ROMANO-BRITISH WATERMILLS
R.J. SPAIN
Watermills, by their very nature, are associated with one of the most
violent factors of land form evolution, namely water-power. Rivers
and streams are the chief agents in the excavation of valleys, and their
powers of erosion and transportation are manifest in the everchanging
interface of land and water. If the landscape was viewed
from a historic time-lapse the water-margins, especially of the rivers,
would exhibit most rapid moving land surfaces. River profiles,
gradients and courses are restless under the interacting influences of
erosion and deposition. It is ironic that the natural force, which
watermills harness for man, is the instrument of their preservation or
destruction.
Watermills, probably more than any other building on the early
landscape, have suffered most from the forces of nature. Even
bridges, although subject to the same forces, have the singular
advantage of their position usually being known to the archaeologist.
Ancient roads do not move; rivers do. What other type of abandoned
building within a valley could be obliterated by the downstream
migration of incipient meanders or other manifestations of river flow?
One wonders whether, in the study of Roman watermills, British
archaeology is at any disadvantage in having a landscape of damp
climate and abundant rain, about half of which finds its way to the
sea. The Roman countryside of forested and soil-mantled slopes has,
in the centuries following, suffered from increasing rates of denudation
and intensive agricultural activities. Elsewhere in the Empire
buildings, among them watermills, still stand in undisturbed Roman
landscapes. However, if we have the disadvantage of climate, it is
more than offset by our advances in archaeology.
In order successfully to study ancient watermills, there must be a
fruitful reciprocal relationship between the archaeologist and engineering
historian. The study of mechanical and hydraulic properties
inherent in historic watermill structures and arrangements cannot be
101
R.J. SPAIN
facilitated without the primary material provided by the archaeologist.
An analogous relationship occurs in our understanding of the
making of the landscape, where the historians' studies spring from
the geologists' work. So, too, must the engineer depend upon the
archaeologist for earth-born historical studies. This is especially true
of Roman studies.
But the archaeologist should recognize additional problems germane
to watermills. He must be familiar with evolutionary studies of
drainage systems and have an understanding of the physical arrangements
of watermills, their effect on the landscape and relationship
with associated water-courses. The archaeologist will, with advantage
to an increased body of knowledge, respect the inter-disciplinary
division of responsibilities. The interpretation of the primary evidence
for the purposes of engineering analysis, is the prerogative of
the engineer.
At the turn of the last century, when Bennett and Elton were busy
producing their impressive study concerning the history of corn
milling,1 Romano-British watermill sites were unknown in these
islands. Apart from a dubious suggestion by an eighteenth-century
historian, that a conduit discovered at Knott Mill near Manchester,
may have been the race of a Roman watermill,2 no archaeological
evidence had been identified to support their existence. Many
millstones were being discovered in Roman contexts and horizons
throughout the nineteenth century, but these were rarely recorded
and invariably cast aside and often lost, many museum specimens
being unprovenanced. Querns were even more common as a staple
artefact of most domestic, military and industrial sites.3 Unfortunately,
the identification of millstones and querns has always caused
problems for the archaeologists, and still does, especially when
fragmented and degraded.4 Perhaps understandably then these artefacts
are largely ignored in site reports, lost among the volumes of
minutiae devoted to the drawings and identification of sherds and
other artefacts. The difficulties of interpreting milling artefacts
together with a lack of understanding of associated hydraulic and
1 R. Bennett and J. Elton, History of Corn Milling, 4 vols, (1898-1904).
2 Whitaker, History of Manchester, ii (1771), 216.
3 It has been estimated that there are from eight to ten thousand pre-Saxon querns
in the north of England alone. Adam T. Welfare, post-graduate research student,
University of Newcastle-upon-Tyne, in correspondence with the author.
4 In this context the word degradation means the mechanical detrition by man
involving the obliteration in part or whole of the original stone by subsequent,
different functions, e.g. rotary milling-reciprocating milling-whetting.
5 Antipater of Thessalonika, Pliny the Elder, Procopius, Strabo, Vitruvius, etc.
102
ROMANO-BRITISH WATERMILLS
CHOLLERFORD
MUSEUM
CARPARK BRIDGE
4/iy Cl
Iv. CHESTERS 3»Z-3V
FORT
STEOF
£> WATERMIU
HILECASTLE 27
CIVIL
SETTLEMENT N
t VALLUM
SOQ tooa ISOO FT
I I i
100 200 300 400 M
Chesters Bridge Ktstaeu.
Fig. 1. Chesters Bridge.
mechanical technologies have inevitably retarded studies of watermills
in the Roman landscape.
Victorian archaeologists were well aware that the Romans employed
watermills; this was attested by the often quoted classical
writers5 and the wealth of codes and edicts associated with Roman
law which mentioned watermills.6
In 1860 Clayton, excavating the eastern abutment of the Roman
bridge just south of Chollerford Bridge, near Chesters fort, discovered
a millrace and millstones associated with the tower (Fig. I).7
During this and other work on the site in the 1860s the evidence was
not identified as a possible watermill.8 At Chesters bridge the
generation of waterpower was facilitated by a stone-lined watercourse
passing through the base of the tower (Fig. 2). The source of
this water must surely have been the North Tyne, which now flows
6 Far too numerous to mention here. As an introduction see O. Wikander,
'Water-Mills in ancient Rome', Opuscula romana, xii, 2 (1979).
