^{page 1 note 1} Curle, A. O., Antiquity, I, 290 ff.Google Scholar

^{page 1 note 2} Watson, W. J., Celtic Place Names of Scotland, 61 ffGoogle Scholar; Brögger, A. W., Ancient Emigrants, 43 ff.Google Scholar; R.C.A.M., , The Outer Hebrides, Skye and the Small Isles, 1928, p. XXXIII ffGoogle Scholar. and Orkney and Shetland, 1946, I, p. 28 ff.Google Scholar

^{page 1 note 3} Childe, V. G., Prehistory of Scotland, 1935Google Scholar; Prehistoric Communities of the British Isles, 1940; Scotland before the Scots, 1946; Callander, J. G. and Grant, W. G., ‘The Broch of Midhowe,’ P.S.A.S., LXVIII, 444 ff.Google Scholar

^{page 1 note 4} Anderson, J., Scotland in Pagan Times, the Iron Age, 192Google Scholar; R.C.A.M., , Orkney and Shetland, I, 31Google Scholar. The numbers in these islands are less conservatively estimated than were the numbers estimated for the Hebrides in the Commission's former book. The most recently prepared list is that published by Mr Angus Graham, the Commission's Secretary, in Ant. J., XXIII, 19 ff.Google Scholar, which would bring the total slightly over 500. This list is generally followed below. The broch area is regarded as extending a few miles south of the Ross-Inverness-shire boundary so as to include the Glenelg and Struy brochs.

^{page 3 note 1} Anderson, , P.S.A.S. XXXV, 116Google Scholar.

^{page 3 note 2} V. G. Childe, op. cit., 1940, 247.

^{page 3 note 3} No ancient sites have been more excavated than those of brochs and, unfortunately, none with less result. The extensive early excavations are inadequately reported, if at all, and were not conducted so as to yield systematic evidence of stratification. The only fully reported modern excavation, that of Midhowe, was purposely not carried down to the primary occupation level. The results of the excavation of Aikerness carried out officially by the Ministry of Works are known only through a brief summary published by the Royal Commission.

^{page 3 note 4} R.C.A.M., (Scotland), The Outer Hebrides, Skye and the Small Isles, 1928Google Scholar. While generally agreeing with the descriptions given by the Commission, the statements regarding these brochs are derived from the writer's observations and must be his responsibility.

^{page 4 note 1} Two of these are recorded by the Royal Commission as cairns. The first (no. 458, Chambered Cairn, Bal na Craig) does not appear to be a denuded long cairn and, in its present state at least, indicates a village site with hut circles and other domestic structures. The second (no. 459, Cairn, Tigh Talamhanter, Cuier) gives no indication of having been a burial site and represents a village or farmstead enclosure by a stream, roughly 200 feet square and surrounded by a wall with an entrance on the north side. Numerous round buildings stand within it and a rectangular one against the surrounding wall beside the entrance. A third similar site, not recorded by the Commission, stands half a mile to the north and on the same stream.

^{page 4 note 2} Another ‘dun’ stood on a rocky knoll, now sand and grass covered, at the foot of the valley. The buildings are wholly hidden and nothing is known of their character.

^{page 5 note 1} Both are classified by the Royal Commission as ‘duns,’ but the writer's observations suggest with confidence that they are brochs. Apart from their siting and dimensions Dun Borve shows specific broch features in a lintel of a door to a guard chamber and in walling, visible in two places, of a gallery in the wall; Dun Rodil shows the latter feature. Dun Borve is included in Mr Graham's list.

^{page 6 note 1} Based, by permission, on the distribution map on p. XXXIV of the Royal Commission's Hebrides volume with some changes.

^{page 8 note 1} The fact that a circular stone ‘dun,’ or ‘fort,’ on Clettraval, North Uist, which is under excavation by the writer, has turned out to be a wheelhouse is a warning that this type of structure may be more widely distributed than has been thought.

^{page 9 note 1} J. Anderson, op. cit., 205.

^{page 10 note 1} In a paper which will appear in P.S.A.S. LXXXI, and to which he very kindly allows me to refer, Mr Angus Graham records a total of 45 brochs, excavated and unexcavated, giving evidence of stairs; and a total, largely included in the former, of 29 brochs giving evidence of galleries. Of the latter, 6 show more than one gallery and 23 one gallery, but he thinks that less than half of the 23 cases provide useful evidence. While, as Mr Graham emphasises, such statistics need to be used with much caution, these figures would suggest that the number of brochs which can confidently be inferred to have had wall-head walks or galleries is less than would be deduced from the data for excavated examples which is given above. It should be noted, however, that Mr Graham would not draw this inference from his data, since he believes on more general grounds that all brochs had stairs and galleries.

