unravel all the complexities of the surface features of the area, with little enough dating-material, doubtless, to help. It is a commonplace of archaeology that rich sites (such as the huge opencasts at Dolaucothi prove this to have been) present the student with palimpsests of multi-period working; with so little certainty as to the date of large areas of this mine, it would be hazardous in the extreme to claim otherwise, and maintain that the surface morphology preserves a Roman landscape tout court. There certainly was Roman mining at Dolaucothi, and from an early date, as indicated by the dewateringwheel; and perhaps it continued until the fourth century in some form or other. Beyond that, prudence compels silence until much more has been done. 38 Penarth (G.C.B.) Newent (D.B.) G.D.B. Jones and K. Maude reply: Our 1991 note under discussion here was published to bring views of the admittedly complex mining development around Dolaucothi into better focus by noting two C14 dates of relevance. One of the above authors (GCB) first percipiently linked the Dolaucothi complex with Ptolemy's Aouevxivov (Luentinum, Luentium) (Geog. xi 2.13) because of its reference to the hydraulicing and washing of gold bearing deposits; but conversely later argued that the aqueduct systems were not of Roman origin. 39 The early Roman date of the water-raising wheel in the National Museum of Wales makes it clear that the development of the underground workings took place in the Roman period as the last and most difficult phase of the development of the main opencast. All mining operations are dependent on abundant water supplies for hushing and processing, hence the existence of the Cothi aqueduct. The other author (DB) denied the antiquity of the 'Annell' aqueduct system, while usefully locating its source some 1.5 km further northeast up the 'Gwenlas teat'. 40 The main purposes of our 1991 note was to publish the two related C14 dates that demonstrate unequivocally that the Annell aqueduct channel at GR.SN 705 426 was in fact already infilled by ad 865120 at the latest (Grn-16553), a point of significance quickly noted by Manning. 41 This significant demonstration of the relative antiquity of the Annell leat is nowhere acknowledged in the above note, and that negates the core of the view proposed. University of Manchester A Reconsideration of Gelligaer. E.W. Black writes: The Roman fort at Gelligaer was excavated from 1899 to 1901 and a complete plan of the internal buildings was made. 42 Further excavations examined the military bath-house and other buildings in a walled annexe. 43 The completeness of the plan (FIG. 5) has meant that Gelligaer has featured in discussions concerning the possibility of identifying the type of unit in garrison from the size of forts and the plan of their internal buildings. 44 However, although fragments of building 38 The reference in Britannia xxii (1991), 210 to 'a satellite industry carving cornelian gemstones' cannot be left as it is. There is a single intaglio from Dolaucothi, a ring-stone in ...