7 J. Clayton, 'The Roman Bridge of Cilurnum', AA2, vi (1861), 80-5.
8 AA2, vi (1865), 86; J.C. Bruce, Roman Wall, (1867), 148.
103
R.J. SPAIN
RXS i&eu.
30 FT
WALL
to M
%c
«c ^^ § &*,
<^7CA
^
TOWfK
~*
* T
Li
1
N>
Chesters Bridge Abutment
AFTER CLAYTON AMD BRUCE
Fig. 2. Chesters Bridge Abutment.
some 25 m. away to the west, but Wilson queried this,9 probably
because it was not proved by archaeological work. The width of the
watercourse varies throughout its length from 1.85 to 1.93 m. just
before it passes into the tower, to 1.65 m. through the mill and the
tailrace.10 Apparently, it was not possible to identify the level of the
original bottom of the course, although the depth of the sides over
the stone courses is approximately 0.84 m. A survey by Simpson
showed the course to be almost level over its length of 39.6 m.,
although the bottom of the walls indicated a drop of 11.5 cm.
An exact dating for the working of this mill is difficult. Although
clearly within the Roman period, for the Military Way passed over
the tailrace, the entry of the headrace through the mill wall is
partially blocked by stones, suggesting a crude earlier attempt to
9 F.G. Simpson, Watermills and military Works on Hadrian's Wall: Excavations in
Northumberland, 1907-13, (Ed.) Grace Simpson, with a contribution on watermills by
Lord Wilson of High Wray (1976), 46.
101 am indebted to Mr. F.J. Fuller for the detailed site measurements of the mill and
its watercourses made during October 1975.
104
ROMANO-BRITISH WATERMILLS
reduce water flow or an abandonment of water-power during the late
Roman period. This possible abandonment could be related to the
change in the river course, which moved towards the west during the
Roman period. Richmond11 and Moritz12 consider the mill to have
been operational during the third century A.D. But the abandonment
theory must be questioned. An important feature of the plan is
the axis of the watercourse through the tower which is skewed and
not parallel with the north-south mill walls. Why did they build it
thus? Apparently the watercourse does not pre-date the tower
structure. The interesting point of the partial blockage is that it was
so constructed as to leave a clearway of 0.97 m. wide by 0.53 m. high,
the stones arranged to leave a neat rectangular hole low in the
watercourse. Of greater significance is the position of this aperture
against the east wall of the headrace, which makes the axis of the
water through the mill more parallel to the walls. Such an arrangement
would have improved the flow conditions through the tower and
the application of water to the undershot wheel. It is therefore
suggested that the partial blockage of the headrace through the north
tower wall existed during the working life of the mill, perhaps
installed soon after the building of the mill.
The size of the undershot wheel can be roughly estimated. If we are
influenced by the aperture in the north wall and allowing that the
skew axis of the water reduces the effective width of application to
the wheel, we could expect a wheel from 0.61 to 0.91 m. wide and a
diameter of say, 3 or 3.6 m. at the most. Several millstones were
found on this site confirming that the mill was used for corn-milling.
An unusual barrel-shaped stone with equi-spaced mortices around its
girth was found here. Various suggestions have been made as to its
use, including a waterwheel hub which it clearly is not, confusing
archaeologists for some while.13 Current opinion favours a mortar
operated by several men.
During 1907/8 Simpson discovered and excavated, during this and
the following year, the third-century Romano-British watermill at
Haltwhistle Burn Head not far from one of the Stanegate forts just
south of the Wall.14 Thomas Smith first noticed some features on this
11 J.C. Bruce-I.A. Richmond, Handbook to the Roman Wall, (1947), 166.
12 L.A. Moritz, Grain-Mills and Flour in classical Antiquity, (1958), 136.
13 J. Liversidge, Britain in the Roman Empire, (1968), 184; S. Holmes, PSAN2, ii,
178 ff.; PSAN3 ii, 283; Bruce-Richmond, op. cit., 79; I.A. Richmond, Roman Britain,
(1955), 171. A similar stone found at Lincoln in 1950 beneath the east side of the High
Bridge was also thought to be a waterwheel hub, which it is not.
14 F.G. Simpson, op. cit., 26. Simpson published a very brief note of his discovery in
PSAN3, iv, 167, 'Discoveries per Lineam Valli'.
105
R.J. SPAIN
: ^ > J
i ^
<& m ^
^=3
% m SHBEP ^ N
^ ^ K
^ 3
s x > 3 *vyy m?/ *.* * . 'HHf/fAr. ?N«<& vUUi<,
a> rpfa ^ ^
>=5 »SeP:I ^ L - ^
"ZiV -V
O 2 0 3 0 FT
1
IO M
-jS*vZf ?Z.^
M/Z.L *xv t^^a 0 Ql° -«Ltr-.-A !3«& -*KL>-
D:
£ VAA, ^ * << *>. "<9AA. *=r
%
V yts:-^ ^ fts-^
%
\
< ^ 7
^
Watermill at Haltwhistle Burn Head
AFTER F. S. SIMPSOM KOS 1934.
Fig. 3. Watermill at Haltwhistle Burn Head.
site in 1897 and thought them to be the remains of a bridge, but it was
Simpson,, prompted by J.P. Gibson, who systematically uncovered
and recognized the true purpose of the remains he had found. To
Simpson goes the singular honour of being the first to uncover and
simultaneously successfully identify a watermill in the whole of the
Empire. His interpretation of the remains was masterly, surely a
reflection of his engineering training, which produced a combination
of disciplines most rare among archaeologists (Fig. 3).
