ROMAN GOLD-MINING IN NORTH-WEST SPAIN

By P. R. LEWIS and G. D. B. JONES

(Plates xvI-xxII)

The Augustan conquest of the Asturias was resisted with all the tenacity native to that
region, but under the combined pressure of no less than three legions, this wild and
mountainous area of North-Western Spain finally capitulated in c. 25 B.c.1 On the Roman
side the prospect of mineral exploitation was a major motive that demanded at times the
presence of both Augustus and Agrippa. The literary references to the Spanish mining-
projects that followed the conquest do not specify particular sites, but indicate instead
general areas where mining was initiated.2 Fortunately, however, the gold-rushes of the
last century in California and elsewhere reawakened interest in other areas of the world, and
particularly this region of Spain, partly as a result of the legendary stories of Roman
successes. The prospectors found many traces of those efforts, although in the main
unsuccessful themselves. Part at least of what they saw was recorded in the current mining
papers and journals of that period, and we are indebted to the work of O. Davies for
abstracting and summarizing much of this information, which would otherwise be difficult
to assimilate, the sources now being unobtainable or very inaccessible. We may be sure
that the twenty or so mines that he noted are an underestimate, and that many more await
discovery. Although Davies' list was made over thirty years ago, none of the sites have
since been surveyed in any detail and no photographic record exists.
The region of interest lies to the west of Le6n, the base of Legion VII Gemina,3 in
the present provinces of Galicia, Orense and Leon. Most of the sites lie either in or close
to river beds, particularly those of the Sil, the Duerna, the Mifio, the Narcea, the Navia
and their respective tributaries. The Sil and the Miiio unite to flow south-west into the
Atlantic, the Narcea and the Navia north into the Bay of Biscay. The watershed between
the two systems, the Somiedo Mountains, forms a natural boundary. The mountains, with
the vein gold they contain, are also the original source for the alluvial and placer-gold
deposits found in the rivers that flow from them.4 We may be sure that the Romans
discovered this at an early stage, for there are records of at least ten deep mines in the
Somiedo Mountains.5 Moreover, natural weathering of the gold veins has produced a
series of secondary occurrences, either as high-level alluvial deposits or low-level placer
deposits in the river valleys. The former are, of course, geologically much older than the
placers, often having gone through several cycles of uplift and subsequent erosion. They are
thus usually easy to find, being in most cases surface deposits and easy to treat in com-
parison with vein gold, since they contain a larger proportion of free gold. From the
analogous history of the great gold-rushes in California and Australia, we might expect that
it was the discovery of gold in these ores that first stimulated Roman interest in the region.
Once they had been exhausted, prospectors would have turned to the primary sources
higher up in the mountains.
The distribution-pattern that emerges from the known sites remains that originally
charted by Davies, and is shown in fig. 23. The main division between the mining sites is
twofold: there are those that lie in the valley of the Sil and its upper tributaries, a group to
which both Montefurado and Las Medulas belong (Nos. i and 2 in fig. 23); and, secondly,

1 For the classical topography of the area, see by classical authors to minerals in Spain is given by
E. A. Schulten, Iberische Landeskunde: Geographie Schulten (Landeskunde 223; Spanish ed., vol. ii) and
des Antiken Spanien (I955), also published in a 0. Davies, Roman Mines in Europe (I935), 94 f.
Spanish translation (I959). For the Augustan cam- Cf. also Sil. Italicus I, 228 ff.
paigns, see D. Magie, ' Augustus' War in Spain', 3
I. A. Richmond, ' Five Town Walls in Hispania
Class. Phil. xv (I920), 323 ff., further discussed with Citerior ', RS xxi (1I93I), 91 ff.
some topographical modifications by R. Syme, 'The 4 For a brief discussion of the nature of gold
Spanish War of Augustus ', Am. J. Phil. LV (I934), deposits, v. P. R. Lewis and G. D. B. Jones, ' Dolau-
293 if., and E. A. Schulten, Los Cantabros y Astures cothi Gold Mines I: the Surface Evidence ', Ant. J
y su guerra con Roma (Madrid, 1946). See also now XLIX (I970), 244 ff. A more extended discussion is
A. Brancati, Augusto e la Guerra di Spagna, Pubblica- given by M. Maclaren, Gold, its geological occurrence
zione dell'Universita di Urbino xvii, I963. and geographical distribution (I908), passim.
2
Pliny, NH xxxiii, 67-78, fully translated and 5 Davies, op. cit., 103.
discussed in the appendix. A list of general references
 , .

I70 P. R. LEWIS AND G. D. B. JONES

those that straddle the watershed of the Somiedo Mountains. To the sites noted by Davies
have been added several where the modern Spanish name (such as Rio del Oro) clearly
implies the working of placer-deposits. These sites should not be accepted uncritically as
representing simply the sum total of gold mines. Very few mines yield a single mineral or
ore, most produce a spectrum of several minerals. The detailed attribution for each particular
mine can only be worked out by individual examination on the ground.

:, : ;
, * .. *
..................'....;...-..........'..
l

.'...: 6.::.:
'T. ,, LEG
'--'........,. ~,.~~,'.::.!...::::~~,?
..).l....!.~:
. .:'..-/..-
<'" ., . 4TO

A1torsa

; 2 *'* *'.'-
....
MONTEFURADO
;. . .: '
:::::::::::::::::::::::::
' . . Cae . .
::..:.
.* . *
:. * . .
.J
^**y. -v-,"'.--.'-' .' --.-' - : *-, .; -.
? ; '. ', ;-- .
--.-- -- ".-.'- '" .'.- ' '.'.-.
": .'.'..
* So * -.* - . *. . ': * *oo
*<7;. - ' . ' 600 *.* * '**- - ** . - . *
- *.* I . ... nt 7.
.. . * .* . * . * *

FIG. 23. DISTRIBUTION MAP OF ROMAN MINING SITES IN NORTH-WEST SPAIN. THE CONTOUR IS DRAWN AT THE 800 M
LEVEL. NOT ALL THE SITES ARE GOLD MINES. TIN, IRON AND COPPER WORKINGS ARE KNOWN AS INDICATED.
Key: (I) Montefurado, (2) Las Medulas, (3) Paradeseca, (4) Menival, (5) Anclares (mines along the valley of
the R. Anclares), (6) Burbia (mines along valley of R. Burbia), (7) Duerna Valley (17 mines, including Los
Castellones), (8) Rio Boeza, (9) Murias, (Io) Paramo del Sil, (II) Salientes, (12) Pontedo (copper), (13) Villamanin
(copper), (14) Pefialva, (I5) Milagro (copper ?), (16) Ablaneda, (17) Sierra Aramos, (i8) Ortguera, (19) Pravia
(lead), (20) Trevias, (21) Fornones, (22) Vegalagar, (23) Puerto del Palo (= Cueva de Juan Rata ?), (24) Valle
del Oro, (25) Naraval, (26) Carcobas de Miudes (iron), (27) Slabe (Castropol) (tin), (28) Via de Foz (copper),
(29) Rio de Oro, cf. no. 24, (30) Punta de la Estaca de Vares (iron ?), (31) Valdovinos, (32) Ferrol, (33) Monfero,
(34) Carballino, (35) Lor Valley.
Drawn by the authors. Copyright reserved

