are mixed in Eld's, Braddock's, Purfoot's, Short's and other italic fonts. Read's (and Eld's early) font included a variant 'A M' in which the swash terminated in a curl, fairly common variants. Francois Guyot cut a third style of capitals (see Low Countries, Fig. 231, p. 301) in which the bowl-stroke of the 'B D R P' begins almost vertically at the x-line to the left of the ascender, the left top-serif of the 'M' hooks over, and the swash diagonal of the 'N' extends below the baseline. The Guyot style was less common than the Granjon.

G. K. Hunter noted of Mal Q1: "One of the problems of printing The Malcontent seems to have been the very large number of italic Ms required for the names, and this put the printer to all kinds of shifts. But at the beginning of sheet F, and incessantly thereafter we find a new kind of italic capital M [Guyot], which we must assume to have been in one compositor's box but not in the other's. That this M was in Sims's possession is shown by its use in QC [Q3] of The Malcontent and in that part of The Honest Whore printed by Sims," "Introduction: The Text", The Revels Plays: 'The Malcontent' (1975), p. xxxii. Simmes's section (F-H) of Q1 also uses the Guyot 'P' in large numbers, and both Granjon styles of 'B M P'. Eld's section (B-E) uses the straight-stem Granjon and the curl-swash variants of 'A M', swash 'B D P', and what appears to be a few Ameet Tavernier 'M' (B3v:34, C1:29, C4v:8, D3:32, E2:19, E3:32). The Guyot 'M P', and the curl-swash 'A M' and Tavernier 'M', are exclusive to the two sections of the book (and Simmes's and Eld's sections of Whore as well). Given the number of 'M' and roman substitutions, the situation suggests two separate printers' fonts rather than two cases, and calls for a survey of a few other books by Simmes to determine whether the letters found exclusively in Eld's sections were ever used by Simmes. Unfortunately, Hunter seems not to have been aware of W. Craig Ferguson's Valentine Simmes, Printer to Shakespeare (1968).

I must note the possibility that the replacement of a large portion of the italic font could be responsible for the shift in these sorts. The italic of Q3 prints much more

crisply than that of Q1-2, a fact which strongly suggests new type. The difference cannot be noted in reproductions.

See Robert K. Turner, Jr., "Printing Methods and Textual Problems in A Midsummer Night's Dream Q1", Studies in Bibliography, 15 (1962), 33-55. The major portion of the study aimed at establishing the setting and distribution sequence by correlating 115 reappearances of identified types with evidence from substitutions. Turner's use of a reproduction for the typographical identifications produced an error rate of 44% which, in combination with my identification of 47 additional appearances of damaged types, discounts his conclusions about the setting and distribution of the text and his interpretation of substitutions; see my discussion of Turner's distributions (pp. 240-242) and the Table of Identifications (pp. 266-268), "Reproductions as a Source".

The distribution of foul-case letters deeper into sort boxes or into a different case, or the setting of a text from alternating cases, one of which is not used after a certain point, can only be surmised from the behavior of foul-case letters which, in any event, is random. Although it can be assumed that types in a sort emerge in a "last distributed, first to be reset" pattern, the pattern is merely probable, since a certain randomness is introduced by the order in which pages of a forme are distributed: the last page to be set is not necessarily distributed first. The genuine exhaustion of a sort prior to substitutions would seem to pull all normal letters from a sort box, but in order to establish a real number, the point in the text at which setting from the sort began would have to be established, and that is possible only through an exhaustive identified-types survey to accurately pinpoint distributions. Hence, this discussion of the behavior of foul-case letters ignores the possible effects on recurrence behavior by setting from two cases and by distributions. Nonetheless, distribution factors seem responsible for recurrence in many instances. For example, the damaged 'ſſ' in the 'ſl' sort appears eight times (A2:9,12[text], A5:11, A6:15, A7v:22, B1:20, B4:11, B4v:11, C1:10) in 3H6 but not later in this book in either the 'ſſ' or 'ſl' sorts; seven times in R3 (H3v:28, I4:35, I4v:29, K2:27, L4:4, M1v:21, M3:27); none in 1H4; then six in Secretary (F3v:20, I1:25, L3:13, T3:36, Ff3v:8, Gg3v:10). The damaged ligatures were purged and newly damaged ones wrongly distributed in the sequence of texts except for the one which reappears (R3, L4:4, M3:27; Sec, F3v:20). It is worth noting that the right shoulder of the foul-case 'ſſ' at Sec I1:25 is not sheared away, but merely bent downward rather severely. This alteration of its appearance was sufficient to occasion incorrect distribution into the 'ſl' sort.

