Time-Frames, Case Books, and the Value of Paper as Evidence

Perhaps the slight nuance of contempt that the student of paper is aware of arises partly from a recognition that some collectors and publishers of watermarks seem not to have known clearly why they went through these motions. As E.J. Labarre has sometimes admonished me, only Briquet was a professional.

Allan Stevenson, ‘Paper as Bibliographical Evidence’, The Library, s. V, vol. 17 (1962)

L’informazione immediatamente accessibile di una filigrana è perciò, per nostra sfortuna, assai inferiore a quella di quella che ci offre l’etichetta di una bottiglia di vino, sulla quale compaiono, oltre al nome e all’indirizzo del produttore, l’anno di vendemmia e il numero di identificazione della bottiglia stessa.

Ezio Ornato [et al.], La carta occidentale nel tardo medioevo (2001), I, p. 109.

We have learnt a lot (if you are still there, that is). Has it been worth it?

Well, we have discovered above all that a sheet of paper is a moulded object and that signs are placed on the mould which allow us to identify it as an individual or as one of a pair of individuals. The matter is complicated by the existence of twin moulds; but twins taken together have a greater individuality as a couple than a single mould taken on its own. So the fact that we are dealing with twin moulds paradoxically improves our chances of making a positive identification.

In every assessment and description of the material object, whether of the sheet in its entirety or just of the watermark, it is important to remember, as with printing type, that what we describe is not the original object, but the impression it left on the surface of quite another object. Paper shrinks on drying, so that the original was larger than the trace it has left; depending on the make-up of the fibres and the thickness of the sheet, paper does not shrink uniformly, so that sheets from the same mould may differ somewhat in dimension and shape.

If we can recognise the idiosyncratic features of each pair of moulds in sheets of paper made thereon, and tie them into other sheets of paper made on these same moulds, we have acquired a powerful, but not necessarily easy to use, instrument for bibliographical research.

But don’t get carried away.

The Lifetime of a Pair of Moulds was … ?

How long did a mould last?

And how sure can we be in making a watertight identification of a pair of moulds?

Briquet, who as a young man worked in a papermaking factory, puts the duration of a set of moulds as being not more than a couple of years (“sa durée moyenne ne dépasse pas deux ans”). Gaskell cites evidence from the archives of the Whatman mill, which bought “an average of 10 new pairs of moulds a year for the six vats of Turkey and Loose Mills in 1780-7, giving an average life per pair of moulds of just over seven months” (New Introduction to Bibliography, cit., p. 63). Of course, put in this way, the statement is ingenuous, and even misleading, since the introduction of new moulds does not mean that older ones were immediately discarded. (To get a perspective, try asking a serious Marathon runner how many pairs of shoes they get through in a year? There is no simple answer. In the first place how many marathons are they running, and in the second they might have new shoes that are being broken in, an in-use pair kept for the races and more testing runs, and an older pair used for short training runs, and perhaps yet another reserve pair in the house at the seaside, so as not to have to remember to take one’s regular shoes, and so on. So, although a new pair may be bought every year, three or four pairs might be on the go at the same time, while a particular, lesser-used pair might last for years.)

The assessment of between one and two years seems reasonable for a pair of moulds undergoing a daily wear-and-tear at the vats. This indication therefore can be taken as true, where Renaissance paper is concerned, for Chancery-size (i.e. what the Bologna stone calls reçute) moulds; if however the moulds are making larger sizes, such as Median and Royal, for which there was less request, they were used less often and therefore they lasted much longer, maybe even generations. How many sheets of paper might a pair of moulds make in this time? Again there are variables: mould construction techniques, in which the wires became finer and more closely set, undoubtedly improved over time, and much probably depended on the skill of the individual mould maker. Nevertheless, a rough calculation can be made along the following lines: if a pair of moulds is used constantly at the vat for two years, three hundred days each year, with a ten hour day, producing 150 sheets an hour, the total is 900,000 sheets of paper, or 1,800 reams. So the individual mould is responsible for 450,000 sheets and 900 reams. How much of that paper has survived in today’s books and archive documents? Hardly any at all.