INTRODUCTION

Our intention is to provide a survey of three mines, Montefurado, Las Medulas and
Puerto del Palo, that are geological type-sites. From a reconnaissance-visit made in I967
by one of us (P. R. L.), supplemented with more recent work by both authors and Mr.
D. G. Bird,6 and from comparable evidence now available from Britain, we now know that
the sites exemplify the three main types of mine, i.e. placer, alluvial and hard-rock, a point
that did not emerge clearly in previous scattered references to the sites. Las Medulas may
certainly be regarded as one of the best preserved alluvial mines of the Roman period.
Montefurado proved to be more complex, showing elements of placer and vein mines that
6
Mr. Bird is currently undertaking a compre- postgraduate thesis.
hensive survey of the mining sites of the region for a
 ROMAN GOLD-MINING IN NORTH-WEST SPAIN 171
have been heavily re-worked at a later period. A similar situation is known at Dolaucothi,
the only Roman gold-mine so far known in Britain; originally thought to be solely a deep
mine,7 it too has been shown to preserve extensive traces of placer-mining.8 Las Medulas
remains a good example of a high-level alluvial mine that never developed the secondary
characteristics typical of Montefurado and Dolaucothi. By contrast, Puerto del Palo presents
an excellent example of an opencast vein mine operating at an altitude of over I,ooo m; like
Las Medulas it was never re-worked after the end of Roman activity. Whatever their
complexities, these must all represent major sites in comparison with the others so far
known in North-Western Spain, if one may judge from the size and elaboration of the
hydraulic systems at Las Medulas and Puerto del Palo, and from the placer-mining tradi-
tion inferable at Montefurado.
The exploratory survey of the three sites involved fieldwork techniques originally
developed at Dolaucothi in South Wales, but which could be applied in Spain, mutatis
mutandis. Corroborative excavation has not been attempted at the Spanish sites, although
at Dolaucothi the evidence from excavation confirmed and supplemented the information
already available from fieldwork. In Spain, however, fieldwork had to take first place, for
a number of reasons. The literature on the Spanish mines is diffuse and repetitive, being
partly based on local folklore rather than the actual remains. Moreover, the visible features
in Spain tend to be related in a much more obvious way to local geology and geomorphology
than those of the British mine.
The discussion of the sites is divided into two parts: description of the primary
remains and of the topographical setting followed by detailed interpretation based on the
internal logic of each site, as well as on the analogous evidence from elsewhere. Finally,
the theory and practical techniques involved are discussed, and the possibilities for future
work.

MONTEFURADO

Description. About a kilometre above the point at which the Rivers Sil and Bibey join
stands the small village of Montefurado. The village itself lies within walking distance of
the group of features and remains from which it takes its name, the gold-mine of Monte-
furado. The immediate topography of the site is shown in pl. XVI and fig. 24.
The mine lies on and around a double-swing meander somewhat more sharp than those
usually encountered at irregular intervals in the river beds of this region.9 Whether or not
local geological factors have been solely responsible for the formation of this feature remains
unclear owing to lack of detailed information.10 Broadly, the bed-rock (as can be seen clearly
in pl. XVII, a, b, c) consists of highly distorted strata of Cambrian slates and mudstones,
which at several points have been subjected to strong mineralization. The village lies just
opposite the undercut slope that dominates the first inswing of the meander, as seen in the
centre of pl. XVI. As is apparent from pl. XVII, a, b, the river runs in a fairly deep gorge
within its own bed from a point somewhere below the two dams (A and B on pl. XVI)
which are situated just at the start of the slip-off slope of the spur. The level of the water
is abnormally low in these plates, partly because of climatic conditions (they were taken in
late summer), but primarily because of the current programme of hydro-electric schemes
in the region.
The second spur of the feature, La Vega, consists of a steep promontory which at its
highest point rises some 65 m above the river bed. The neck of the spur contains the
primary auriferous intrusions (pl. XVII, a), at which point the rock face falls almost vertically
6o m to the river below. The south-western side, by contrast, has been severely gashed
7 A recent general survey by W. H. Manning, an Introduction to the Study of Landscapes (I939).
Antiquity XLII (I968), 301, over-emphasized this A broad survey of the physical geography of North-
aspect, following G. C. Boon and C. Williams, West Spain is available in M. Terain, Geografid de
JRS LVI (I966), 122, n. 6. Espana y Portugal (Barcelona, 1952) I, ch. I2.
8 Lewis and Jones,
op. cit., 246. A more detailed 10A description of the geology in the province of
discussion is given by Jones and Lewis, Carmarthen Le6n is given by J. A. Jones, Trans. Fed. Inst.
Antiquary vi (1970). Mining Eng. xx ( 900-I), 420 if. It is not known if
9 Geomorphological terms, if not immediately any more detailed geological surveys of the region
obvious, may be found in several textbooks currently have been made subsequently.
available, for example A. K. Lobeck, Geomorphology:
172 P. R. LEWIS AND G. D. B. JONES

from a point about 50 m above the line of the river bed, and consists of a sheer, if not over-
hanging, rock face. This western outcrop is largely pyritic, while virgin gold ore appears
to outcrop further east (pl. XVII, c). This forms the most striking evidence of bed-rock
mineralization. The river at present flows through a gap at the base of the escarpment,
which is shown in detail looking downstream in pl. XVII, b. The rock arch is clearly of
some antiquity, hence its survival in the name Montefurado (mons foratus, the pierced
mountain) and in its second name, Boca do Monte, the hole in the mountain. The cut-off
has thus isolated a considerable portion of the meander, of which the larger part consists of
dried-up river bed (shown at right in pl. XVI), although there is a small oxbow lake at the

FIG. 24. GENERAL PLAN OF THE MONTEFURADO SITE. T = TUNNEL

Drawn by the authors. Copyright reserved

farthest end of the loop (the tip of the lake can be seen in the extreme right of pl. XVI, a).
The old river bed lies some 8-9 m above the present river gorge.
The most obvious archaeological feature of the site is the deep canal (c. 3 m deep by
3 m wide) that runs from the pair of dams, closely following the contours of the slip-off
slope, and eventually disappears into the old river bed (fig. 24). Of the two dams the upper
one, dam B, is of recent origin.1l Its massive slate-and-concrete construction contrasts
markedly with the remains of another dam (A) slightly downstream. The latter is relatively
slight in construction and consists simply of a coarse concrete matrix.
Apart from one section in the bend of the inswing, where river scour has revealed its
full section, most of the canal is now choked with detritus. Approximately 40 m beyond the
point where a breach has occurred in the outer wall of the canal, the channel disappears
11 Probably built at the end of the last century by tective overlay of river debris (now removed) may
a German company based on Ponteferrada: M. best explain the survival of this feature in its founda-
Breidenbach, Zeit. prakt. Geol. 1893, I6. A pro- tion courses.
 ROMAN GOLD-MINING IN NORTH-WEST SPAIN I73
altogether into the old river bed. But fortunately one onward continuation (on a line
rather different from the former alignment along the river bend) can be traced as a line of
stones bonded with concrete, that projects a few centimetres above the present silt level.
This leads to a tunnel that has been driven straight through the rock-face of La Vega, about
20 m south of the river-gap described above; the size of the arched tunnel (c. 2 m wide by
3 m high) compares with the dimensions of the canal. It is I80 m long and contains shot-
holes, a feature that was not found anywhere along the canal. If the tunnel was designed to
explore a southward branch of the principal area of mineralization, then no veins of
significance were found, and very little scour is evident at the western exit. However, to
the south the main branch of the canal must have continued some way along the edge of
the dried-up meander. At one point here, the derelict washing tables of nineteenth-century
exploitation can be seen (pl. XVI, b). No clear evidence for an earlier period was traced
in this area, but placer-pitting along the banks of the meander suggests that similar working
of placer-deposits had occurred in an earlier phase.

