University of Virginia Library

WATERMARKS AND THE DETERMINATION OF FORMAT
IN BRITISH PAPER, 1794-CIRCA 1830

MY concern here is with British paper produced in what may be described
as a `transitional' period, extending from 1794 to about 1830. That
concern has been stimulated by the recent article by G. Thomas Tanselle,
`The concept of format',[1] and it is an interest in the relationship between the
characteristics of various papers and the format of volumes in which they
appear that underlies my observations.

For the bibliographer there are two major developments in the production
of paper in Britain in the period under consideration which are of immediate
relevance. One is the change, resulting from legislation which came
into force in 1794, in the location of watermarks within the traditional hand-held
mould. The other is the transition from hand-made paper (a process by
which paper is produced, one sheet at a time, from hand-held moulds dipped
into vats of stuff in suspension in tepid water) to machine-made (a process by
which paper is produced by machine in a continuous web, to be cut into
sheets of the required size at some subsequent stage).

I

By the beginning of the 1790s much British paper (still all hand-made)
was being produced in wove moulds, which had been gradually supplanting
laid moulds from the late 1750s. Wove paper—made, it is assumed, by James
Whatman snr.—appears to have been used for the first time in a British publication,
John Baskerville's edition of Virgil's Bucolica, Georgica et Æneis,
published in 1757 though the paper itself was presumably available as early
as 1754.[2] The use of wove paper had become the norm by the end of the
century, even if the demand for laid paper was never to be eclipsed entirely.
Not only were moulds increasingly wove; there was a tendency towards the
use of moulds without watermarks. At all periods a certain proportion of

hand-made paper—both laid and wove, and not always paper of the poorest
quality—was produced without watermarks; for volumes printed on wove
paper without watermarks it may therefore be impossible to determine their
format, given the lack also of chain lines and especially in the likely absence
of such alternative determinants as deckle/cut edges, point holes, press figures
or sheet numbers.

The dilemma for the modern bibliographer is, however, largely resolved
by the paper-makers' adherence to the provisions of `An Act for repealing
the duties on paper, pasteboard, millboard, scaleboard, and glazed paper;
and for granting other duties in lieu thereof, (34 George III, c.20), which
came into effect 5 April 1794. Among other things, it provided that British-made
paper used for writing, drawing and printing (`first class' paper) was
to be taxed at the rate of 2½d. per lb. For present purposes the key provision
of the Act was that printed books—whether bound or unbound—were, on
export, eligible for a drawback (a refund of part of the duty already paid)
amounting to 2d. per lb., with the proviso that `any such printed books . . .
shall have visible in the substance thereof a mark commonly called a Water
Mark, of a date of the present year of our Lord in the following figures 1794,
or in a like manner of some subsequent year of our Lord.'[3]

The `or' in this provision of the Act of 1794 is ambiguous, with the result,
it is sometimes maintained, either that some paper-makers misunderstood
the requirement to use the current year or that it allowed them to continue
using moulds dated `1794', thus saving themselves the bother of changing
the date in their moulds at the beginning of the new year.[4] Indeed, Thomas
Balston notes of the Whatman mills, the largest and most highly regarded
paper-making concern in the country, that

In addition, D. C. Coleman observes[6] that at most times wholesale stationers
were willing to buy virtually all the paper that the paper-makers could
produce, in the knowledge that, despite fluctuations in demand, they would
soon be able to sell it in a market which was continuing to expand—in other

words it is possible that a batch of paper could have remained in the stationer's
warehouse for an indeterminate time.[7] Hence (a) the date appearing
in a sheet of paper (whether `1794' or some other year) cannot be taken as
unequivocal evidence of the date of manufacture of the paper containing it,
and (b) paper supplied for a particular job may have been made (and accurately
dated) some years earlier than the date of printing. Nonetheless, despite
the qualifications just adverted to, observation of the typical intervals
between dates in title-page imprints and those appearing in the paper used
by James Ballantyne in Edinburgh in printing the works of Sir Walter Scott
suggests that by and large the paper was used within a year or two of its
manufacture. It might also be noted that there was a `mechanical' reason
why moulds were likely to incorporate a current date: a mould in regular use
might need replacing in as little as six months, by one presumably bearing
the current date.[8] Broadly speaking, therefore, it can be asserted that the date
found in British paper used for the printing of books in the years from 1794
is, after all, likely to be not far removed from the year of manufacture. The
Act of 1794 was repealed in 1811 (51 George III, c.95),[9] meaning that there
was henceforth no requirement that a date be included, but individual paper-makers
continued to produce dated paper for some years beyond 1811, possibly
until the time came to replace their vats and hand-held moulds by
machines.[10] For the purpose of claimng the drawback paper-makers placed
the date close to the longer edge(s) of the mould, so that it would be visible
close to the edge(s) of the resulting sheet and, ideally, visible in a margin of
a printed leaf, though, truth to tell, the watermark has often enough to be
read through the letterpress, particularly in volumes in octavo and smaller.

