University of Virginia Library

It is always disturbing when logic outruns judgement. We feel obliged to accept an argument but remain, inexplicably and therefore silently, unconvinced by it, troubled by a gentle but patently unfair scepticism based on no more than a conviction, bred of our own imperfect and mixed experience, that things can't have been quite like that.

Take, for example, these two propositions: first, one compositor may be distinguished from another by his manner of spacing punctuation; second, a compositor will, more often than not, set type from the same pair of cases, so that recurring types may be used to plot his work. Since the propositions are simple, and the evidence adduced to exemplify them wholly inferential, it is not difficult to achieve a logical formulation. When there are parallel structures conceived to the same end they will, syncretically as it were but not logically, reinforce one another. So the identification of compositors by spelling, or the division of work between compositors according to a pattern of skeleton-formes, may extend the series of tests and, where they match up, convert correspondence into truth.

Because such tests are only syncretically or accretively, but not logically, related, it is no great loss to the argument if any one of them fails. One merely shifts ground. Were they logically dependent upon one another, since they are directed to the same end ('There were three men at work; here, there and there; namely Tom, Dick and Harry.'), the failure of any one test might be thought, if not to refute, at least to bring seriously into question the validity of the others.

Well aware that Gertrude thought the study of empty space a sure sign of madness, and an eye bent to that purpose a deformity in Nature, I nevertheless thought it might be of some service to ask: can the first proposition above be tested? Has it the reliable status we expect of theories which have preoccupied so many editorial projectors, commanded so many hours of scholarly time, transmuted so much vacancy

into learned articles?[1] Thinking that one solitary but incontrovertible instance would give some philosophical dignity to disbelief, I found that I could test a comparable method on a book of some substance in which the compositorial stints and press-work were well-documented. But before I report the results, a few words on the spacing of punctuation as a compositorial practice may not be out of place.

The great merit of spaced punctuation as evidence is its purely typographic, if not wholly arbitrary, nature. It is largely independent of copy-influence and indifferent to chronology, of a frequency high enough to imply developed habits, and it appears in texts of several languages, set by compositors of any European nationality. Justification, and a medial or terminal position, may affect frequency, but these are discernible and so discountable influences on the statistics. There are of course various ways in which marks of punctuation may be spaced: the spaces may be placed before or after the mark and may vary in thickness. But such variation merely strengthens the basic assumption that the variety observed indicates the idiosyncratic practices of the men who transferred those palpable, leaden spaces from case to stick.

Those who have recently used such evidence have not, I think, offered any 'rationale' for the spacing of punctuation marks; and indeed that does seem a singularly inappropriate demand of a practice so apparently personal, insubstantive and 'accidental'. And yet experience prompts. 'Should I put a space before a comma?', every new student asks. What did an old hand say to apprentices in those early habit-forming days of their youth? 'Lard out your line, lad; fat formes fill fast'? Or 'Tight-setting, son, sees the job soon done'? Moxon must have assumed some such gnomic instruction at case for he has nothing comparable to Fertel's discussion, 'Des Espaces, & de leur usage', at pp. 16-18 of La Science Pratique de l'Imprimerie (Saint Omer, 1723):

So Fertel says, in effect, that there are rules and he hints that they have a purpose. There is a good practical reason for keeping a line easy and open (apprentices take note), and that in turn implies at least one cause of variation in a pattern—later adjustment to take in corrections.[2]

For a sophisticated formulation of the rules, however, it would be hard to go beyond that of Bertrand-Quinquet whose own book, printed by his own firm, also exemplifies them. His Traité de l'Imprimerie (Paris, An VII = 1799), pp. 126-128, specifies seven marks of punctuation and the 'manière de les placer dans la composition'. I cite only his comments on the comma and full point:

The extension of the principle from one compositor to a whole nation of compositors has its own theoretical interest, for it is, in form, precisely that—an extension of the same fundamental assumption that compositors, the wordiest of men, may paradoxically give themselves away in, to coin a Pascalian phrase, 'le silence infernal de ces espaces finies':

Logically of course, Bertrand-Quinquet's proposition may offer difficulties. For example, distinct national patterns within the same text could be taken to attest division of work between countries as well as between compositors; conversely, it could be taken to evidence a cosmopolitan compositorial companionship in the same shop. Lest either of those extensions of the argument be thought frivolous, one need only recall the invaluable work of R. A. Sayce along similar lines, the complexities of clandestine printing on the Continent, and the high proportion of Continental workmen in the English printing trades in the sixteenth and seventeenth centuries.[3]