Analysis. The features described present three problems. First and foremost, how can the
ancient phases of mining be recognized with any confidence ? Second, how are the extant
archaeological remains related to the visible geomorphological features ? Last, does the
interrelation of the two kinds of evidence gives us a coherent picture of the way in which
exploitation proceeded ?
The mine does not at present exhibit any direct evidence of working in Roman times,
such as an inscription or graffito on a rock-face. The indirect evidence, however, is strong.
Montefurado was known in local tradition prior to any modern attempt to re-exploit it.
Moreover, among the mining areas of the Sil valley, Montefurado appears to have been one
of the richest and could therefore hardly have been ignored. The explanation given by
Beuther12 and repeated by Davies 3 for the tunnel-which has tended, like the underground
features at Dolaucothi, to monopolize attention-seems obscure and unsatisfactory. The
difficulty lies in a confusion of the archaeological and geomorphological features, which
must be analysed separately before any reconstruction. Davies gives two possible explana-
tions for the tunnel: either it was used to divert the river in order to attack deposits in its
bed at unspecified points, or it was used to bring a supply of water to a mine lower down the
valley. The latter suggestion can be eliminated immediately, for there is no trace of a
continuation of the canal at any point beyond the tunnel exit, and there are no known workings
between this point and the confluence of the Sil and Bibey, a full kilometre away.
The occurrence of primary mineralization in the neck of La Vega and the gash which
occurs in an upstream direction strongly point to opencast working of the rock-face. While
at first glance one might suspect a natural origin, the depth of the gap, its direction and the
relatively rapid rate of erosion to be expected from a young river all argue for an artificial
origin for the present course. Once the rock-gap through which the Sil runs is understood
to be a man-made cutting, then it is possible to explain how the site has come to have its
present shape and appearance. As opencast working from the rear of La Vega progressed,
the point when collapse of the eastern wall (preceded by extensive seepage through the
mudstone) occurred would eventually be reached. The situation would have been even
more critical if the richer veins were near or even below river level, as implied by the depth
of the river gorge. For these, a coffer dam of some kind would have been essential; it
could not, however, have survived the winter torrents for which the Sil was notorious
before it was regulated by the present hydro-electric schemes.
While the development of an opencast on the downstream side of La Vega can thus
be postulated, there are problems concerning the remaining features. Neither of the two
dams is demonstrably ancient, but if the Roman miners worked the opencast through to
the upstream side of La Vega and attacked the actual bed of the Sil (as seems to have
occurred from the 8-9 m drop in level, p. i72) then a dam must be postulated upstream at
roughly the same point as the surviving examples. While the straight tunnel through La
Vega exhibits shot-holes, the canal itself does not. Likewise, as already suggested (pp. 172-3),

12
W. Beuther, Zeit. Berg-Hiitten-Salinenwesen 18 Davies, op. cit., 102.
preuss. Staat. xxxIx (I891), 55.
I74 P. R. LEWIS AND G. D. B. JONES

the canal does not lead directly to the tunnel and the lateral feeder channel may be secondary.
Possibly, therefore, the various features are not contemporary, but the decisive evidence
now lies buried beneath the tons of silt choking the old river bed. By itself the canal is a
useful reminder of the major role of placer-mining in the Montefurado complex.14 Thanks
to the nineteenth-century workings, it is impossible to point to placer-workings of specifically
Roman date, much of the evidence for which would, in any case, disappear with each
winter's flooding. Indeed, upstream every major meander of the Sil exhibits the tell-tale
rock piles of placer-workings. Although many of the details escape us, at Montefurado we
are dealing with a more sophisticated stage in which Roman miners worked not merely the
placers but also traced and attacked part of the parent ore body from which they derived.

LAS MEDULAS

Description. About 40 km from Montefurado up the Sil valley lies one of the largest and
most distinctive outcrops of the quaternary alluvial deposits of the region. The exact
extent of the formation has not been charted,l5 but the characteristic bright-red deposits
may be seen either as a terrace along the Sil basin and other adjacent valleys, or as isolated
masses in the case of Las Medulas. The formation here is very extensive, lying at its
highest point c. I,ooo m above sea level, and stretching for some kilometres around the
village of Las Medulas. The salient topography is shown in pl. XIX, a, looking north-west,
the village lying at the extreme left of the panorama. The fantastic ' badland ' weathering,
and particularly the knife-edges and pinnacles that have formed in front of the retreating
scarp to the right, illustrate well the erosive nature of the deposit.'6 The older slates upon
which the deposits rests can be seen in the lower plate (XIX, b), taken at a point 6 km to the
south-east on a tributary of the Sil, the Cabrera. The site lies just over the ridge to the
extreme left beyond the village of Pombriego in the foreground.
Closer inspection of the escarpment, 30-40 m high (pl. XIX, b), shows that it consists
of numerous irregular bands of fine sand and clay dispersed in coarse conglomerate. One
result of the weathering process is evident in the caverns existing in its face (pls. XVIII, b;
XIX, b). It may be suggested that the formation of caverns, and their enlargement by
internal erosion, followed by roof-collapse, eventually gives rise to the much-dissected
'badland' scenery.'7 In such a deposit, placer gold may be expected to occur only in
certain bedding-planes. Much of the deposit, whatever the exact proportion, may have
had no gold at all, and merely constituted waste overburden.'8
The extant archaeological features were noted several times at the turn of the century
by mining engineers, and have been summarized by Davies.19 Of these accounts those of
Del Mar 20 and Longridge 21 (both extensively quoted by O'Reilly 22) and lastly J. T. Jones 23
are probably the most perceptive. As for Montefurado, there is little earlier information in
Spanish sources.24 The date of the site is attested by the Roman artefacts discovered by
Longridge and the find of a Neronian aureus in the course of a sluice.25
The most important surviving remains are the three large aqueducts, 2-3 m broad,
entering the site from the south-east at about the 950 m level (fig. 25). The lower two
systems skirt the southern ridge of the deposit (between spot heights 1,012 m and 973 m),
ending in the series of tanks and reservoirs (shown in fig. 25). Judging from the numerous
large and small opencasts in the area, it is clear that a considerable portion of the ridge has
been removed by hydraulic action (pl. XXI, a, b). The interpretation is reinforced by the

14M. Maclaren, op. cit., and R. W. Paul, Cali- techniques by the good offices of Dr. J. M. Anketell,
fornia Gold: the Beginning of Mining in the Far West Dept. of Geology, Manchester University.)
(University of Nebraska, 1947), I53 ff., explain the 19op. cit., I02.
practice of hydraulicing placers. 20 A. Del Mar, Australian Mining Standard,
15 J. A. Jones,
l.c., fig. i. 25th April, I906, 399, concluded in issue of 2nd May,
16 R. W.
Fairbridge ap. The Encyclopedia of I906, 429.
Geomorphology (Reinholdt, i968), s.v. Badlands. 21 C. C. Longridge, Mining Journal, 29th January,
17 Lobeck, op. cit., 388 ff., gives an extensive I898, 139.
discussion of the subject of loess weathering by a 22 S. O'Reilly, Proc. Roy. Irish Acad. Ser. III,
similar mechanism. vol. vi (1901-2), 63 ff.
18 A sample taken from the roof collapse of the 23 loc.
cit., 427.
cavern shown in pl. XVIII, b showed zero trace of 24 For
example, Revista Minera XLIV(i893), 36.
gold. (The analysis was performed using X-ray 25 Del Mar, loc. cit.
 ROMAN GOLD-MINING IN NORTH-WEST SPAIN I75
existence of an extensive group of tanks sandwiched between the two largest opencasts
directly below spot height 973 m. All the tanks at this point compare well in overall dimen-
sions with the more easterly ones, not exceeding I00 X 20 m in size and roughly I 5 m
deep. In several cases the stone sluice-gates are very well preserved, and in at least one
example (the highest tank below S.H. 973 m) the inflow channel can easily be traced. This
set of tanks is now completely isolated from the primary inflow aqueducts, as illustrated in
pl. XXI, a. The photograph was taken from the ridge overlooking the middle opencast to the
east, and the two aqueducts can be traced along the hillside shown in the right-hand side
of the plate. The most westerly open-cast is seen in pl. XXI, b, with the corner of the scarp
in the extreme left-hand corner. The series of cone-shaped rock piles along the edge of the
working is very similar to some features observed at the mine of Los Castellones in the
Duerna valley.26 Finally, the scarp face of the middle opencast contains a series of short
tunnels (many now in an advanced state of collapse), a northward-facing example of which
can be seen from the modern village below. The interior of the tunnel presents a two-phase
sequence, one tunnel having been driven directly over an earlier adit. The appearance of
the lower tunnel presumably gave rise to erroneous reports of aqueduct-channels within
the deposit.27
The aqueduct that follows the northern side of the ridge is shown in pl. XX, a, at a
particularly interesting point where several phases are involved. The earliest period is
represented by the aqueduct-ledge at the extreme right of the picture, which is now cut
by a modern track running almost parallel to its course. The ledge ran to a long 'hush-
gulley ' that dropped down the north face of the ridge (fig. 25). The same technique was
used at a later stage as shown by the two deeply-incised gullies running from the aqueduct
course in the right of pl. XX, a. The final phase is represented by the superimposed course
of a lower aqueduct in the deeper of the two ravines. It is clear that this system fed the rest of
the site to the north with an adequate supply of water. The aqueduct can actually be traced
as a line of stones in the modern track leading to the village of Orellan. By contrast with the
collecting systems on the south ridge, the corresponding reservoirs at this point (A) and
further north (B) are much larger, measuring 200 m x 40 m x 3 m and i60 m X 40 m
x 5 m respectively (fig. 25). The latter tank was fed from tank A (shown in the right fore-
ground of pl. XX, a) by an aqueduct circling spot height 951 to the rear of the main escarp-
ment. Long, deep gulleys run down from each tank, the more impressive series being from
tank B where some of the channels follow the contours to the edge of the opencast (fig. 25).
The tunnels of the south ridge are paralleled by similar features in the main escarpment,
particularly by a tunnel under tank A leading directly to the face beneath tank A, and others
radiating inwards from caverns elsewhere in the face of the main opencast. In cross-section
the arched tunnels normally measure roughly 2 *5 m high by 2 m broad. Further traces of
mining are not evident.
Longridge and Del Mar noted that the site was served by a series of aqueducts of
varying length (claimed to be between 25 and 40 kilometres) but unfortunately left no
details. The aqueducts number in fact no less than seven, the largest group associated
with any Roman mine. The complete series are intermittently visible as lines on the upper
slopes of the Cabrera valley in pl. XIX, b, cf. fig. 25 (upper). There is a height difference of
c. 400 m between the upper and lower systems, that must be related to the development
stages of the mine itself. It seems clear that the upper aqueduct never reached the site at all,
being rather designed to explore or work deposits on the slopes of Cruz de la Peiia east
of Las Medulas. The next three systems, however, served the site in the form in which it
survives, and have already been described. The three lower aqueducts were probably
intended to explore and work the lower slopes of the south ridge, but mining activity at the
higher level has obliterated their final destination.
Although the aqueducts have not been surveyed in detail, local tradition suggests that
the sources lay in the area of Castrillo de Cabrera in the valley of the Cabrera proper, over
twenty kilometres distant from the mine. There is no reason to doubt this report, which,