What follows is a series of illustrations of the disposition of dates (and
other watermarks) in paper made in wove moulds between 1794 and the
1820s.[11] All examples are identified by details of the publication and a note
of the gathering/sheet in which the particular watermark is located,[12] with,
where appropriate, brief comments on the volumes in which the particular
examples have been found and more general observations suggested by the
particular examples (the location of the copy from which the example is
drawn is recorded in a footnote).

First there are examples of the date being attached to a laid mould which
also contains a watermark and countermark in the conventional positions,
resulting in a sheet of paper such as that illustrated in figure 1,[13] which is
found in: "A Lady" [Mary Julia Young], A sketch of modern France. In a
series of letters to a lady of fashion. Written in the years 1796 and 1797, during
a tour through France, ed. C. L. Moody (London: Printed for T. Cadell jun.
and W. Davies, 1798), 8°: a4 B-2K8 2L4 χ1, gathering F. This volume is a little
unusual in that all the paper, whatever the watermark, is laid. Here the date
`1794' is likely to be `genuine', in the sense of having been attached to the
mould in that year, rather than the consequence of a misunderstanding or
misrepresentation of the provisions of the Act—i.e. gathering/sheet F appears
to be the remnant of a paper used in an earlier publication. It is found
in Monash copy 1 in gatherings F-H only and in copy 2 in gatherings R 2A
2D only; the major run of paper in both copies has the `fleur-de-lys/IV'
accompanied by `97' in all four corners, while the final four gatherings
(2H-2L) and the preliminary gathering (a) have a paper without watermarks
except for `1796' in two corners. Whether the dates were added to existing
moulds or not cannot be known, though the `1796' and `97' dates would
imply not, I think, given that an undated mould would have been unlikely
to survive in use for the two years beyond 1794. The `W' within the fleur-delys
suggests that all but four sheets (if not all thirty-three) came from the
same paper-maker, whose practice it was to follow the spirit of the Act. More
frequently, however, dated laid moulds do not have watermarks in the traditional
location(s)—i.e. in the centre of one half of the mould—and from
my limited observation I can cite no instance of a laid mould with a maker's
name in addition to the date.[14]

As has already been noted, by the early 1790s much British paper was
wove, and often enough without watermarks, thus lacking the traditional
indicators of format. Hence the Act of 1794, in requiring the inclusion of a
watermark, in the form, at a minimum, of a date, has provided the modern
bibliographer with the means of undertaking the analysis of printed volumes,
in particular the means of determining format. The Act had the effect of
encouraging the migration of all watermarks (now increasingly paper-makers'
names or initials) to the edges of the sheet, often accompanying the date. I
am not aware, however, that there has been any systematic illustration of
the form of watermarks in wove paper from around the turn of the century
or their disposition within the sheet.[15] But since a familiarity with the various
patterns may be an aid in determining format (or in undertaking other forms
of analysis), I have selected a number of examples from the period for illustration
here.[16] (It is likely that further patterns will be revealed with the use
of a differently constructed sample.)

Throughout I have illustrated the outer forme of the paper (not the imposed
type); dotted lines represent folds, solid lines cuts. The watermark
would be sewn or welded onto the mould in reverse so that it could be read
in the finished sheet from the mould (or `right') side, and in some instances—
depending on whether the watermark lay to the vatman's left or right on the
mould—the watermark, though correctly placed in the illustration, would
be read from the other side of the sheet from that shown and sometimes
actually run in the opposite direction. To give some idea of the resulting
location of watermarks within gatherings I have added type-pages to the
sheet, using the signature `A' to represent sheets, regardless of how in fact
signed, rather than the generally applicable `$' or the actual signature of the
sheet from which the example is drawn. Particularly in the earliest years the
date might be fairly crudely formed by a single wire, but thereafter it is usual
to find a much finer open lettering, an approximation of which I have used
throughout.