A stock of paper, after it was made, had to go through a series of further processes and would only go on sale several months after its fabrication. At this point it was subject to a variety of practices, especially in printing shops, so the analysis never rests on paper alone. It always has to take account of all available data.

In the employment of paper as bibliographical evidence, especially in the study of printed editions, three basic situations nevertheless occur with a certain frequency and it is worth describing them in order to chart the relevant research procedures.

In the first scenario we are seeking a pair of watermarks, or several pairs of watermarks, in order to establish the relationship between them and the watermarks found in other books. Most often the purpose is to give a date to an undated printed book, or a book in which the date furnished by the printer is open to challenge; sometimes it is also a question of discovering where it was printed.

The principle was set out with exemplary clarity and simplicity by the Italian scholar Roberto Ridolfi in 1957 [10]: if we have a dated book from a known press and an undated book from the same press and we find the same watermark (or, better, pair of watermarks) in both, the same date can be assigned to both imprints. If we take account of the time-frame represented by the life of the watermarks in the papermill, the period is a couple of years; if, however, we consider the fact that paper was bought and used by a printing shop in terms of its immediate needs, the two books are probably printed at much the same time. The whole hypothesis of course requires much caution. Without venturing into the labyrinth of “printers of the mind”, early shops often had several books going at the same time with significant overlaps in production; the same shops were also using considerable quantities of paper for ephemera and other materials that have not survived. The best known example of this principle, which furthermore identifies different states in the lives of several different pairs of watermarks, is Stevenson’s superlative 1967 study of the Constance Missale Speciale, believed on the basis of its typography to be possibly anterior to the Gutenberg Bible, but which the paper evidence unequivocally assigns to c. 1473 [30]. Another is the demonstration, begun by W.W. Greg, continued by Stevenson, and more recently rounded off by Carter Hailey, that a group of Shakespearian quartos, variously dated 1600, 1608 and 1619, printed by William Jaggard for Thomas Pavier, and thus known as the “Pavier Quartos”, were all printed together in 1619 [30].

In the second scenario it is not necessary to spot individual watermarks or pairs of watermarks. It applies in a more general fashion the principle that printing shops expended something between 50% and 70% of their budget for a book on the purchase of the paper and therefore tended to buy only enough for the work in hand. Papermills were generally placed out of town, in hilly areas of difficult access, so printers bought from a middle-man, the paper merchant, who obtained his supplies from several different sources. In the press output therefore it is common to find different lots stratified by their watermarks, because each time a supply of paper was obtained the source changed. Since mould sizes, as we have seen, were more or less standard, there was no problem in matching up supplies of paper from different sources and in any case, as the printer knew perfectly well, any small discrepancies would be evened out by the binder’s plough.

Every analysis and description of the distribution of the paper structure inside a printed edition must be done with as many copies as the bibliographer can get their hands on and therefore takes time and plenty of shoe leather. It is not to be undertaken by the hasty, the imprudent, and the impatient. A single sheet, showing a different watermark, in one copy may be a genuine anomaly revealing some change in the presswork, but – as Stevenson warns on many an occasion [29] – it might just be a remnant from an earlier job slipped into the new run. Only the extensive comparison of multiple copies can establish the true pattern.

