University of Virginia Library


In the article that motivated me to undertake this study, Paul Needham pre-
dicted what a more complete investigation of the Pavier quarto paper stocks
might discover:

Under an awareness of twinship, the situation obviously changes markedly, though Greg's
argument from paper retains all its strength. Greg recorded 27 "different" marks, mostly
Pots, in four sets of the quartos: thus four copies each of the 81½ edition sheets, or about
310 watermarks (subtracting unwatermarked sheets). Simple probability suggests that
within these parameters various consequences become highly likely. First: some of the
recorded marks will surely pair off as twins; the number of stocks represented by the
drawings must be less than 27. Second: there will very probably be twin marks either that
did not turn up in this relatively small sample, or that Greg overlooked because their dif-


Page 166
ferentia are inconspicuous. Third: there could well be fugitive stocks that do not happen
to show up in these four copies. After all, nine of the watermarks recorded by Greg appear
in his sample in three or fewer examples each.[35]

Needham's predictions have turned out to be remarkably accurate. Indeed
several of Greg's drawings do pair up as twins; I have been able to identify twins
for all marks illustrated by Greg that are not themselves twins of other Greg
drawings; and there are indeed quite a number of "fugitive stocks," watermarks
that Greg had not encountered, including Stevenson's "dated" watermarks. An
additional and offsetting phenomenon that Needham could not have predicted
is that, perhaps paradoxically, Greg's drawings represent, at least in a sense, not
fewer than twenty-seven pairs, but considerably more. This is so because, as I
have noted, paper manufacturers often repeated the "same" design in a series
of mould pairs. (I use the word "same" guardedly, because of course the designs
are never identical [36] and similar pairs can be distinguished by their chainspace
models.) Thus for example Greg's drawing No. 4 turns out to represent four dif-
ferent mould pairs, all with the "same" POT: w/fleur top design.

I will now describe in more detail the several situations I encountered in
comparing Greg's drawings with the results of my own investigation. The most
unproblematic situation is where one of the drawings represents a single pair;
that is the case with numbers 2, 3, 5, 8, 9, 13, 17, 18, 21, 22, 23, 25, 26. As can
be seen in the summary table in appendix 3, some watermark pairs, like 2 and 5,
occur quite frequently and appear in several of the quartos (these were thus
among the marks that were particularly valuable to Greg in exposing the false
dates), while others, like 3, 8, 9, and 13, occur in only one or two gatherings of a
single play. Other categories, however, present considerable complexity.

Ten of Greg's drawings pair up as twins: 1\14; 10\11; 12\19; 15\16; 24\27
(the fourth of these pairs had been identified by Stevenson.) How can I be certain
of these pairings, especially in the case of 1\14, where Greg's I shows no initials
in the pot's belly while his 14 bears the initials LE, and my own drawing of I has
vertical lines in the belly, my 14 with indeterminate initials (see figure 1)? After
the gathering of mugshots and fingerprints, the most important evidence for
pairing twins is their distribution in a printed book or books. In the simplest situ-
ation, if two individual marks of similar design, with similar average chainspaces
and wireline density, appear in a number of consecutive gatherings—arun—it
is highly probable that they came from twin paper moulds. One should not of
course expect them to alternate regularly or even necessarily to appear in a 50/50
ratio. Although the workings of probability dictate that in a long book printed on
a run of a single paper stock each of the twin marks will appear roughly half the
time, it is not at all uncommon in shorter runs for one of the twins to dominate.
I have occasionally found eight or nine examples of one twin before encountering


Page 167
the other. Chance is chancy, though as Tom Stoppard wittily illustrates in the
opening scene of Rosencrantz and Guildenstern Are Dead, no matter how many times
in a row heads has been thrown, the odds on the next throw remain 50/50. On
the other hand, in dealing with books that are printed on remnants, as are most
of the Pavier quartos, where changes in paper stocks are frequent and individual
gatherings may appear on more than one stock, one must examine multiple cop-
ies to begin reliably to pair twins. But again, if two marks of similar design and
measurements regularly appear in a particular gathering or gatherings, the prima
evidence is that they are twins.