Recurrences of Y-face capitals across three gatherings in the verse text of Sejanus: 'S': B3:38, B3v(5), C1v:16,24, C2v:24, C3:10, D1:16, D1v:9,11, D3:2, D3v:2,3; 'T': B2:4,16,20, C3v:14, D3v:28, D4:20,21,30; 'A': B4:31, C2v:24, C4v:17,22, D3v:33, D4:18. Despite the extremely low demand on the sorts in the prose text of Spider, it was sufficient to produce the following appearances in the first gathering of Eld's section: 'S': H1:13[text], H1v:4; 'A': H2:30, H4:29; 'T': H4:20. In addition, new fouling occurs with Y-face capitals in 'F' (H4v:13) and 'G' (H3:4); 'C' also appears later (I1:3). Incidentally, the mixture of S- and Y-face 'I' occurs in both texts.

See Sejanus: 'g': B1:32, B1v:6, B2v:9; 'k': B2v:18; 'p': B3v:37. In Spider, the prose text requires a greater number of lower-case letters, but this does not seem to increase the appearance rate of low-density fouled letters: 'g' appears at H1:6[text], H2:28, 'p' at H2:28, but 'k' is lacking until I1:25, I1v:23.

I think it fair to observe that consistent performance in the purging of fouled speech prefixes, emphasis names and stage directions (roman capitals with italic lower-case) exhibits no real significance with respect to compositorial skill: the roman capitals preceding 2-10 italic lower-case letters present a minimal challenge, if any. The capitals are almost invariably distributed back into the roman pica cases as numerous compositorial analyses have shown. However, the practice of automatically interpreting roman substitutions (capitals) in speech prefixes irrespective of the italic font's history merits comment. The possibility of fouling the italic cases with roman capitals calls for caution in invariably assuming that substitutions are transient fouling. The S-face 'S' substitutions in Creede's printing of Henry V Q1 (1600) may represent an exception. The ratio of S-face 'S' to italic 'S' is: D:

4:14, E: 6:17, F: 7:10, G: 0:4. When Creede printed R&J Q2 in the previous year, the italic 'S' sort set 20 speeches prefixes in A and 10 in B; the first apparent substitution with S-face 'S' occurs at C3:2 in a gathering which used only 8 italic 'S'. Then in Pastor Fido STC12415 in 1601, a single S-face 'S' appears among 24 italic 'S' in B, and another at K4:15 amidst a few italic 'S'. I am at a loss to explain how the C3:2 and K4:15 appearances could be viewed as substitutions in view of the quite light sort-pressure. Even so, it is possible that Creede's italic 'S' sort was severely depleted between 1599 and 1600, and then replenished to par by 1601 (or was being used simultaneously in another text). However, resident fouling with S-face 'S' is an equally plausible explanation. A lapse in the compositor's attention cannot be ruled out either, especially in view of the inexplicable shift of the text typeface from roman to italic in most of B4 and parts of B4v in R&J Q2.

The density is fairly consistent throughout 3H6 which was printed in eights. Dividing into units of eight pages (equivalent to a quarto in fours) gives the following numbers of black letter queries: 0, 11, 9, 14, 5, 8, 6, 9. The numbers in 1H4 are: 0, 4, 8, 14, 16, 1, 9, 7, 8, 6. In Short's section of R3 (H-M3v): 2, 1, 15, 2, 1. The Y-face and italic query numbers are omitted to emphasize the variation that occurs in R3, although the pressure on the '?' sort in the three texts is not significantly different. What remains, I believe, is a reflection of compositorial intention either to purge or to tolerate obvious foul-case letters. Since both cases were used to set R3, Short-Y as a whole was imperfectly purged of the black letter query prior to setting.

See MacD. P. Jackson, "Two Shakespeare Quartos: Richard III (1597) and i Henry IV (1598)", Studies in Bibliography, 35 (1982), 173-190. Jackson did not consider the possible connection between foul-case evidence and the cases from which each compositor set his respective section of the text.

Foul-case italic T1 occurs in 3H6 (D5:19, E6v:3?) and R3 (I1:24, L4v:9), but not in the two later texts, suggesting a purging of the text of R3, but not the cases. The 'T' in 3H6 and Secretary do not correspond to any seen in R3 with one probable exception (R3, L4:26, M2:33; Sec, Ll4v:29).