We need not, however, rely wholly upon French manuals in search of a 'rationale'. John Smith's The Printer's Grammar (1755) is the first in English to be explicit, suggesting an en-quadrat after a comma, semicolon, etc., and an em-quadrat after a full point. He calls such practice a 'rule' but quickly adds 'it is no law neither' (p. 113) and offers some golden advice on the relation between the spacing of punctuation and the troublesome business of justification: 'putting nothing at all after a Comma, Semicolon, or even after a Full-point, in composing, shews more readily [towards the close of a line] how much more or less may be taken in; and what space may be allowed after a Point or Points in a line' (p. 114). The thought may be fanciful, but that remark rings true and suggests to me a trick of the trade of proven worth and real antiquity. Normally of course a line would be regularly spaced until it was almost full, and then modified to fit. The practice Smith describes would make the spacing more contingent. Writing of word-spacing, he says that 'most Compositors chuse to put a thick Space, called The Composing Space, after a word' but adds that this practice makes the lines less easy to justify: '. . . therefore those who put Spaces as they come up, have a better chance to justify the contents of their lines to equal distances.'[4]

Justification is one concern, correction (as Fertel implies) another. Smith ingeniously suggests a way of making both problems easier: if all punctuation were cast to a uniform thickness, it could be changed without rejustifying the line. It would also save spaces, for punctuation cast centrally on bodies of uniform size would, when printed, appear to be spaced. He goes on:

Smith's mid-century suggestion, we might note, together with Fertel's and Bertrand-Quinquet's explicit directions, indicate that spaces before and after punctuation marks must have been separately set, not cast on the body. This was certainly so in type set in the seventeenth and eighteenth centuries and still standing at the Plantin-Moretus Museum in Antwerp.[5]

Thus documented, the trade's attitudes to the spacing of punctuation are seen to be related to several different practical concerns. Bertrand-Quinquet's authority supports and extends the general theory that compositors may be distinguished from one another by the ways in which they space such marks; both Fertel and Smith offer sound advice (though in different forms) about setting and using space on either side of punctuation marks to justify or correct a line more easily. They do not, however, offer a 'rationale'. Bertrand-Quinquet shows a faith perhaps too touching in the efficacy of education, forgivable though that might be in Year Seven of a new dispensation. Fertel and Smith complicate matters a little by recommending practices which would obscure a presumed pattern. In sum, however, their comments reinforce the idea that any one compositor should and would have a practice of his own, and his companions perhaps different ones, and that in a book where the practices seemed to differ from one another, those compositors might be separated out. The status of our inferences, if they are not to derive only from a simple juxtaposition of the extremities of the evidence,

depends crucially of course upon which of several possible motivations and methods we select to make sense of our analysis, and also upon the extent to which we can define the limits of identity and discount correspondence.

Unfortunately, the theory by definition functions only where the distinctions are in fact extreme, and it collapses into complete inutility at mid-point, where the evidence of distinction declines into spasmodic eccentricity or disappears altogether. Nor is that inutility simply a regrettable inadequacy: it is profoundly subversive of the whole theory. For a theory can hardly be thought sound which predicts one way (division) but not the other (integrity). So, if it anywhere fails to predict (that is, confirm the hunch . . .) that work was or was not divided; or if it fails to accord with the facts as soundly established by other evidence; then, logically at least, even the statistically impressive figures of difference cannot be trusted either. They appear significant only by virtue of the simple assumptions we bring to them, and they appear deceptively attractive because we find it easier to divide than unify.

Such a preamble presupposes a case which proves (i.e. tests) the rule, and disproves (i.e. falsifies) it. The one I offer has involved, in a very literal sense, much soul-searching, for the book is the second, revised edition of Joseph Beaumont's Psyche, a folio in fours printed at the Cambridge University Press in 1701-2. The work is a long one in 24 cantos, set in Pica roman, double column, with eighteen stanzas per page. Semicolons, colons, question and exclamation marks almost invariably have a space before them (see Fertel's comment above on 'le point-virgule'), but the practice with commas, of which the following illustrations are typical examples, varies in statistically significant numbers.