The simplest way of meeting the requirements of the Act of 1794 was to
place a single date at one corner of the mould. However, it is certainly commoner
to find the date appearing twice (at diagonally opposite corners) or
four times (at all corners). Often enough, the paper-maker's initials accompany
one of the dates, as in figure 2:[17] Walter Scott, The lay of the last
minstrel, 4th ed. (London: Printed for Longman, Hurst, Rees, and Orme,
and A. Constable and Co. Edinburgh; by James Ballantyne & Co. Edinburgh,
1806), T/B14Ag;[18] 8°: [A]8 B-U8 X4 Y2. This sheet is another remnant,[19]

found only in gathering/sheet Q in two of the three Melbourne copies;
otherwise the paper is uniformly watermarked `D & A C | 1805 | 1 [2]', gathering/sheet
Q in the third copy being on the paper with the vat number `2'.
(On vat numbers see further below.)

Common, too, is to find the maker's full name plus date in one corner—
see figure 3:[20] Anthony Hamilton, Memoirs of Count Grammont [ed. Walter
Scott], new ed., 2 vols (London: Printed for William Miller, and James Carpenter,
1811), quarto issue, T/B53Ab; vol. 1, 4°: π4 2π1 a-d4 e4(-e4) A-2K4,
gathering 2H. This example also shows how part of the watermark may be
lost in cutting, here at the head of A1, as it is in all gatherings/sheets in the
copy examined, with nothing surviving beyond the `C'. In this publication
the dates in the paper may suggest that the bibliographer should at least entertain
the possibility that there was an interruption in the printing of the
volume. Gatherings A-2C are watermarked `1802' (no maker's name); gatherings
2D-2G `1804' (no maker's name); gatherings π-e and 2H-2K in the way
illustrated—that is, only the last three gatherings of the text and the six
gatherings of the prelims (presumably printed last) are on the paper dated
`1810'. Volume 2 collates π2 A-2X4 2Z2, and gatherings A-2A are watermarked
`1802', 2B-2P `1804', with the `Dickinson 1810' paper in gatherings π and
2Q-2X only, thus implying that the two volumes were going through the
press simultaneously.[21] In this instance the order of the dated papers within
the volume provides a strong indicator of the progress of the work through
the printing house, an incidental virtue which suggests that, where feasible, a
precise record of the incidence of particular papers should be given for at
least those volumes on dated paper. The `1804' paper is employed throughout
for the octavo edition (in French and published concurrently with the two
issue of the English edition),[22] except that gathering e is on the `Dickinson |
1810' paper. On the basis of the paper one might therefore hazard that the
printing of the French edition began when the English edition was well
under way but that the French was finished first, though no doubt other
explanations could be advanced, particularly as the papers may be different
(T/B records the English edition as on Royal, the French on Demy).

The value of dates as watermarks in providing clues to the progress of a
work through the press may be further, and more amply, illustrated by Scott's
nineteen-volume edition of The works of Jonathan Swift, commenced in mid

1808 and finally published 6 August 1814, nearly four years after the date
originally projected. The progress of the work can be followed in Scott's
letters, as well as in the volumes themselves—in printers' imprints, press
figures and the form of signatures, but more obviously in the paper, where
the dates are a ready guide to the breaks in production. Take, for example,
volume 9, which collates π4 A-2I8 2K4(-2K4) *2L8 *2M4 *2N1. Most of the
volume was worked off by 1809: gatherings/sheets A-2I are dated `1808', and
2K `1809'. It is clear from the asterisks that 2L-2N constitute an addition,
while the date in the paper (`1813') suggests that the interval may well have
been four years.[23]

Other paper includes the maker's name and the mill name (plus the date),
as in figure 4:[24] Sir Walter Scott, Provincial antiquities and picturesque
scenery of Scotland (London: John and Arthur Arch; Edinburgh: William
Blackwood, 1826), T/B132Ab; 4°: [a]1 b-g4 h1 A-H4 I2 x I1 K-M4 N1 χ1 O-Q4
R2 x R1 S-T4 U2 X-Y4 Z2 2A1 2B-2E4 2F4 [2G]1, gathering Y. This gathering is
the only one to contain the mill name; the remaining gatherings are watermarked
`J WHATMAN | 1817 [1818, 1819, 1820, 1821, 1822, 1825]'; the sequence
of dates is to be explained by the fact that the work was published in
ten parts, the first 1 May 1819, the last 14 December 1826, before being issued
in volume form. It would seem that the inclusion of a mill name was unusual:
in the several hundred volumes of Scott in Melbourne[25] there is apparently
only one other publication containing paper bearing a mill name (`J WHATMAN
| TURKEY MILL | 1822').[26] Perhaps a different sample would provide
a different picture; for example, I have noticed `IVY MILL | 1812' as the
watermark in all but one gathering in George Colman the Younger, Poetical
vagaries (London: Printed for Longman, Hurst, Rees, Orme, and Brown,
1814).[27]

