University of Virginia Library

A Technique of Headline Analysis, with Application to Shakespeares Sonnets, 1609
by
Randall McLeod

Since the 18th century collation has been a standard procedure in editing Shakespeare. At first it sought lexical differences between editions; but recently variation within the same edition has come into focus, largely through the work of the late Charlton Hinman.[1] His mechanical collator and its adaptations[2] have revolutionized collation. Instead of comparing words

one at a time in two different exemplars of an edition, it compares whole pages simultaneously as visual gestalts, making obvious even minor differences between them, such as unequal inking and change of type or damage to it.

Let me illustrate the principle with an example from Shakespeares Sonnets, 1609 that has hitherto escaped notice, and in the process introduce a photographic method of collation that does not employ expensive mechanical collators. The title page of the Q1 of 1609 has long been known to have two imprints. In the account of Hyder Rollins, the authoritative Variorum editor, the title pages "represent an identical setting of type except for the imprints."[3]

What happens when we move beyond the unaided eye and superimpose the images, however, contradicts Rollins' assessment, as the following plates will show. Before I analyze their evidence let me describe the simple process by which these images are obtained. I began with photocopies of the title pages of the two imprints generously supplied by Curator Carey S. Bliss of the Huntington. As the Library photographed them in slightly different magnification, one of the images was copied again, and this served to bring them to size. (Photocopiers rarely conserve the size of the original.) This experimental process of changing the magnification was repeated with the photocopier stocked with transparencies rather than opaque paper.[4] The resulting transparent image of the one imprint was then placed over the opaque image of the other supplied by the Library, and the two images were tested for alignment. This securing of alignment is cumbersome in mechanical collators, especially as in the present case when different parts of the page align on different axes, but hand-held images offer no such problems. Once a significant alignment had been found, the two pages were placed transparency-down on a photocopier to make an image of the superimposition. In the first illustration (Pl. 2) the alignment is of the ornament and 'SHAK' at the top of the page. The alignment in the second illustration (Pl. 3) is toward the bottom of the page (as in the colinearity of the rules).[5]

The alteration of the imprint, long known, seems to have been accompanied by something never guessed at. First there is horizontal and vertical displacement of rows of type between the two images, although identical letterpress is found in common text. The horizontal movement might, to some degree, be explained by application of different pressures in locking up the forme after the change of imprint (for collation of the surviving title pages reveals that the gross differences observed here correspond to the change of imprint). But the vertical displacement of type can be explained only by the transposing of a reglet from above to below 'SONNETS.'—or vice versa. This alteration occurs in conjunction with another, revealed by the fact that the two images of 'SHAKE-SPEARES' do not fully align with each other (see Pl. 2). This evidence can be readily understood by another manipulation of the photoimages (a procedure impossible on mechanical collators). The following plate is made as the former two except that the transparency is turned over and rotated 180° with the result that the two imprints can be

contrasted as mirror images along some useful axis, here the base line. The images can be shifted slightly along this axis in order to align mirror images of particular types or type groups. Plate 4 shows two such different alignments.

The explanation of the misalignment is now obvious: the first and third E's in the name have been interchanged; this feature, like that of the transposing of the reglet, coincides with the change in imprint. In the Wright title the first E (that in 'SHAKE') has a slightly narrower set than the two E's in 'SPEARE' and may represent a different fount. In the Aspley title this E with the narrower body is transposed with the second E of 'SPEARES'. The difference in the bodies of these transposed types throws off the alignment of '-SPEARES' up to the final S by altering the spacing before the hyphen. Of course, there is no change in length of the whole line.

The order of these changes, which coincide with the unlocking and adjustment of the page to alter the imprint, is not to be determined by strict bibliographical evidence. It is possible that in the unlocking of the page the

first and perhaps the second line of type 'SHAKE-SPEARES | SONNETS.' pied and was reassembled, in the process the two E's being interchanged and the reglets rearranged. However, since the position of the narrower-bodied E in 'SHAKE-' in the Wright state of the title produces a more pleasing typographical effect than when it is the penultimate letter in 'SPEARES' in the Aspley state, it is also possible that the change was deliberate, in which case the Wright was the later imprint. The cause can remain only a matter of opinion: it is sufficient to state the evidence and to testify to the ability of mechanical or photocollation to isolate such minute differences between texts, differences unknown a generation ago when the Variorum Sonnets was published, a model of editing in its day.

