48 PROCEEDING E SOCIETYTH F O S , 1946-48.

V.

SOME OBSERVATION E BROCHS . TH GRAHAMA N Y O SB . , M.A., F.S.A.SCOT., F.S.A. Read May 12, 1947.

SUMMARY OF CONTENTS. PAGE . INTRODUCTOR1 Y ...... 8 4 . . NUMBER2 DISTRIBUTIOD AN S BROCHF NO S ....9 4 . . ARCHITECTURA3 L FEATURES: (i) Entrances ...... 54 (ii) Guard-cells ...... 56 (iii) Mural CellBasad san l Galleries ....7 5 . (iv) Upper Galleries and Stairs ...... 62 (v) Scarcements ...... 67 (vi) Voids ...... 71 (vii) Roofing and Wall-heads ...... 71 (viii) Wells and Tanks ...... 76 (ix) Hearths ...... 7 7 . 4. SHAPE AND DIMENSIONS: (i) Ground Plan ...... 7 7 . (ii) Diameter ...... 78 (iii) Height ...... 0 8 . . EXTERNA5 L DEFENCES ...... 7 8 . . RECAPITULATIO6 N ...... 0 9 . APPENDIX (List of Brochs, etc.) ...... 91

1. INTRODUCTORY. The main features of the typical are very well known, and general informatio e subject difficulth no n s o i ncomo tt t e by.1 Further detailed descriptions of individual brochs are to be found here and there

1 E.g. Anderson, Pagann i Times: the Iron ff.4 ; 17 Age, Antiquity,. pp ff. 0 , 29 . volpp . i . K.O.A.M., Inventory of ,d an . ff.1 Nosii 3 ; . vols. pp 263 . i . , 553; iii. Nos. 1149, 1206, 1246. SOME OBSERVATIONS ON THE BROCKS. 49

literature,e iinferencee nth th d an s regarding their origi datd nan e that

ca mose drawe lighe nth b th f to tn n receni 1 t evidence have been discussed Professoy b leavre s soow Childe.e reala s eth t na f generalitie mo Bu 2 s we encounte considerabla r e bod factf yo s whic muce har h less familiao t r antiquaries thesed an , , thoug t leah no importano dt they y ma con w -ne t clusions, are yet relevant to any attempt to fill in the outline sketch and same th et a time posses a certais n intrinsic interest purpose e Th th . f eo present paper is accordingly to make some of these facts conveniently accessibl o futurt e e student e subjectth f o s ; they relate mainle th o t y distribution and to the physical features of the brochs, while a list of structuree th s themselve gives Appendix.sn i a n i 3 The material on which the paper is based has been taken for the most part fro publishee mth d description brochse th f t thesso bu , e descriptions have been eke t witou d h information gaine n somi d e case personay b s l visits to the monuments themselves, and in others by discussion with observers whose experienc s beeha en wider tha herd owny nm I eAn . desire to express my particular indebtedness to Professor Godfrey Thomson, D.C.L. , e statisticaD.Sc.th r fo ,othe d an l r calculation goos s wa thad e h t enough to make on my behalf; to Mr A. O. Curie, C.V.O., LL.D., F.S.A., . HP .. M WatsonG r , F.R.I.B.A., R.S.W.. T . ,S F.S.A.Scot.. C r M d an , Calder, A.R.I.A.S., F.S.A.Scot., who have given me the benefit of their intimate acquaintance wit ha ver y large numbe f brochso r Professoo t ; r Childe. VG . , D.Sc., D.Litt., F.B.A. ,Richardson. F.S.A.S . J r ,M , F.S.A.Scot., H.M. Inspecto Ancienf ro t Monuments, Professo Piggott. S r , B.Litt., F.S.A., Mrs Piggott, F.S.A. d other havo an , wh se give e informatiom n a n o n e Royavarietth o f pointst lo y Commissiod an ; e Ancienth d n o nan t Historical Monument f Scotlano s r permissiodfo f certaio o maknt e eus n unpublished data.

. NUMBER2 DISTRIBUTIOD AN S BROCHSP NO . It will be convenient for the purposes of this paper to regard Scotland as being divided into the following six regions, shown on the map in fig. 1: 1 E.g. Archceologia Scotica, vol. v. pp. 71 if., where references are given to eighteenth-century notices of certain brochs; ibid., pp. 341 ff. and 365 ff.; Beveridge, and , pp. 73 if.; Proceedings of the Orkney Antiquarian Society, numerous articles on the Orkney brochs; Proceedings of the Society of Antiquaries of Scotland, numerous papers, of which the most important are: vols. xxxv. pp. 122 ff.; S.;0 ff.11 4 ;iWiii d 44 R.C.A.M. an . pp . if ff.1 3 ;8 24 . .Iv. Countpp I.pp . y Inventories, especiallf yo , , Orkney and Shetland, and the Outer , Skye and the . These Inventory articles contain many references to descriptions in P.O.A.8., P.S.A.S., and elsewhere; while for areas which have not yet been inventoried some literary references have been given in the Appendix thif o sd paperen e th . t a 2 Prehistory of Scotland, pp. 197 if., and Map IV; Prehistoric Communities of the British Isles, ff.6 . f ; 24 8 Scotland. 12 pp d an . f before 9 Scots,e 8 th . pp 3 This list bring dato t earlien p ea su publishe e ron Antiquariesde inTh Journal, vol. xxiii. , Nos2 , 1 . embodyiny ff.9 b 1 , . pp e resultgth f field-worso k done sinc e earlieeth t r alsI s drawlis o. wa t nup includes a number of "comparable structures." 4 VOL. KXXXI. 50 PROCEEDINGS OF THE SOCIETY, 1946-48. . ShetlandI . OrkneyII ; ; III. Northern Mainland, mainlani.e.e th d lying north of a line drawn from Gruinard to Tain; IV. West Coast and Inner Islands, comprising the western coastal areas from Gruinard to Kirkcud- bright with the islands of Skye, liaasay, Tiree, Mull, , etc.; V. Outer Islands . CentraVI ;d Easteran l n Mainland r virtuallo , e resf th yo t