The third scenario involves the recognition of cancellans sheets or leaves printed to make a correction or a substitution in the text of a book at a later stage. Quite often a different paper supply is employed, though the evidence always has to be assessed with care. A cancellans may however take the form, especially in middling and small formats, of a half or a quarter sheet, which leaves the rest of the unit intact. In this prospect paper evidence comes into its own with a vengeance, though collecting it and interpreting it may not be a straight-forward matter. Again the information garnered from the first copy to happen to hand will not give a complete picture; only the examination of a large number of copies allows the bibliographer to build up a full portrait, which means travelling to look at them. A cancellans of this kind, involving any unit less than a full sheet, constitutes a physical disturbance in the structure of the same. On the basis of this fact, discovering its presence is a straightforward matter of mathematics and probabilities. If the substitution involves a half-sheet of paper, there is one likelihood in two, even if the supply of paper is the same, that the watermark will either be duplicated or vanish, i.e., assuming that the sample of copies reflects the distribution of the original, 50% of copies will appear normal, 25% will present two watermarks in the said gathering, and 25% will be without a watermark (or present two countermarks, or some more intricate equation). It is also worth, where one suspects that a disturbance has taken place, attempting a more complex level of analysis and identifying, with the help of a raking light, the distinction between the mould and felt-side of the sheet [14]. Again a partial substitution of a sheet has one probability in two of disturbing the original relationship. (Given that we are here explaining bibliographical wizardry, the third trick, which does not directly involve paper, in recognising a physical anomaly is to spot the difference – again it is a good idea to employ a raking light – between the “first” and “second” formes in the printing, wherein the latter pushed the first indentation back the other way. Again a substitution has a 50% probability of being different from the original. The method does however require a certain amount of bibliographical savoir faire and it is better to attempt it on a copy in pristine condition, which has not been excessively pressed and hammered in rebinding.)

The “Runs and Remnants” Principle

Our understanding and application of dating on the basis of finding the same watermark in two different books from the same press has, however, to be tempered by Allan Stevenson’s brilliant definition of the principle of “runs and remnants” [29].

For reasons of cost, early printers tended to calculate the amount of paper required for any job very precisely and to order an exact quantity from their supplier. Most of this paper is used up in the printing of the same and constitutes the “run”. Often however there are residues, perhaps of slightly damaged or imperfect sheets, perhaps recovered from the “cassie” quires, which are set aside for proofs or other usages, such as cancellantia or sheets reprinted to make up a short-fall in the original press-run. Sometimes these random sheets or “remnants” will drift into later books and thus appear as isolated witnesses, perhaps even years after acquisition of the original supply.

Stevenson expounded the principle several times in different writings. Here is how he memorably put it in 1962: “These observations lead me to point out a pair of fallacies concerning longevity and time-lag, the first of which Briquet himself falls into, as many have fallen since. We note that most of his tracings were gathered from manuscript sources, often from individual letters or short documents. Now, it is obvious that some scribes or accountants or notaries might use a supply of paper over a period of years, especially if the paper were of large size or high quality. Ordinary small paper would be used up faster. But a printed book differs in its manner of using paper. When a printer secures a supply of paper (from publisher, author, merchant, or mill) for the printing of a book, and then prints an edition of (say) 250 or 1,250 copies, he tends to use that supply up before going on to the next. The consequence is that we often find runs of a single paper for many gatherings; or perhaps a rhythm of two papers when two presses have been used. And then the printer proceeds to runs of other papers, usually papers of similar quality. After study of many such runs, in the light of warehouse or printing records such as those of William Bowyer, I suggest that it can be inferred that the paper was usually bought expressly for the printing of the book and probably not long after manufacture. In early days paper was such a costly commodity in terms of the economy that it cannot have made sense to buy bulky bales of it years ahead. As a corollary it needs to be pointed out that the random occurrence of a particular watermark in a book may represent some remnant of stock used for an earlier book and possibly overlooked in the printer’s warehouse for a matter of years. Here is a reason why a Pot watermark dated 1598 can turn up in the second quarto of Hamlet dated 1604 or 1605: once in the Huntington copy, twice in the British Museum copy. Running paper is highly relevant to the dating of a book; random paper is unreliable for the purpose. This I call the Principle of Runs and Remnants. Scholars should no longer assume that printed books use paper in the manner of pieces of manuscript or that an intrusive paper is as significant as a main stock”.

Just for the Record: Some Case Studies

These are all useful little tricks. How do they work in practice?

Every manuscript, or printed edition, on paper from identifiable moulds is a law unto itself, so there are no fixed rules, only previous experiences.