Such features can be readily seen in other books also set in Pica roman and printed at the Press in 1701-2: Thomas Bennet's A Confutation of Popery; the second, reset edition of the same work; and the second, reset edition of his Answer to the Dissenters Pleas. In each, spaced commas appear throughout the text. William Whiston's A Short View of the Chronology of the Old Testament (1702), also set in Pica, has none; but in The Harmony of the Four Evangelists included with it, spaced commas appear in the extensive sections set in Long Primer. In all cases, the compositors and pressmen are reliably known from their personal claims for payment, claims which specify the work units (half-forme, forme, sheet, quire). The function, variety, dating, and range of men involved, inhibit the entertainment of any conspiracy theory, whether to defraud the Press or mislead future scholars, and their evidence is still extant and examinable in the Cambridge University Archives.

Drawing inferences of the kind which mark the use of spaced commas to track compositors in Shakespearian texts (though the assumptions underlying such inferences have not I think been made clear), one would come up with wholly respectable results. For the 224 pages in the quires examined (A-2E), there are, plus or minus perhaps 2 per cent, 13,777 commas, or an average of 61.5 commas per page. A glance at the appended table will show an unmistakable and statistically acceptable distinction between pages in which the ratio of unspaced to spaced commas varies between 120:0 and 64:14 (say 4.6:1), and those in which spaced commas equal or outnumber the others in ratios varying from 27:27 (or 1:1 to 8:60 (or 1:7.5). As the text is in verse, justification is not often a factor; as it is a revised reprint of the 1648 edition which has no such pre-spaced commas, copy-influence can be discounted. The theory, cautiously applied, might then support either of two propositions: (a) that two men were involved, one who scarcely ever set a space before a comma, and another who did so more often than not; (b) that two groups of compositors were involved, one of which could not be further divided by this test. A third proposition, however, will doubtless occur to those who accept the second: (c) that while the common habits of one group disguise their separate identities, the second group might certainly be further sub-divided (within statistically reliable limits) by gradations in the incidence of spaced commas (the statistics tolerating distinctions within the seven-fold difference between 1:1 and 1:7.5). In the latter case, some fine distinctions would undoubtedly remain to be made at border-line points, but the success which has attended the resolution of such problems in the Shakespeare First Folio gives excellent grounds for thinking they could be resolved here too. Some sixty years ago Thomas Satchell settled for A and B; by a simple but exacting process

of sub-division, they have now been splintered into no fewer than nine different men. With little fear that we run counter to current assumptions and scholarly practice, we may however settle for proposition (a) while conceding, with a proper openness to fresh evidence, that the compositor who set spaces before commas might yet turn out to be one of twins or triplets. With that proviso then we may reasonably infer that the same compositor set the following pages, but no others, in this book:

• D1, D3v, D4
• E1, E3v, E4
• F2v-F3, F4
• G2, G3, G4v
• H2v-H3, H4
• I1-I4v, I2v-I3
• K2v-K3, K4v
• L1, L2v-L3
• P1, P1v, P2
• Q1, Q1v
• U3, U4
• X2, X3
• Y2, Y4
• Z1, Z3v
• 2A2

The pattern is itself seductively attractive: three pages in each of nine consecutive formes (with a hiccup in I), then a decline into two pages only in the last few before the compositor takes leave of the entire book with one page only in 2A. It must be conceded, however, that the inclusion of certain pages in the above list is questionable: E4, U3, X2, X3, Z3v, 2A2, for example, have only between 16 and 19 spaced commas each (exact figures are given in the appendix). Nevertheless, as a clear case of a slightly imbalanced composition: presswork ratio it will certainly support a complementary theory of concurrent printing to keep our less productive man occupied. Bennet's three books could be claimed to show the same compositor at work on those texts, and quite independent proof that one of them was indeed in exactly concurrent production sustains the inference.

One other element in our analysis may lead us to complicate the theory. I refer to the concurrent use, through the quires examined, of four skeleton-formes. These are often evidenced to buttress a theory of divided work, on the further theory that, if work was divided, the press-crew would need to save as much time as possible to keep up with the increased rate of setting. Or, were there not such clear evidence of two men, we might posit a larger edition than normal as another explanation of this time-saving practice of using several skeletons. In this case, however, we have independent proof that the edition numbered only 750, a figure which happily reinforces our inference from skeleton-formes that the setting must have been shared. It is rare of course to be able to establish so easily and indubitably the exactly concurrent printing of another book with matching features and a firm figure for the press-run. It would have been too much perhaps to hope for an exact correlation between a particular skeleton-forme and a particular compositor. A recurrent skeleton might then have allowed us to corroborate

the theory that, under certain circumstances, a skeleton might substantiate the man; but the subordinate role of the one who spaced his commas in Psyche (one-and-a-half formes per quire) really—and reasonably —rules that out. At best, we might explore the significance of his setting both pages of the same forme in F2(i), H2(i), I1(o), I2(i), K2(i) and L2(i), but the other 27 pages give strong evidence of shared setting as the norm in each forme when a second compositor was involved.