Let me now turn the techniques used here to a specific kind of problem that arises in close bibliographic scrutiny of any book—analysis of running headlines.

The Headline

The first scholarly attention to headlines came over a century ago,[6] but the realization of their analytical importance should be dated from the work of Professor Bowers, beginning in the late thirties.[7] The usefulness of headlines in deducing aspects of presswork—one of its primary uses in bibliographic scholarship—has been called into question by D. F. McKenzie in his "Printers of the Mind." [8] But, his criticism is merely a salutory reminder that conclusions should be warranted only by evidence; the bibliographical analysis of the headline still remains valid as a technique.

Now, headline analysis entails collation, but collation of this kind is not primarily between surviving copies of an edition (though this is still important), but within any copy of that edition. To save a compositor the labour of continually resetting them, headlines are usually repeated within their skeletons at varying intervals throughout a book. In fact, the earliest English printing manual, Moxon's Mechanical Exercises, 1683-84, stressed the importance of keeping intact the configuration of the skeleton (which includes the headline) when it moves from one forme or signature to another and is imposed about new letterpress. The recurrence of headlines can be readily

ascertained when large type is used, as in Shakespeare's Folios, for damage or idiosyncratic features render the types identifiable. Gradually it was realized that, besides the evidence of typeface, the nonprinting types that space and justify the headline also contributed evidence—the lengths of their segments. In fact, such information proves invaluable in the present case, Shakespeares Sonnets, 1609, because the typeface is too small and the inking too irregular always to permit identification of single types.

Traditionally headlines have been analyzed by the painstaking method (attributed to Charlton Hinman)[9] that entails exacting measurements of various parts of the headline arbitrarily chosen and then comparison with similar measurements from other headlines. As there are in the present quarto four new skeletons employed in the first four sheets, the analyst must accumulate a mass of data blindly before even a tentative analysis can emerge. Even so, as there are no exact margins to measure from, and as the inking is variable and sometimes makes two impressions of the same type slightly different in size or placement, one cannot be sure that the correct distance is being measured. Even with the discovery of a repeated headline in different formes, one is still far from knowing the correlations of the headlines in the forme mates.

It is possible to collate headlines on the Hinman collator, but only if one possesses two copies (or photocopies) of the text and, of course, has access to a collator. Although individual headlines can be matched by keeping one copy fixed and turning the pages of the other copy one at a time while searching for correlations, the going is necessarily slow. In the quarto format headlines as part of the skeleton come in groups of four and are printed on one of the formes of a sheet; ideally one needs a method that compares such groups simultaneously, but the binding process forbids this formal simplicity. On the collator one fumbles with atoms when a molecular approach is required. To obviate these difficulties I have devised a method of photo-collation. This method has the benefit of displaying its evidence readily in the formats used in the plates of the present paper. The procedure is simple. A photocopy is made of the original text, or even, as in the present analysis, of a well-made photo-facsimile of the text. Before copying begins, the copy machine should be checked for distortions, and the copying should be done all at one time with the book in the same orientation and location. As light weight paper as possible should be used, for translucency is a necessity. The copied pages are now cut apart, trimmed where necessary, and loosely taped together in groups of four in the arrangement by forme they had on one side of each original sheet. This returns them to the molecular format mentioned above, and undoes the folding and cutting of the binder.

The comparison of headlines grouped in formes can now proceed at a rapid rate. One simply superimposes any two of these reconstituted formes one over another on a light-table or sunny window pane. Then, with a slight

shifting of one forme relative to the other, one aligns margins, individual words, letters or spaces in the headline. If no significant pattern emerges, one forme should be rotated 180°, and the matching repeated. Observation is often facilitated by interchanging the top and the bottom formes. Also useful is turning one forme over to compare headlines (one of which will now be in mirror image) along their base lines. These procedures will quickly reveal any molecular correspondence present.

