¹ JRS xli, xliii, xlv, xlviii, li, lv, lix.

² The photographs illustrating this article (plates XV–XVIII) are all from the Cambridge University Collection: of the text-figures, 10–13, 15–19 and 22–23 were drawn by D. R. Wilson, 14 and 20–21 by B. Thomason: Copyright The Committee for Aerial Photography of the University of Cambridge.

³ JRS lv, 1965, 74–6 and Fig. 2.

⁴ JRS xlv, 1955, 82.

⁵ AA ² xxv, 1904, 193–300.

⁶ JRS xlv, 1955, 83–4; for sketch-plans and sections see Bellhouse, R. L., CW ² lvi, 1956, 28–36.Google Scholar Troutbeck was examined in 1953, and again in company with Professor S. S. Frere in March 1972 and April 1973. The earthwork forming the W. rampart of the fort was first noticed in March 1973 by Mr. R. A. H. Farrar, of R.C.H.M. England.

⁷ CW² xxxviii, 1938, 32–41.

⁸ CW² xlviii, 1948, 202–3, with references to earlier accounts.

⁹ McCord, N., AA ⁴ xlix, 1971, 120Google Scholar, pl. xi.

¹⁰ JRS li, 1961, 121.

¹¹ JRS xli, 1951, 56; Richmond, I. A., Proc. Soc. Antiq. Scot. lxxxv, 1951, 138–145.Google Scholar

¹² JRS xlv, 1945, 85.

¹³ JRS xlviii, 1958, 87–8; li, 1961, 121.

¹⁴ JRS xlviii, 1958, 88; lix, 1969, 107.

¹⁵ JRS lv, 1965, 80.

¹⁶ JRS xli, 1951, 38; xlv, 1955, 85; lv, 1965, 79.

¹⁷ JRS xli, 1951, 58–9; Trans. Dumfries Antiq. Soc., 3rd ser. xxxiv, 1957, 9–21.

¹⁸ JRS xli, 1951, 59; xlv, 1955, 85–6.

¹⁹ JRS xlviii, 1958, 89; li, 1951, 122.

²⁰ This is probably the site observed by Pennant two hundred years ago. Pennant, T., A Tour in Scotland: 1772, 1774, 100.Google Scholar In an account of the line of Roman road between Kaims Castle and Ardoch ‘… a little further … is a small round area, like those on Gaskmoor, but considerably stronger, being surrounded by not fewer than three fosses’.

²¹ Crawford, O. G. S., Topography of Roman Scotland, 1949, 59.Google Scholar

²² Macfarlane, W., Geographical Collections relating to Scotland, i, 1748, ed. by Mitchell, A., 1906, 121.Google Scholar

²³ Feachem, R. E., Antiquity xliv, 1970, 1204.Google Scholar

²⁴ Agricola, ed. Ogilvie, R. M. and Richmond, I. A., 1967, 64, 251–2.Google Scholar

²⁵ JRS xli, 1951, 63; xlviii, 1958, 91; lv, 1965, 82; lix, 1969, 110–11.

²⁶ JRS lix, 1969, 116–18.

²⁷ JRS lix, 1969, 109–110.

²⁸ Correcting the measurements given in a previous account.

²⁹ JRS lix, 1969, 109; xl, 1950, 92; xli, 1951, 64.

³⁰ JRS lix, 1969, 111.

³¹ J. Ainslie, Map of the County of Angus, 1794, scale 1 in. to 1 mile. I am grateful to Dr. W. Dally for drawing my attention to this map.

³² JRS lv. 1965, 83; lix, 1969, 111.

³³ JRS lix, 1969, 111–12.

³⁴ Agr., ch. 25.

³⁵ JRS lix, 1969, 112–13.

³⁶ JRS xlv, 1955, 87; xlviii, 1958, 93; li, 1961, 123.

³⁷ As will be seen on reference to the Ordnance sheets NS 79 and NN 70, 1: 25,000 scale, on which contours are drawn at 25-ft. intervals.

³⁸ Excavations at these camps could not have been carried through without generous assistance from many quarters. For help with digging, thanks are particularly due to Mr. N. Clayton, Dr. R. S. Hardie, Mr. G. Maxwell, Dr. R. M. Ogilvie and boys from Tonbridge, Mr. J. J. Robertson and Mr. D. R. Wilson. Grants from the Administrators of the Haverfield Bequest, and from the Society of Antiquaries of Scotland towards the cost of the operations are gratefully acknowledged.

³⁹ Agr., ch. 22.

⁴⁰ It may be explained that the symbol on the map (Fig. 17) for a temporary camp at Inchtuthil indicates the 45-acre camp there, which presumably held the legion that began the construction of the fortress. Within the camp, rows of rubbish-pits, related to tent-positions, suggest a more prolonged occupation than would have been in question at a camp accommodating forces on campaign.

⁴¹ JRS lix, 1969, 113–14. There may now be added the evidence from excavations at Ardoch, which has shown that camps (2) and (3) there (see ibid. Fig. 4) are earlier than the 63-acre camp (4): Britannia i, 1970, 163–171.

⁴² Agr., ch. 25.

⁴³ JRS lix, 1969, 114–19.

⁴⁴ The suggestion of a Flavian crossing of the Tay at Carpow arises from the recognition, since these campaigns were last discussed (JRS lix, 1969, 118–119), of three large camps, Dunning, Abernethy, and Carpow. Evidence for date at Abernethy is provided by a fragment of Samian, at Carpow, by the position of the camp in the sequence of structures there. As to the fact that the considerable engineering feat involved in crossing the Tay finds no mention in the Agricoia, it may be said that this was perhaps regarded as a lesser achievement than, say, the forcing of the Menai Straits.