Analysis shows that Haltwhistle Mill had an undershot wheel 0.35
m. wide and 3.6 m. in diameter which powered a single pair of
millstones. The remains of several millstones were found varying
from 0.63 to 0.93 m. in diameter, together with the remains of six
querns. An interesting find was a basalt footstep-bearing stone which
had supported the millstone spindle. This flattish piece of stone had
several circular, tapered-section cavities produced by circular motion
106
ROMANO-BRITISH WATERMILLS
of the iron spindle end. The mill building was rectangular in plan with
walls 0.83 m. thick faced on both'sides with well squared stones.
Simpson found from three to five courses in situ above a footing
course. The floor area was 21.75 m.2 which compares closely with the
Chollerford Bridge Mill area of 23.0 m.2 Wilson thought the mill was
built of timber on low stone walls,15 but the external structure could
equally have been all stone.
This watermill was well-sited for the headrace gradient increased
over its length of 17.5 m. as it approached the wheel position and the
later part of its bed was solid rock.16 This was probably no accident of
siting for the scouring velocities that existed in the race necessitated
revetment of the headrace banks for most of its length. For the last
two metres the water was carried to the wheel by an inclined wooden
trough, some remains of which Simpson found in the bottom of the
watercourse beside the mill. The wheelpit, with ample clearance for
maintenance of the wheel and land-side bearing, was cut out of solid
rock as low as possible without causing tail-water under the wheel.
Analysis shows that the theoretical impact velocity of 4.5 m/s
generated a maximum of 1.1 hp. at the driver gear.17 Optimum
efficiency would have occurred with a gear ratio of four or five to one,
which suggests that the driven gear took the form of a lantern gear
(Fig. 4). With this arrangement the heaviest top stone, weighing 165
kg., would have revolved at close to 50 rpm. Alternative lighter top
stones or well worn stones would have resulted in a balance of power
occurring at a stone speed of approximately 90 to 100 rpm.
Palladius, a Roman writer on agriculture, recommended that when
there was an abundant supply of waste water from public baths it
should be used to drive watermills,18 thus giving future archaeologists
a clear indication of where to look. Evidence of watermills within
Roman baths has been found elsewhere in the Empire19 and Watts
has rightly suggested that we should be alert in the city of Bath.20
15F.G. Simpson, op. cit., 31.
161 am indebted to Dr. Grace Simpson for the information she kindly provided from
her father's site note-books.
17 A hydro-mechanical analysis undertaken by the author, part of a thesis, Imperial
College, London.
18 Palladius, Opus agriculturae, i, 41, (42).
19 The remains of building fabric associated with two vertical waterwheels exist
under the court of the Baths of Caracalla, Rome; see T. Schoiler and O. Wikander, 'A
Roman Water-Mill in the Baths of Caracalla', Opuscula romana, xiv, 4 (1983), 64-7.
There is slight evidence that a watermill may have operated in the Baths of Mithras at
Ostia; see T. Schoiler, Roman and Islamic Water-Lifting Wheels, (1973), 138, Fig. 99,
no. 6.
20 In private correspondence with the author, a letter dated 30th November, 1983,
from Mr. Martin Watts.
107
r ^ f e r ^
Section through the millstream
View from downstream
The arrangement of the machinery
- M&LSTOUe SPttJOLE
U i
-LANTERN GEAR
-FOOTSTEP BEARJN6
k. S FT
Reconstruction of the Romano-British watermill at Haltwhistle Burn Head
Fig. 4. Reconstruction of the Romano-British Watermill at Haltwhistle Burn Head.
ROMANO-BRITISH WATERMILLS
Aware of the permanency of mill sites through the centuries, he
points to the existence of an earlier mill in Leland's time powered by
the baths' overflow.21
During 1923 and 1924, Shaw working on the Roman bridge over
the Irthing at Willowford, uncovered evidence of another probable
watermill, the third associated with Hadrian's Wall.22 Running parallel
to the 1.7 m. wide channel between the bridge pier and landside
abutments, he found two stone-lined water channels, which were
understandably interpreted as sluices (Fig. S).23 The suggestion that a
watermill operated here was strengthened by the finding of millstone
fragments.24 Also found was a reputed spindle-bearing stone,25 since
identified as a socket for a vertical timber. But this interpretation of
the Willowford evidence is not without problems and needs further
analysis. The sluices, one or both, would have had a negligible effect
in relieving the water pressure and flow as a bypass to the wheel
because their position and face area on the whole of the fabric normal
to the river flow is insignificant. Furthermore, as Shaw noted, the
lower courses of the Wall would have been in danger of being
undermined in times of flood.26 To effect control over the volume of
water approaching the watermill and to provide a measure of
protection, building fabric had to be extended from the pier northwards
to create a bifurcation. This would have been terminated at a
convenient point upstream where sluice gates could be positioned
and operated thus creating a separate water channel and head race
for the mill wheel. It is unfortunate that this area north of the pier
was not dug by Shaw, however his work gives us a few tentative clues.