Fig. 1. Sketch-ma-p of Scotland, showing subdivision into Regions.

the country. The quantity of material available, and its distribution among thes regionsx esi shows i , Tabln ni , whileI e detailed information, with references to hterature, will be found in the Appendix. e tablth en I opposite, "brochs e "structure ar (col ) 6 . s positively identified as brochs and still showing remains; "broch sites" (col. c) are SOME OBSERVATION E BROCHSTH N O S . 51 e site th f vanisheo s d structures know havo nt e been brochs; "uncertain examples" (col e .structure d)ar s som f whico e h probabl d otheran e syar possibly ma brochse yb , though none have been identifie s sucha dd an ; "comparable structures" are not brochs but embody certain features which also appear typically in broch architecture. With the unimportant excep- tions noted below under (i) and (ii) the foregoing classification follows the published accounts of the structures, and in particular no attempt has been TABLE I.—NUMBERS AND DISTRIBUTION OF BROCHS.

Totals of Uomparable Region. Brochs. Broch Uncertain cols. sites. examples. (6) to (d). structures.

(«) (&) (c) (d) (e) (/)

I. Shetland 51 30 14 95 2

II. Orkney 42 20 43 105

III. Northern Mainlan. d 169 17 41 227 2

IV. West Coas Inned an t r Islands 28 21 49 28

V. Outer Islands . 8 20 28 6

. CentraVI Easterd an l n Mainland 6 2 8

Totals . 304 67 141 512 38

mad discuso et r appraiso s distinctioe eth t presenna t recognised between the brochs and the broch-like "galleried duns," or to discuss how, if at all, they are related. For this far too little evidence seems to be available as yet, and the most that can usefully be done is to sound some notes of warning. The followin gnotee b point n connectioy i d ma s n wit e foregoinhth g table:— (i) While the figures are not exactly the same as those given by Childe

or by the Ancient Monuments Commission,2 the differences are small an1 d Prehistorye Th 1 of Scotland,. IV p 2 Ma Inventory of Orkney Shetland,d an . 31 . volp . i . 52 PROCEEDINGS OF THE SOCIETY, 1946-48.

affect only point f detailso wil s explainee A b l . e nexth tn d i sub-section , rather more "uncertain examples" have been recognised in the Inner and Outer Islands than were include Commission'e th n di s estimate. (ii) It would be wrong to pay too much attention to the fact that the figure n columni s s regulay t stan(6)an no , n o d i (d)(c)d r an rati, o ot one another, or to their totals (col. e}, in the several regions. This apparent irregularity is probably due in great part to special circumstances that affected the compilation of the records. Thus the comparatively high proportion of "uncertain examples" recorded in Orkney may be due to the possibility—always present to the minds of the Ancient Monuments Com- mission's officers—of unexplored mound e chamberesb turnino t t ou gd cairns rather than ruined brochs; this was probably felt to be of greater urgency in Orkney than in either Shetland or the Northern Mainland. The proportion of "uncertain examples" is also high in Regions IV and V, but here again possible explanations exist. Region IV, for example, contains islane th f Tireedo , wit hseriea monumentf so s which have bee poorlo ns y described that their true nature is doubtful, but which are far more likely to be brochs tba.n a.nything else: and also Mull, where some structures exist which canno positivele b t y identified without excavation n RegioI . , nV again, discrepancies appear between the Ancient Monuments Commission's recorn o listdat d Captait y db san apu n Thomas. t seemI 1 s clear thae th t Commission's surveyor certain i , n cases, classifie s "dunsa d " ruins which becomd ha e quite featureles s visi t whice dathi bu th tf eo y hb s Thomas, more than fifty years earlier seed ha ,n reaso identifyo nt s brochsa , . Some "uncertain examples, t recognise"no Commissione th y db , have accordingly been admitted here. The absence of "broch sites" in Regions IV and V likewise suggest sa differenc e method th e Commission' n th i e f o s s survey from those employed in Regions I, II, and III. The "total" figures, however s givea , n columni n som (e),o towardy g ewa s evenin t thesgou e accidental discrepancies. (iii) All the totals given are likely to be below the numbers that originally existed. For one thing, brochs are particularly liable to destruction by farmerbuildersy b d an s thesa , y provide large quantitie f usefuo s l stone e oftear nd situatean n gooo d d agricultural lands r thifo s; reasoe nth rate of their destruction will have been exceptionally heavy in the better agricultural districts, such as parts of Orkney and Caithness. Again, no brochs are recorded on the long stretch of coast between Loch Broom and the Kyle of Loch Alsh; this area has not been surveyed by the Ancient Monuments Commission and is also difficult of access to casual antiquaries; therefory ma t i thae eb t some further examples remai discoveree b o nt n di this district. (iv) While nothing woul gainee db attemptiny db expresgo t densite sth y ' Arch. Scot., vol 36. . v5.pp ff. SOME OBSERVATIONS ON THE BROCHS. 53