26 Recently surveyed by Mr. R. F. Jones of 27Davies, op. cit., I02.

Manchester University, in collaboration with Mr.
D. G. Bird.
 .

I76 P. R. LEWIS AND G. D. B. JONES

allowing for contours, would give the aqueducts a length of over fifty kilometres. P1. XIX, b,
shows the aqueduct channels above and north-west of Pombriego. The most spectacular
section observed lies six kilometres further east and below the summit of Guiana (1,848 m),
where the Cabrera valley has been deeply indented by a tributary. P1. XX, b, shows some of
the aqueduct channels running across cliff, scree and mountainside. Nothing could better
illustrate the spectacular problems of aqueduct construction described in Pliny, NH xxxIII,
75, and discussed in App. I (p. I83).

Analysis. The problem presented at Las Medulas is in one sense much simpler than that at
Montefurado. There can be very little doubt that the extant archaeological features are

Carucedo '. ' .'.'.'. ..

j '-'.' MONTES
L AQUILIANOS

Las
c. Medulas . '
'. ' ' ?'.
...
:,.':"'":' . . ' . . ..' '
.'Gu iana
r. /.. MULTIPLE . . . . 194r.

.
I-',d9 . d
..M ..... _____c ? ' ?. .'

FIG. 25. LOCATION MAP AND GENERAL PLAN OF THE LAS MEDULAS MINE

Drawn by the authors. Copyright reserved

 -

MONTEFURADO
VILLAGE

~~,"-
,-- - \

,CE ^\- j WASHINGTABLE

. -.^.-,,.:_
'--
'. .--_-- " ~ ~
--' S /_.:_I :~ ; ,
:.l
? -
ROCK~/
CUT?
r
'. '.'-'
CANAL ..~. _ . . ~..__ .

a, b, PANORAMA AND EXPLANATORY DIAGRAM OF THE SIL VALLEY AT MONTEFURADO LOOKING FROM EAST TO SOUTH

Photograph by authors. Copyright reserved

MONTEFURADO: (a) GENERAL VIEW OF THE SIL SHOWING THE RIVER TUNNEL AND OLD RIVER BED, TAKEN FROM NORTH-EAST. (b)
UPSTREAM (EAST) SIDE. (C) THE RIVER TUNNEL AND OPENCAST FROM THE DOWNSTREAM (WEST) SIDE SHOWING THE PRIMARY

Photographs by authors. Copyright reserved

JRS vol. LX (1970) PLATE XVIII

(a) MONTEFURADO: DETAIL OF THE RIVER TUNNEL SHOWING THE ROCK-CUT CORNER ON THE EASTERN SIDE.
(b) LAS MEDULAS: PART OF THE ESCARPMENT IN THE MAIN MINE SHOWING HORIZONTAL BANDING OF CLAY, SAND
AND GRAVEL (see p. 174 ff.)

Photographs by authors. Copyright reserved

LAS MEDULAS (see p. I74 ff., fig. 25): (a) PANORAMA OF THE SITE, LOOKING WEST. (b) TRACES OF AQUEDUCT CHANNELS (ARROWED) ON
THE MINE LIES OVER THE MOUNTAIN RIDGE AT THE EXTREME RIGHT

Photographs by authors. Copyright reserved

JRS vol. LX (I 970) PLATE XX

LAS MEDULAS (see p. I74 ff., fig. 25): (a) THE MAIN AQUEDUCT LEADING IN FROM THE SOUTH-WEST. THE GULLIES
LEADING FROM AQUEDUCT LEVEL REPRESENT DIRECT HYDRAULICING OF THE SOFT AURIFEROUS DEPOSIT. THE REMAINS
OF TANK A CAN BE SEEN IN THE FOREGROUND

(b) THE LINES OF THREE ROCK-CUT AQUEDUCTS (ARROWED) SERVING LAS MEDULAS ON THE SLOPES BELOW THE PEAK
OF GUIANA (see fig. 25). TELEPHOTO VIEW FROM SOUTH-WEST

Photographs by authors. Copyright reserved

LAS MEDULAS: (a) GENERAL VIEW OF THE WESTERNMOST OPENCAST ON THE SOUTH RIDGE. THE AQUEDUCTS SERVING THE WORKING
UPPER RIGHT. (b) ROCK PILES FROM WASTE CLEARANCE AT THE FOOT OF THE SMALL OPENCAST BELOW SPOT HEIGHT 973 METRES,

Photographs by authors. Copyright reserved

PUERTO DEL PALO (see p. 178 ff., fig. 26): (a) GENERAL VIEW OF THE MAIN OPENCAST FROM THE SOUTH. A HUMAN SCALE IS SHOWN ON TH
MONTEFURADO (NR. LAGO) ALONG THE VALLEY OF THE RIO DEL ORO. THE MAIN AQUEDUCT SERVING PUERTO DEL PALO IS VISIBLE ON TH
AT PUERTO DEL PALO WITH (ABOVE) THE TWIN RESERVOIR OF THE NORTHERN AQUEDUCT

Photographs by authors. Copyright reserved

PUERTO DEL PALO: (a) GENERAL VIEW OF WESTERN RIDGE. THE AQUEDUCT EXTENSION TO TANK C (see p. I78 ff., fig. 26) CAN BE SEEN AN
RUNNING DOWN THE SLOPE AND CROSSING AN EARLIER AQUEDUCT LINE. (b) THE INTERIOR OF TANK A. (c) THE OUT