These are (i) the pavement appeared to extend northwards upstream
from the sluices, (ii) the north end of the pier was much disturbed and
it was not possible to confirm the original upstream profile,27 (iii)
large stones were found with checks on one surface like those
covering the sluices,28 suggesting that the sluices had been more
extensive at some previous period. Such evidence, although helpful,
is insufficient to confirm the existence of upstream bifurcation.
21 P. Rowland Smith, The Baths of Bath, (1938), 46; E.H. Bates, 'Leland in
Somersetshire 1520-42', Proc. Som. Arch, and N.H. Soc, (1887), 69.
22 R.C. Shaw, 'Excavations at Willowford', CW2, xxvi (1926), 429-506; see especially
450-77.
23 Ibid., 467.
24 Ibid., 485.
25 L.A. Moritz, op. cit., in note 12, 136, 7; J. Liversidge, op. cit., in note 13, 184;
J.C. Bruce-I.A. Richmond, Handbook to the Roman Wall, (1966), 16.
26 R.C. Shaw, op. cit., in note 22, 469.
27 Ibid., 473-4.
23 Ibid., 474.
109
R.J. SPAIN
CONJECTURED ELEMENTS - - TOWER.
::; THE-:;-:
iwAli'Z
• ABUTMENT
.WlUS OF.
ABUTMENT
PAVEMENT
SiUICES EE ;• ABUTMENT
PAVEMENT
r - - —
i
i - . . WATERWHEEL;
RIVER IRTHIMG
/
' - - - - - ^ , . , i . : ^ ) . v i Y ^ r ^ ^ r ^
«£:.'•••':'•••.•• PIER'.':.•'.'•'•'••.:'•••'•?
DISTURBED
PAVEMENT &B> a
Willowford Bridge — East Bank
AFTER R.C.SHAW RTS 1931*
Fig. 5. Willowford Bridge - East Bank.
If the sluice were to operate successfully as a bypass, separate
water gates would have been required on both the sluice and wheel
channel.
Shaw found a pit or depression at least 5 ft. (1.5 m.) deep in the
suggested wheel position, filled with cobbles and loose gravel.29 This
was clearly not a wheel pit for such an arrangement was impracticable
and inefficient, so that we must conclude that the cavity was created
by natural forces of erosion combined perhaps with stone-robbing.30
The mill structure would have spanned between the pier and
abutment two and might have carried a footbridge above as a
continuation of the rampart walk on the Wall. Mortices for vertical
™ Ibid., 475.
30 Shaw noted that stone robbing had occurred around the pier, see ibid., 473,
although he would probably disagree with this conclusion. See ibid., 475.
110
ROMANO-BRITISH WATERMILLS
timbers can still be seen on the abutment, although bolt holes are
absent, suggesting that the bearings were carried by timber sole
plates strapped to the uprights. Allowing for underside clearance, a
2.7 m. diameter wheel is suggested, which conveniently fits the length
of the abutment so that it could be enclosed by the mill building. The
driver gear would have been at least 1 m. in diameter and mounted
on the wheelshaft close to either the abutment or the pier depending
on which side the millstones were mounted. It is tempting to suggest
that there was one pair of millstones above the pier and another
above the sluices operated from a common wheel having a width of
between 1.0 and 1.3 m. Tandem drives are unknown in Roman mills,
though not impossible, and so if two pairs of stones existed here, it is
more likely that they were driven by independent narrow wheels
operating within the main water channel. It is not worth going into
more detail within the mill itself, it would be conjectural and
speculative. Suffice to say that the potential area for the mill is 3.0 m.
wide by nearly 7.0 m.31 Access to the mill would have been from the
roadway immediately outside on the downstream side.
During the early 1950s evidence of another Romano-British mill
was found at the site of a modern watermill, Spring Valley Mill, just
off the old Harwich Road, near Ardleigh in Essex (Fig. 6). Pettit,
working with local archaeologists, found numerous Roman artefacts
within the tail race, bypass channel and the millpond, including
pottery, tiles and millstones. The stones were reputed to have been
passed to Colchester Museum, and it is interesting that there are
three unprovenanced, typical Romano-British lava topstones on
display measuring 64.5, 54.0 and 41.0 cm. in diameter. Unfortunately,
no archaeological report appears to have been made of this
discovery, but a first-hand eyewitness account is reassuring, confirming
that Spring Valley was probably the site of a Roman
watermill.32
The archaeological evidence which points to the existence of a
watermill can take several forms, such as water-courses, wheel fabric
or evidence of emplacement and rotation, building remains, millstones,
etc. Ideally, all of these primary elements should be present in
juxtaposition, but it is not always so. At many sites the evidence is
much less, even scant and tantalizing. This is hardly surprising; as
time passes we gain a better understanding of watermills, becoming
more aware of their existence and the different forms that they and
31 The length over the pier reaching as far as Wall C in Shaw's drawing.
321 am indebted to Mrs. Hilary Dean Hughes for bringing to my notice the
information concerning this mill site. Mrs. Dean Hughes lived much of her childhood
at Spring Valley Mill, which her father owned and occupied.
I l l
R.J. SPAIN
IOO eoo »oo AOO FT i
So loo M
N
Mill
ie»r
O OQP
Qr*& P
£>
-OryO
*>£>*
near Ardleim Essex las :aeir
Fig. 6. Probable Romano-British Watermill Site near Ardleigh, Essex.
the evidence can take. It is almost certain that evidence has been
uncovered in the past but never recognised or alternatively, considered
as not worthy of recording. Searching extant archaeological
records for likely Romano-British watermill sites is never-ending.