of brochgivey an nn s i regio basia n f theino so r number r squarspe e mile, the heavy concentrations occurring in the Northern Mainland, in Orkney, and in Shetland immediately leap to the eye. The concentration in the Outer Islands will likewise appear fairly heavy if account is taken of the poor qualit e e landWesf mucyo th Th f .tho Coas d Innean t r Islands, however, show a considerable falling-off; while Region VI, which com- prisecountrye wholresth e e f sth th o tf e hardlo s i pictur ,e th n y i t allea . (v) If it is assumed that any type of structure is to be found in its greatest numbers and highest development at or near its place of origin, the foregoing facts would be enough to prove that the broch was a product of Regions I, II, or III, or possibly of Region V. They would, however, hardly provid ecriterioa thesf preferrino r ne fo othersee on fouth y o t r g.an But in view of the broch-builders' admitted ability to move about freely by sea, it would clearly be unsafe to be guided by any such assumption in the present case, as the broch might easily have been transferred, by sea- going people, at an early stage of its history, from its place of origin to another accessible locality—in which, again, it might later have enjoyed a long and eventful life. It will therefore be best simply to note the distribution without venturing to define its significance. (vin contrasI ) whao sais t t dwa t above under (iv), e nearlth l al y "comparable structurese th n o y "sa occutha, o t V s Regionn i ti rd an V sI Western Mainland coast or in the Inner or Outer Islands. Consequently, if any of these structures were assumed to be ancestors of the broch, we should hav supposo et e that this latte firss trwa evolve weste th n d,i perhaps e Hebridesinth importes , thawa t i t d thence mainlane th th o et d dan thad t an ,subsequentli t y attaine greatess dit t numerica- de l velopmen thesn i t e localities. This assumption, however, canno t presena t t be based on any more solid grounds than superficial resemblances in points of structural technique e proximit vien th i f wd o an , Irelandf yo n , witow s hit schoo f gallerieo l d building t wouli , unwise db drao et w premature conclu- sions about the broch's place of origin from mere considerations of typology and distribution. (vii) The total number of brochs is sufficiently large to make it clear that, in the area and period of their most general use, they must have been an extremely important social factor. It must not, of course, be assumed that all brochs are of the same age—indeed, brochs are sometimes placed so close to one another that the opposite conclusion seems probable, while the technical skill shown by their builders is so great as to argue long experienc dry-stonn ei e construction Howd Wesd Mi ean . t n HoweDu r o , TelveTroddan Du ,d mighnan suggestee b t cases da pointseriee n si th r so , occurrin e nortth n h o g e Loc shorth f Harrayf ho o e . Moreover, thaa t broch might collaps earln ei y time proves swa excavatioy db t Gurnessna ; while Mid Howe had to be shored up, though at a late stage in its career, 54 PROCEEDING E SOCIETYTH F O S , 1946-48. rebuildind an g seem havo t s e taken plac t Kintradwellea . Oxtro, again, must have already been reduced to the condition of a grass-covered mound by the date at which Norsemen used it as a place of burial; and other similar examples could probably be cited. On the other hand, there is nothin n theii g r respective state f preservatioo s o suggest n t that, e.g., and are not more or less contemporary. However be—any thisma answee e probledth th o onln t r mgivee ca yb n by the spade—it is clear that the broch is not to be regarded as a mere architectural freak but, as Professor Childe has already shown,1 possesses an historical significance which is comparable, in its way, with that of the mediaeval castle. 3. ARCHITECTURAL FEATURES. (i) Entrances.—Some sort of information, not always very full, is available abou entrancee tth s of ninety-six broch ''d uncertaisan n examples''; t nothinbu g like this numbe goon i e d ar rpreservation lintele th f d so an , entrance-passage th stile epositioar n li lesn i s tha ndozea n cases. Normally thera singl s i e e entrance only e entrance-passagth ; e passes radially through the wall and has a "guard-cell" on one or both sides; there are checks in the sides of the passage to receive a door, often with stone slabs for door-jambs and a transverse kerbstone for threshold. The door-fittings e frequentlar y t beindoublese e g on place, d outsid e doorwayeth e th f o s guard-cells and the other inside. The doors were evidently secured by massive wooden bars, sliding in holes of which the inner ends sometimes bacn ra k int guard-cellse oth passagee Th . clearle ar s y intendee th r dfo strict control of entrants, being rarely over 4 feet in width and frequently s narroa fee2 inche6 s wt a r les so t thei sa r outer ends, beind 2an g subject to further constriction where the door-jambs project from the passage walls. The foregoing normal arrangement is, however, subject to modification certain i n casesfirse thesf o Th t . e consist possessioe th f so morf no e than one entrance entranceo tw s a , recordee sar eacn di sevef ho e n caseson d an , broch (Clickhimin threes t thiha ) Bu s .departur e fro generae mth l ruls ei more apparent than real, as the secondary character of one of the two entrances is certain in twTo brochs (Ness and Yarrows), and seems to stand o reasot thren ni e others (Brounaban, Keiss Keisd an , n s Road)Du t A . Fhiadhairt, too secone th , d entranc possibly ema secondarye yb , thougo hn trac disturbancf eo e abnormalls existsi t i s ,a y smal alsd olan lead t passou t the base of the stair where the wall could have been pierced with least difficulty and with least danger of collapse. Nor is it easy to see why a 1 The Prehistory of Scotland, p. 204. 2 A number of samples taken from the "certain" brochs gave the following average breadths: Shetland fee2 , t 3i inches; Orkney fee3 , inches7 t ; Caithness fee3 , t J inch; Sutherland fee2 , t 11 inches; Oute Inned ran r Islands fee2 , inches8 tpassag e Th . t Bverleea recordes yi havins da g been narrowe foo1 inche6 o tdt s betwee door-jambse nth . SOME OBSERVATION5 5 E BROCHSTH N O S .