In gathering 2A of Pericles, sixteen of twenty-three copies examined have
either the 1 or 14 mark; in gathering H of Merchant of Venice, thirteen of fourteen
copies have one or the other. Despite their variable appearance both from each
other and from different examples of each individual, average chainspace mea-
surements combined with the distribution evidence indicate that 1 and 14 are
a pair. Greg was rightly perplexed by these (and several other) watermarks, as
indicated by his notes to the drawings: "It is not absolutely certain whether Nos. 1
and 14 and 24 and 27 are really distinct or not. On the other hand Nos. 18 and 20
may be capable of being resolved." In 1 and 14 Greg had encountered variant
states of both watermarks, with the wires that had formed whatever letters origi-
nally appeared in the bellies of both pots bending, breaking, and in number 1
finally falling away entirely. On the other hand the basic designs of the two were
relatively stable and similar. Without the knowledge of watermark twinness or a
means for recognizing the same watermark in variant states, the situation was un-
derstandably baffling. Yet Greg's instincts were correct that both 1\14 and 24\27
were somehow connected. And if his use of the word "resolved" with reference to
numbers 18 and 20 means "separated, broken up, analyzed" (OED, sense 7), he
appears to have intuited that these individual marks might actually be two. Greg
seems to have been very close to discovering that watermarks are twins.

Greg's 10\11, both fleurs-de-lis, are uncomplicated, but the other pairs repre-
sented in Greg's drawings—12\19, 15\16, and 24\27—require additional atten-
tion since for each of the three there is in fact a second set of marks with similar
designs. In considering these pairs it is worth returning to Stevenson's statement
that "The RG Shields (Greg 15 & 16) are themselves twins; [and] the RG/D Shields
may be quadruplets."[37] Particularly notable is his characteristically offhand but
provocative suggestion that watermarks might, in some instances, be quadruplets.
Stevenson apparently encountered, as I have, a third pair of SHIELD: RG water-
marks in copies of Sir John Oldcastle. Because of their size, he noted their similarity
to 18, and, perhaps playfully, suggested quadruplets. (Greg states in the notes to
his drawings that "18 is really considerably larger" than 15 and 16.) The new set
of marks is, other than size, actually closer to 15\16, sharing the distinction of
being the only other Pavier quarto mark to be centered on chainlines rather that
between them. I have therefore designated the two as 15\16.1 (oSHIELD: RG.1)
and 15\16.2 (oSHIELD: RG.2). Does the distribution evidence suggest quadru-
plicity? Probably not. The pair 15\16.1 appears in some copies of H5 gatherings
E, F, G, and in most copies of Oldcastle gatherings A and B; 15\16.2 begins to


Page 168
appear in gathering D of Oldcastle and is also found in gatherings E, F, H, and I,
always mixed with 15\16.1, and always in a distinct minority. The distribution
thus suggests not quadruplets but a second set of twins, perhaps odd sheets used
as cording quires, possibly left over from a pair of moulds that had worn out.

Greg's 12\19–BPOT: GG.1 and BPOT: GG.2—not only presents the ques-
tion of quadruplets but also illustrates nicely a problem that I had encountered
before in the Crowley Piers Plowman papers. As can be seen in figure 3, each pair
consists of one POT with a floppy top and wide base, and one with a skinny top
and narrow base. If one were to judge by appearance alone, a reasonable as-
sumption would be that the two floppy tops and two skinny tops are the twins.
But the average chainspace widths show this not to be the case. The distribution
of these two pairs is intriguing (see the sequence table of watermarks in individual
plays in appendix 2). In Henry V gathering A, I encountered seventeen examples
of Pair 1 and only two of Pair 2. But in Lear there is a greater balance; in the
examined copies gathering F had Pair 1 in seven, Pair 2 in thirteen, and in gath-
ering H thirteen of Pair 1 and six of Pair 2. In total I found thirty-one instances
of Pair 1 and twenty of Pair 2, a relatively even balance considering the size of
the sample. It is of course possible that the distribution in H5 gathering A is a
statistical anomaly, and that were the sample great enough that there would be
a relatively even mix of the two pairs in all three gatherings. But this can be
only speculative, because sufficient evidence is not extant. And even if it were,
watermarks are not literally quadruplets. Although in the present example gath-
erings F and H of Lear certainly give that appearance, the measurements indicate
two distinctive pairs of moulds. How then might such mixtures come about?
Although the predominant sources of paper for the English printing trade had
only one vat, there were presumably a few larger establishments that possessed
two or more.[38] Papetieres with more than one vat might on occasion have used
two pairs of moulds with similar designs simultaneously, indiscriminately mixing
the sheets as they were dried, sized, and packed for shipment. It is even possible
that in some mills the product of two vats was turned out onto on a single post,
a procedure that would have distinct advantages in completing the post more
quickly and would result in the appearance of watermark quadruplets. But even
in such cases, average chainspace widths might distinguish the original twins.

Greg's numbers 24 and 27 are the last of the twins represented in his draw-
ings that turn out actually to be two pairs, both shapely two-handled pots sur-
mounted by an arrangement of small balls in a diamond shape, and surmounted
by a crescent. Greg's 24 shows the initials JI in the belly, his 27 bearing the initial
C/LG. In no example I have seen are these initials clear, or even clearly initials,
though there is some sort of design in the belly of all four individuals of these
related pairs. Here distribution is a key to correctly pairing the twins, since, with
the exception of a single example of 24\27.2 in MND, 24\27.1 appears only in
MND, 24\27.2 only in MWW.