Learning about paper as bibliographical evidence took myself a long time, and the process is far from over, but on many occasions it has proved to be the rabbit that pops out of the conjuror’s hat. Here therefore is a little summary of items from my own personal casebook. This display might well seem, and certainly is, narcissism to the nth degree; on the other hand none of these books and articles declare in their titles that they have anything to do with paper evidence (to be honest, they use lots of other kinds of evidence as well), and so they have been wholly ignored by repertories such as the IBP [0]. (Such blindness even in state-of-the-art bibliographical resources is anything but rare: I recently noted that extraordinarily high-powered databases such as Medline, and its sister resource in nursing sciences, both ignore an excellent bibliographical article by Victor Skretkowicz on the early printing history of Florence Nightingale’s Notes on Nursing, first published in 1859 and kept in standing type for well over twenty years. Why? No proper reason, or simply because the journal concerned, The Library, in 1993 is not among those regularly considered by their sources. Notes on Nursing, however, even a century and a half after its first publication, is still a cornerstone of the profession, and an enjoyable read, if you ever pick up a copy, while ‘History of Nursing’ is part of the MESH Tree String [K01.400.608]. In bibliography parameters are never a simple issue, but outreach, or the ability to go off the beaten track and find minor material, rather than the standard well-known monograph is the most precious part of the discipline. End of sermon.) The other reason for these potted summaries is that all these items were written and published in Italian, which I do not think it reasonable to presume all the readers of the present text to have mastered.

An early instance, in which the paper analysis provides part of the answer to a bibliographical mystery, is the 1472 Venetian edition of Boccaccio’s Filocolo. This substantial folio has a curious and at first sight puzzling feature: the very last page, which should be blank, has a text printed upside-down from an earlier part of the book. Finding a correct explanation of the phenomenon obviously requires a study of the text, but paper-analysis is the key to showing that the book was printed on a one-pull press, i.e. a primitive version of the process only able to execute the equivalent of one folio page at a time [28]. During the composition of the text of a gathering in the middle of the book, a mistake was made: half-way through the setting of the page, the compositor turned a leaf too far forward in his manuscript copy-text without noticing and thus jumped three pages forward instead of one. Since only the first of the sheet’s four pages had been printed, when the mistake was noticed, the ruined paper was set on one side to be used for proofing and other operations. Towards the end of the edition, however, the printer found himself short of paper (and probably of money); rather than spend for a new supply, he had the crafty idea of recovering the waste sheets by turning them upside-down, so that the first page became the last (a bit like our modern habit of recycling photocopies for drafts and such like in the home printer). As well as the spoiled sheets, in at least one copy an uncorrected proof sheet, containing the page from a different part of the Filocolo, was recycled in this strange manner. The edition employed six different paper supplies, stratified in the surviving copies, with the recycled sheets clearly recognizable as “remnants”, while the bibliographical analysis also reveals that copy was divided in half, with one compositor starting at the beginning of the book and the other starting in the middle, with the printing being done simultaneously on different presses.

The first book printed in the shadow of the towers of San Gimignano (very metaphorically speaking), the folio 1510 De Cardinalatu by Paolo Cortesi, takes the Oscar for the most nightmarish collational formula ever written [30]. Just for reference purposes, it has recently been expressed as: π²  a¹⁰  A-E⁸  2F⁶  F-H⁸  I²⁴ [i.e. I⁸(I5+<I6.7.8.9.10.11.12.13>¹⁶)]  K-L⁸  M⁸(-M1; -M2.7, +<M1>.7)  N¹⁴ [i.e. N⁶(N3+<N3.4.5.6>⁸)]  2N⁸  <O3.4>⁴  p²  P⁸  Q⁸(-Q6,7,8)  <Q6.7.8.9.10>¹⁰  q-r⁶  [s]1,2 (=Q7,8)  R¹⁸ [i.e. R8(R3+<R4.5.6.7.8.>¹⁰)]  S-X⁸. Which, unless you are an experienced analytical bibliographer, is going to ruin your day. Establishing the precise nature of the book’s structure depended a great deal on the paper evidence acquired in the examination of over a dozen copies. The most valuable contribution paperwise came through defining the relationship between the original gathering Q and the two unsigned leaves that followed on from gathering r, which are denominated [s] in the collational formula. By matching up the position of the watermarks and the relation of mould/felt sides in the various copies, it was demonstrated beyond all reasonable doubt that [s]1 and [s]2 were originally printed as Q7 and Q8. A very large insert, consisting in the elimination of the original leaf Q6 and its substitution with the ten leaves of <Q6.7.8.9.10>, followed by the twelve leaves of gatherings q-r, shifted the original Q7 and Q8 into quite a different place in the book. Although this fact could have been suspected on the basis of the typography, proof (that marvellous word!) came through the paper evidence.

