¹ Photograph as pl. 3, i in Nash-Williams, op. cit. in note 4.

² It is hoped that a programme of research will shortly be initiated by Mr. W. H. Manning of University College, Cardiff.

³ Smyth, Warington, Mem. Geol. Surv. 1 (1846), 480–4 and pl. 8.Google Scholar

⁴ Bosanquet, R. C. in Roy. Comm. Anc. and Hist. Mons. (Wales): Carmarthen Inventory (1917), 25–32.Google ScholarNash-Williams, V. E., Bull. Bd. Celt. Stud. XIV (1952), 79–84.Google Scholar

⁵ Add the material cited in note 7, and the coin-hoard noticed on p. 196 of this Journal.

⁶ Jones, G. D. B. and others, Bull. Bd. Celt. Stud. XIX (1962), 71–84.Google Scholar It has however escaped notice that the workings occupy the north-west side of a saddle between Allt Ogofau and Allt Cwmhenog. The rock-cut bed of the aqueduct is clearly traceable beyond (inspected 1966 by G. C. B. and his colleagues Messrs. D. Morgan Rees and D. Emlyn Evans, also by Mr. Wm. Manning) and proves that the aqueduct discharged into a side-valley of the R. Annell to the south-east (cf. Williams, E., Cambrian Register III (1818), 47Google Scholar) and did not enter the workings: the suggested reservoir (‘Melin y Milwyr’) is probably only another flooded opencast. The intensely hard nature of the quartz-reef also makes any suggestion of the use of hydraulic methods of excavation impossible. The aqueduct was therefore required for washing milled ore, and possibly to some extent for driving simple machinery.

⁷ Including part of a Dr. 37 bowl mould-stamped by Frontinus of La Graufesenque, and a coin of Trajan. The fort would lie between Llandovery and Llanio-isaf.

⁸ Loc. cit. in note 4. A recent survey (Arch. Camb. CXII (1963), 93) claims two mottes. One near Ogofau Lodge looks like a pile of mining-débris. The other has no bailey but remains a possibility.

⁹ Loc. cit. above. Cited hereafter as ‘Davies’.

¹⁰ The documentary sources for this period are of extreme poverty. Cf. North, F. J., Mining for Metals in Wales (1962), 30–2.Google Scholar

¹¹ North, op. cit., 33 ff.

¹² Williams, op. cit. in note 6, 42.

¹³ Owen, G. W. and Davies, V. C., Journ. Mer. Hist. and Rec. Soc. IV. 1 (1961), 60–71Google Scholar; D. M. Rees, Journ. Mer. Hist. & Rec. Soc. VI, forthcoming.

¹⁴ Mine and Quarry Engineering, Jan.-Mar., 1944 (esp. Jan., hereafter cited as ‘Nelson’).

¹⁵ A gallery following a vein; a cavernous upward or downward enlargement of this, designed to remove a larger ore-body.

¹⁶ Forbes, R. J., Stud. Anc. Techn. VIII (1964), 155Google Scholar, gives this limit as 0.188 per cent. The figure given by Georgius Agricola, De re met., ed. 1562, p. 76 ad fin., is 3 unciae to 100 librae, i.e. 72 oz. 18 dwt. to the ton according to Hoover and Hoover in their translated ed. of 1912 (p. 108), which I make 0.219 per cent (G.C.B.). But the economic extraction-rate depends very much on the method of costing used. Were forced labour and its upkeep costed by the Romans ?

¹⁷ As is well known, the metal-content of a lode is variable, and is normally richest near the surface of the ground, owing to the solution of other matter.

¹⁸ Nelson, 3.

¹⁹ ibid. 6–7, fig. 3.

²⁰ ibid.

²¹ Palmer, R. E., Trans. Inst. Mining and Metallurgy XXXVI (1926–1927), pl. 16Google Scholar; cited hereafter as ‘Palmer’; Gossé, G., Ampurias IV (1942), pl. 12Google Scholar, cited hereafter as ‘Gossé’.

²² Davies, 43; Nelson, 4.

²³ Davies, ibid.