Doubtless, this could be extended to other parts of the Empire but
few provinces are as rich in records and have the same intensity of
archaeology that Britannia enjoys.
112
ROMANO-BRITISH WATERMILLS
Large industrial sites, such as Heronbridge astride Watling Street
on the west bank of the Dee in Cheshire, often invite the suggestion
that water-power might have existed. From 1929 onwards excavation
work has taken place more or less continuously, revealing traces of
workshops involving bronze- and iron-working, a dock flanked by
quay sides,33 corn drying on a large scale34 and a stream bed artificially
deepened and revetted. With all of this activity, in addition to
domestic premises on this large site, it is not surprising that a
millstone fragment was found.35 Symptoms yes, but still no positive
identification of a watermill.36
A smaller industrial site was located on the south side of Muncaster
Head in Eskdale, Cumberland, where evidence of iron-working was
found. Many such Roman sites are known usually with smelting
furnaces and smithing hearths, slag and sometimes roasting pits and
mines.37 At Eskdale additional evidence is provided by the traces of
two ponds and water-courses.38
Other sites have been identified by archaeologists as possible
watermills such as Holeywell Hill, close to the River Ver where a
foundation was considered as a possible mill,39 or at Kimpton in
Hampshire where a large Roman building was found beside a
dried-up river bed.40 At Kenchester in Herefordshire, a secondcentury
stone-tower granary and timber-framed building was found
close to a stream which showed evidence of being re-cut to a square
section with a firm hard base. Within the timber-framed building
three millstones were found placed at regular intervals between the
posts.41 Wilmott concluded that the millstones were in store and not
33 B.R. Hartley and K.F. Kaine, 'Roman Dock and Buildings', Journ. Chester Arch.
Soc, xli (1954), 15-38.
34 (Ed.) F.H. Thompson, J.J. Bagley, Roman Cheshire, (1965), 60-5.
36 W.J. Williams, "The Roman Ditch at Heronbridge', Journ. Chester Arch. Soc.,
xxx (1933), 111-7.
36 There are two sources suggesting that a watermill may have existed here; see T.
Gariick, Romans in Cheshire, (1973), 52; and I am informed by Dr. G. Lloyd-Morgan
of Grosvenor Museum, Chester, that Sir Ian Richmond apparently thought that this
was possibly an undershot mill - letter to the author, dated 28th August, 1975.
37 For a definition of terms and a list of iron-working sites, see Andrew R. Aiano,
'Romano-British Ironworking Sites, a Gazetteer', Historical Metallurgy, xi, 2 (1977),
72-82.
33 A.C. Parker and M.C. Fair, 'Bloomery Sites in Eskdale and Wansdale (Part 1),'
CTV2, xxii, 22 (1922), 90-7; H.R. Schubert, History of the British Iron and Steel
Industry from c. 450 BC to AD 1775, (1957), 48-9, 133.
"JRS, lix (1969), 221.
40 Hampshire Field Club, Newsletter, 4 (Sept. 1976), 9 ff.; R. Goodburn, 'Roman
Britain in 1978", Britannia, x (1979), 331.
41 Britannia, ix (1978), 438; Britannia, ix (1979), 298; West Midlands Archaeological
News Sheet, 20 (1977), 33-6; ibid., 21 (1978), 69 ff. I am most grateful to Mr. B.
Phillips for bringing this site to my attention.
113
R.J. SPAIN
in their working position and that a watermill had worked close by.42
No evidence of a mill frame or wheel-pit was found, although it
should be noted that the archaeologists were not able to excavate all
areas of the site, including the stones. However, we should not
preclude the alternative explanation that these stones may have been
animal-powered. This suggestion is strengthened by the fact that the
top millstones were apparently overdriven.
Millstones, usually fragmented, occur at most watermill sites,
confirming that water-power was applied almost solely to the milling
of corn and pulses. Such finds often give the archaeologists a strong
clue as to what he may be uncovering, but care has to be exercised.
Fragments of millstones were found beside the remains of a Roman
building on the south bank of the River Mithram at Dickets Mead,
Hertfordshire in 1963, but there is doubt concerning the function of
the building which has tentatively been identified as a bath house.43
Between 1978 and 1981, an interesting second-century A.D. site
was unearthed near Littlecote Park villa at Ramsbury in Wiltshire. A
large timber building 10 by 30 m. containing a corn drier, oak-lined
tanks, quern-stone fragments and ovens suggests that a bakery
existed there.44 Outside the building another fragment of a very large
quern (millstone ?) was found and in the adjacent stream part of a
millstone. The scale of operation combined with the proximity of a
large stream makes this a potential watermill site but, as yet, no
direct evidence of water-power generation has come to light.45
Such a combination of millstones and a suitable stream in close
proximity occurred at Barton Court Farm villa near Abingdon in
Oxfordshire. In the course of digging a well of the fourth-century
A.D. fragments of four millstones were found built into the stonelined
shaft. These stones were obviously brought together for a
different function following abandonment for corn-milling purposes.