Yet other paper includes, in addition to the maker's name and date, a
vat number, such as that illustrated in figure 5:[28] [Walter Scott], Tales of my
landlord [First series], 4th ed., 4 vols (Edinburgh: Printed for William Blackwood;
and John Murray, London, 1818), T/B98Ad; vol. 1, 12°: π2 A12 (±A1)
B-P12 Q2, gathering B. The appearance of these numbers is explained by
Balston: `When more than one vat was engaged in the making of a paper,

FIGURE 1. Laid paper; watermark and countermark centred in half-sheets, and date
in four corners.

FIGURE 2. Wove paper; maker's initials and date in one corner, commonly with date
also in the diagonal corner or in all corners.

FIGURE 3. Wove paper; maker's full name and date in one corner.

FIGURE 4. Wove paper; maker's name, mill name, and date in one corner.

FIGURE 5. Wove paper; maker's name, date, and vat number in one corner.

FIGURE 6. Wove paper; maker's name in middle of one long side, and date in one or
more corners.

FIGURE 7. Wove paper; maker's name (or initials) and date centred in separate half-sheets.

FIGURE 8. Wove paper; maker's name in middle of one long side, and date in one or
more corners; imposed for three gatherings in 18°.

the vat's number was frequently included in the countermark in order to
identify the maker of each sheet.'[29] Just as the range of press figures found in
the work of a particular printing house will suggest the number of presses in
that establishment so will the range of vat numbers suggest the number of
vats in the particular paper-mill, though in both instances caution must be
exercised in concluding that all—presses or vats—were always in use. For
example, during a slump in demand in 1809 the paper-makers Keys commended
Balston for having `discharged 3 [of his ten] Vats',[30] which I take to
mean that he had temporarily taken them out of commission. If the `M' preceding
the date in this paper also stands for `Mavor' it is redundant, but the
inclusion of the maker's name suggests that other paper of the period, with
a watermark in the form `M1816' plus a vat number but no maker's name,[31]
may well have also been made by Mavor. In all three Melbourne copies of
volume 1 of this fourth edition the paper is a mixture of `MAVOR | M1816 |
5' and `MAVOR | M1817 | 6'.[32]

More unusual is it to find the paper-maker's name in the middle of one
long side, plus dates in one or more corners, as in figure 6:[33] Joseph Addison,
Addisonia, 2 vols ([London:] Printed for Richd. Phillips, by T. Davison,
[1803?]), vol. 1, 8°: b8 c4(-c4[=R1?]) B-Q8 R1 χ1, gathering B. The same paper
is used throughout, except that what are presumably the last four gatherings
to be printed (P Q b c) are on a paper watermarked simply `1802', the date
appearing about a quarter of the way along a long edge—i.e. it appears in the
inner margin of $2.7 in those four sheets in the copy seen.

Occasionally sheets of wove paper contain an additional number the explanation
for which is not obvious: in two publications there are sheets containing—as
well as the watermark `W BALSTON | 1812'—a further number
below a vat number, in the form `5 | 2022'. One of these publications is the
fourth edition of Scott's Rokeby, 1813 (T/B64Af),[34] in which the additional
number appears only in gathering/sheet T. The other is Richard Porson,
Tracts and miscellaneous criticisms (London: Printed by Richard and Arthur
Taylor, for Payne and Foss, 1815),[35] which collates [a]4 b-g8 [B]8 C-2C8 2D4
and in which the additional number appears in gatherings/sheets B-S Z. At
first sight the additional number suggests itself as a mill excise number. These

numbers were first allocated in 1712, when a duty was imposed on paper;
but by 1861, when the last remaining tax on paper was removed and the
practice of allocating excise numbers abandoned, the number had reached
only 694.[36] Alfred H. Shorter provides a possible clue to the interpretation
of higher numbers when he reports that `Some mills which were founded
after 1861 have long been known by a distinctive trade number which is
obviously not of Excise origin; examples are Broad Dumers (888) and Withnell
Fold (3009).'[37] What I have not established is whether `trade numbers'
were assumed before 1861, whether `2022' was a Balston trade number. Nor,
it should be said, have I found a number below 695 as part of a watermark—
i.e. I do not know whether excise numbers were in fact ever added to a mould.

