University of Virginia Library


To deal effectively with the profusion of paper stocks found in the Pavier quar-
tos requires a methodology that was unavailable to either Greg or Stevenson. Some
twenty-five years ago, in an article that I believes stands with Stevenson's "Water-
marks are Twins" in its importance for the study of handmade laid paper, David L.
Vander Meulen provided a new method for achieving certainty of recognition.
In "The Identification of Paper without Watermarks," Vander Meulen demon-
strated that the measurement of chainlines could provide a unique identifier for
the product of individual paper moulds.[23] Unlike sewing dots[24] —or watermarks
themselves for that matter—chainlines "are present in every leaf of handmade
paper manufactured before the introduction of wove paper in the later eigh-
teenth century."[25] Since the spacing between chainlines is rarely if ever entirely
regular, chainspace models—produced by the careful measuring, recording, and
ordering of chainspaces to reflect their original arrangement across the length of


Page 157
the sheet—can serve as a fingerprint for the mould. This is so because there is
nearly always somewhere in the sequence of spaces some distinctive pattern of
narrower and wider spaces that can serve as a sort of genetic marker (as can the
highly unusual pattern of absolute or near regularity). Chainlines result from the
impression of the fine sewing wires that tie the laid wires together. Since they are
secured at intervals of approximately 5–10 mm to the supporting wooden ribs,
chainlines are much less likely to shift their position or deteriorate than are wa-
termarks, and hence are much more stable over the life of the mould. The truth
of this can be seen by comparing regular chainlines to tranchefiles—"an extra
wire between the final chain wire and the frame of the mould."[26] Unlike the other
chains, tranchefiles are not attached to a rib, and anyone who has measured
chainspaces extensively will have noticed that tranchefiles wander more from
top to bottom of the sheet, and are subject to much greater variation, than their
more firmly anchored fellows. Chainlines may also, though less conspicuously,
wander from each other as they traverse the sheet. For this reason chainspaces
should consistently be measured across what was originally the center of the
sheet, whatever the format of the book. Folios are measured across the middle
of the leaf, quartos in the gutter, octavos along the top edge, etc. The state of the
chainlines themselves will remain relatively stable, even as the watermark goes
through a number of successive states because of damage or wear. I have found
chainspace models to be particularly valuable in helping to establish the identity
of watermarks in variant states, and thus to avoid the misidentification of water-
marks in different states as different watermarks. While Vander Meulen devel-
oped his method for the identification of unwatermarked papers, the chainspace
model is perhaps even more powerful when combined with a watermark image,
a combination I have dubbed the "mugshot and fingerprint" method, following
the apt suggestion of David L. Gants.[27]

Before I move to a discussion of the Pavier quarto papers, I will explain
in some detail my method for collecting and displaying the data that I use to
describe watermarked papers. And let me first reassure would-be paper detec-
tives that when stalking the wild filigrane they need arm themselves with only
a thin, flexible, plastic ruler with a millimeter scale. Or perhaps two—I find it
handy to have both a 6″ and 12″ caliber ruler available. Frequent reference to
figure 1, which illustrates Greg's marks 1 and 14 (a single pair), will make the
following discussion clearer. First, I believe that it is essential always to view and
reproduce watermarks from the same side of the sheet, regardless of whether
initials, if present, are right reading or not. Consistency helps to avoid confusion.
While arguments are sometimes made for viewing from the mould side on the
analogy of a printed sheet being the mirror image of the forme, my own prefer-
ence is to view from the felt side, which is the equivalent of looking down on the
paper mould.[28] It is thus first necessary to determine the mould and felt side of


Page 158

FIGURE 1. Watermarks 1\14; BPOT: LE (?).

the sheet by examining the leaves with a raking light to find the side that bears
indentations from the chain and laid wires and the wireform of the watermark.
This may on occasion be a simple task, with the chain indentations appearing
almost like furrows from a plow; at other times, especially if the book has been
washed and pressed, it may be considerably more difficult, even unachievable,
though an experienced researcher will grow more sensitive to subtle clues. Once
the felt side has been identified it is necessary to determine the leaves and sides
of the leaves which must be measured to yield the full chainspace model (I call
these the complementary leaves). Most watermarks have a definable right-side-
upness, which helps to establish the correct orientation from which to measure
chainlines.[29] (In order to follow my explanation, it may be helpful at this point to