Not far away are several streams whose gradients and flows during
Roman times would have been sufficient for the generation of
water-power. However, buildings have not been located and whilst
421 am indebted to Mr. A.R. Wilmott for passing to me extracts from his
forthcoming publication, S.P. Rahtz and A.R. Wilmott, 'An Iron Age and Romano-
British Settlement outside Kenchester (Magnis), Herefordshire: Excavations 1977-79',
Trans. Woolhope Naturalists Field Club, for 1983 or 1984.
43 R.P. Wright, 'A Graeco-Egyptian Amulet from a Romano-British site at Welwyn
Herts', Antiq. Journ., xiiv (1964), 142-6.
44 Britannia, xiii (1982), 387-8; xii, (1981), 360, Fig. 16; x (1979), 329, Fig. 17.
451 thank Mr. B. Phillips for his patience and the care which he gave to my lengthy
communications.
114
<~f\
PROBABLE. MILL
I—I OUTLINE. \
POST HOLES
—1 EXCAVATED
—' AREAS
MILL STREAM
Romano-British site at Fullerton Hampshire
Fig. 7. Romano-British Site at Fullerton, Hants.
R.J. SPAIN
the collection of millstones invites the suggestion of a possible
watermill in the vicinity we have no site located.46
Corn-milling sites have been found where water-power was
obviously not used to drive millstones. Two salient features to watch
for at such sites are (i) millstones where positioned on plinths, usually
indoors, often displaced equidistant so as to provide space for men or
animals to walk around driving the stones via a lever and (ii)
millstones providing evidence of being overdriven, that is, an
arrangement where the millstone rynd cavity is in the top surface of
the runner rather than let into the grinding face. In modern corn-mills
overdriven stones are not uncommon, especially in windmills, but no
example has yet been found related to a Roman watermill. Another
feature which might be taken to indicate an overdriven millstone is
when the rynd aperture, normally a twin dovetail shape, passes right
through the top stone.47
An example of an overdriven millstone with a rynd cavity on the
upper surface of the top stone was found among millstone fragments
at Orton Hall Farm at Peterborough and confirmation of an animate
power source was provided by the remains of three stone bases found
in one of the buildings. These were made of selected stones arranged
to create plinths of 0.75 by 0.6 m., standing at least two courses high.
They were between 3.05 and 3.25 m. from each other and all three
1.5 m. from the wall of the building, strongly indicating that in this
case man-power was the prime mover.48 A very similar arrangement
existed at Silchester where one establishment was thought to have
been devoted to corn-milling on a commercial scale. In one area six
masonry plinths, each between 1.2 and 1.4 m. in diameter and 0.6 m.
high stood in two rows, each 1.5 m. from the wall.49
Sometimes the tracing of a watercourse can reveal other evidence.
This occurred when Whitehouse decided to follow, during 1964, an
ancient stream course some 7.6 m. wide and 1.2 m. deep running
away from Fullerton villa near Wherwell in Hampshire. He came
across a bifurcation and on the bank of the man-made branch the
remains of a flinty rubble platform some 5 by 7.5 m. long with traces
46 Correspondence from D. Miles of the Oxfordshire Archaeological Unit to the
author dated 11th September, 1979 and 15th October, 1979. Report by R.J. Spain,
Millstones from Barton Court Farm Villa, dated November 1979. (Oxford Archaeological
Unit).
47 Rahtz and Wilmott, op. cit., in note 42.
48 Report by R.J. Spain to the Nene Valley Research Committee, dated August
1981.
49 G.C. Boon, Silchester : The Roman Town of Calleva, (1974), 289. Millstones
found elsewhere on the site measured 0.71 m. in diameter and c. 0.19 m. thick.
116
ROMANO-BRITISH WATERMILLS
Section I
l^-IHCLIUED
SLUICE GATE
V .s~ CURVED srove
PRESENT RIVER LEVEL
1
I STEPS TO PLATFORM ; V L—S-' ^2 LEVEL OF
Q _ PLATFORM
DPOCC
CDCDB
Plan
SPLAYED
MTflKB
o
1
1
0
II II
1
II
s
.1
\
FT
M
_JC3C5<£J
a
Tq/LRflCE
|OC3000acaopCZDC^C7Cc;Oc30
J PLATFORM
cf
Possible Romano-British Watermill site at
Nettleton, Wilts.
AFTER WEDLAKE
KT» 198V
Fig. 8. Possible Romano-British Watermill Site at Nettleton, Wilts.
of seven post-holes (Fig. 7). Fragments of at least two millstones were
found here which suggests that this building was most likely a
watermill.50 Unfortunately, time did not permit a more detailed
examination of the site.
50 J.R.S., Iv (1965), 199-220; 'Hampshire Archaeologists to search for Roman Water
Mill', Archaeological Newsletter, 7 (1965), 261.
117
R.J. SPAIN
SANK
3 0 FT
IO M
TTTTrr
1MII1II.
RIVER. -> j ^ T ^ ^ ^ ^
mfirfntrIIJfirffffnfrfff(
1nu^|r|mmi1M
ROMANO-BRITISH WATERMILLS
course. Analysis of the linear and group post clusters has facilitated a
probable reconstruction of the machinery and building frame.56 This
mill frame was not identical to traditional mill hursting, for the
support functions of the bridge tree and millstone platform were not
completely integrated (Fig. 10). This small rural corn mill was
probably extended at least once during its life and modifications to its
structure have been tentatively identified as attempts to arrest
settlement. It was ill-sited close to estuarine land and suffered from
deposition in the tail race. Its foundations were laid out rather
inaccurately, the post and space-frames associated with the machinery
were surprisingly asymmetrical and the selection of the scantling
in relation to their function somewhat haphazard. All this suggests
that the military or Roman involvement was limited and the most
probable explanation is that it was executed largely by native labour
under a minimum, perhaps even an absence, of Roman supervision.