The former may simply skip Fawne in the manner characteristic of low-density fouling. However, the high density of the 'I T' fouling precludes the possibility that it could skip gatherings or texts. Furthermore, both cases were fouled since the 'I T' appear in both formes and halves of each gathering A-E. Deliberate purging of large-scale fouling is clearly indicated. The 'I' in Fawne Q1 C2v(5) and D2v(5) is probably fresh transient fouling. DC was entered 26 June, 1605, and Spider on 4 February, 1606, about six weeks prior to Fawne. Appearances of Purfoot-Y1 italic cluster members: 'B': DC, C3v:16, D1v:11; Fawne Q1, C4:21. 'C': DC, D3:24, E4:13, E4v:12; Spider, C4:4; Fawne Q1, C1v:13, Q1-2, D2:10, D3:36. 'G': DC, D2v:14, E1:16; Spider, B3:37, B4v:34, C2v:9, E3v:20, F2v:17. 'L': DC, A2v:9, B3:21, 25, C1:29, C2:18,20, C3v:15,17, D1v:38,39, D4:12,34; Spider, G1:5, G2: 31, G4v:22; Fawne Q1-2, D2v:34. 'S': DC, B2:24, B3:3, B4:3, C1:28, C1v:1; Spider, C4v:20, D3v:31; Fawne Q1-2, D2:30.

Appearances of common low-density cluster members in Whore: 'C': D1:5, D1v:37; 'F': D4v:25; 'G': C1v:5, D1v:31; 'swash-G': C3v:13, C4v:7; 'P': D2:23; 'W': D2:32,33; Y-face 'M': D4v:31. Appearances in d'Olive: 'C': B3v:28,30, B4:2,26; 'F': B1v:30, B2:35; 'G': A4v:29, B4:17; 'swash-G': B2:15,27, B3v:15, B4:5; 'P': A2v:6, B1v:17, B4:21[2]; 'W': A4v:22; Y-face 'M': A3v:22. Incidentally, Windet-F prints Ee1-Ff4r of Regiment STC1827.

Windet-S2 sections: G1:1-8, G1v, G2, G3:1-28 except most of 17-18, G3v, G4, G4v; F1v, F2, F3; E1, E2v, [E3?], E4v.

The Fawne was entered 12 March, 1606, five days before Sophonisba, but this does not definitely indicate the order in which the two were printed. In any event, the resident roman fouling is the same in both and was not purged between books.

S-face k1: Mal Q2, C3v:24; Q3, I1v:14; Whore, G1:19, G2:18, G4:23, Iv4:8; Fools, A4:14[text], B3:29, G3v:37, K1:7; Eastward, B1v:15, B2:28, C2:4, H2v:11, H2v:35. S-face g1: Mal Q1, B2:16, C4v:28; standing type Mal Q1-2, D4v:16, E4v:11; Q3, H2:24, H2v:10, H3:24, H4:26, I1:34, I2:22, I2v:4,8, I3v:27; Whore, G1:8, G2:18, H1:5, H1v:30, H2:33, I2:6,30, I4v:7, K1:32; Fools, H1:5, H2:28; Eastward, D4v:37, E1:14, E2:5, F3:27, H3v:17,29, H4:33. Expanded S-face x2: Mal Q3, H1:8, I3:29; Eastward, E1:4, F3:20, H4:20.

Errors and omissions in the list of the 'k w' appearances noted in "Reproductions as a Source" (note 39, p. 264) are corrected as follows. Windet-F 'k': Q1, E3:8; type k2: Q1, F4v:30, H3:4, Q2, E3v:14, H3:14; both Q1-2: H1v:17, H3v:27, H4:17, I1v:29, I2:23, I3:6. Windet-F 'w': Q1, E2v:8,36; Q2, E3v:11, E4:15; both Q1-2, H1v:17, H3v:27.

The recurrence of the S-face A1 in MND Q1 has been noted earlier. The number of appearances (28) of S-face 'A' indicates that replenishment introduced this letter along with the S-face 'H S' (with a few 'Q R V') and 'ſt' to form the S-face cluster in Braddock-Y1. See also later discussion and the Braddock-Y1 Tables in the Appendix.

This 'W' also appears regularly in 3H6 (B6v:12,32, C3:15, C5:24, D1:30, D2v:26, E2v:12) but does not seem to occur in 1H4 or Secretary.