primary entrance should have bee placeo ns rendeo t s dstaire a th r , which was presumably an important defensive feature, so easily accessible to intruders.1 The remaining-example, on Freswick Links, has now completely disappeared, but a photograph 2 preserved in the Library of the Society of Antiquarie Scotlanf so d support idee sth a tha entrance th tbas e th et ea of the stair was secondary. Of the two extra doorways at Clickhimin, one, as at Dun Fhiadhairt and in some of the Caithness examples, passes out bottoe staire th t th a mf ,o whil othee eth r opens fro mpassaga e where eth thickness to be pierced is not great. Both these last are also above ground- level, and neither possesses any of the normal features of a broch entrance.

There3 is therefore nothing in any of these cases to modify the accepted view that the typical broch possesses one entrance only. Other divergence e lesar ss important n ninI . e case e passagth s s i e not on a strictly radial line, but either runs somewhat obliquely through the wall or is slightly curved (e.g. Coldoch and Tor Wood).4 Frequently, again, the passage is not of equal width throughout, but expands slightly behind the checks (e.g. Netlater), and this plan may be'varied further by a contraction at the inner end (e.g. Dun Telve). At Dun Ard an t-Sabhail passage th broadess ei outes it t a trapparentl e endTh . y pointless arrange- men f placino t e checkgth s insid e guard-cellseth , whic e s notehi th t da Castl f Bothicaeo t Daia d l nan Langwell explainee b n ca , supposiny db g tha outen ta r door existe "wells da , outsid checkse eth t tha frams ,bu tit s ewa wooden and has perished. This suggestion is supported by there being a cavit woodea r yfo n door-jam e Hil Worksf th t o woulli t ba d dan , alst ofi the cases in which there are no checks at all (e.g. Dun Ard an t-Sabhail) or only one (e.g. Lingro). Some miscellaneous feature interesf spacee so th e sar t left betweee nth lintels of the entrance-passage (e.g. Mid Howe), perhaps to provide for defensive action from a room over the passage; drains running out from the court, under the passage (e.g. Nybster); and the use, as outer lintels, of massive blocks of stone triangular in elevation (e.g. Culswick). Four such lintels survive a fift s recordei d h e vanishean , th t Alascaign a d dDu ; 5 and while no doubt partly ornamental, they may also reflect the idea that a lintel's maximum bending centr momene s span th it thad t f ea t o an , i t s i t

1 This last consideration will also appl Brounabanyo t , Keiss Keisd ,an s Koastaire th questiof dn i si n• themselvee ar s regarde. primary65 s d a . p thin e O sse . positioe Th 2 Freswicf no k Housbackgroune th n ei thif do s photograph proves tha norte th t h point "bees ha n reverse e publisheth n di d plan (R.C.A.M., Inventory of Caithness, . 14)p . s Gordon appears (Itinerarium Septentrionale, . 166p havo )t e entered Dun Telve throug holea h" " secone levee th th f t lo d a gallery (of . whil82)d p . an ; e wely thislma have been mad previoua y eb s explorer, or have resulted from the decay of the structure, it is just possible that it may have been a subsidiary entrance like those see t Clickhiminna . 4 This latte locaa r e featurb l t beepeculiarityy no ema ns noteha t i ds ,a elsewher e than in theso etw examples, whic onle har y some thirteen miles apart. s The triangular block found outside the broch at Keiss had no doubt originally served as yet another example. 56 PROCEEDINGS OF THE SOCIETY, 1946-48. consequently e strengtheneneedb o t s t thida s pointdiscovere Th . f yo socketed stones in some excavated entrance-passages shows that in these cases the doors turned on pintles. Though broch doorways are of course larger than those of the Stone Age settlement at Skara , it is remarkable to see how they reproduce what is virtually the same arrangement of jambs, sills, and bar-holes, (ii) Guard-cells.—The positions of the guard-cells are known in eighty- four cases, and analysis of these shows that what is often thought of as the normal arrangemen e cellsth eithen f ,o o t i.e. e e rentrance on th sid f eo - passage, in fact occurs in only eighteen cases (21 per cent.); a single cell occur forty-foun si r cent.)—thirty-fou casepe 2 s(5 e r cent.th pe n 0 o ) (4 r left—whils r cent.entrant'hi pe n 2 o ) (1 twenty-twn n sei te righ d tan o cases r cent.(26pe ) there .arguard-cello n e t alla s . Provisio s thuni s made, where guard-cells exist r engaginfo , g attacker thein o s r ope r shieldlesno s side in fifty-two cases (62 per cent.) and on their covered side in only twenty-eight'(3 cent.)r geographicae 3pe Th . l distributio f thesno e several arrangement shows si Tabln ni . eII

TABLE II.—GUARD CELLS.