Page 159
get out a sheet of ruled paper and fold it as a quarto.) In quarto, measurements
can be taken from four possible orientations of the sheet, as viewed from the felt
side and with the watermark right-side up: 1v/2r; 2v/1r; 3v/4r; 4v/3r. Once I've es-
tablished the felt side of a quarto sheet, I turn the book sideways and find the leaf
that has the top half of the watermark. If, for example, the top of a POT appears
in 4r when viewed from the felt side, I know that by measuring from left-to-right
in the gutter first of 3v, and then, having rotated the book 180 degrees, left-to-
right in the gutter of 4r, I can put the two halves together in the sequence 3v/4r
to produce a chainspace model of the whole sheet before folding (less whatever
portion is lost to trimming), as viewed from the felt side and reading from left to
right. Similarly, if the top of a HAND appears in 2v when viewed from the felt
side, I know to measure the sequence 2v/1r. A bit of reflective origami will show
that, oriented from the felt side, if the top of a mark appears in the verso of a
leaf it will be a left-hand mark, if in the recto, a right-hand mark. Measurements
are conveniently done to the nearest half-millimeter, since, as Vander Meulen
notes, chainlines are approximately one millimeter wide and an easy discrimina-
tor when measuring is whether the chain falls on or between millimeter markings
on the ruler.[30]

Whenever possible I gather chainspace data from seven to ten examples of
a given mark and transfer the measurements to an EXCEL spreadsheet that
I use to produce a composite model by averaging the measurements for each
individual space. I include in my descriptions the number of exemplars that
have gone into making up the composite model (note this is not the total number
encountered of a particular mark; these totals will appear in the summary tables
in the appendixes). While a model made up from seven or eight exemplars will
be somewhat more accurate than one with only two or three, even a model built
with only a few exemplars will usually permit certainty of recognition. Once a
composite chainspace model has been built, it is easy to determine how many
ribs the mould had by counting the number of non-tranchefile chainlines. In the
present example the tranchefile at the right end of the "a" twin is excluded, and,
since the measurements on either side of the gap in the middle of the model add
up to nearly the average chainspace width, it is apparent that no chainline has
been lost. Thus the BPOT: LE (?) moulds had eighteen ribs.

Having formed the composite model, I then establish the average width of
all full non-tranchefile chainspaces. The average chainspace width for a pair of
moulds is generally exceptionally close, in the present example, BPOT: LE (?),
a difference of only .08 mm. The reason, upon reflection, is fairly obvious. A
mouldmaker must produce twin moulds of identical size so that so that a single
deckle will fit both. Although the spacing of the supporting wooden ribs will
generally vary slightly in the two moulds—and that is their great evidentiary
value— they will both have the same number of ribs, and thus their average
spaces will usually be quite close. (Average widths for a pair of moulds will tend
to be closer when the composite models from which they are extracted have a


Page 160
larger number of exemplars.) Average chainspace widths are of particular value
in correctly pairing twins when a watermark design, say a POT with the initials
IM on its belly, has been repeated in a series of mould pairs.[31] In displaying the
composite chainspace model I use the following conventions. Vertical strokes
indicate chainlines, and the numbers between them are the chainspace measure-
ments; I indicate the position of the watermark in the sequence by bolding and
italicizing the measurement of the space between its attendant chainlines. If the
mark is centered between chainlines, as are nearly all of the Pavier examples, a
typical sequence will appear like this:[32]
5.5 | 20 | 23.5 | 22.5 | 21.5 | 19.5 | 21 | 21.5 | 21 | 14}
{17.5 | 20 | 22.5 | 21 | 20 | 21 | 20 | 20.5 | 20
If a mark is centered between chainlines, the typical sequence appears like
19 | 18.5 | 18.5 | 20 | 18 | 19 | 19 | 19 | 20.5 | 17}
{18.5 | 18.5 | 20 | 20 | 19.5 | 20 | 19 | 20 | 18 | 17
Braces indicate a gap or gaps in the model. In quarto, the gap in the middle of the
model represents the slight loss occasioned by the opening and trimming of the
top edge during binding. (Very occasionally the combination of copies allows for
a quarto model with no center gap.) Chainspace models for folios will also have
a single gap, in this case representing the part of the sheet that is hidden in the
gutter. Octavo models will have three such gaps, two from gutter loss, one from
opening and trimming the fore-edge. The measurements before and after the gap
are the widest encountered at those positions, and the measurements at either
end of the model are from the largest copy examined; if a deckle edge is pres-
ent in any copy, it can be designated by a percent sign (%).[34] I include one fur-
ther measurement in my watermark descriptions, the wireline density per 3 cm.
Wireline density is a measure of the number of fine laid wires that run perpen-
dicular to the supporting ribs and form the sieve of the mould. I have found
papers with as few as 22 and as many as 40 wirelines per 3 cm, though as with
average chainspace widths, the two individuals in a pair of moulds will exhibit
similar densities.