⁴⁵ The proven length of ditch at Logie suggests that a camp larger than some 50 acres is in question. It can hardly be a member of the 130-acre series, since there is such a camp at Balmakewan, only 2¾ miles away. Logie might prove to belong to the 63-acre series, though it lies at a distance of only 5½ miles from Oathlaw, another member of that series, with no natural obstacle between them.

⁴⁶ Such close correspondence would be more likely if only a year or less intervened between the operations represented by the Stracathro camps (? built in the sixth season of Agricola's campaigning, or even early in the seventh season) and by the 110-acre camps, than if a century and a quarter separated the construction of the two series.

⁴⁷ JRS lix, 1969, p. 112, Fig. 3.

⁴⁸ The obstacle created by these mosses is well illustrated by the position of the 63-acre camp at Lunanhead. From Marcus direct to Eassie is only 12 miles, but this interval was divided in two by a camp at Lunanhead (6¼ miles march from Marcus, and 8½ from Eassie) most easily reached from Marcus by rounding the W. end of the hill of Finavon, and turning eastwards to the low ridge on which the camp is set.

⁴⁹ The principal historical sources are Dio, lxxvii, 10–13, and Herodian, iii, 14, but Dio is known only in epitome, and Herodian is not the best of historians.

⁵⁰ JRS li, 1961, 123; lv, 1965, 84; lix, 1969, 104.

⁵¹ JRS lix, 1969, 104.

⁵² For these see Antiquity xl, 1966, Fig. 1.

⁵³ JRS li, 1961, 123.

⁵⁴ JRS lv, 1965, p. 77 and plan Fig. 3.

⁵⁵ JRS lix, 1969, 105.

⁵⁶ JRS xliii, 1953, 54; 5½ acres is the area within the inner ditch.

⁵⁷ The site was seen first by Mr. Arnold Baker in early July 1970, when the fort (under wheat) was visible, but the camps (under roots) were not. By the last week of the month when the photograph (PI. XVII, 1) was taken, crop marks had developed in the roots, but the outline of the fort had all but vanished as the wheat ripened.

⁵⁸ JRS lv, 1965, 85.

⁵⁹ JRS lv, 1965, 85; lix, 1969, 120.

⁶⁰ JRS lix, 1969, 119–120.

⁶¹ JRS lix, 1969, 119.

⁶² JRS xliii, 1953, 85 for earlier references. See also Nash-Williams, , The Roman frontier in Wales, (ed. Jarrett, M. G.), 1969, 85–88.Google Scholar

⁶³ As at Ambleside, , Inventory of Westmorland, (RCHM) 1936, 1–3.Google Scholar

⁶⁴ JRS xliii, 1953, 86; xlviii, 1958, 95; Nash-Williams, V. E., The Roman Frontier in Wales, (ed. Jarrett, M. G.), 1969, 77–80Google Scholar, where it is suggested that Clyro was abandoned at (or before) the Flavian settlement of Wales.

⁶⁵ JRS lix, 1969, 121.

⁶⁶ For the pottery, etc., and for the trial-diggings see Trans. Radnorshire Society, xxvii, 1958, 68–9, recording the finding of coarse pottery including a storage jar, an amphora, and a fragment of Samian. I am grateful to Professor Jones for telling me of this reference. The excavations yielded ‘… an abundance of broken tile fragments—hypocaust, curved and flanged tiles mixed together—filling an artificial clay basin …’ The description suggests remains of a bathhouse: the site of the digging was pointed out to me by the farmer, Mr. H. Price. Professor Jones' account of the fort, with a report of more recent finds of pottery, including ‘south Gaulish Samian of Claudio-Neronian date’, is published in Mont. Coll. lxi, 1971, 41.

⁶⁷ O.S. Map of Roman Britain, 3rd ed. 1956.

⁶⁸ See Ordnance sheets SN 95 and SO 05, 1:25,000 scale.

⁶⁹ Castell Collen lies 6½ miles to the NNE., Caerau 5¾ miles to the SW.

⁷⁰ They lie just beyond the left-hand margin of plate XVIII, 1.

⁷¹ Inventory of Caernarvonshire, RCHM (Wales) iii, 1964, 114–16; Hogg, A. H. A., Arch. J. cxxv, 1969, 101–192.Google Scholar The fort has now been destroyed by gravel quarrying.

⁷² There are, for example, six camps at Greensforge and four at Old Radnor, The practice camps on Llandrindod Common, and a similar camp near Caerau have been omitted. Nantmel, identified by Nash-Williams from a vertical air photograph taken by the RAF, is here treated as a camp. While its Roman character seems not in doubt, the degree of permanence of the work remains to be determined. The fort at Trent Vale is known from a partial section of the defences (Britannia ii, 1971, 259), that at Whitchurch from excavation (Arch. J. cxxv, 1969, 193–254).

⁷³ They lie only 4 miles apart but the shortest effective marching-distance is 6 miles which includes a crossing of the Wye and a steep, 1,000 ft. climb.

⁷⁴ Perhaps at Westbury, as suggested by Jones, G. D. B., Mont. Coll. lxi, 1971, 96.Google Scholar

⁷⁵ Antiquity xlv, 1971, 224–5, pl. xxxv.

⁷⁶ Atkinson, T. D., Camb. Antiq. Soc. Proceedings viii, 1894, 229.Google Scholar

⁷⁷ JRS lv, 1965, 88, pl. xiii, 2.

⁷⁸ Antiquity xlvi, 1972, 224–6.