^{page 12 note 1} As calculated by Mr Graham. I avail myself in the summarised statement above of the very careful analysis made by him of the data for these four brochs in the forthcoming paper mentioned in the preceding note.

^{page 12 note 2} J. Anderson, op. cit., 184.

^{page 12 note 3} Wester Broch, Keiss, P.S.A.S. XXXV, 121–2Google Scholar.

^{page 13 note 1} Ramps sloping at 30° to the horizontal to the top of each side of a wall 6 feet high and 12 feet broad would, if composed as to half of stones and half of earth, contain jointly sufficient stone to raise the height of the wall by 2.6 feet.

^{page 13 note 2} P.S.A.S. LXVIII, 444 ff.

^{page 13 note 3} P.S.A.S. LV, 110 ff.; R.C.A.M., The Outer Hebrides, Skye, etc., no. 479.

^{page 14 note 1} The numerous early excavation reports give data for calculation in a surprisingly large number of cases and the fact that the excavators did not make the calculation, but tended to refer vaguely to an ‘immense mass’ of fallen material, must be due to a general ignorance of mensuration. There were notable exceptions. Dr Goudie, reporting the excavation of a broch on an island near Lerwick, expressed the opinion that ‘the wall never reached an elevation approaching the supposed normal altitude of brough buildings (say 40 feet). The quantity of debris, little of which could ever have been removed, does not correspond with any safe assumption of a large cubic mass having existed on the spot’ (P.S.A.S. XXIII, 248). See also SirDryden, H. on Clickhemin, in Arch. Scot. V, 200, 205Google Scholar.

^{page 14 note 2} Over 10,000 tons of stone would have been required in the case of Cockburn Law and the work would have occupied a labour force of 60 men working a 70-hour week for three years (following Capt. Thomas's calculation below).

^{page 14 note 3} R.C.A.M., The Outer Hebrides, Skye, etc., XXXV.

^{page 14 note 4} Egils Saga, c. 32, 33; Orkneyinga Saga, c. 93.

^{page 15 note 1} Giraldus Cambrensis, Conquest of Ireland, Book II, Chap. 38 and 39 (Rolls Series edition).

^{page 15 note 2} Comparison with the mottes was first made by the Royal Commission in their Wigtownshire volume (p. XXXII) and a fact there noted shows how misleading such a comparison may be. There were 54 mottes within a radius of 60 miles from Caen; the average area controlled by each was therefore over 100 sq. miles. There are the same number of brochs within a radius of 16 miles from Wick; the average area controlled by each broch is only 7 sq. miles, and of this only a fraction is cultivable. The amount of really useful land controlled by Barra No. 1, the largest broch in Barra, did not much exceed one square mile.

^{page 16 note 1} Prof. Childe has indicated (op. cit., 1946, p. 88) the points of agreement with the brochs of the ‘murus gallicus’ fort of Rahoy in Morvern; this small and abnormal round specimen of the type was only 40 feet in internal diameter and was, he believes, roofed. He has also suggested that the use in these forts of transverse wooden beams to tie together the inner and outer faces of the wall might provide the source of the broch practice of building with hollow walls tied by transverse slabs. There are however other sources whence the broch builders might have derived this idea, and, in the light of the facts given above as to the character of the two cultures and as to their distribution, it does not seem possible to regard the ‘murus gallicus’ fort as parental to, or markedly influencing, the broch.

^{page 16 note 2} The uncomfortable looking cleft beside the broch of Rudh' an Dunain in south-west Skye was, however, in fact used at some period, as certain improvements made to it testify. Its use can only have been rendered practicable by the existence of an alternative landing in contrary winds in a bay half a mile to the north-west.

^{page 17 note 1} This is Mr Graham's list with the addition of two more brochs in Islay, made on the authority of Prof. Stuart Piggott, and the omission of the Tay estuary examples, which seem more than doubtful. The Tiree ones are Mr Erskine Beveridge's ‘semi-brochs.’