Once I have measured the chainspaces I make a careful freehand drawing
of the watermark, and in the description I note the copy, gathering, and ori-
entation from which the drawing has been made. If beta-radiographs or digi-
tal photographs are available, so much the better, but a reasonably accurate
drawing will suffice since it is the combination of mugshot and fingerprint that
permits positive identification. When a watermark appears frequently, such as


Page 161
Greg's 23, which is present in every gathering of every copy of The Whole Conten-
I can be selective about the exemplar I use for the drawing, searching for
a less tightly bound copy and for leaves in which the watermark is relatively
clear. With a rarely occurring mark, I simply draw whatever I can see, and hope
to run across a better exemplar in another copy. Since quarto watermarks are
inevitably divided by the gutter and are usually at least partly obscured by the
type, it is possible to produce a more complete drawing by conflating images
from several exemplars that are positioned slightly differently in respect to the
gutter. I find this, however, to be a dangerous and unnecessary expedient, since
it is all too easy to conflate drawings from various states of the watermark, thus
falsifying the evidence. As with the chainspace model, a bit of indeterminacy in
the middle of a drawing is no bar to positive identification with the mugshot and
fingerprint system. When I have created the mugshot and fingerprint for a paper
stock, including a composite chainspace model from at least seven examples of
each twin, I can move from collection to identification. Once the product of a
mould pair can be positively identified, there is little to be gained by collecting
additional chainspace examples, though additional drawings may be useful if a
particular mark is encountered in clearly variant states.

I have developed a system of reference notation for naming watermarks, es-
sentially to serve as a shorthand mnemonic for categorizing the large number of
watermarks that are encountered in books printed on heavily mixed paper stocks.
These references are not intended to serve as formularies, since no formulary
I have encountered can be used by itself to identify a particular watermark in
another setting. That is the job of the full mugshot and fingerprint description.
A sample reference, in this case for Greg's mark 2, would read: BPOT: R/LM.
The initial superscript indicates whether the watermark is centered on (O) or
between (B) chainlines (in the Pavier quartos all but Greg's 15\16 are centered
between chainlines). The second element identifies the basic design of the water-
mark, following which I add particular distinguishing features like the initials
R over LM that appear on the belly of this POT. If the initials are uncertain,
I follow them with a query. Such is the case with Greg's 22, where he sees the
initials as BP but where in my own drawings they appear more likely to be PR.
Sometime there are what appear to be initials, but they are so indistinct as to
be unidentifiable. Here I simply substitute queries for the (presumed) letters. In
other cases, especially when no initials are present, I may add several words of
description; for example one of the new marks I encountered is dubbed "BPOT:
neck w/vertical stripes." As will be shown below, some of Greg's drawings actu-
ally represent more than one pair in circumstances where the manufacturer has
repeated a design in a second (or more) set of moulds. In this situation I mimic
software manufacturers by adding a .1, .2, etc., to the descriptor; for example,
Greg's mark 6 turns out actually to be two pairs bearing a similar design: BPOT:
GL.1 and BPOT: GL.2. If I need to refer specifically to one or the other twin
rather than the pair as a whole, a final superscript "a" or "b" distinguishes the
two twins that make up the watermark pair: BPOT: R/LMa and BPOT: R/LMb.
Since the default for watermark pairs is to have one twin's mark centered in the
right half of the sheet, the other in the left (see note 26 above), in the absence of
a following discriminator, the "a" twin may be assumed to be a right mark, the


Page 162

FIGURE 2. Greg's watermark drawings from "On Certain False Dates in Shakespearian


Page 163


Page 164

TABLE 1: List of watermarks

0?  unidentified unmarked stocks 
0.1  UNMARKED.1 
0.2  UNMARKED.2 
0.3  UNMARKED.3 
0.4  UNMARKED.4 
1\14  BPOT: LE(?) 
4.1  BPOT: fleur top.1 
4.2  BPOT: fleur top.2 
4.3  BPOT: fleur top.3 
4.4  BPOT: fleur top.4 
BPOT: BP crescent top 
6.1  BPOT: GL.1 
6.2  BPOT: GL.2 
7.1  BPOT: C/DV.1 
7.2  BPOT: C/DV.2 
7.3  BPOT: C/DV.3 
7.4  BPOT: C/DV.4 
10\11  BFLEUR 
12\19  12\19.1  BPOT: GG.1 
12\19.2  BPOT: GG.2 
13  BLEAF(?) 
I am uncertain what this design is meant to represent; Briquet has no
similar examples. 
15\16  15\16.1  oSHIELD: RG.1 
15\16.2  oSHIELD: RG.2 
17  BPOT: CL cross top 
20  20.1  BDIAMOND/FLEUR.1 
22  BPOT: BP? 
In my drawing the letters seem rather to be PR. 
23  BPOT: PA 
24\27  24\27.1  BPOT: ?? balls and crescent top.1 
24\27.2  BPOT: ?? balls and crescent top.2 
For both these pairs it is unclear what the initials are. Greg's drawings
of 24 show the initials J1 and for 27 the initials C/LG; both were made
from copies of MND. My own drawings, also from MND, are more am-
biguous. The similar pair 24/27.2 appears, with one exception, only in
MWW; again the initials are unclear and ambiguous. 
Greg's drawing is upside-down; he sees the V as an A. 
29  BPOT: neck with H 
30  BPOT: neck with vertical stripes 
31  BPOT: MP 
32  BPOT: C/DC 
33  BHAND: PA 
34  BPOT: MV 
35  BPOT: LDB 