Photographs by authors. Copyright reserved

 ROMAN GOLD-MINING IN NORTH-WEST SPAIN I77
Roman in date, and probably belong to the early imperial period on the evidence of the
Neronian aureus. There is no evidence of later working to the scale implied by the aqueduct
systems. The site is an archaeological fossil formed at the moment when the aqueducts
ceased to flow, for the height of the Medulas site prevented later large-scale exploitation
unless conducted by hydraulic means on a similar scale. For any assessment of the mine,
the amount of natural weathering after the Roman abandonment therefore becomes of
paramount importance.
The problem of reconstructing the site in its pre-Roman state remains daunting,
primarily because of the difficulty in assessing how much material has been removed by
hydraulic mining in the Roman period (v. App. II). It is apparent that the c. 34 million
litres of water a day delivered by the three major aqueducts was used in at least two distinct
ways.28 First, it was played directly on to the deposit itself in a continuous stream, in exactly
the same way as ground sluices were used in the nineteenth century.29 The stream has thus
cut the numerous gullies in the northern half of the site, as already described. This type of
work is known as hydraulicing, by analogy with the modern practice of mining metalliferous
gravels and placer-deposits with jets provided by a head of water.30 Secondly, the numerous
smaller tanks on the south ridge were used in the practice of hushing, or the use of a wave of
water released from a tank, as described by the Elder Pliny and discussed in App. I (p. 183),
to remove waste overburden, a particular problem on the site of Las Medulas. The tanks
are perched on the lip of the opencasts, in ideal positions to work the deposits below.
Certainly the character of the gold deposit at Las Medulas excludes alternative uses, such
as driving simple machinery. This probably did happen in the different geological circum-
stances of the Dolaucothi mine,31 but at Las Medulas an efficient series of sluices would
have been quite sufficient to process the auriferous effluvium from the hydraulicing.
From the evidence at present available, it is possible to suggest a three-period sequence
with several sub-periods for the overall mine development. This does not, of course, take
account of the earlier exploration or working at a lower level that is implied by the three
lowest aqueducts (p. 175). (i) The south ridge being more directly accessible, working
probably started at the western extremity, using the two aqueducts already described.
Following exhaustion of this part of the deposit, the western hydraulic system was abandoned,
and the middle part of the ridge attacked by means of a second set of tanks, thus isolating
the earlier features. (2) A third aqueduct was constructed in the second period to attack
the northern slope of the south ridge, not easily accessible from the other side, although
small-scale working had perhaps been attempted. The same method of working back along
the ridge appears to have been adopted, although ground sluicing rather than hushing was
used. (3) The final period was probably the most difficult of all, because of the problem of
providing a levelled water supply from the Cabrera valley. Since this was clearly impossible,
a compromise was tried which involved tapping the aqueduct via the lower gulley (shown
in pl. XX, a). At some stage attempts may have been made to construct a lower aqueduct,
as is suggested by the discontinuous line that appears below the main aqueduct-level.
Whatever the exact method, it became possible to attack the northern part of the site,
where the provision of an adequate head of water was of great importance. For this reason
reservoirs A and B are of a quite different scale and style compared with the smaller ones
on the south ridge, and were partly designed, perhaps, to collect run-off rainwater.
From comparable data on hydraulicing operations elsewhere in the last century,32
and the calculation of the amount of water delivered by the comparable Dolaucothi aque-
duct,33 it is possible to estimate a minimum time-span for each phase. The full details are
reserved for an Appendix (pp. 184-5). The three opencasts that lie along the aqueduct line
on the south slope of the south ridge took at least eighteen years to be excavated; the second
period probably lasted no longer than a season or so, judging from the size of the accom-
panying gullies; the last period must have extended for a considerable time owing to the

28 The Elder 31 Lewis and Jones, loc. cit., 258.

Pliny classifies such hydraulic gold-
mines as alutiae in a passage relating to lead: 32 C. C. Longridge, Gold and Tin Dredging (I914),
' (plumbum) invenitur et in aurariis metallis quae 233 f.
alutias vocant', NH xxiv, 157. 33 G. D. B. Jones, 'The Dolaucothi Aqueduct',
29 R. W. Paul, op. cit., Bull. Board of Celtic Stud. xix (196o), 71 if.
I5I ffi.
30Maclaren, op. cit., pl. 25.
 A

178 P. R. LEWIS AND G. D. B. JONES

different style of working. Earlier observers 34who focussed attention on the main' badland '
area below tanks A and B (pl. XIX, a), suggested that the whole complex was an artificial
creation. Yet no fossilized hydraulic features were found in the isolated masses, and this
negative evidence points to a largely natural origin for this part of the site.
PUERTO DEL PALO

Description. The repetition of the name Montefurado was the clue to the location of another
site that is even higher and rivals Las Medulas in its spectacular character.35 The geology
of the site is very different. It lies on the northern watershed of the Asturian mountains,
amid the upper tributaries of the Rio Navia. One major tributary in Oviedo province is
the Rio del Oro; it flows in a half-circle from the watershed known as the Puerto del Palo,

? , [.-

PUERTO DEL .. . -
PALO ,..>.-,,L:
GDBX *. .. ; } , OR ,

..'* ' " ''

; i'-'', : ..fr - '".' i' '

* '
/ ' . '<: 'j
1. ... . ...
'
( " " ' "''
_ .ONTEFUAO

FIG.26. GENEAL PLAN OF THE PUERTO DEL PALO MINE

Drawn by the authors" Copyrghtreserved

-Drawn
by theauthors. reserved

34 Notably Longridge and Del Mar.

company of Mr. D. G. Bird, to whom we are grateful
36 The site was examined by the authors in the for comments and information.
 ROMAN GOLD-MINING IN NORTH-WEST SPAIN I79
which divides the river systems of the Navia and Narcea. Its name, Rio del Oro, is a possible
indication of early mining, and there is a tradition of Roman activity in the valley.36 On
this score nothing can be now established, for both this tributary and the Navia valley have
been flooded to form a reservoir. The modern route to Puerto del Palo climbs eastwards
from the reservoir at Grandes de Salime reaching the g99 m ridge on which rests the small
village of Lago, three kilometres short of the final ascent over the Puerto del Palo at c.
1,I70 m. (fig. 26). At the foot of the final climb lies a feature traditionally known as Monte-
furado. As its name implies, it is a narrow tunnel cut through the ridge along which the
present road approaches the pass. The small, level tunnel is surely an aqueduct channel
rather than any kind of mine adit, and many parallels to it in size and function could be
found amongst the cuniculi or aqueducts of the Roman campagna.37 The Montefurado
tunnel serves the lower of two aqueduct systems that can fairly be claimed to represent one
of the most outstanding examples of hydrological engineering known in the Roman world.
The headwaters of the Rio del Oro consist of two tributaries, flowing in from north and
south in diametrically opposite directions, presumably along the line of a geological fault
(fig. 26). The southern branch, running from the peak known as Orrua, is the highest
substantial stream in the Navia-Narcea watershed, rising at an altitude of over I,200 m. The
major aqueduct (pl. XXII, b) tapped the headwaters of this southern tributary and, passing
over the ridge of the Montefurado tunnel, carried a head of water from the Rio del Oro
valley on to the western flank of the Sierra del Palo. The minor aqueduct flowed through
the Montefurado tunnel, following a parallel but shorter course at a level of about 20 m
lower. The tributaries of the Rio Pumarin on the western side of the mountain only reach a
height of 900 m. So carefully engineered an aqueduct system, which tapped the water from
one river valley and passed it into another, must have been designed to serve a mine at the
1,000-1,100 m level, i.e. at too great a height to be served by the Rio Pumarin system.
The mine itself, which is to be associated with the traditional name of Cueva de Juan
Rata,38 can be located by the white mark of a vertically bedded quartz outcrop at the
north-western tip of the Sierra del Palo. In geological terms, it probably forms part of a
north-south strike of the primary quartz outcrop that runs across the Asturian mountains.
Another such outcrop can be seen at about the same height of i,ooo m over I2 km to the
south-east. The mine itself consists of a major opencast and a minor one immediately to the
north. Past its foot there ran another aqueduct, flowing southwards from the highest branch
of the Rio Pumarin. It appears to be smaller in scale than the Rio del Oro aqueducts, but
more detailed work is required before a definite description can be given.
The main opencast is so massive in scale that there is an initial temptation to dismiss
the site as a natural formation (pl. XXII, a). But the presence of an exceptionally well-
preserved hushing-tank (A), just beneath the southern summit of the opencast (pl. XXIII, b)
makes this impossible. The opencast proper measures just over 200 m vertically.
While the floor of the lower part has been covered by a deposit of silted material
(pl. XXII, c), the upper half is a nearly vertical wall of highly-weathered quartz that reaches
up to the summit of the mountain ridge, the rock becoming somewhat looser on the
southern side. It is there that the hushing tank (A) is located, 55 m long by 5 m wide, with
the walls standing over 3 *5 m high. The profile of the tank has been exceptionally well-
preserved, since the altitude ensures a thin vegetation (the top of the mine area is under
snow throughout the winter months) and a lack of erosion (pl. XXIII, b). The tank is set
back c. 15 m from the present edge of the opencast, and its outfall shows a clearly preserved
two-phase sequence. At first the outflow was channelled downhill through a V-shaped
sluice (pl. XXIII, c); when this was abandoned, the head of the channel was blocked and
another massive sluice cut in an arc on the edge of the opencast. The tank is thus a classic
example of hushing or hydraulicing from the head of an opencast, a principle implicit in the
Elder Pliny's account of mining (App. I) and first attested by Tank G in the Dolaucothi
mine. Further down the southern flank of the opencast, below the tank, there are further
uncertain indications of mining activity that include a stepped washing table, served either
36 Mining Journal (8/2/X896), I99. 38 O'Reilly, op. cit., 65.
37 G. D. B. Jones, PBSR xxxi (I963), 140 f.,
cf. 74-99. For the background, see T. Ashby, The
Aqueducts of Ancient Rome (1932).
I80 P. R. LEWIS AND G. D. B. JONES