The following theoretical information and practical example show the great precision of photo- or mechanical collation in this kind of analysis. An individual typeface that is indistinguishable from others of its sort can sometimes be identified by the effect of its typebody (and the typebodies next to it). Printers have long been aware of the individual set of each piece of type even in the same sort in the same fount. Writing in The Printer's Grammar, 1755, Smith showed that the same words "composed out of the same Cases, without picking or chusing the Sorts" exhibit "a small difference in the thickness of the same Sorts in one word," hence "a greater might be discovered in a long line."[10] Smith's interest, of course, is in justification and in driving out and getting in, but the principle serves well those who wish to identify individual types. However much the two headlines of Plate 5 look alike by virtue of similar typeface and centering, the internal spacing of the typefaces as a function not only of their sets but also of any letterspacing present reveals decisively that they are not printed from the same array of types.

Headline analysis of Shakespeares Sonnets, 1609

This text collates [A]2 B-K4 L2. Jackson's analysis of the headlines revealed ("by careful inspection and measurement") that for B—K the inner and outer formes of each sheet were imposed by the same skeleton, but that several skeletons were recycled through the gatherings as follows: #1 for B and H; #2 for C, F and I; #3 for D and G; and #4 for E.[11] Overlay collation verifies these findings quickly. However, the real problems with the headlines of the quarto have not been touched: the headline changes from 'Shakespeares Sonnets' to 'A Lovers Complaint' in K; and L exhibits variation in the headline, as L2v reads 'The Lovers'.[12] To understand the complete picture of the imposition of formes we need to consider bibliographic, not textual, units.

As the solutions to the problems of K and L will prove to be interconnected, let me describe the problems of L briefly, and then proceed directly to K. Since signatures [A] and L contain only four pages each, we may suspect half-sheet imposition. The two common methods of such imposition either work-and-turn all of [A] in one forme, all of L in another; or work half of [A] and half of L together in each forme. Can we determine which method was used? To answer this question we need only remember that if headline material is shared between L(o) and L(i) we have positive proof that L was imposed by the second method.[13]

Overlay collation and base-line mirror-image comparison quickly reveal decisive information to show that inner and outer L were imposed separately in two different formes. First, the headline 'Complaint' in L1r and L2r is identical. Second, the letterpress 'Lovers' is identical in both headlines L1v and L2v. Unmistakable are the large o and the slight misalignments of the letters. However, as noted before, the wording of the headline is changed, one headline having 'A', the other 'The'. In addition, the spacing between these different articles and the following word 'Lovers' is also different (see top Pl. 7). In any case, common types in L(o) and L(i) indicate that L was printed by twin half-sheet imposition (likely with [A]), using, as did the other sheets of the edition analyzed thus far, the same skeleton for inner and outer formes. The next question, simplified now that we know the imposition of L, deals with the provenance of its skeleton. Should we suspect derivation from that used in K, the only other setting of the same headline text? (If this suspicion proves true it would be an atypical practice, for alphabetically sequential formes elsewhere in Q do not use the same skeleton.) Let us enquire, then, about K.

K(i) and K(o), reports Jackson, use the same skeleton. More accurately the skeleton of one is derived from the other, for as the following figure shows, only 50% of its headlining is textually the same. Overlay collation quickly reveals that the headlines common to both formes, whichever way the derivation went, are the non-conjugate pairs on the right side of the diagram of each forme. The headlines to the left of each forme do not contain common text, and so naturally we find no correspondence here.

Now, if the imposition of L began when the latter forme of K had been machined, rinsed and unlocked, the compositor would have found his two required headlines for L ready-made in the conjugates K2v-K3r or K3v-K2r. But examination shows that this easy course of imposition was not followed. Let us assume, then, that one forme of L was imposed while the latter forme of K was not yet available for unlocking, and ask, Which forme of K was machined first? If K(i) were machined first, then the 'Complaint' of K2r would be freed for use in L, as it would be replaced by 'Sonnets' in K1r. If K(o) were machined first, then 'A Lovers' in K2v would be released for use in L, its place being taken by non-printing types, spaces or furniture in K1v. But obviously a complete running title could not be freed for use in L while either forme of K was locked up.