The most difficult situation I encountered was that in which a single Greg
drawing was found to represent multiple pairs: 4.1, 4.2, 4.3, 4.4 ("a" twin only);
6.1, 6.2 ("a" twin only); 7.1, 7.2, 7.3 ("a" twin only), 7.4 ("a" twin only); 20.1,


Page 169

FIGURE 3. Watermarks 12\19.1 and 12\19.2; BPOT: GG.1 and BPOT: GG.2.

20.2. As with the other multiple pairs I was able to identify and distinguish simi-
lar watermarks and pair twins on the basis of average chainspace measurements
and distribution. All four pairs of the Greg 4 POTs have similar fleurs-de-lis
chapeaux balanced precariously on top. The most frequently occurring pair, 4.1
(BPOT: fleur top.1; see figure 4), appears regularly in gatherings C, D, and G of
MWW(with a few strays found in MND and MV); the newly discovered 4.2, 4.3,
and 4.4 all appear in either/both MWW gatherings C and D. These seeming wa-
termark octuplets would be nearly impossible to distinguish without chainspace
measurements and their averages:

  • 4.1 average 21.13 and 20.88;
  • 4.2 average 23.27 and 23.27;
  • 4.3 average 19.50 and 19.84;
  • 4.4 average 22.88.


Page 170

FIGURE 4. Watermark pair 4.1; BPOT: fleur top.1.

Fortunately, the first three have such distinct averages that the pairings are clear.
On the basis of average alone, the single example of 4.4 might possibly pair
with one of the 4.2s, but watermarks are not triplets and its twin remains a

The four pairs represented by Greg 7, a crescent-crested pot bearing the
initials C/DV, present a slightly different situation. Pair 7.1 occurs regularly and
predominantly in Pericles gathering X and MVgathering C; the other three pairs,
7.2, 7.3, and 7.4, are more scattered, lurking mostly by twos and three in the
gutters of assorted sheets of MND, MWW, MV, and Lear. In contrast to the Greg
4 POTs, I have found no gathering in which all four pairs are present. But again
distribution (for 7.1) and average chainspace width permit pairing, and while the
average for 7.3 is close to that of 7.2, its chainspace model is distinct:

  • 7.1 average 22.09 and 21.95;
  • 7.2 average 21.57 and 21.65;
  • 7.3 average 21.45;
  • 7.4 average 24.23.

The two remaining cases where a single Greg drawing represents multiple
pairs are, by contrast, fairly straightforward. The only examples of a second pair
of Greg 6 watermarks (BPOT: GL.2), both of the same individual, were found in
gathering F of the Trinity Cambridge and National Library of Scotland copies of
MWW. Greg saw the Trinity copy, but understandably did recognize that it was
different from the examples he had seen in five other Pavier quartos. The chain-
space models of 6.2 are, however, quite distinct from the 6.1 pair. The two pairs
represented by Greg's 20 are distinguished not only by their chainspace models,
but, in contrast to the Greg 4s and 7s, by a mutually exclusive distribution pat-
tern: 20.1 is found only in gatherings I and L of Lear, 20.2 only in gatherings B
and C of H5.


Page 171

Not only do some of Greg's drawings represent multiple marks, but Greg o:
"NO WATER MARK" comprises several identifiable unmarked stocks. Using
Vander Meulen's system of chainspace measurement—which was developed spe-
cifically for identifying paper without watermarks—and the evidence of distribu-
tion, I distinguished four distinct pairs of unwatermarked papers, all appearing
in Lear, H5, or both. Correctly pairing twins in unmarked papers can present a
greater challenge than pairing their watermarked cousins. In marked papers the
default probability is that two marks of similar design that appear either in runs
or in the same gathering or gatherings of multiple copies are twins (though there
may be the complicating factor of multiple pairs of the same design as discussed
above). Unmarked papers lack this obvious visual discriminator, and without
the cue of the watermark, it is easy to measure exemplars of the same unmarked
paper in opposite directions, producing chainspace models that are mirror im-
ages of each other. The only expedient is constantly to be alert to this possibil-
ity, and to compare individual models backwards and forwards when trying to
find matches. When a run of a single unwatermarked stock appears in a series
of gatherings, pairing the twins is perfectly straightforward. But this was not the
situation I found in Lear and H5, though eventually the pairings became clear.
Unmarked stock 0.1 was the most prevalent, and, in the copies examined, the
only one of the unmarked papers to appear in both quartos: in gatherings C, D,
E, and G of H5 and I and L of Lear. Stock 0.2 ("a" twin only), appears in Lear
gatherings C and E; stock 0.3 exclusively in Lear gathering G; and stock 0.4 in H5
gatherings E, F, and G. Only in gatherings E and G of H5 does more than one
of the unmarked stocks appear, so that while proper pairings were a bit murky at
first, distribution and average chainspace evidence brought eventual clarity:

  • 0.1 average 21.25 and 21.32;
  • 0.2 average 23.85;
  • 0.3 average 21.73 and 21.61;
  • 0.4 average 24.23 and 24.31.