by the first-phase downfall from the tank, or by a second small aqueduct running back
along the mountainside. To the north, on the opposite side of the opencast, there are traces
in the scrub of the ledges carrying two further leets towards the lip of the mine. If they
originally fed tanks, then these have been eroded away. Towards the foot of the slope there
is the bowl of the minor opencast. The Pumarin aqueduct, which appears to be double-
channelled in its final stages, ran across the lower bowl of the opencast just as the Cothi
aqueduct at Dolaucothi traversed an earlier opencast.39 It then delivered into a double
reservoir (B), seen from above in pl. XXII, c. This had two functions: first, it supplied water
to the northern foot of the main opencast (by contrast the final stages of the Montefurado
aqueduct, including much of the terminal reservoir, have been eroded away); and secondly,
it carried an extension of the aqueduct system on to the western spur, where a third tank (C)
suggests a northward-facing opencast. A particular interest attaches to the continuation
of the aqueduct system beyond this last-mentioned tank. It was carried in a deep trough
(pl. XXIII, a) a further 75 m or so along the scarp, and then turned through a ninety-degree
angle to discharge directly down the mountainside to the valley floor (fig. 26). Its channel
is still crisply preserved, without significant signs of lateral erosion, thus indicating a very
brief period of use. The strike of the main quartz outcrop might, if projected, reasonably
be expected to reappear at this point on the mountainside though in fact this is not the case.
The likeliest explanation of the feature, therefore, is that it represents a prospecting gully
that drew a blank.
The gully must have been preceded by another aqueduct, also visible on the hillside
(pl. XXIII, a). It ran eastwards along the western ridge, and appears to have divided in two
close to the foot of the mine (fig. 26). Its potential head of water was limited owing to its
position, and it is best viewed as an early attempt, discarded after the major aqueducts were
built.

Analysis. Puerto del Palo, as suggested in the introduction, is an example of a hard-rock

mine designed to attack primary quartz deposits, and at a greater altitude than either of the
other two gold-mines. The exploitation of such mines is critically affected by the problem
of water supply. The site lies at the southern end of an area that, in its geological form, is
probably a dissected dome that contains the sources of several river-systems. The height
of the Puerto mine, however, was too great for any of the local streams to yield a continuous
supply. In the first instance, the short aqueducts draining the surrounding spurs must have
been utilized. Nor should one underestimate their yield, which could have been considerable
after the melting of the winter snows; and the snow-falls from late October to April would
in any case make mining a seasonal operation. Nonetheless, the aqueduct system was
extended by a remarkable hydraulic system which tapped the southern headwaters of the
Rio del Oro (fig. 26). General probability suggests that the smaller aqueduct of the two
using the Montefurado tunnel-whether deliberately planned or a mistake in levelling-was
the earlier. To it was added the larger, second aqueduct on a slightly higher line, which
took the same principle to its furthest point by tapping the main stream yet further south
on the slopes of the Orrua.
Only one adit is visible on the site itself, an important point that implies that the
greater part of the extraction process must have been conducted by hydraulic means.
Tank A is a classic example of its kind, perched at the head of the main opencast; it shows
us how hydraulic mining could proceed along the lines described by Pliny (App. I), and
the melting snows and autumn rains at this height (I,200 m) would provide a sufficient head
of water at least for seasonal operation. The discovery of several leets on the northern
side of the opencast implies the existence of other such tanks that have possibly been eroded
or mined away. Lower down on the north side the development of the minor opencast
occurred relatively early, because the northern aqueduct can be traced crossing the bowl
of the opencast. The development of this hydraulic system, tapping the various streams
rising from the dissected dome to the north, now made it possible to explore the western
ridge below the main mine, where tank C was used to work a probable opencast to the
north and explore any continuation of the quartz deposit by means of hushing along the

39Lewis and Jones, op. cit., 259.

 ROMAN GOLD-MINING IN NORTH-WEST SPAIN 18I

southern slope. Such was the general framework, and such the basic stages of the develop-
ment of the Puerto del Palo mine.

CONCLUSION

Of the many problems presented by ancient mining sites, attention may be drawn to
three in particular. These are, first, the determination of the origin and development of the
site in question; second, the techniques used to extract the ore; and last the character of
the associated social and economic organization. Partial answers to these problems may be
obtained by exploratory fieldwork, if we bear in mind the peculiar nature and large scale
of the features normally present at mining sites.40 Selective excavation may then clarify
various aspects of the site, as has been done in the second-stage programme at the Dolaucothi
mine in Britain, wherein excavation was used to test the most probable hypotheses suggested
by fieldwork, largely in relation to hydraulics, and it became possible to build up a more
detailed picture of the mining techniques employed.41
An heuristic approach to important sites remains an essential first step in a rigorous
study of Roman and ancient mining. More extensive exploitation of other sites in north-
western Spain, for instance, will undoubtedly focus attention on detailed problems within
the threefold division discussed above. We may wish to determine, for example, how
quickly the sites were developed following the conquest of the region, and how long it took
for the particular deposits to be exhausted. While a rough estimate can be made from the
calculations set out in App. II, further evidence is clearly needed. A related problem
concerns the extent to which the principal gold-bearing deposits were exploited before the
conquest, and what techniques were employed. A priori one might expect the easier
placer-deposits to have been exploited during the prehistoric period, and their existence
undoubtedly would have initiated Roman interest in the area. Such might have been the
case at Montefurado, but seems very unlikely at the difficult high-level mines worked at Las
Medulas; this site could only have been exploited with the use of large-scale hydraulic
techniques depending on aqueducts, since the height of the deposit excludes any working-
technique using rivers or streams directly. By analogy, at Dolaucothi it was possible to
suggest that in prehistoric times a small-scale attack on the vein-system stimulated extensive
Roman exploitation; the use of exploratory hushing (v. App. I, below), however, exposed
other and possibly richer ore-bodies close by, and it was around these that the main develop-
ment occurred.42 The pre-Roman working areas thus became obsolescent owing to the
introduction of a much more efficient technology. In Spain, the essential preliminary to any
estimate of the extent of pre-Roman gold-working must be an adequate distribution-map
of the hill-forts on which pre-Roman settlement was based, as a pointer not only to the
likeliest areas of pre-Roman placer-working, but also to a solution of some of the topo-
graphical problems still awaiting explanation in the course of the Augustan conquest.43