Overlay collation and base-line mirror-image comparison of the letterpress of 'Complaint', common to L1r and L2r, show that it at least does not derive from K. The other headline presents a different and more complicated story. Collation of K2v and L1v shows immediately that, first, the word 'Lovers' in K2v has all the characteristics that we saw in L1v (and L2v). Most noticeable is the large o; second, the 'A' of 'A Lovers' in K2v does not correspond with that of L1v; and third, the spacing between 'A' and 'Lovers', and also between 'Lovers' and the left margin differs between formes. This is all shown in Plate 7, which begins with the comparison of common letterpress from headlines in L(o) and L(i), and ends with comparison of the same letterpress in L(i) and K(o).

We now have enough information to conjecture securely on the sequence of imposition of these formes and the provenance of the headline error in L2v. Let us construct two hypothetical paths of the headline from K through L. The type of the headline in question must follow one of two paths: Path 1, K2v→L1v→L2v, or Path 2, K2v→L2v→L1v. Each hypothetical path contains two re-impositions. The evidence does not allow complete certainty, but the second path is much more likely. The first hypothesis strains credulity with its train of loss, accurate correction and subsequent miscorrection of the same part of the headline. The second hypothesis necessitates only one loss, erroneously reset, and then a subsequent correction, and thus does away with the awkward miscorrection.

The following represents a probable reconstruction of events. The compositor unlocks K(o) and sets aside the headline from K2v ('A Lovers'), or perhaps only part of it, as it is incomplete when we see it again. K(i) is then imposed and sent to press, where it remains when the compositor goes to impose L(o). The only headline already set and available that can be used in L(o) is that from K2v, now incomplete. The compositor takes the remainder ('Lovers') and sets spacing and 'The' to go with it. We see in the choice of the article, perhaps, an easy confusion of the singular possessive with a plural form, an easy error, as the contemporary spelling "louers" does not distinguish them. The mistake is easy to set in this forme, as there is no setting of the correct title elsewhere in the skeleton. The proofreader can miss the error for

the same reason. (The compositor also sets a new headline, 'Complaint', for the conjugate (L1r).) By L(i), however, the mistake in L2v is noticed and corrected. Thus we can assert confidently that the order of formes through the press is likely K(o), K(i), L(o) and L(i).[14] This sequence argues strongly for cast-off copy in "A Louers Complaint."[15] Another source of information, recurrence of distinctive types in the text, can be utilized to show the probability of cast-off copy throughout Sonnets.[16]

It remains to enquire whether Jackson is correct when he states, "naturally a new skeleton was constructed for Sheet K." By "naturally" I assume he means that as seven of the eight headlines in the two formes of K are different textually from any preceding headlines, those parts of the skeleton must have been newly composed. But could not a chase and part of its skeleton used in the printing of some earlier forme have been reemployed in K? There has been labor-saving recycling of skeletons between sheets ever since F, and one of the headlines in K does read 'Sonnets'. When applied to this problem, overlay collation shows that the headline of K1r seems identical to that of E1r. Normally when we compare recycled quarto skeleton formes we have four repetitions of identical letterpress, and uncertainty of identification of any one headline can be overcome by the corroboration of the other three. Nevertheless, the evidence of overlay collation calls in doubt Jackson's ascription, and suggests the recycling of skeleton #4 from E to K, whence it derived to L (see Plate 8).

The argument has necessarily been preoccupied with the specific text analyzed. In conclusion let me stress the main interest, the general technique of analysis. The means of collation described in this paper is inexpensive and rapid. Its photographic format provides durable information ready for

double-checking results and reporting evidence. Its reconstitution of the forme restores relevant spatial relationships of imposition that are lost in binding and cutting. Conveniently the use of photocopiers means that illustrations of bibliographic analysis, as in the present article, can be assembled rapidly and inexpensively from research materials. The method should be of use to anyone engaged in headline analysis. Finally, as the collation of the title pages of Shakespeares Sonnets demonstrates, it can double in a pinch for a mechanical collator.

Notes