Page 165
36B   POT: P/AV 
37  BPOT: ?? 
This pair is very tentatively identified from two marks found in the L and
L18(2) Oldcastle, neither of which is wholly visible. The average chain-
spaces are nearly identical. 
38  BPOT: O/DD 
39  BPOT: IM 
40  BPOT: BB(?) 
41  BPOT: flower and crescent top (bottom not visible in only exemplar) 
42  BPOT: CM/L 
43.1  BPOT: PD with spike top.1 
One of the twins is the mark Stevenson believed bore the date 1617 or
43.2  BPOT: PD with spike top.2 
45  BPOT: R/OD 
46  BPOT: MD? 
This is the other of Stevenson's "Shakespearian Dated Watermarks." 

"b" twin a left mark. In cases where both marks are centered in the same half of
the mould I add a [L] or [R] to the twin that does not match this default: i.e. in
a pair with both twins centered to the left 1 add an [L] to the reference for the
"a" twin: BPOT: BLa [L]; in a pair with both twins centered to the right 1 add a
[R] to the reference for the "b" twin: BSHIELD: RGb [R].

For ease of comparison with Greg's watermark drawings, which are repro-
duced in figure 2, and simply to conserve space in the sequence and summary
tables that appear in the appendixes, I have retained and extended Greg's system
of numbering. In table 1 I first give Greg's number, followed by my own refer-
ence. Where two of Greg's drawings turn out to be twins, I indicate the pair with
a backslash between the numbers: 1\14, etc. Where a single Greg drawing or
pair of drawings actually represents multiple pairs with similar designs I add the
.1, .2, etc., to the number(s). Numbers higher than 27 indicate marks not found
by Greg, nearly all of which appear very rarely in the Pavier quartos. The only
significance to the numbers themselves is that they indicate, for the most part,
the sequence in which the watermarks were encountered.


David L Vander Meulen, "The Identification of Paper without Watermarks: The Ex-
ample of Pope's Dunciad," Studies in Bibliography 37 (1984): 58–81.


Stevenson had discovered that for incunabular watermarked papers the unique pat-
tern of sewing dots—the points at which the wireform was fastened to the mould with relatively
coarse sewing wire—persisted even as the watermark deteriorated around them, providing a
ready means for identifying the same mark in variant states. But sewing dots are rarely visible
in later papers when the sewing wire had become finer and eventually the wireforms, rather
than being attached at discrete points, were sewn continuously to their moulds. See The Problem
of the Missale Speciale,
p. 35.


Stevenson, Problem of the Missale Speciale, p. 60.


Stevenson, Problem of the Missale Speciale, p. 64.


"Identifying and Tracking Paper Stocks in Early Modern London," Papers of the Biblio-
graphical Society of America
94.4 (2000): 531–540 (p. 532).


I initially made this determination from an examination of the mixed stocks used to
print Robert Crowley's three editions of Piers Plowman (STC 19906, 19907, 19907 a, all dated
1550). Of the seventy-eight individual watermarks encountered, a majority of those which bore
initials were right reading from the felt side. Examination/description from the felt side is also
preferred in the Watermark Registration Standard of the International Association of Paper
Historians, Section 3.0.17: "It is recommended to collect data with the wire side facing down."


Not all watermark designs have a definable up and down, though POTS, which pre-
dominate in the Pavier quartos, clearly do. The watermark can appear in either half of the
sheet. In watermark pairs the marks are frequently placed in opposing halves of the sheet, but
pairs with both marks to the left or both to the right are not at all uncommon. In the three
Crowley Piers editions, of thirty-nine watermark pairs, thirty-four had one twin with the mark
to the right (R), the other with the mark to the left (L); two pairs were R/R; three pairs were
L/L. In the Pavier quartos, of the thirty-six pairs where I have identified both twins, twenty-
three were R/L; six were L/L; seven were R/R.


Vander Meulen, "Identification," pp. 60–61.


While the two individuals in a pair of moulds will have nearly identical average spaces,
the spacing of ribs, and thus of chainlines, can vary widely from mould pair to mould pair. I
have found pairs with average spaces as narrow as 18 mm, and as wide as 35 mm.


Note that in this model we can determine from the widths of the partial spaces in the
middle of the model that a chainspace has been lost; like the BPOT: LE (?) pair this mould had
eighteen ribs.


And the principle can be extended for larger watermarks, which may spread them-
selves over a number of chains.


Of the more than 1600 Pavier quarto sheets examined, I have seen deckle edges on
only six occasions.