There is even the occasional wove paper with a maker's name (or initials)
and a date in the traditional (pre-1794) positions of watermark and counter-mark,
an arrangement quite surprising in emanating from a large mill (D.
and A. Cowan) and in being so late (1806)—see figure 7:[38] Joseph Strutt,
Queenhoo-Hall, 4 vols (Edinburgh: Printed by James Ballantyne & Co. for
John Murray, London: and Archibald Constable & Co. Edinburgh, 1808),
T/B32Aa; vol. 2, 8°: π1 2π1 A-P8 Q4, gathering A. The four volumes are all
on the same paper, except that the cancellantia and preliminary leaves are
on a paper in which the watermark appears to be `C S | 1806'. In the uniformity
of its paper a publication of this extent is unusual in the output of James
Ballantyne, though it should be allowed that the particular form of watermark
may well obscure any number of vats (and moulds).

Shorter reports that `All the available evidence as to the size of paper mills
in England during this period [1701-1800] points to the conclusion that the
small mill, equipped with only one or two vats, was typical.'[39] He calculates
that in 1800 there were just over 400 mills in England,[40] more than eighty
per cent of which had only one vat; fifty-five had two vats, eight had three,
three had four, and two had five. The two largest were both in the Maidstone
district, the Whatman Turkey Mill and Clement Taylor's Upper Tovil Mill.[41]
Numbers increased in the early years of the nineteenth century; among large

establishments with known complements of vats were D & A Cowan (eleven
vats, 1808)[42] and T. Balston (ten vats, 1811).[43] The highest vat number that
I have seen is `14', a single instance, found in gatherings/sheets A and B in
some copies of volume 1 of Scott's edition of Swift, this volume printed by
George Ramsay in Edinburgh. Elsewhere, in Ballantyne's printing of Scott, vat
numbers of `10'-`12' are found, in paper dated `1811'-`1814';[44] it would seem
that by the later date many of the largest paper mills were about to become
mechanised, given that vat numbers did not, on the evidence of the paper
used by Ballantyne, go any higher. Unfortunately none of this paper contains
a maker's name, so that we cannot know who had such a large complement
of vats. The fact that a maker's name was not included in some
paper confirms that vat numbers (whether or not accompanied by the papermaker's
name) were included in the watermark as a means of internal `quality
control', of concern to the paper-maker only, in identifying the individual
workman or team.[45]

As stated at the outset, the underlying interest in this study has been in
the determination of format, by implication of volumes printed on wove
paper. For a worked example of using the evidence of watermarks for determining
the format of a post-1793 volume printed on wove paper I have
chosen Jonathan Swift, The works, new ed., ed. Thomas Sheridan, rev. John
Nichols, 24 vols (London: Printed [by T. Bensley] for J. Johnson [and 23
others], 1803), vol. 11.[46] With a cut leaf measuring 155 × 95 mm it collates
[A]6(A1+1) B-2H6 2I4(2I3+1),[47] suggesting that the format is likely to be 12°,
18° or 24°, but not 16°. There are press figures only on E6r H4r Q6v X6v 2D6v
—not enough to be useful for deriving format—and ploughing has removed
all point holes, along with all deckles save for a trace at the foot of a few
leaves. Watermarks, however, are sufficient to determine the format. Gathering
A has no watermark, but from B to T (after which a different paper is
used) `L MUNN' appears in CS3.4 (i.e. leaves 3.4 in gatherings C and S) and
FIMP1.6, with `1801' in BEH3, DGK6, LOR1 and NQT4. On the basis of
the distribution of the watermarks the volume may be identified as an
eighteenmo, with a typical printed sheet arranged as in figure 8. Since the
sheet was imposed in such a way as to produce three gatherings, actual signatures
(B C D) are used for clarity, but in the illustration the sheet has been
turned 180° from its actual orientation in gatherings B-D in this copy in

order to show the effect that would result from inverting it.[48] Two further
variant patterns to that illustrated would be achieved by turning the paper
over on its long axis, so that the date would appear in B3 and D6 and the
maker's name in either C3.4 or C1.6.[49] That the scheme derived from the
evidence of the watermarks is correct is confirmed by a second, uncut, copy
in the Monash collection,[50] which exhibits these characteristics:

• a deckle on the outer edge of B1-3 and D4-6;

• a deckle at the foot of BCD1 3 4 6;

• a tear on the outer edge of B4-6, C1-6 and D1-3;

• a cut or a tear at all heads and at the foot of BCD2 5;

• point holes in the outer edge of B4 and B6.