^{page 18 note 1} This summary is based on Curle, J.'s Inventory in P.S.A.S. LXVI, 277 ff.Google Scholar, supplemented as regards coins by SirMacdonald, G., P.S.A.S. LII, 203 ffGoogle Scholar. and, as regards armlets, by Kilbride-Jones, H. E., P.S.A.S. LXXII, 366 ff.Google Scholar

^{page 18 note 2} As suggested by Curle, J., P.S.A.S. LXVI, 290Google Scholar.

^{page 18 note 3} Tacitus, Agricola, c. 10. The fleet circumnavigated the British Isles and subdued ‘previously unknown islands called the Orcades’ in A.D. 83 or 84.

^{page 19 note 1} The lower limit may be given by a scrap of incised pottery of broch type (but without any raised decoration) from Bow Broch on the Gala Water, which probably went out of use about A.D. 100 (vide supra). The upper limit is provided by the end of the pottery sequence at Jarlshof Village (pp. 25 and 28 below); this despite the incised sherds from Wiltrow, near Jarlshof, which Dr Curle would interpolate into that sequence. TheWiltrow site comprised a bloomery and an adjacent courtyard house, which, however, had been reconstructed. The pottery, which ‘in its ornamentation bore no resemblance to any pottery found at Jarlshof,’ came ‘for the most part’ from the bloomery building. The ‘few pieces of coarse black pottery’ from the house were ‘not identical with’ that from the bloomery and it is not stated that they included incised sherds; even if they did, it is not demonstrated that they belonged to the house in its original state (P.S.A.S. LXX, 153 ff.). In the face of the very extensive pottery sequence at Jarlshof Village, it seems wise to admit that the bloomery and the incised pottery may have been coeval with the reconstructed house and after the end of the Village sequence; and correspondingly unwise to rely on the sherds as evidence of incised decoration at a stage when courtyard houses were being built, or even the earlier wheelhouses.

^{page 19 note 2} The loop-rimmed pot, which comes from a North Uist wheelhouse, is unpublished but noted by Prof. Childe, op. cit., 1940, 247.

^{page 19 note 3} Curle, J., P.S.A.S. LXVI, 307Google Scholar and fig. 55, and A Roman Frontier Post, Pl. LIII; Déchelette, , Manuel, II, iiiGoogle Scholar, fig. 649.

^{page 19 note 4} Curle, J., A Roman Frontier Post, 246Google Scholar and figs. 25 and 28. A substantial minority of the rims from 1st century B.C. Breton sites in the Institute of Archaeology (not yet published) are of this type.

^{page 19 note 5} Wheeler, R. E. M., Maiden Castle, 204 ff.Google Scholar

^{page 19 note 6} Bullied, and Gray, , Glastonbury Lake Village, II, 519Google Scholar; M. E. Cunnington, All Cannings Cross, Pl. 37.

^{page 20 note 1} From the oppidum of Bibracte in Central France (Déchelette, , Fouilles de Beuvray, 1904Google Scholar, Plates, and Manuel, II, iii, fig. 679; Bulliot, , Fouilles de Beuvray, 1899, vol. 3Google Scholar, Atlas, Plates passim).

^{page 20 note 2} I am greatly indebted to Miss H. Pincombe of the Royal College of Art for experiments to identify the technique employed.

^{page 20 note 3} Radford, C. A. R., J. Roy. Inst. of Cornwall, XXIV, App. 11, 80Google Scholar.

^{page 20 note 4} Bulliot, op. cit., Pls. XXIII, 11 and XXXVI, 1 and 2.

^{page 20 note 5} Du Chatellier, La Poterie… en Armorique, Pl. 14–17; Déchelette, , Manuel, II, iii, p. 1467 ffGoogle Scholar. and figs. 654, 663 and 664.

^{page 20 note 6} Curle, J., P.S.A.S. LXVI, 297 ffGoogle Scholar. The brochs were Okstrow and Midhowe.

^{page 21 note 1} The purposes of souterrains and of chambers in the thickness of the walls of wheelhouses and brochs have never been determined but may be conjectured to have been storage, particularly storage at a uniform temperature: compare the storage of food in ‘stone pyramids’ recorded by Martin, Martin, A Late Voyage to St. Kilda, 1697, pp. 43, 114Google Scholar.

^{page 21 note 2} Machair Leathann (Beveridge, Erskine, North Uist, 121 ff.Google Scholar) This wheelhouse was as much as 33 feet in internal diameter and the use of stone for forming the architrave may have been due to the greater weight of the corespondingly large roof.