I leave it to the appended details to collapse this house of cards. Here I need only observe that the pattern of spaced commas has no significant relation to division of work among the compositors. Six different men worked on the book and they rarely shared a quire, let alone a forme. Four of them (Bertram, Crownfield, Délié and Michaëlis: English, Dutch, French-via-Oxford, German?), each set some pages with very high, and other pages with very low, ratios of unspaced to spaced commas. The statistics are impeccable; the assumptions, and therefore the inferences, are nonsense. As a question of logic, it matters not that compositors in 1600 or 1623 did or did not set spaces after commas instead of before them. Nor can this challenge to the theory be respectably met merely by affirming that 1701 is not 1623 or that Cambridge is not London. The case for the efficacy of a modified theory, reliable for London shops in 1600 or 1623, positing different conditions, motivations and methods, must be rigorously argued as a matter of history and meet the appropriate standards of historical scholarship.

The point was made some years ago but it is probably worth stressing once again: there is no relationship of an easily statable kind between compositors, skeleton-formes and press-crews. The formes for Psyche as a whole were printed by six different pressmen in eight different combinations including, on three separate occasions, work at half-press.

For all that, I must strongly and explicitly affirm the importance of such theories—and affirm also the most exhaustive testing of them as an essential complement to their entertainment. Our scholarship is deficient, not in the formulation of interesting hypotheses, but in our failure to test them rigorously, and in the superimposition of one fragile theory upon another, as if two crutches gave us healthy legs.

In that spirit, therefore, let me before I close propose another theory. It is this: notwithstanding the date of Smith's comments on a uniform size of body for punctuation marks, or Fertel's and Bertrand-Quinquet's earlier and later assumptions of compositorial free will, the Cambridge University Press (unusually for 1701-2) might have had a fount of Pica roman which contained a high proportion of punctuation with cast-on spaces. We see it travelling from quire to quire as it was set, distributed, and reset. This theory accounts for its recurrences, its successive use by

four different men in the same book, and its concurrent use in others. Taken no further that theory might well salvage some of the others.

In testing it, however, we have at once to confront the absence of historical evidence for such a practice at such a date. The Plantin exhibits show independent spacing before commas, although the Cambridge type, it is true, came from Amsterdam. A new and therefore perhaps distinctive fount of type did reach Cambridge from London on 28 March 1701 and Clement Knell was paid for papering and filling two pair of Pica cases on 17 May 1701, a date which neatly coincides with the start of work on Psyche. Yet more evidence, however, shows that it would be wrong to conclude that this new fount contained spaced commas. In the first place, they are extremely rare in Psyche before quire D (claimed on 21 July 1701); in the second place, such new types would not have been discarded as early as quire 2A (set by 30 August 1701). Finally and conclusively, such spaced commas had already appeared in two books set in Pica roman and printed the year before.

So let's try another theory: the fount containing spaced commas was an old one and was discarded once the new fount had been brought into effective use. Whiston's A Short Chronology may be invoked to support this line of argument, for there are no spaced commas in the main text of this later book which is set in Pica roman. But since external evidence proves that the fount thought to contain spaced commas was no more than two or three years old and had had little use, we must abandon that explanation too.

Perhaps then only the spaced commas were discarded—although they survive in the Long Primer of Whiston's Harmony set in 1702. If that were so, the new theory of punctuation with cast-on spaces might be saved and a contribution offered to the history of typefounding (Moxon, for example, has no mention of the practice in his section on typecasting). Let's pursue this form of the theory a little further. It assumes (a) that pre-spaced commas were cast as such; (b) that they formed a distinctive element in at least one pair of cases; (c) that their presence in the text proper was in fact anomalous; (d) that they had a different typographic function which we have still to explain; (e) that after quire P they were gradually segregated and removed from the fount until, by quire 2B, it was wholly cleansed of them. We have now to specify a function: the spaced commas, let us say, were so cast specifically for use with small caps but had been dissed into one lower case, either mistakenly or to make good a deficiency. Fertel's discussion of the spacing of caps and small caps is suggestive:

But before we can accept that Fertel has led us at last to an incontrovertible theory, we must in all honesty note that the passage just cited makes no mention at all of punctuation, and that bills of fount treat caps, small caps, lower case, punctuation, and figures as discrete categories. More seriously, we must note further that the spacing of the prespaced commas is not uniform as we should expect it to be if it had in fact been pre-cast. That objection might, of course, be met by compounding a theory of distinctive types with one of compositorial practice: the cast-on space was a fine one and might be variously expanded by the addition of others as need arose in justifying, correcting, or merely as the mood took each of Bertram, Crownfield, Délié or Michaëlis.

But now Whitehead's admonition, so respected by Greg, bursts through: 'seek simplicity—and distrust it'. The theory of cast-on spacing cannot be quite trusted, and any theory of compositorial practice (given the four compositors and the extreme variations within the work of each) returns us to the initial problem—why such distinctive groupings in such a regularly recurrent pattern? Which is also to say that we are indeed back where we started: in a cleft stick. Lest it lead me into self-parody then, let me formally abandon even my own theory of commas with cast-on spaces. It is possibly unhistorical, and in any case it would have been seriously damaging to proposition two: if such distinctive types from the same pair of cases recurred with such well-defined regularity, it would have been reasonable—or at least orthodox—to suppose that they had been set by the same compositor from 'his' cases; and that, demonstrably, was not so.

I was tempted to call this paper 'Amoebic Scholarship: Or, The Counsels of Duessa?' because I believe that there is a question of deeper principle to be faced, and perhaps also a moral issue in how best we should spend our bibliographical time and space. The deeper principle relates of course to our use of division as a function of analysis. Evidence of difference is observable and countable; by contrast, what is common or coherent is thought to be inert and uninformative. The computer, which is becoming indispensable in the service of such analyses, is the child—indeed, the supreme expression—of the binary system. Its virtue lies in the separation of sheep from goats, of chalk from cheese. If it finds it harder to tell the difference between scholarship and pseudo-scholarship, that deficiency we must attribute to our

own inability to teach it.[6] Is it then altogether surprising that the same two 'compositors' should have been divided into three, then five, then six, then nine? Or that a refined eye for linguistic distinctions, and a superior technical facility to store and report them, should now lead to a revival of disintegrationalist theories of divided authorship? Or that observed differences between two substantive texts should, within the parameters of each, be offered as evidence of inviolably distinct structures whose division into two their conflation would destroy? Or, where the structures seem inconveniently interlocked, that the 'text' should be divided yet again by the ingenious invocation of an intermediate (but lost) manuscript to create a third (or fourth, or fifth . . .) structural unit into which one might syphon off inconsistent evidence? The formalities of such analyses, but especially the limited assumptions that underlie them, are quite inadequate to the diversity of the human behaviour which created the 'evidence'. Yet the paucity of reliable historical evidence makes it difficult to improve them. More ominously, however, such procedures threaten to redefine bibliography as an essentially disjunctive tool and to distract us from the greater challenge to discern the unity in human variability. ('Two Distincts, Division none'—Shakespeare's own succinct statement of one of Love's mysteries—is not without a certain admonitory force here.)

Having used that phrase 'human variability' I must add that I should be sorry if it became a license for relaxing the rules of evidence in bibliography, any more than it might in a court of law. The pressure to prove our theories (again, that is, to test them rigorously) by what is historically knowable must in no way slacken. A theory which collapses from book to book, like a chain with broken links, is nowhere sound. In the present instance, the forms of two theories (rarely I confess so nakedly exposed) have been stripped and shown as impotent to explain what actually happened in printing just one book used to test them: spacing as a compositorial practice, and founts as a compositorial trace. Even their skeletons are uninformative. Statistics merely compound the

errors. And, most damning of all, it is perfectly clear as a matter of fact and of logic that, far from one theory assisting the other where either is weak, the two theories in this case are mutually exclusive.

My anatomy of Beaumont's Psyche is now concluded. I assure the reader that 'no levell'd malice Infects one comma in the course I hold'. But since it takes courage to make a fool of oneself by counting 13,777 commas in public, I can only hope that this example may serve as at least a caution to others doing likewise in their pious efforts to discover how many comps can dance on the point of a bodkin.