Cell on Total defence on attackers' No cells on Totals Region. entrance- of cols. open covered both open covered passage. side. side. sides. side. side. (b) (c)+ (a) (6) (c) (d) (b) + (d) (c) + (d) (e) (d)(e)+ +

I. Shetland 2 1 3 1 4 7

. OrkneII y. 5 1 5 10 6 2 13

III. North Mainland 21 2 8 29 10 10 41

'IV. West Coast and Inner Islands 4 5 2 6 7 4 15

V. Outer Islands . 1 1 1 2 2 •• 3

VI. Central and Eastern Mainland . 1 1 1 2 2 2 5

Totals . 34 10 18 52 28 22 84 SOME OBSERVATIONS ON THE BROCHS. 57 e Oshap nth d siz f guard-cellsean eo , however, evidenc s onlei y forth- coming from thirty-eight brochs, and the impression that it gives is very generavariedn i celle e Th sar .l round, oval, elongated-oval, sub-oval with straightish sides and inner end, or squarish; but there seems to be no ruling cello system f differentw o s d an , t siz d shap e samoccuy ean th ema en i r broch. Squarish cells, however, seem to occur only in Orkney, with the possible addition of one at Torwoodlee; and in Orkney likewise are the only certaio tw n cases,1 HoweGrurnesd Mi guard-cellf d ,o san s whic internalld hen y in complete mural galleries. One broch in Shetland (Clumlie) has a guard- cell wit hsecona d door opening int courte oth ; elsewhere this arrangement occurs only at in Skye, at Ousedale Burn in Caithness, and at East Kinnauld in Sutherland.2 Otherwise the cells in both Outer and Inner Islands are chiefly oval or round, and are often small; and the oval or elongated-oval plans prevail in the other districts. At Hillhead (Caithness) both cells are long and slightly curved, tapering at their inner ends. (iii) Mural CellsBasald an Galleries.—The arrangemen f murao t l cells, other than entrancethose th lengthf t o ea d gallerf ,an so y where these occur at ground-floor level, is so varied that it defies summary description. Moreover, manbroche th f ywhico n si h trace cellf basasd o san l gallerien sca still be made out are so ruinous that no detailed measurement or planning of these feature s beeha s n possible, whil n additioi e n e oldesomth f ro e description e regrettablyar s ' imprecise t I mus. t therefore y sufficsa o t e that the mural cells are generally of beehive form, and are rarely lintelled; that they are usually round, oval, or elongated to a club-like form; and that a single entrance from the court may often give access to two cells, or to a cell and a staircase, opening to right and left. The cells are unlighted, and a few contain aumbries. Some details, however givee b n n ca ,regardin openinge gth s leadino gt cellse th thess a , more ear e adequately recorded tha celle nth s themselves. Thusconfine w inquirr f i ,presen e eou th r yfo brocho t whicn si h cells alone occu t ground-levela r r cello , s with length f gallere o sb o shoro t ys s a t comparable with cellsd postponan , e discussio f thoso n n whici e h true gallerie poin n founde ca sar fifty-threo t e t w , e structures—of whicw fe ha are '' uncertain examples''—in which the number and positions of openings leading to mural cells can be discovered or confidently inferred. "Position" is here taken as meaning position in the circuit of the wall relatively to the broch-entrance d accordinglan , y those cases—mentione e previouth n i d s section—in whic e originahth l positio e broch-entrancth f no itsels ewa n i f doubt have been omitted from the count. The number f cell-entranceo s r brocpe s s theha y e severaoccuth n i r l Regions are shown in Table III.

1 Sandray anothee (Barrab y structure r)ma th case ruinout o to bu , s certaintyer i sfo . 2 It can be paralleled in Ireland, in the fort on Lough Doon, near Ardara, Co. Donegal. 58 PROCEEDING E SOCIETYTH F O S , 1946-48. TABLE III.—NUMBER ENTRANCEF O S MURAO T S L CELLS.

Numbe f cell-entranceo r s per broch. Number Region. of brochs considered. 1. 2. 3. 4. 5.

I. Shetland 2 1 1 1 5

. OrkneII y 2 1 2 5

III. Northern Mainland . 15 10 1 1 27

IV. West Coast and Inner Islands 4 3 3 10

V. Outer Islands 1 1 2

. CentraVI Easterd lan n Main- land .... 2 1 1 4

Totals . 25 16 7 4 1 53

From this tabl efea w miscellaneous facts firse appearth t n placeI . , cleas i t i r tha mose th t t usual arrangemen singla s i t e entrance; this very commonly leads to a little lobby off which the stair opens to the right and cela somf o l lefte th sor . o t t This arrangemen tr cent accountpe 7 , 4 r sfo of all the cases shown in the table. Another common arrangement is that which provides two entrances, as this accounts for a further 30 per cent. of the total number of cases. It will be noted, too, that the weight of these figure Northere s th lie n si n Mainland entranceo tw r eithes o a , e s on roccu r twenty-fivn i twenty-sevef o t eou n examples drawn from this region, while the West Coast and Inner Islands, with a total of ten cases, show this arrangemen n onli t y seven. Other arrangement f entranceo s e fairlar s y evenly distributed, thoug e fivhth e cells foun t Feranacha d , which leave only an aggregate length of 47 feet of solid walling in a circuit of 155 feet, seems to be something of a freak. The fact that three openings occur in only seven cases deserves to be noted, as this sometimes tends to be regarded as the normal arrangement in consequence of its occurrence in the much- described Broch of . SOME OBSERVATIONS ON THE BROCKS. 59 The manner in which the cell-entrances are disposed about the courts s showi nooooo diagrammaticall e numbe th n figyi , 2 . f caseo r n whici s h each o oooooB o ooH o oooM N o Q S oooRP o W V U

Pig. 2. Disposition of cell-entrances. The figure in the centre of each diagram indicates the number of case whicn si correspondine hth g arrangemen founds i t .

arrangemen s founi t d being indicate e e centre figurth th th f n y i o edb corresponding diagram. These diagram e inferenceth drawe d b an s o t sn from the e subjecm ar a certai o t t n degre f inaccuracyo e publishea s a , d description will often state simply tha n entranca t s "oi e left, th n " 60 PROCEEDING E SOCIETYTH F O S , 1946-48.