^{page 22 note 1} Capt. Thomas, F. W. L., R.N., P.S.A.S. VII, 165 ff.Google Scholar

^{page 23 note 1} This site is distinct from the Jarlshof Village site, where it is admitted that the wheelhouses are pre-broch. The grounds for regarding the wheelhouses in the broch's outer courtyard as secondary are that the northern wheelhouse cuts across a wall alleged to be the outer defensive wall of the broch and contemporary with it (R.C.A.M., Orkney and Shetland, no. 1149). The wall, however, does not continue beyond the wheelhouse and round the broch, and it is only four feet thick on the average. Prima facie it would seem not to be an outer defensive wall to the broch but part of the same complex of building as the southern wheelhouse, to which it is linked throughout its length by a slab roof covering a passage round the wheelhouse. Broch, wheelhouses and wall are of masonary of similar quality. The excavations yielded no finds which would determine their relative ages.

^{page 23 note 2} As was pointed out as long ago as 1935 by Prof. Childe in his Prehistory of Scotland. He states on p. 90 of his Scotland before the Scots, 1946, that the unreported excavations of Aikerness, Orkney, have proved the lowest stratified buildings accessory to that broch to be contemporary with it.

^{page 23 note 3} A series of Roman coins going back to Vespasian (p. 18 supra). The coins were scattered among the outbuildings and were not a hoard (J. Anderson, II, 244n.).

^{page 23 note 4} Curle, A. O., P.S.A.S. LV, 83 ffGoogle Scholar. It was Dr Curle who first pointed out that the wheelhouses provided the clue to the roofing of the brochs, idem p. 93.

^{page 24 note 1} Arch. Scot. V, 207 ff.Google Scholar; P.S.A.S. LVI, 172 ff.; R.C.A.M., Orkney and Shetland, No. 1206. The Commission tentatively classify the radial piers as tertiary on the ground that they are not bonded into a facing wall, which in turn is not bonded into the wall of the broch and, on that ground, is classified as secondary. This argument would prove equally that the piers are primary, and it is reasonable to suppose that they are so if indeed they are founded on the original floor as drawn. The fact that the tip of one overlaps the cover of a. sunk tank is not decisive, since we have knowledge elsewhere of radial piers being extended and modified, no doubt when a roof was being replaced. But certainty is not attainable on the evidence. Other brochs showing substantial survivals of radial walls are: Clickhemin, Shetland, where there is some evidence of their primary character (Arch. Scot. V, 205Google Scholar); Levenswick, Shetland (Anderson, op. cit., fig. 207); Freswick Links, Caithness (R.C.A.M., Caithness, no. 34). At Ousedale Burn, Caithness (P.S.A.S. XXVI, 355), there were remains of corbels at eight foot distances the one from the other at the level of the scarcement and the stump of a wooden post in the original floor was thought to represent a roof support.

^{page 24 note 2} Supra, p. 23, note 1.

^{page 24 note 3} Op. cit., 1935, 237 ff. and 1940, 247 ff.

^{page 24 note 4} The original dating of Chysauster to the 2nd century B.C. was questioned at the time of the excavation by Mr Radford (as recorded by Dr Hencken on pp. 266 and 277 of his report in Arch. LXXXIII) and Mr O'Neil has shown (Ant. J., 1936, 300Google Scholar) that subsequent work on courtyard houses in Cornwall and Wales makes it probable that, in these areas, the type belongs entirely to the Roman period.

^{page 24 note 5} All the hearths at Porthmeor and at Caerau were in the rooms (J. Roy. Inst. of Cornwall, XXIV, App. II and O'Neil, op. cit. supra). The evidence from Chysauster was indecisive. At Wiltrow (see below) they were at the entrances of the rooms.

^{page 26 note 1} Wheeler, R. E. M., Maiden Castle, 55 and 96Google Scholar.

^{page 26 note 2} Radford, C. A. R., Castle Dore, Interim Report, Cornwall Excavations Committee, 1937Google Scholar.

^{page 27 note 1} Bersu, G., P.P.S., 1940, 78 ffGoogle Scholar. For the plan, see figs. 21 and 25; for the ultimate reconstruction, see Antiquity, June, 1946, p. 81Google Scholar and Plates V to VII.

^{page 27 note 2} Some such radial framing seems structurally necessary to tie these two rings of posts to resist the outward thrust of the roof.