What this exercise has done is to show that the watermark evidence is sufficient
of its own in determining format, that familiarity with the patterns of
watermarking dated paper should enable the bibliographer to establish format
independently of any other evidence—just as, of course, the patterns of
watermaking pre-1794 wove paper may.

Determining the format of a volume in hand on the basis of the disposition
of dates and names in the paper may be achieved by reference to the
illustrations. The process is to some extent circular, but the resulting determinations
will, in my experience, withstand scrutiny, except that a single
date in a gathering in a cut copy could just as well indicate a half (or quarter)
sheet as a full sheet, though in such instances size and shape, as well as the
number of leaves in a gathering, should be a good guide to the relationship
between the original sheet and that portion of it represented by the gathering.

II

The second development of the 1790s of obvious interest to the bibliographer
is the beginning of the mechanization of the production of paper,
a process initiated probably in 1796, by Louis Robert, at Essonnes, south of
Paris.[51] The realisation of Robert's ideas was achieved in Britain, financed

in part by Henry and Sealy Fourdrinier, stationers and paper-makers of
Huguenot descent, by whose name the paper-making machine is still known.
The machine was perfected in 1807, when a patent was granted to the
Fourdrinier brothers and John Gamble—`perfected' in the sense that modern
Fourdriniers operate on the same principles, even if accompanied by
numerous refinements and huge increases in scale. The machine comprises an
endless moving wire screen onto which the stuff (or `pulp') is dribbled from
the stuff chest (or `vat') via the breast box;[52] the paper is then wound up at
the `dry end' of the machine to form a continuous reel (or `web').

Such paper is conventionally known as `machine-made'; but it is necessary
here to consider briefly a distinct variety of machine-made known as
`mould-made', which could be produced in one of two ways. The earlier way
was the `chain-mould' method,[53] which comprised an endless belt of conventional
moulds (and deckles), without the underlying wire screen, onto
each of which a measured amount of stuff was poured. It is not certain, however,
that any machine constructed along these lines was ever put into use
commercially. The later way was to use a composite machine, combining
chain-mould and Fourdrinier;[54] whereby conventional moulds could, when
desired, be attached to the wire screen; with the moulds removed, the machine
was then available again for producing a web. The moulds could be either
laid or wove, but obviously the objective of the exercise was to mechanize the
production of laid paper, since until 1825 only wove paper could be produced
on the normal Fourdrinier. The continuing demand for laid paper
was met rather by hand-mades, sheets produced by the traditional method of
dipping mould and deckle into a vat of stuff, so that mould-mades may be
regarded as little more than a historical curiosity. In any case, as far as the
bibliographer can probably determine, there is nothing to distinguish a sheet
of laid paper produced by either of the chain-mould methods from one produced
by hand.

A distinction, however, can tentatively be made between a sheet of wove
paper produced on a Fourdrinier and one produced by hand. Before 1839
any sheet containing a watermark, whether or not it includes a date, must
have been produced by hand, for it was not until that year that the dandy
roll was first employed with a watermark design on its surface.[55] As first employed
on being patented in 1825 the dandy roll was no more than a roller
with a laid finish passed over the surface of the web as it was being formed,
its purpose being simply to press out more water. From 1828 dandies were
also made with a wove finish, though without a watermark.

Paper produced between 1825 and 1839 with a laid appearance but without
a watermark was probably made by machine, an assumption which it
may be possible to confirm by determining which side of the sheet the indentations
from the dandy roll are in. Machine-made `laid' paper (i.e. that
created with a dandy roll) differs from hand-made (and mould-made) laid
paper by the fact that in machine-made the `chain lines' appear in, and the
watermarks are read from, the side which was not in contact with the wire
screen, the reverse of the procedure with hand-made. In hand-made paper
the side which came in contact with the mould is described as the right (or
`mould' or `wire') side, the other side the wrong (or `felt') side. With machine-made
paper, however, the terms are reversed: according to Labarre, `the upper
side is the `right' side, i.e., the side on which the couch-roll acts and not
the wire'.[56]

It is also of some bibliographical interest to note the methods by which
the web was reduced to sheets of conventional sizes. In the period preceding
the introduction of dandies bearing watermarks the web might simply be
drawn across a table and cut to the length desired. After 1839 the web might
be `cut to register', a process whereby the circumference of the dandy corresponded
with the length of the shorter edge of the intended sheet, so that
the watermark would appear in a consistent position in the resulting sheets.
Unless the sides of the paper can be distinguished it may not be possible to
tell a laid hand-made sheet from one cut to register from a web. Conversely
paper with a watermark appearing in a varying position within the sheet—
as in modern superior paper (variously, and often misleadingly, described)
used both for printing and for writing—must be machine-made.