In other circumstances it is necessary to extend research on paper to all the editions printed by a particular press in a certain period, hoping always that its output will not be too large! Again the principle involved rests on the assumption that a printer will exhaust one supply of paper before buying in a new one. One instance in which the paper evidence provides an important part of the answer to a bibliographical conundrum involves the first edition of Francesco Berni’s rifacimento, or rewritten version, of Boiardo’s Orlando Innamorato [30]. The princeps of the work, to all intents and purposes a new poem, is known today in a double issue: one from the Giunta family in Venice dated October 1541 and the other from Andrea Calvo in Milan dated 1542 (with the date 1 January in the dedication). The difference between the issues lies in the fact that the preliminaries, the first gathering of the text and the last gathering of the same, including the colophon, are in two different settings; otherwise the body of the book is all from the same setting of type. Internal evidence shows that the main part of the book was printed in Milan and that the Venetian issue was necessarily predated. Paper analysis, especially the systematic comparison with all the other editions printed by the Calvo shop in the period 1539-42, as well as modifications to the italic type used in the Milan shop, show that the edition had really been executed a couple of years earlier, at the end of 1539 and beginning of 1540, by Francesco Calvo, who, due to illness, later handed the running of the firm over to his brother Andrea. The delay in publication and the reissue of the edition in this strange guise was the result of the declared enmity of Pietro Aretino towards his enemy Berni.

Another nice little instance of Renaissance publishing jiggery-pokery occurred in Venice in 1551, when Gabriele Giolito reissued a number of copies of the Italian and Latin poems of Agostino Beaziano, originally published by Bartolomeo Zanetti in 1538 [30]. The edition was an octavo: in order to substitute both the title-page and the colophon Giolito ran off two half-sheets, i.e. A1.4.5.8 and N1.4.5.8. The paper analysis of some ten copies, assisted by the circumstance that the sheets included both a watermark and a cornermark, duly showed that the two halves were printed together as part of the same full sheet and subsequently divided. The accumulation of the data even made it possible to show the imposition of the two formes involved. Although perhaps not terribly important from a bibliographical viewpoint, in this instance the paper evidence gave a result that could not have been reached by any other means.

Something very similar occurred in Venice in 1600 for the publication of the Vaticinia by Girolamo Giovannini [30]. In quarto format, it was decided to insert a half-sheet cancellans in the first gathering and to add an analogous two-leaf index at the centre of the second gathering. The two half-sheets were printed together: quite apart from the fact that, due to a mistake in signing, the binders often placed the whole sheet at the centre of the first gathering, watermark evidence incontrovertibly shows that in any one copy the two halves came from the same original sheet.