Again as predicted by Needham, I discovered a substantial number of water-
marks not encountered by Greg, several of which had been previously identified
by Stevenson. Needham's description of these marks as "fugitives" is particularly
apt, since, of the twenty new marks I rounded up, a dozen were loners repre-
sented by only a single individual. The fugitive kind is by nature elusive, and
fittingly enough, half of them were rousted from their hideout in the Oldcastle,
the last of the Pavier quartos to be printed. A few of the new-found marks were
more gregarious: four examples of No. 29 = BPOT: neck w/H; six of No. 33 =
BHAND: PA (the only HAND marks encountered); and seven of No. 31 = BPOT:
MP. And I eventually found both twins for seven of the new marks. There are a
few fugacious unwatermarked papers as well. These are the loneliest of the lon-
ers. I cannot posit a separate stock for each individual since without additional
distribution evidence there is no way of determining which might be twins. Paper
stocks, prolific as they already are, should not be unnecessarily multiplied.

By far the most interesting of the fugitive watermarks are Stevenson's "Shake-
spearian Dated Watermarks" (see note 9). It was impossible not to feel a certain
frisson on first looking into Huntington's Oldcastle and H5 and recognizing the


Page 172

FIGURE 5. Stevenson's watermark drawings from "Shakespearian Dated Watermarks," p. 161.

little pots that Stevenson had scrutinized on that long-ago July morning. It was,
however, a distinctly chill and rainy March day when playing at bowls was not an
option. And it is with no little sadness that I report Stevenson's supposition that
these marks are dated to have been erroneous. For ease of comparison, I have
reproduced Stevenson's drawings (figure 5) as well as my own (figures 6 and 7),
which I believe to be slightly more accurate (my drawing of the H5 mark [figure 7],
from the felt side, is the reverse of Stevenson's.) I am fairly certain that the mark-
ings in the neck of the Oldcastle POT are not in fact a date. It is unclear that
there is in fact an initial digit 1 distinct from the line that forms the side of the
pot's top. What Stevenson saw as a 6 is not at all clearly so, and could be more
plausibly seen as an 8 or an X; the design may well have originally been XOX.
And it seems highly unlikely that even had Jaggard swept out his stockroom he
would have found a sheet of paper that had lain about for a decade.

The arguments against the H5 example being dated are fairly definitive. The
last figure, if indeed it is a figure, is a 9 not a 7. And Stevenson somewhat exag-
gerates the dot between the putative 6 and 9, which he believed to the stub of a
1 that has broken off; it appears more likely to be a small sewing dot. Stevenson's
argument became even more tenuous once I had discovered the twin of this
mark in gathering E of the Victoria and Albert copy of H5. Here the figure that
Stevenson supposed to be a 9 is actually reversed and there is insufficient space
for a 1 in the middle of the design, though it would in any case be singular (pun
intended) if the same digit had coincidentally broken off in both twins. Any doubt
was dispelled by the discovery of one of the twins from a second set of moulds
bearing a similar design (figure 8). Again there are only three figures, which are


Page 173

NOTE. Though it is the only one of the twins encountered, I call this the "b" twin because
it is centered in the left of the sheet as viewed from the felt side. The likelihood is that if I
found its fugitive twin it would be a mark centered on the right. The same is the case with
the mark illustrated in figure 8.

FIGURE 6. Watermark 46; BPOT:MD?b.


FIGURE 7. Watermark 43.1; BPOT: PD w/spike top.1.


Page 174

FIGURE 8. Watermark 43.1; BPOT: PD w/spike top.2.

probably lines and curlicues rather than numerals. There are no Shakespearian
dated watermarks in the Pavier quartos.


Needham, "Allan H. Stevenson and the Bibliographical Uses of Paper," p. 38.


It is thus risky at best to try to identify a watermark by saying that it is "like" a mark
in the Briquet, Churchill, or Heawood watermark catalogues. Is the mark in question literally
identical to the one pictured, is it a twin of that mark, or from an entirely separate set of moulds
perhaps produced years before or after the mark adduced? The utility of these catalogues for
identification is thus extremely limited.


Stevenson, "Twins," p. 80.


Stevenson, "Twins," p. 59 and p. 59, nn. 9 and 10.