²⁴ The two adits described by Smyth (loc. cit. in note 3) were cut for drainage and haulage. They debouch on the hillside well above the level of the Cothi, as does a third adit below these which he does not mention, presumably because it was inaccessible. As the growth-rings of an oak-stump at the entrance of this ‘field adit’ (cf. Nelson, 8) show, it was driven many years before the rediscovery of gold at Dolaucothi. The field adit is now blocked.

²⁵ Davies, 57, mentions a Rio Tinto wheel which had twenty-four pots around it. This must have been a temporary expedient to increase capacity. The British Museum wheel in the Gallery of Greek and Roman Life has spokes pierced near the rim. Possibly additional buckets were attached by cords passing through these holes. Tags of rope have been found in other instances (cf. Stevenson, A. S., Arch. Ael. n.s. VII (1876), 280Google Scholar) and have been claimed to represent the means whereby the wheels were turned. It is more probable that these fragments represent cords attaching extra buckets.

²⁶ Used in the Sotiel Coronada mine: Gossé pl. 13. We add for completeness that the ‘Archimedean’ screw has been found in several mines, e.g. in Santa Barbara (Gossé, pls. 8, 9, 10) or Centenillo (Palmer, fig. 87). The ‘Ctesibian’ force-pump appears at Sotiel Coronada (Gossé, pl. 14).

²⁷ E.g. Buffet-Évrard, , L'Eau potable à travers les ages (1950), 71Google Scholar, fig. 56: fig. 55 for the smaller tympanum.

²⁸ Cf. the numerals on the staves of the Silchester, and barrels, London, JRS LI (1961), 195–6.Google Scholar

²⁹ Stevenson (op. cit. in note 25, cited hereafter as ‘Stevenson’), between pp. 280–1; Gossé, pl. 11. I suspect that the well-known generalized section given by Palmer, 303, fig. 74, which has been widely copied (e.g. in Hist. Techn. II, 7), is laterally compressed to omit much of the length of the galleries intervening between the successive sump-chambers: but may be corrected on this point. G.C.B.

³⁰ The Tarsis wheels are shown by Stevenson (loc. cit.) as having a dished form, and ports therefore on one side only. This is unusual, most wheels being of the balanced type adopted in our own reconstruction (cf. fig. 6, B). The purpose of the dished wheel would be to enable the launder to be set more vertically beneath the ports at the point of discharge than would be possible with the standard arrangement, but since only the spokes are outwardly inclined, the Tarsis type of wheel could never have been so strong as the normal type.

³¹ E.g. Stevenson, loc. cit.; Palmer, figs. 69, 74. One revolved clockwise, the other (like ours) anticlockwise.

³² Dacia: Pošepný, F., Mitt. anthr. Ges. Wien XII (1892), 44Google Scholar (no illus.); id., Öst. Zeitschr. f. Berg- u. Hüttenwesen XVI (1868), 153–4, 165–8, with fig.; XXV (1878), 391–3 (Verespatak). The reconstruction sketch cannot be correct in the central position of the ports, cf. XXV, 392; the shute running to the base of the sump was probably the fallen launder. The Verespatak wheel had half-jointed side-boards like ours, but instead of paired spokes had a single spoke expanding into a shovel-like head which served to divide the compartments and had chases for the edge-boards.

³³ For Rio Tinto see Rickard, T. A. in Palmer, 326–7 and more generally JRS XVIII (1928), 129–43.Google Scholar For Verespatak there are the famous ‘Dacian wax tablets’ (Pošepný shows site of find). It can only be assumed that the wheels were used at the same time as such dateable material accumulated in the mines.

³⁴ Stevenson, pl. between pp. 280–1.

³⁵ Palmer, 302–4. Another calculation gives the efficiency of the wheel as 61 per cent.

³⁶ Kindly examined by our colleague Dr. Brian Seddon.

³⁷ Davies, 56. Forbes says wheels from Leon are 20 feet in diameter.

³⁸ Palmer, figs. 69, 82. For Verespatak, see note 32.

³⁹ Pošepný, Öst. Zeitschrift XVI, 166: ‘die Welle war aus weichem Holz [softwood] etwa 3 Fuss lang.’ We are indebted to Mr. K. S. Painter for obtaining Xerox copies of Pošepný's papers.