^{page 27 note 3} For the three houses in the Isle of Man, see Ant. J., 1944, 152Google Scholar and J. Manx Museum, 1945–1946, 177 ff.Google Scholar; for the house at Scotstarvit near Cupar, Fife, see P.S.A.S. LXXXI (forthcoming). Site A in the Isle of Man produced a clay floored house 38 feet in diameter with a wall of close stakes, two internal rings of concentric posts and a central hearth. This stood in an irregular enclosure some 85 by 75 feet surrounded by a palisade banked externally with clay; the site was marshy and the enclosure was paved with logs. Dr Bersu, while regarding the inner area as designed for dwelling space and the outer for stock, believes that the whole was covered with a single roof. It seems unnecessary to draw this rather strained conclusion, if the postholes in the outer area can be explained on the assumption of detached buildings for the stock. The Scotstarvit house had, it is understood, double walls of timber suggesting the internal and external stone facings of broch walls.

^{page 27 note 4} O'Riordain, S. P. (P.P.S., 1946, 164)Google Scholar points out the analogy between the Isle of Man houses and some raths. As he indicates, however, it is premature to claim that Iron Age raths necessarily exemplify this house form; some indeed had several houses within the outer wall. It is now generally recognised that the small ‘ring-forts’ of Ireland were farmsteads rather than forts, despite their outer ‘ramparts’ –as it is argued in this paper the brochs also were.

^{page 28 note 1} For Dr Curle's final conclusions on the long and complex culture sequence at Jarlshof, see P.S.A.S. LXX, 249–51.

^{page 28 note 2} Op. cit., 1940, 185, 211 and 249.

^{page 28 note 3} V. G. Childe, op. cit., 1935, 187.

^{page 28 note 4} This is hardly a paradox, for travelling bronze workers can have had little incentive to follow their trade to remote and inhospitable islands so long as more profitable markets were open to them in the richer lands. In the latter part of the first millennium their trade in such lands declined and, like travelling tinkers today, they sought a living corresponding with their lower status in peripheral areas. Outside Jarlshof the Royal Commission can cite only two bronze implements from Shetland (R.C.A.M., , Orkney and Shetland, I, 59Google Scholar).

^{page 28 note 5} Dr Curle regards the earliest pottery as of Iron Age type (P.S.A.S. LXIX, 104) and this is certainly the impression gained by examining the pottery as a whole. Prof. Childe has called attention to the similarity of steatite urns from Dwelling no. V to an urn from the two-tier cremation cist at Little Asta in Shetland; he suggests that these small urns ‘fulfil the function’ of Food Vessels, and accordingly that the first stage of Jarlshof was ‘systadial with the Food Vessel culture’ on the mainland (op. cit., 1935, 16, 17). However that may be, there is no evidence for dating the Little Asta cist, or urns of the Little Asta type, and it does not seem that they can be of any service for dating the beginning of Jarlshof.

^{page 29 note 1} Watson, W. J., Celtic Place Names of Scotland, 59, 60Google Scholar. Some crannog finds argue for a settlement of Iron Age B peoples in south-west Scotland but, in the absence of dating evidence, a suggestion that the crannogs represent a step in the migration of the broch people would be mere surmise. The Royal Commission have suggested (Orkney and Shetland, 31) that the non-circular broch, their ‘galleried dun,’ ‘formed part of a typological series of galleried structures of which the broch is the most elaborate form.’ This view, which is based on the supposition that the circular brochs were all broch towers, has no excavation evidence to support it and seems to be rendered improbable by the facts of distribution. The non-circular brochs are a handful of structures confined to the extreme southern tip of the broch area and on present knowledge can best be regarded as peripheral variants.

^{page 29 note 2} Craw, J. Hewat, Proc. 1st Int. Congr. of Prehist. and Protohist. Sciences, 1932, p. 285Google Scholar. The classification of the many unpublished querns from brochs would be valuable for chronology. Adopting the classification proposed by DrCurwen, in Antiquity (1937, 135 ffGoogle Scholar. and 1941, 15 ff.), the Scottish rotary querns would fall into the Hunsbury type with fiat grinding faces. The examples Dr Curwen gives from brochs are of shallow beehive shape and generally analogous to those from the Roman stations of Newstead and Castlecary; they look late in the Iron Age B series, as do those from the Road Broch of Keiss and Cam Liath illustrated in P.S.A.S. XXXV, 138 and Arch. Scot. V, Pl. XVI. English examples of this series begin well back in the 1st century B.C. and the examples from the Roman stations must date between A.D. 80 and 180. On this information we can only say that the latest likely stage for the introduction of the rotary quern into the brochs is the 2nd century. Prof. Childe has however proposed a derivation of the broch rotary querns from an independent Spanish family which might have arrived in the North as early as A.D. 1, and such a dating is almost required to explain the querns with flat grinding faces from the ring forts at Cush, Co. Limerick, excavated and discussed by Prof.O'Riordain, (Antiquity, 1943, 25Google Scholar; P.R.I.A., 1939–40, 160–4).