APPENDIX I: THE ELDER PLINY AND GOLD-MINING

While prehistoriansoften view evidence from a technologicalstandpoint, Roman archaeologists,
with an embarrassmentof materialremainsto assimilateagainstan historicalbackground,tend to pay
only lip-service to the giant technologicalstrides that were made in the period. There is a tendency
in recent years to take the massive achievementsof Roman technology-dams, bridges, aqueducts
and mines-for granted without bringing the details into focus. This partly reflects the silence of
the literarysources on subjects relatingto technology. One notable exception is the Elder Pliny who
in NH XXXIII, 67-78 gives an account of gold-mining that is clearly much influenced by informa-

40 For the kind of problems concerned, i.e. the Almagro and Dr. Figuerello of Madrid University
geomorphological effect of human agency in the and the Cultural Attache of the Spanish Embassy;
classical period, v. C. Vita-Finzi, Mediterranean Mrs. V. A. Jones, Dr. J. P. Wild, Mr. D. G. Bird
Valleys (I969), passim. and the Needham Hall Fund of Manchester Univer-
41 G. D. B.
Jones and P. R. Lewis, Ant. J., forth- sity; and the Society of Antiquaries of London for
coming. access to the Gowland Bequest. The authors are
42 Lewis and Jones, op. cit. 246.
deeply grateful to Sir William Mansfield-Cooper,
43 The authors would like to acknowledge help formerly Vice-Chancellor of Manchester University,
from many sources in Spain and Britain: Dr. A. P. and to the Dept. of History in the same university
Masia and Prof. A. G. Alvarez of the Consejo for subventions towards the cost of travel.
Superior de Investigaciones Cientificas; Prof.
I82 P. R. LEWIS AND G. D. B. JONES

tion from Spain, although the actual source is not known.44 The account is therefore of particular
interest in the context of this article. Despite a few errors, Pliny's text can be analysed as a com-
mentary on Roman mining practice in general and is particularlyvaluable for its emphasis on the
role of water power, not merely for washing but also hydraulicingand hushing, the latter practice
now being confirmed from archaeologicalevidence at the Dolaucothi gold mine in South Wales.
No translationcurrentlyavailabletakes account of standardmining practiceor terminology; so the
following somewhat free translationis offeredwith a parallelcommentarythat drawson comparative
material from nineteenth-centurysources.
Text of NH xxxIIi, 67 ff. Comment
67. ' Gold prospectorsbegin by collectingsegullum, This is the commonest form of prospection,
that is placer-depositsindicating the presence i.e. tracing vein gold from derivative placer-
of vein gold. After the auriferousmaterial is deposits, cf. Westgarth Forster, A Treatise
washed, the sediment allows an estimate of on a Sectionof the Strata from Newcastle-upon-
the actual vein to be made ... Gold found by Tyne to CrossFell (3rd. ed., Newcastle, 1883),
this method is called talutium, a term that I6x ff. (hereafter Forster). Placer-mines were
covers placer-depositsin general. It is, in fact, not, of course, the only source of gold in
the wealth of placer-deposits that brings Spain. Montefurado,for instance, was both a
prosperityto the dry, barrenmountainsof the vein and placer mine (pp. I73-4).
Spanish provinces.
68. 'Vein gold may be obtained by underground Distinction between opencast and under-
or opencast working. It can be found in a ground working. Geological occurrence of
crystalline matrix . . . Traces of the veins gold.
appear here and there along the walls of the
underground galleries and the overburden is
supported by wooden props.
69. 'The mineral extract is crushed, washed, Smelting.
fired, and ground to a fine powder. Ground
ore is called scudes, the molten metal, when
tapped, the sudor(= " sweat "). Whateverthe The refinement of smelting methods would
metal in question, the molten waste is called suggest reworkingof slag. Such a process was
scoriaor slag. In the case of gold, however,the noticed by Charles Darwin during exploration
scoriacan be crushedand firedagain. Crucibles of Chile, Journal duringthe Voyageof H.M.S.
are made of tasconium,a white fire-clay. This Beagle aroundthe World, I93 ff., v. Lewis and
alone can withstand the forced draught and Jones, Ant. J. XLIX, ii (I970), 253, n. 4.
intense heat of both furnaceand molten metal.
70. 'The third gold-extractionmethod rivals the The two other methods are placer and opencast
projects of the legendary race of Giants. By mining.
the light of lamps long galleries are excavated
into the mountain. The lamps measure the
shifts, and the men may not see daylight for
months on end. This class of mine is termed
arrugiae. In them sudden collapse can crush A general descriptionof deep mining.
the miners, so that diving for pearls or dyes
now seems a safer job-so much has mining
been made a more hazardousoccupation. As a
protectionthe overburdenis supportedby rock
7I. arches at frequent intervals. Whatever the This seems to be a descriptionof underground
undergroundmining methods employed, hard stoping of ore-bodies.
quartzite masses will be encountered. These
can be split with the help of firesettinginvolving For firesetting at Dolaucothi, v. Lewis and
use of acid. More often firesetting in adits Jones, Ant. . XLIX, II (I970), 256.
makes them too hot and smoky; instead, the
rockis split by crushingmachinesincorporating
150 lb. iron weights. The miners then carry
the ore out on their shoulders, each man
forming part of a human chain working in the
dark,only those at the end seeing the daylight.
If the quartzite seems too large a mass, the
miners divert the drive to go round it.
44 The source v. R. Syme, 'Pliny the Procurator', Harv. Stud.
may be personal knowledge. For a
discussion of the Elder Pliny's career in Spain, Class. Phil. LXXVIII(1969), 2z8 ff.
 ROMAN GOLD-MINING IN NORTH-WEST SPAIN i83
72. 'Yet rock of this kind is considered relatively This is not easy to interpret, as a conglomerate
easy in comparison with gangadia, a form of rock should not present such difficulty.
conglomerate,that is almost insuperable.
'The method used in this case is to attack it
with iron wedges and the crushing machine
mentioned above. Gangadia is considered
the hardestof things except the greed for gold
which is the stubbornest of all. When the
work is completed the miners cut away the
roof props, beginning with the last. A crack, From the mention of the watchman it appears
seen only by a watchmanperched on the slope that this is not a description of mining under-
73. above, is the prelude to the collapse. With a ground galleries at deep levels, but of the
shout or wave he orders the miners away and creation of adits that are collapsed in the early
leaps for safety. The overburden falls apart stages of an opencast.
as it collapses with an incredible crash and
blast of air. Like conquering heroes the A modification of this method would explain
miners comtemplatetheir triumphover nature. the development of the opencast at Monte-
Yet even at this stage no gold is apparentnor furado.
did they know of its positive existence when
they began to dig. Mere hope was sufficient
inducement to counterbalancethe danger and
the costly processes involved.
74. 'Equally laborious and more expensive is the Pliny now turns to hydraulics.
associatedproblem of running aqueducts mile
after mile along mountainridges to wash away This would imply the process of hushing, see
mining debris. The aqueduct channels are Note below.
called corrugi,a term derived from the word
conrivatio or confluence. The problems are Reference to the aqueducts serving Las
innumerable; the incline must be steep to Medulas (pp. 174-6) shows that this descrip-
produce a surge ratherthan a trickle of water; tion may not be as rhetoricalas it first appears.
consequently high-level sources are required. The initial stages of the Cothi aqueductserving
Gorges and crevassesare bridged by viaducts. the Dolaucothi mines must have been cut by
Elsewhere protruding rocks are cut away to this method. Jones et alii, Bulletin of the
allow the placing of flumes. Board of Celtic Studiesxix (I960), 77 ff.
75. 'Workmen cutting the rock face are suspended For the other Dolaucothi mine-aqueduct, v.
on ropesso that from afarthey look not so much Lewis and Jones, 'The Annell Aqueduct',
like a herd of strange animals as a flock of BritanniaII, forthcoming.
birds. Usually the men are suspended to level
the projected route; man thus brings water
through places where he himself cannot pass.
' The washing process is ruined if the water is Washing. This passage may seem to contradict
full of silt or uriumas this is called. To avoid Pliny's earliercomment about the incline of the
this the water is made to flow over gravel or aqueduct, but two separate processes are
pebbles. On the ridge above the minehead, involved. Tanks of this kind are known not
reservoirsare built measuring200 ft. each way only at Las Medulas, at La Leitosa, Paradaseca
and ten feet deep. Five sluices about a yard (= No. 3 in fig. 23), and Puerto del Palo
across occur in the walls. When the reservoir (Tank A), but also at Dolaucothi, (Tank G).
is full, the sluices are knocked open so that v. Lewis and Jones, Ant. J. XLIX(I970), fig. 6,
the violent downrush is sufficient to sweep and p. 266. The explicit reference to the
away rock debris. removal of rock debris makes it clear that
Pliny is describinghushing.