It may be impossible to determine whether the wove paper without
watermarks used in a British book printed in the early nineteenth century is
hand- or machine-made, but an uncut copy may help in the determination.
The hand-made sheet will have deckles on all four edges, whereas the
machine-made will have either three or four cut edges, depending on how
the web was reduced to sheets. The remaining edge (either a shorter or a
longer) will exhibit what may be called `deckle strap thinning', created by
the stuff seeping under the deckle straps (the revolving leather straps which
confine the stuff to the wire screen)[57] and resulting in a deckle characteristically
more uniform than the deckle edge created in a sheet of hand-made paper.
Though it may sometimes be difficult to distinguish a thinning from a tear,
viewing the edge against a light should identify a thinning. Of course cut
edges may also result from producing a sheet of the required dimensions
from, for example, a double sheet. Nonetheless, whatever the difficulties, it
may be worth examining the edges of an uncut copy in attempting to determine
format or in undertaking other varieties of analysis.

If James Ballantyne in Edinburgh is typical of major printers elsewhere
in Britain machine-made paper was not in widespread use in substantial

bookwork until about 1820, but by the mid-1820s the transition from handmade
was practically complete. In other words, and discounting the small
amount of laid paper still being used, up to about 1820 paper used in bookwork
was usually wove with a watermark (i.e. hand-made), but from the mid1820s
it was mostly wove without (i.e. machine-made)[58] Thus, in printing
the works of Sir Walter Scott, Ballantyne appears to have first used machine-made
paper in 1819, for the Novels and tales (T/B269Aa, published December
1819) and the Poetical works (T/B262A, January 1820), both in twelve
volumes; thereafter some works are on hand-mades, others on machine-mades,
but beginning with Woodstock (T/B190A, April 1826) all the major works
are on machine-mades exclusively. This transition within Ballantyne's
printing-house parallels the transition within the British paper-making industry:
1824 was the first year in which the output of machine-made paper
exceeded the output of hand-made, 14,459 to 12,750 tons. The progress of
the transition (and the increase in production) can be gauged by the figures
for 1840 (33,463 to 9,937) and for 1860 (95,971 to 3,839).[59]

Wove paper was considered preferable to laid because of its greater
smoothness and uniformity, and machine-made paper was ready for shipping
in five days, rather than five weeks for hand-made. In 1806 the Fourdrinier
brothers estimated the annual cost of operating a seven-vat hand mill at
£2,604.12s., as against the cost of operating an equivalent machine at
£734.12s.; at the same time the cost per hundredweight of making paper by
hand was reckoned to be 16s., as against 3s.9d. by machine.[60] These advantages
notwithstanding, paper-makers were slow to install Fourdriniers. Clapperton
notes that `Owing to the want of enterprise among the paper-makers,
and the opposition of the trade to the introduction of machinery, the business
did not at first progress rapidly.'

However, we should not overlook the expense involved in acquiring and
installing a machine. In March 1808 William Balston was quoted £1500 by
the Fourdrinier brothers for a machine capable of producing the equivalent
of eight vats, with an annual licence fee of £380.[61] Despite the assurance that
installation of a machine would save him £1000 a year Balston chose to retain
his vats. According to Bryan Donkin's `Prospectus of the patent machine for
making of wove paper, 1813'[62] even the smallest machine, a mere thirty
inches between the deckle straps and producing the equivalent of three or
four vats, cost £715 if driven by straps, £750 if driven by wheels, with an

annual licence fee of £200 (three vats) or £300 (four vats); Donkin's licence
fee for an eight-vat machine was £940 or £980. In other words the slowness
of implementation may well have been due to the capital costs involved, not
necessarily to any dissatisfaction with the machine itself or to any hostility to
the principle of mechanisation.