Spectres was entered 11 January 1605; Fawne on 12 March 1606.

The damage is of the kind that a compositor or corrector could not ignore. It is not clear whether Antony Hammond meant to imply a general principle about compositorial behavior in noting that "such types are likely to be noticed by the compositor or proofreader, and discarded rather than distributed," "The White Devil in Nicholas Okes's Shop", Studies in Bibliography, 39 (1986), p. 160. To be on the safe side, it seems wise to take his statement as true of Okes's workmen; it does not apply to every shop.

Implicit throughout this discussion is the categorical necessity of the survey as a prerequisite to compositorial and presswork analysis. Unless this principle is clearly understood, the potential for error is limited only by the inferential ingenuity of the scholar who misunderstands or ignores font evidence. For example, an extraordinary sequence of inferences in Akihiro Yamada's "Q1-3 of The Malcontent, and the Compositors" leads to the rhetorical question: "But was it in fact practical to use only one copy [printer's copy] when two compositors were working at different shops, as must be the case with Q3 according to the implication [actually an unqualified conclusion] of Ferguson's statement that the type used in H-I of Q3 is one 'which Simmes is not known to have used'. Or should it be believed, in order to settle this question . . . that this particular type was in fact in Simmes's possession?" (note 14, p. 130). The implicit answer is absolutely wrong. The "tail" of compositorial and presswork analysis, however ingenious and cogent, simply cannot "wag the dog" of a survey of nearly all the books (in the original) by a printer. Ferguson's survey indeed proved that Simmes did not use this type. The extreme improbability that he nonetheless possessed it is disproved by my survey of Eld-Y1 which yielded about 450 appearances of 85 identified types from The Malcontent in a dozen other books by Eld. Such simple facts as are yielded by the survey can, in effect, "trash" studies of printing which overlook font evidence. (I was tempted by insatiable curiosity to tabulate the number of pages "trashed" by the font evidence noted herein, but the thought of adding even one more statistic tipped the scale to pity, leading doubtlessly to a much-deserved mutual sigh of relief).

A few of the previous isolated examples can be combined to illustrate the point. Given the fact that both Purfoot-Y1 and Eld-Y1 are used in two-font play-texts 1605-1606 by John Marston (Fawne, Eastward Hoe! etc.), a comparison of the two same-face fonts would be the logical first step in determining whether either printer appears in both books. A-D of Fawne exhibit high-level replenishment of upper-case sorts with S-face letters, a foul-case cluster of italic 'B C L S', both erect and leaning italic query (D4:5), colon and exclamation point (A4v:17), while the Y1-font sections of Eastward show a low-density (foul-case) cluster of S-face 'D E G H K M O P S T g k p x!?', black letter and leaning italic'?', and damaged 'ſſ' ligature in the 'ſl' sort. Or, Windet-S in E-I of Fawne could be compared to Eld-S, since Windet printed Sophonisba, another Marston book. In this instance, the fouling is in contrasting typefaces: Eld-S is fouled by Y-face 'g k y', while Windet-S is fouled by Windet-F 'a k w x y' and 'ſt' ligature as well as italic capitals. One final qualification is necessary in concluding this discussion of the random behavior of wrong-face letters. The fonts have been treated as a whole while ignoring the probability that two sets of cases fed the texts included in the examples. It seems likely that the use of statistical methods for analysing the frequency distributions by pages will distinguish the cases from which texts were set, given clusters of appropriate dimensions. However, more work is needed in this area.

Although the group of fonts considered in this discussion represents but a small

fraction of the pica roman fonts used during the period, I believe that the cluster phenomena noted in these fonts suggests a definite possibility that two same-face fonts could include clusters with some lower-case duplicate members, although I have not encountered such an instance. The commonness of italic 'I S T' fouling and wrong-face replenishment in sorts such as 'g k p y' certainly point in that direction. Replenishment of the upper-case suggests a second hypothetical consideration. The progressive replenishment with wrong-face capitals in an increasing number of sorts, for example, could conceivably transform a font with S-face capitals in just a few sorts (Braddock-Y1, 'A H Q R S V') into one with S-face capitals across the upper-case (Purfoot-Y1), given an adequate time interval. This is implicit in the continuum of proportions of normal and wrong-face capitals seen in mixed fonts.

See my discussion of the inadequacies of reproductions for typographical study of the kind needed to demonstrate printing from standing type and to identify running titles, "Reproductions as a Source", pp. 242-245.

standing from Whore