^{page 29 note 3} Dun Troddan, : P.S.A.S. LV, 90 and 92Google Scholar; Dun Telve, : P.S.A.S. L, 252 and 254Google Scholar; Mousa, : Arch. Scot. V, 210Google Scholar and P.S.A.S. LVI, 182.

^{page 30 note 1} It may be noted that the hitherto unexplained high-level second scarcement at Dun Telve, which cannot have supported a roof, would be intelligible on the view put forward. A continuous scarcement, it may be added, is not needed merely to support the ends of rafters, which could rest in the normal manner on corbels or in sockets in the wall.

^{page 30 note 2} Dr Curle regards Dun Troddan as of 2nd century date at earliest on grounds of comparison of the hearth form with Traprain and of a yellow bead with similar beads with 2nd century associations at Traprain and Castlehill (P.S.A.S. LV, 93).

^{page 31 note 1} Seven if we include Culswick and Burraness which, as has been said, are more probably of Class II. Of other candidates for inclusion in Class III the strongest is Clickhemin in Shetland; compare the Royal Commission's plans (Orkney and Shetland, no. 1246) and Sir Henry Dryden's plans of 1855 and 1866 (Arch. Scot. V, Pl. XVII–XX). The upper part of the broch has been twice restored and judgment is difficult, but the balance of evidence is that it belonged to Class II.

^{page 31 note 2} By Capt. Thomas R.N., in the case of Dun Carloway, (Arch. Scot. V, iii, 414)Google Scholar. He assumed a 70-hour week and a height of 34 feet; for higher broch towers, and with walls thicker than the 12-foot walls of Dun Carloway, the effort involved would be considerably greater.

^{page 31 note 3} For a professional military opinion on the defensive limitations of the broch towers, see the remarks of Gen. Lefroy in a report on Cockburn Law broch (supposed to be a broch tower) in Berwickshire Naturalists' Club, 1879–1881, 81 ff.Google Scholar

^{page 32 note 1} The two Glenelg towers may serve in illustration. They are on a coast even now almost uninhabited where high mountains descend directly to the sea. They lie some two miles up a narrow glen and a quarter of a mile apart at opposite ends of the only strip of cultivable land, an area insufficient to support their inhabitants who must have depended on hunting and fishing or on sea raiding. For the latter purpose the retreat is well chosen, being closed from landward by the high mountains, open to sheltered water in the Sound of Sleat, but totally hidden from seaward by a high moraine and a sharp bend at the foot of the glen. The area is wooded and, against these broch towers, attack with scaling ladders would have been possible.

^{page 32 note 2} P.S.A.S. LXVI, 282 and LII, 249.

^{page 32 note 3} R.C.A.M., Caithness, no. 459.

^{page 32 note 4} P.S.A.S. LXVIII, 513.

^{page 33 note 1} Brögger, A. W., Ancient Emigrants, 63 ff.Google Scholar; Adamnan, Vita Sancti Columbae, Lib. II, Cap. 43.

^{page 33 note 2} Castletown Broch, near Thurso, a grassy mound 7 feet high, contained an unburnt burial with brooches dated by Brögger to the latter half of the 10th century (R.C.A.M., Caithness, no. 320, and A. W. Brögger, op. cit., 132). The large, low mound which Okstrowbroch in Orkney had become was used as a cremation cemetery at a date which cannot be later. It contained ‘a great number’ of small flag cists enclosing ‘burned bones and ashes’; one with a steatite cremation urn and one with a bronze Viking ring brooch (R.C.A.M., Orkney and Shetland, no. 11, and Brögger, op. cit., 131). Other brochs reduced to mounds and used for burials were: Kintradwell, Brounaben, Yarhouse, Thrumster and Dunbeath (Anderson, op. cit., 223–6) and Wester Broch, Keiss (supra).