Estimates of the number of machines installed in the 1820s and '30s vary
(one estimate for 1837 is that 105 machines were then in operation, another
279); by 1842 there were 356 machines and 372 traditional vats, with the
machines calculated as on average equating with five vats.[63] As can be seen
from the figures already quoted, by 1860 hand-made accounted for under
four per cent of the paper produced in Britain.[64]

Though published volumes are generally printed throughout either on
hand-made paper or on machine-made, that homogeneity almost certainly is
merely a reflexion of the paper-merchant's supply, not a desire for uniformity
on the part of the printer. Indeed, some volumes were printed on a mixture
of papers, apparently indicating an indifference within the trade. The three
volumes of Scott's Quentin Durward, an octavo, published 20 May 1823,[65]
provide a convenient example. One set,[66] which differs only in the odd sheet
from other sets in the Melbourne collection and so may tentatively be taken
as representative of the edition as a whole, is predominantly without watermarks,
but thirteen of the twenty full gatherings/sheets in volume 2 and six
of the twenty-two in volume 3 do have watermarks. Hence of the sixty-two
sheets making up the three volumes nineteen are clearly hand-made. That
at least some of the remaining forty-three are machine-made is confirmed by
the occasional presence of a seam mark, a mark in the paper produced by
the seam joining the two ends of the wire screen on a Fourdrinier.[67] That
all forty-three sheets without watermarks are machine-made, derived from a
web, is suggested by the patterns of cut and deckle edges—for example, in
the three uncut sets of Quentin Durward in the Melbourne collection the
fore edge of $1-4 in the sheets without watermarks is invariably cut, whereas
those sheets with a watermark invariably have a deckle edge in that position
(i.e. those sheets with a watermark were produced in a traditional hand-held
mould). From copies which have been cut this degree of certainty cannot be
arrived at: it will not be possible to distinguish sheets produced by hand in
moulds without watermarks from sheets produced by machine in a web, also
without watermarks. In the transitional period (and specifically in the years
around 1820), therefore, it is possible that three sorts of paper may be found
in the one volume.

As a counter to the suggestion of indifference just made is the curious

case of the eighth edition of Scott's Waverley (published, according to Lockhart,
in an edition of 2000 in April 1821),[68] of which there are two issues;[69]
copies of both are in the Melbourne collection. Pace Todd and Bowden, all
three volumes of 77Ak are watermarked `1819 | 2', whereas 77Al has no
watermarks at all; that the paper in 77Al is at least in part machine-made is
again demonstrated by the presence of an occasional seam mark, though the
paper is palpably not uniform. 77Al is distinguished from 77Ak in having
`†' in the direction line of $1r, but the two agree in press figures, demonstrating
that they were printed in one continuous run (Todd and Bowden
mistakenly regard the latter as a reimpression). The siglum clearly serves to
distinguish the two issues, but why would the printer (anonymous) or the
publisher (Constable, in Edinburgh) wish to make a distinction between
hand-made (or watermarked) and machine-made (or without watermarks)?
Were the two papers considered as of different qualities (and therefore sold
at different prices)?[70] —Lockhart makes no distinction.

The change from hand- to machine-made paper corresponds also with
the transition from printing with a hand-press to printing with a machine,
and perhaps it could be claimed that it was the development in printing
technology which encouraged the transition in the process of paper-making
by increasing the demand for it.[71] In Ballantyne's printing-house the two
changes do indeed run in parallel. From a Scott letter it appears likely that
Ballantyne introduced machines (how operated is uncertain—perhaps manually)
in 1823,[72] and Jane Millgate reports that two steam-powered machines
were acquired in early 1830 for use in printing the magnum edition of Scott's
collected works;[73] over that seven- or eight-year period the transition—both in
paper-making and in printing (the latter reflected in the abandonment of
press figures)—was complete. The widespread adoption of machine-made paper
by the end of the 1820s may be just one element in the transition in book
production from the hand-press period to the machine-press, but for the bibliographer
it is the element which most emphatically marks that transition.

To return to my original concern: the determination of format in British
books printed on locally-made paper in the 30-odd years from 1794. Until
1811 much of the paper used for printing was hand-made, containing a
watermark comprising a date and often other elements, and the practice of
dating paper was continued well beyond that date, seemingly as long as
paper was being made by hand in a mould. Uncut volumes on dated paper
should readily reveal their origins in a hand-held mould, and hence their
format. Cut volumes on dated paper may be less revealing; nonetheless the
illustrations presented here suggest that a familiarity with the patterns of
placing watermarks on moulds from 1794 onwards should enable the bibliographer
to determine the format of such volumes. However, over the course
of the 1820s there is a clear decline in the capacity to determine the format
of the volume in hand, given that by 1830 the volume may well be among
the majority printed on machine-made wove paper without watermarks of
any kind and cut in the process of edition binding, thereby removing all the
supplementary evidence which might have been brought to bear on the task,
notably the original edges of the sheet. From this decade onwards alternative
sources of evidence will need to be sought if the bibliographer is to determine
the format of the volume in hand.