Note. The applicationof this technique, both as a means of prospectingand of clearingrock debris,
has been doubted by some who are unawarethat the processwas extensivelyused in this country and
elsewhere, even on sites with heavy top-soils, during the last century. In particular, Forster's
descriptions(o.c. I64-5) are worth quoting extensively to correct some present misconceptions:
'When float, or shoad, ore is found in a small flat spot in the middle, or near the bottom, of a
sloping piece of ground, hushingis a very good method of securing it. Where circumstancesare
favourablefor it, this is undoubtedly the most effectual,and, at the same time, the most economical
method of trial. There should be a slope of considerabledeclivity and there should also be a suitable
place for makinga dam head, or reservoir,for collectingwater; the more elevatedthe dam the better
it will be.... The dam should be madeupon the slope of the declivityin the form of a half-circle....
i84 P. R. LEWIS AND G. D. B. JONES

Where the ground which is to be hushed is of any considerable length, the reservoir must contain
a considerable quantity of water, otherwise its force will be spent too soon and will prove ineffectual
for carrying down to the bottom of the soil, earth etc. ... It is necessary to observe that the breast
of this reservoir should be strong in proportion to the quantity of water we wish to collect, and that a
sluice of necessary strength be used.
'The writer has seen stones of several tons weight, and as big as little huts, carried several
hundred yards down a large hush-gutter. The torrent and the rush of stones wear out, not only the
surface and the soil, but also a considerable depth of the superficies of the rock itself, and thus they
discover and wash clean all the veins.' (For the confirmation of this evidence, cf. C. S. Fox, The
Geology of Water Supply (London, I949), pls. xviii and xix, facing p. 119.)

76. 'Another task awaits on level ground. Water Washing. Pliny conflates two varieties of
conduits, the Greek name for which means washing processes: (i) stepped washing-tables
" leads ", are cut in steps and floored with cut in the ground, preferably rock, as found
gorse, a plant resembling rosemary, that between Tanks C and E at Dolaucothi (Lewis
collects gold particles. The conduits are and Jones, o.c., 258); and (ii) wooden washing-
boarded with planks and carried over steep tables with boarded sides, generally termed
pitches. Thus the tailings flow down to the 'Long Toms' in nineteenth-century miners'
sea and the mountain is washed away. parlance,especiallyin California. The systems
could in fact be confused with the washing-
tables (or more elaborate variations, called
riffle-boxes,set in a stepped series). For these
and other comparable nineteenth-century de-
velopments, R. W. Paul, California Gold:
The Beginning of Mining in the Far West
(Univ. Nebraska, I947), 151 ff.
It is because of this that currentlythe Spanish This is a confusion with the normal processes
land mass has encroacheda long way into the of silting that create delta fans, etc.
sea.
77. ' Gold mined in arrugiae does not require This is incorrect, being perhaps a confusion
smelting but represents pure gold. Nuggets with nuggets found in placer deposits: see
are found in deep mining and also in pits, below. The reference to pits relates to small-
some weighing more than ten pounds. They scale working of placer or surface deposits.
are called palagae or palacurnae; gold dust is
called baluce. The gorse is dried and burnt and Another reference to washing. This second
its ash is washed on a bed of turf sods, so that method is comparablewith the use of gorse in
the gold is collected. ? 76. Why the process is repeated remains
78. 'According to some sources Asturia, Galicia obscure.
and Lusitaniaproducetwentythousandpounds Direct comparisons with Roman production
weight of gold in a year, the former supplying figures are difficult because nineteenth-century
the largest amount. Spain has long been the operations were almost exclusively in placer-
main gold-producingarea. By an old senatorial mines. An I855 pamphlet estimated the gold-
decree Italy was protected from over-exploita- production of North-West Spain at 35,000-
tion, for otherwiseit would have been the most 40,000 durosper annum. Towards the end of
productivearea,as it is in agriculturalproduce. the nineteenth century the duro was worth
An extant decree of the censors concerning approx. three shillings at the exchange rate of
Victumulae near Vercellae prohibits publicani the time: O'Reilly, Proc. R. Irish Acad. xx
from employing more than five thousand men (I900-I), 61; cf. Mining Journal, Feb. 8th.,
in mining.' 1896, I71.
Restrictivetrade practices are rarely known to
us from the Roman empire. Another example
is possibly to be found in the limitation on
British tin productionin favour of Spain in the
first two centuries A.D. The resurgence of
Cornish tin-mining in the third and fourth
centuriesis otherwisehardto explain.

APPENDIX II: ESTIMATES OF THE TIME-SPAN IN OPENCAST WORKING

It is possible to estimate the lifetime of particularopencasts from a knowledgeof three factors:

the amount of materialremoved from the opencast, the volume of water delivered by the aqueducts
concerned, and the volume of water needed to hydraulicea unit volume of auriferousgravel. The
 ROMAN GOLD-MINING IN NORTH-WEST SPAIN I85
first factor can be roughly estimatedfrom the size of each opencast(fig. 25), and their dimensionsare
shown in the tables below: it has been assumedthat the opencastsmay be representedby a triangular
or rectangularbox as is more appropriatein each case. The third and fourth columns give the volume
of materialremovedin millions of cubic metres (V) and the estimatedamount of water deliveredper
day (H). The figure of 2 5 million gals. per day comes from an earliercalculationfor an aqueductof
similar size at Dolaucothi, South Wales.45 The life-spans of the opencasts (T), shown in the last
column, have been calculatedfrom the formula:

T 7V
8'7V
H
The constant has been calculatedusing informationsupplied by Longridge,46first that 25 cubic feet
of auriferousgravelweighs one ton and second that I2 5 tons of water (i.e. 2,800 gals.) are needed to
remove one ton of material.

Volume Water
Opencast removed delivered T 8 7V
dimensions (millions per day H
Period (metres) of cu. ms.) (gals.) (Years)
I. a i x 500 7'5 I3
X 500 x 60
b 300 X 300 2'7 c. 5 x Io6 5
x 30
c Ioo X 50 o-I 2 months
X 20

II. a, b ? 2-5 x Io6 Ayear

or two ?
III. ? ? 2-5 X 0o Decades?

The uncertainties of using such a simple argument are manifold but may all be expected to
increase, rather than contract, the time-span. The dimensions are no more than rough estimates,
neglecting subsequent natural erosion, which should be relatively small across the 2,ooo-year span,
but the most important source of error concerns the hydraulicing equivalent of I2-5 tons of water
for every ton of material. This estimate is based on water supplied at a high-velocity by modern
pumps, and an equivalent figure for a relatively slow-moving stream is not available. It may increase
the expected lifetime by a factor of two, if not more. It has also been assumed that the water flow
was continuous over the given time-span but one might expect seasonal factors such as the freezing
of streams at the head of the aqueduct, for example, to affect consumption. Nonetheless, if the sub-
ject is to be further refined, it seems worth attempting (with these provisos) the first assessment
of the life span of a Roman opencast working.

University of Manchester

46 G. D. B. Jones, Bull. Board

of Celtic Studies xix "6C. C. Longridge, Gold and Tin Dredging (1914).
(I960), 71 ff. For a similar attempt to calculate flow
in a closed aqueduct v. G. D. B. Jones, PBSR xxx
(1962), 200 f.