"opposite, "or right,"e o nth whed "an n such recordt accompanieno e sar d by plans the reader is apt to interpret them more formally than the writer may have intended. This tendency may result in the exaggeration of the numbers of entrances given as in the "IX o'clock," "XII o'clock," and "III o'clock" positions. Subject, however suco t , harisy errorma e s a s

from this cause, the figure1 s suggest that, except in the case of diagram F, there is little to choose between one arrangement and another in respect of frequency of occurrence. That one-half of the twenty-five single entrances (A to F), and one-third of all the ninety-nine entrances considered, should o'clockX "I e occuth " n positioi r , howevernis , wortha f notingi d an ; simpler comparison wer e singlth e l madeal "left-handf eo "e th wit l hal single "right-hand" entrances resultine th , g proportio seventeef no e th f no forme lattee fivth o rt f reo woul evee db n more striking preponderance Th . e of "left-hand" entrance f courso s si e tie witn di e commo hth n practicf eo placin e stai th fooe thi n f i rgth o tse circumference parth f o t alludo T . e . once more to Mousa, the arrangement of cells found there and represented by diagra moccurP thren i s e cases only, Edinshal e Easth td Brocan l h of bein othee gth r "IItwoe Th I. o'clock" position occurs twenty "XIe timeth Id o'clocksan " position fifteen times, wit resule hth t thae tth o'clock,X "I " "XII o'clock, "IId "Ian o'clock" positions, taken together, account for 67 per cent, of all the openings noted. For basal galleries the evidence is at once less plentiful and less easy to interpret, owing to the. difficulty of determining whether lengths of gallery observed in a ruinous structure are really fragments of an originally continuous gallery, or separate lengths of gallery, or even long, gallery- like resul a cellss thesf A o t. e uncertaintie classificatioy san bouns ni o dt be somewhat arbitrary, but, subject to this qualification, the material may be distributed among the several Regions as in Table IV—a distinction being made between galleries whic knowe har havo nt e been complete, i.e. continu- ous all round the circuit, those which may or may not have been complete, and those which are known to have been partial, whether associated with cells or not. A number of uncertain examples have had to be ignored altogether. Scant thes ya y are figuree table ,th th en i s suffic shoo t e w that celld san basal gallerie notsometimee s i sar s ,a s thought, regular alternative features; basal galleries are, in fact, a much less common arrangement. They further suggest that basal galleries, whether complete or partial, were chiefly in favour in the Inner and Outer Islands, where fifteen out of a total of twenty- five examples occur—a point to which attention has already been drawn by the Royal Commission on Ancient Monuments.2 Of the galleries known havo t e embrace whole dth t Gurnes a ee s circuie courton th wa s e f th o ,t entered at either end from one or other of the guard-cells; the one at Mid

1 In the diagrams the broch-entrance is placed in the position of VI on the clock-face. 2 Inventory of the , Skye and the Small Isles, p. xxviii. SOME OBSERVATION E BROCKSTH N O S . 61

TABLE IV.—BASAL GALLERIES.

Galleries Partial Galleries which may galleries, known Region. or may with or Totals. to be not be without complete. complete. cells.

I. Shetland 1 2 1 4

. OrkneII y 3 1 4

III. Northern Mainland 1 1 2

IV. West Coast and Inner Islands 1 4 4 9

V. Outer Islands 1 5 6

VI. Central and Eastern Mainland .

Totals 5 8 12 25 guard-celothere e th d on s f a an ,o r t quitd Howfa no l en s fron ea e era m th could also be entered by an opening from the court slightly above ground- t Eedlana e on levelbees de ha n th ; demolished detailo n knowe d sar an , n except that an "encircling gallery" was found by the excavators; for Houbie we are dependent on Low's plan, which shows the gallery in a

diagrammatic style; and for Kingsburgh detail1 s are likewise lacking, but the gallery was entered from the court in the "II o'clock" position. It is perhaps worth noting, in conclusion, that three of the foregoing five brochs appear to have lacked stability—Grurness fell down at an early stage in its history; 2 one sector of Mid Howe had to be buttressed and thickened, and the basal gallery filled up ;while Redland had to be propped

up with an outer facing-wall, though not3 apparently until long after its first

TourA through Islandse th of Orkney Schetland,d an . 169p .

1 2 B.O.A.M., Inventory of Orkney, No. 263. 3 Ibid., . 553No . 62 PROCEEDINGS OF THE SOCIETY, 1946-48. construction.1 If, therefore, the Royal Commission on Ancient Monuments is right in suggesting 2 that the hollow wall may have been a structural device rather than a provision for sheltering the broch's inhabitants, the

inference might possibly be drawn that the device proved to be a bad3 one when used in the basal storey, and that the comparative rarity of complete basal galleries was due to a timely recognition of this fact by builders. (iv) Upper Galleries and Stairs.—The upper storeys of nearly all the brochs have suffered so much damage that our knowledge of the galleries and stairs that they once contained, as well as of the other features originally

existin uppee th n gri wallse part th necessarils f ha so , inferree b o yt d from 4 comparatively few examples, as shown in Table V. TABLE V.—SURVIVA UPPEF LO R GALLERIES.