An even more cogent example of the utility of paper evidence, in the absence of any other sort of typographical proof, comes from the first edition of Alessandro Manzoni’s Promessi sposi, published in three volumes, in octavo format, in Milan in 1825-26, though actually put on sale only in June 1827 [30]. It just happens to be the most important novel in the history of Italian fiction – an equivalent of all Jane Austen and all Dickens rolled into one – and so has attracted a corresponding amount of critical and bibliographical attention. In 1970 an example was noticed of the survival of a two-leaf – i.e. quarter-sheet – cancellandum in a single copy in the Brera library in Milan. Between 2003 and 2006 the Centro Nazionale di Studi Manzoniani organised a large-scale collation, in which 68 copies were compared on a McLeod collator. The research – conducted with exemplary patience by Emanuela Sartorelli – did not discover a significant number of press-variants (but certainty on this point is always a positive result); however, as well as confirming that the previously-identified cancellandum was the only extant example in gathering 9 of the first volume, it uncovered another instance of a cancellandum in gathering 10 of the same volume in a copy in the Ambrosiana library. The problem evidently lay in the efficiency with which the quarter-leaf substitutions had been introduced by the printer and subsequently by the binder (bibliographers much prefer the binder in particular to be inefficient or, better still, lazy, so that examples of the cancellandum survive in a certain number of copies). It was a plausible hypothesis therefore that in the sample there were cancellantia in other gatherings, which had been introduced in all 68 copies and thus were invisible, as far as mere textual comparison was concerned. Paper-evidence is more truthful on the other hand. The watermark of the paper of the edition comprises an eagle set over the letters GFA, which stand for Giovanni di Faustino Andreoli, founder of the Toscolano mill that supplied the paper. The eagle and the said letters are placed in mirror writing in a corner of the mould – the following table indicates whether the G or the A is the outermost letter – in other words the position occupied by a Renaissance countermark, while the ubication of the watermarks in the gatherings, almost always in the first four leaves, showed that the imposition of the typographical forme was of the type denominated “common octavo” by Gaskell. Now the law of averages suggested that the substitution of two conjugate leaves in order to introduce the cancellans had a one in four chance of generating an anomaly, i.e. either the presence of two watermarks in the same gathering or their absence. So checks began. For reasons of space, the full table of the watermark distribution in the three volumes of the edition and in the nearly thirty copies verified is not given here. Instead, just as a simple demonstration, the table shows what emerged in the first volume in the first four copies looked at in libraries in Bologna and Florence, which immediately confirmed the workability of the hypothesis (the second column nevertheless indicates all the cancellantia that were eventually discovered).

 

gathering	cancellans	Copy 1	Copy 2	Copy 3	Copy 4
1	3.6 4.5	2A	2A	-	-
2		2G	1G	1G	4G
3		4A	3A	4A	1G
4		1A	1G	1G	2G
5	3.6	4A	2G	4G	4G
6		2G	1G	3G	1G
7		2G	2G	4G	4A
8		2G	3G	3G	4A
9	1.8	1G+2G	2A	2G	1G
10	1.8	4A	3G	-	2A
11		1G	2A	3G	4A
12	3.6	1G+3A	-	-	1G
13		3G	4G	2G	4G
14		3G	4G	1G	4G
15		3A	4G	3G	3A
16	3.6	2A	1G	2G	-
17		2G	4G	4G	4G
18		4G	2G	3G	4G
19		1G	1G	1G	3G
20		2A	3A	2A	1G
21		2A	4G	1G	2G
22		-	2A	1A	1G

Figure 1. Distribution of the corner-placed watermarks four copies of volume 1 of Alessandro Manzoni, I promessi sposi, Milan, Ferrario, 1825-26.

What does this signify? Well, although we already knew about the cancellantia in gatherings 9 and 10 from the survival of the cancellanda, it was nice to obtain a confirmation from the paper evidence. The sample unquestionably showed that a previously unsuspected cancellans had been introduced in gathering 12, since two watermarks appeared in the gathering in Copy 1, while the absence of watermarks in two copies in gathering 1 and in one copy in gathering 16 pointed to something going on. A more extensive check, comparing a further 23 copies, raised the total of identified cancellantia in this volume to seven (eight were subsequently discovered in the second volume and none in the third). This same evidence made it clear that the first gathering contained not one, but two separate cancellantia (three copies emerged with three watermarks in this gathering), and identified a further example in gathering 5, unrevealed in the first assay. But a word of warning, there is a rogue card in the pack. The unwatermarked sheet in gathering 22 in Copy 1 is perfectly whole and genuine, i.e. sheets of paper without a watermark can creep in and upset any system that tries to be too rigid about the way it evaluates the data. It should be emphasised yet again, therefore, how much good paper evidence in analytical bibliography depends on the study of multiple copies of the same book, with all the trekking around libraries that that involves. The other intriguing information, thrown up by the analysis of the paper, was the existence of a proof-copy, printed on a different paper-stock, presumably to allow the printer to follow the progress of the edition and eventually intended to be thrown away at the end of the operation; instead it was taken over by the author and used as a canvas for the definitive 1840 edition.

In all these instances the watermark and paper evidence gave a result that might have been suspected, but otherwise could not have been obtained.