The small undershot waterwheel, 2 m. in diameter by 0.6 m. wide was
totally enclosed within the mill.
During the same summer another watermill was found and excavated
by Young, some 140 m. upstream from the other mill.57 This
proved to be a third- and fourth-century structure with a similar
earth-fast timber frame beside a revetted channel of the Stour. An
improved arrangement is evident here, for the c. 3 m. wide mill-race
was lined with planks on the bottom and sides over a length of at least
28 m. and the remains of a substantial wooden sluice gate was found
in a bypass channel, revetted with hurdles, to the south of the mill.
No machinery remains were found, except millstones, but the
building length of 13 m. alongside the stream suggests that more than
one waterwheel may have existed here. Some of the millstones found
at this site were dressed for clockwise operation and others the
opposite direction, and although alteration to the gearing arrangement
during the life of the mill could explain this, it might also
support the suggestion of two or more waterwheels in simultaneous
operation.
In addition to corn-milling considerable evidence of metal-based
industrial activity was found on this site and an interesting find was a
large iron hammer-head with mechanical deformation on one face
indicating a possible water-powered trip-hammer. Large quantities of
pewter, iron and bronze, some of it in the form of belt buckles and
56 R.J. Spain, 'The second and third Century A.D. Romano-British Watermill at
Ickham', in (Ed.) N.A.F. Smith, History and Technology, forthcoming.
57 My thanks to Dr. C.J. Young for providing me with information concerning this
and the earlier watermill. A structural and mechanical analysis and reconstruction of
the late mill is currently being attempted.
121
R.J. SPAIN
fittings, ballista bolts and lead seals, has prompted the theory that this
was an official depot or works associated with the Roman forts of
Reculver and Richborough, situated at either end of the nearby
Wantsum Channel.58 Some eighty fragments of millstones and querns
from both sites were found, most from the later mill, including three
bearing stones. No whole stones were recovered and many exhibited
varying amounts of degradation. The degree of fragmentation and
degradation, including evidence of reciprocal grinding, leads one to
suggest that these stones continued to be used long after the
abandonment of water-power.59
Although we know that the Romans apparently used water-power
for non-milling purposes,60 no mill sites have yet been discovered that
prove this beyond doubt. The late Roman mill at Ickham may have
used water-power for iron-working, but further analysis of the
evidence needs to be done to support the suggestion. When considering
other industries mining should be given some attention. Numerous
man-powered drainage wheels have been found in Roman mines
throughout the Empire61 including a fragment from a Welsh gold
mine at Dolaucothi.62 Here the Romans brought water to the mines
by four aqueducts including the Cothi of 11 km. length and the
Annell of 7 km., which together with the water-courses and storage
tanks, querns and a mortar associated with crushing and grinding, are
mute reminders of the power-intensive processes involved.63 It was
earlier thought that hydraulic methods of excavation had been
58 C.J. Young, 'The Late Roman Water-Mill at Ickham, Kent, and the Saxon Shore,'
in (Ed.) A. Detsicas, Collectanea Historica, Essays in Memory of Stuart Rigold, (1981),
32-9.
59 R.J. Spain, An Analysis of the Millstones and Quern Fragments from Ickham,
Kent, (1977), an unpublished report.
60 For a late fourth-century A.D. reference to water-power saws for cutting marble
see Ausonius, trans. H.G. Evelyn White, (1919), 1:253, bk. 10, Mosella, lines 361-64;
see also Vitruvius, On Architecture, trans, F. Grainger, (1931), Book x, chapter v.
61 Rio Tinto, Tarsis, Logrno (Spain); Leon, Tharsis (Huelva); San Domingos, Minos
dos Mouros (See note 65)(Portugal); Rudo, Verespatak (Transylvania). See R.J.
Forbes, Studies in Ancient Technology, vii (1963). The well-preserved remains of a
specimen from Rio Tinto, at the British Museum (accession no. 1889 6-22.1) is worth
close attention.
62 G.C. Boon and C. Williams, 'The Dolaucothi Drainage Wheel', JRS, lvi (1966),
122-7.
63 A.E. Annells and B.C. Burnham, The Dolaucothi Gold Mines, (undated but
between 1980-1983); see also A.E. Annells and K.P. Williams, The Dolaucothi Gold
Mines, Guided Tours, (1983), both booklets available from the Department of Mineral
Exploitation, University College, Cardiff. Royal Commission on Ancient and Historical
Monuments, 5, Carmarthen. I am grateful to Mr. Nigel Clark of Canterbury for
bringing these publications to my notice. P.R. Lewis and G.D.B. Jones, 'The
Dolaucothi Gold Mines 1: The Surface Evidence', Arch. Journ., xlix (1969), 244-72.