Highest storey of which anv traces survive, not counting ground-floo. r . 1st 2nd 3rd 4th 5th 6th Total

Numbe brochf o r . s 23 1 3 1 1 29

Actually the material is even scantier than the table suggests, as fewer than half of the examples in the lowest class provide any useful evidence. We really depene followinth n do g structures alone: floor)h Mous6t o ,(t a Telvn 5th) o Du CarlowaTroddan en (t Du ,Du 3rd)d o n(t yan , Clickhimin (to 2nd), and Burray (E.), Carrol, Caisteal Grugaig, Dun Cromore, Knowe o' Burristae, Levenwick, Loc n Duna ha Howed d YarrowMi , an , o 1st)(t s . Our present-day accountTroddan Telvn Du Du d ef n s5o an 6 can, however, be eked out with the descriptions recorded by Gordon in 1720 7 and by Pennant in 1772,8 from which some facts—or at least some strong prob- abilities—ca extractee nb criticay db l reading.9 e comparativelTh y perfect conditio e Brocth e f f Mousth o hno d an a frequency with which it has been described combine to create the impression that its arrangements constitute the norm, any divergences from which

are to be classed as exceptional. It is true that at Mousa the stair does not B.C.A.M.1 , Inventory of Orkney, No. 320. 2 Inventory Outer e oth f Hebrides, etc., . xxxixp . whicn O 3 h . poin als71 e . otse p reasone Th believinr sfo g tha l brochal t s originally stood high enoug contaio ht n several galleries consideree ff0 ar .8 . 4pp n do 6 P.S.A.S., vol. 1. pp. 241 ff. 6 Ibid.,. volff 3 8 . . Ivpp . ' Itinerarium Septentrionale, 16. . 6pp f 8 A Tour of Scotland, vol. ii. pp. 339 ff. 9 The special importance of these accounts in their bearing on the former height of the towers will appear when this subject is discussed in Section 4 (iii). SOME OBSERVATIONS ON THE BROCHS. 63

first-flooe th groun e begin th i t n no rd bu gallery, while other stairs beginning markedly above the ground are noted in only six other cases ; but, in so

maie th fans a rconstructiona l feature concernede ar s singla , e stair rising 1 througp regularlto e th hserie a o t y gallerief so s which remain,e eveth t na

top, large2 enough to serve for passage is ordinarily accepted as typical of

brochs as a class. The rest of the evidenc3 e can consequently be best studied by means of comparisons with Mousa. A first point of divergence occurs at Dun Telve and . It has been suggested that the rough condition of the inside of the Mousa

galleries, whic s parallelehi 4 a singl n di e lengt f passagho t Ch'ckhimin,ea

points to their not having been intended for use5 as passages ;but however

galleriese th bee s , ha be n facs n thiy a said,i te s ma largar , e enoug sucr hfo h 6 same th t euse A tim. et sho theno wregulao a y d r diminutio sizn i e froin below upwards firse th t s lease a galler, th ts headrooyha e th allf d mo an , fourth and fifth are both narrower, in places, than the sixth. At Dun Carloway and Dun Telve, however, different conditions obtain. At Dun Carlowa battee outee yth th f r o r pronounce o wals s i l d that whil firse eth t galler fee2 inche6 s yti s floore leve it widthire th f th o ,t l e a d measures only 12 inches and 8 inches in width at floor and roof respectively. The third gallery is also obstructed by two bonding-slabs which span it at about hals heightit f . Similarln Telve e intra-muraDu th ,t a y l space narrows markedly between the floor of the first gallery, which is 2 feet 6 inche seconde sroo e th broad f th fo d , whic,an onlhs i fooy1 inches.6 t e Th 7 second gallery is still large enough to have been used as a passage, and the face botf neatlso e har y finished uppee th t r bu thre, e gallerie onle sar y about 12 inches wide and have their inner faces rough—facts which suggest that galleriee th s abov e seconeth d wer t meaneno r ordinar t werfo t bu ee yus primarily structural features whild An e .Gordon' s statement tha walle tth s e fourtth f ho Troddann gallerp "closedto Du e t ya th , t jusa " s thea t y "joined together" at the top of Dun Telve, is very unlikely to be true seeing that the third gallery is to-day over 2 feet wide at about half its height,8 it does suggest that the vanished portion of this tower must, like 1 Clickhimin in Shetland; Burrian (Russland), Gurness, and Mid Howe in Orkney; Carrol in Sutherland; Wester Broch in Caithness. 2 The sections appearing in R.C.A.M., Inventory of Shetland, figs. 533 and 534, show that neither of the two minor landings that do in fact interrupt the even rise of the stair is more than about 4 feet in length. 3 They actually var breadtn yi h fro mfoo1 inche6 t fee3 inches6 to s t id n an heigh, t fro mfee4 t to 5 feet 6 inches.