122
ROMANO-BRITISH WATERMILLS
practised, but this is now considered most unlikely.64 The scale of
some Roman undertakings is so large as to make it diffiult for us to
believe that they relied solely on men and animals, especially when
copious supplies of water were bought to most sites.65 At Dolaucothi
no positive evidence of a watermill has yet appeared although
symptoms exist and much archaeological work remains to be done.66
When evidence of more than one millstone is found this strengthens
the possibility that they are close to their working position, but
care should be taken to confirm, if possible, that they are not present
in a workshop, especially a smithy.67 The presence of whole millstones,
in apparently good condition, within a workshop suggests that
a new mill-rynd was perhaps being fitted or its emplacement being
deepened. Another possibility is that the grinding face was being
re-dressed. This involves re-cutting the furrows with hardened and
tempered mill bits and picks, serviced by a furnace close by. In such
circumstances corroborating evidence should be sought and by far the
most important factor would be the presence of a water-course.
When well-worn millstones are found within an ancient watercourse
this suggests that their rejection follows either fragmentation
in use or their being too thin for work. In either event the likelihood
of a mill nearby is greatly increased.
In their exploitation of native stones millers no doubt became
aware of the properties required for corn milling such as durability,
hardness, homogeneity, coarseness, porosity, etc. Such stones were
useful for other purposes as bearings, whetstones and for grinding
other materials. When the miller rejected his stones, other trades
often made use of them. The analysis of such fragments is not easy
especially with the subject confounded by the terminology peculiar to
millstones and their dressing. Essentially, the classification of fragments
depends on the identification of the surfaces and determining
whether or not their shape and symmetry have resulted from
dressing, reciprocal or rotational wear, or natural erosion subsequent
to its abandonment. Additionally, the geometry of the grinding face
furrows, handle, rynd and other cavities need definition. Even when
a fragment is identified as having evidence of rotational wear and its
64 Boon and Williams, op. cit., in note 62, 122.
65 Consider for example the scale of ancient mining at Mouras along the Gralheira
mineralized zone in Portugal; see F. Harrison, 'Ancient Mining Activities in Portugal',
Mining Magazine, xiv (1931), 137-45. Another example of large-scale working is Las
Medulas.
66 A suggestion of a possible watermill has been made by P.R. Lewis and G.D.B.
Jones, 'The Dolaucothi Gold Mines', Bonner Jahrbuch, 00 (1971), 297.
67 An example appears to have occurred at Caerwent. See T. Ashby et al.,
'Excavations at Caerwent', Archaeologia, lxii (1910), 1-20.
123
R.J. SPAIN
original size is known, it still may not be possible to decide whether it
is a quern or millstone. Their diameters and weights overlap to a
degree. The archaeologists should also be aware that querns are often
found in watermills,68 although we are not sure as to what use they
were put. Perhaps for processing small quantities of grain for some
customers, or testing grain for hardness and moisture content preparatory
to millstone work. Other suggestions are the further refinement
of a product or to provide a quality control datum for the main
milling work.
The mechanics of millstones are abstruse and the artisan skills
embodied in their design and working little understood. A satisfactory
study of the typology and evolution of Romano-Britsh millstones
has yet to be published,69 and the relationships between querns and
millstones require analysis. To facilitate these studies full records
must be kept and these stones should receive the meticulous attention
given to other artefacts.
As a valuable re-usable material, wrought iron is rarely found in
situ on Roman sites. We have much to learn concerning the design
and fabric of early watermills, but we can anticipate that iron was
probably used for wheelshaft journals, hoops (both on shafts and
gears), millstone spindles and rynds, staves in lantern gears in
addition to nails, bolts, straps and brackets. Very few mill sites have
yielded identifiable iron-work. Some iron was found in the watercourse
beside Haltwhistle, thought to be straps, but it was much too
corroded to have a positive identification.70 Ickham yielded masses of
iron which is hardly surprising in view of the site's industrial nature.
A great deal of slag and corroded conglomerates of iron were found,
some of which appeared to be collections of bolts and nails - perhaps
the raw material for re-processing.
Four iron millstone spindles have so far been found in the Empire,
three of them with two- or three-winged mill-rynds attached (Fig.
11). Two were found at the bottom of wells in Zugmantel, one of the
forts on the Germania-Superior Limes, one of the spindles complete
with a gear attached.71 These forts are high in the Taunus mountains,
68 This occurred at Haltwhistle, Ickham and the Athenian Agora sites.
69 See D. King, Petrology, Dating and Distribution of Querns and Millstones: the
Results of Research in Beds., Bucks, Herts and Middlesex, unpublished B.A. dissertation,
Institute of Archaeology, London, (1982).
™F.G. Simpson, op. cit., in note 9, 35.
71 Both spindles, complete with two-winged rynds, came from Zugmantel. The one
with a gear still attached was found at the bottom of a well of a vicus house and has
been dated to the second half of the second century A.D. I am grateful to Dr. D. Baatz
of the Saalburgmuseum for helping me with enquiries. See H. Jacobi, 'Romische
Getreidemuhlen', Saalburg-Jahrbuch, iii (1912), 75-95, 89, Fig. 43; also ibid., 'Kastell
Zugmantel, die Ausgrabungen', 54, Abb. 17 and 18.
124
Roman Millstone Spindles
to
Great
Chesterford
B
f
Silchester Zugmantel
FOS I98AFig.
11. Roman Millstone Spindles.
Romano-British Watermill Bearing Stones
zx$%
itig
I 2 INCHES
i
- r
£T CM
Haltwhistle
l\ ••.•-:Y:>-,3£te$8g^rai
SCAL£ APPROX
!^S RJi-L SIZE
r
""'v"*"«