4 B.C.A.M., Inventory of Orkney and Shetland, vol. i. p. 33. 5 B.C.A.M., Inventory of Shetland, p. 70. 6 Ibid., p. 52. ' P.S.A.S., vol. 1. p. 247. The photograph on p. 249 shows how noticeable the contraction is in the second gallery. 8 P.S.A.S., vol. Iv. p. 86, fig. 3. But the debris of a ruined fifth gallery piled on the lintels of a considerably narrowed fourth might well create this impressio inexperiencen a n ni d observer's mind. On this question see also p. 81. 64 PROCEEDING S E SOCIETYOTH F , 1946-48.

the upper galleries of Dun Telve and Dun Carloway, have been very difficult of access. From all this it seems to emerge that the upper galleries at Mousa, notwithstanding their internal roughness, lend themselves to use as thapassagey t wa othe those a th n si t r ea three broch questioa notd o sd an ; n consequently arises as to whether this difference is something more than an accident actualld an , y mark distinctiosa n betwee differeno ntw t varieties of structure. A subsidiary but connected problem may be mentioned here. Gordon states that at Dun Troddan the stair went "to the top," but he does not carries wa explait di througw nho e continuallhth y narrowing galleries— let alone out on to the wall-head above the supposed convergence of the walls. This could perhaps have been providelocaa y b l r widenindfo f go the intra-mural space along the spirally ascending course of the stair; or else we must suppose that, notwithstanding Gordon, the stair actually ended somewhere much lower than the wall-head and that its functions were taken laddea oy n b woode r ro n step difficultse walle affixeface th Th f th .e o o dt y suggest sa furthe r distinction from Mousa, wher t woulei d presumablt yno have arisen, resulting fro e differencm th e treatmen th e uppen i eth f o rt galleries that has been mentioned in the preceding paragraphs. A third point of difference between the and certain other examples is perhaps to be seen in the logical simplicity of the former's structural arrangements. At Mousa all parts of the intra-mural space are immediately accessible from the stair, the galleries being open and unobstructed. Some other brochs, however designee ar , lesa n sdi straight- forward manner. At Dun Telve, for example, the first gallery is blocked solia y ofdb f wall fro space mth e abov entrance-passagee eth s terminait , l portion being also divided from the remainder by transverse slabs and having to be entered from the gallery above by two openings in its roof. Howed Mi t , A althoug arrangemente hth s have been obscure dilapiday db - tion, there seem havo st e bee nkina entresof do l galler south-southe th n yi - east sector, intermediate between the ground-floor and first-floor galleries. At Dun Carloway, in Thomas's time, the second gallery, or according to his numbering the third, seems to have been stopped by a sloping wall of transverse lintels, whic mistooe hh blockea r kfo d stair,1 instea openinf do g on to the back of the stair as at Mousa; the third gallery is also barred by two slabs extendin Cromorn Du g t acrosA e Thoma . sit s appear havo st e found an inaccessible space between the under-side of the stair and the sloping, lintelle e guard-cellth df o end-wal e on —unlesss f o l , indeed, this

end-wall was in fact, as2 he stated it to be, a flight of stairs descending from e first-flooth rBaravan galleryDu t e noteA h t . d even more complicated

arrangements, t thesbu deale b nee t t dno wit Baravan h herDu s s ea i t 3 1 Arch. Scot., vol. v. p. 385. 2 Ibid., vol. v. p. 380 and pi. xlix. 3 Ibid., vol. v. pp. 391 f. and pi. xlviii. SOME OBSERVATION5 6 E BROCKSTH N O S . listed as a "probable" broch only.1 At Levenwick, at Dun Troddan, and at Caisteal Grugaig the stair does not rise steadily, as it does at Mousa, interruptes i t i t levea bu y db l stretch of galler t first-flooya r level; while t Caisteaa l Grugaig, further sectioa ,e first-floo th f no r gallery must have been blin t eitheda r end beine ,on g othee underth staie rd th blocke an r d by a chamber overlying the entrance-passage. Finally at Clickhimin the arrangements are entirely abnormal, being correlated with the exceptional syste f entrancesmo . Her estaia r rises fro mpassaga e which piercee th s north-west sector 5 feet above the level of the court and leads out through o subsidiartw e th on f yo e entrances; while froe othemth r subsidiary entrance, in the north-east sector, a passage slopes gradually upwards, levels off into a gallery, and ultimately opens on the back of the stair above the passage in the north-west sector. As providing another important variation of the arrangement familiar at Mousa, attention must be called to four brochs, all in Caithness, which possess two stairs, one on either side of the court. The fact that, in three of these cases (Brounaban, Keiss Road, and Keiss), one of the stairs rises from beside a duplicated entrance has raised a doubt as to whether both f theso y e stairan structures r e primaro ar s l al re-examination A yi . f no e sitesth , mad 1946n ei , cas freso n t h ligh thin o t se questionth s a t bu ; insertion of a secondary stair in a standing broch would seem to be technically impossible safs i assum o t et i , e that both stair primare sar everyn i y case— the more so as a double stair is in itself a very reasonable arrangement. Thus ample evidence exists to show that the supposed "norm" of Mousa was in fact frequently varied, and to give grounds for believing that brochs were less fully standardised t leasa , mattern i t detailf so , thaappeay nma r from summary descriptions. Ther gooo n s dei reason, however supposr fo , - ing tha broca t eves hwa r constructed withou staiy t staia an allta r s a ,r was necessar t onl accesr ywall-heae no yfo th o st t alsodbu , most probably, e supplfoth r e builderf stonyo th o et s durin procese gth construction.f o s 2 e absencth Non f stair-treadca reo s fro e ruin a brocmth f so regardee hb d negativs a e evidence eas e vien i ,th ef w o wit h whic hdestroyede thesb n eca , or covered up by debris.3 To conclude this section somethin e baseth sw g ho mus o t saie b s t da staire oth e disposef ar s d abou e courtsth t e stai ,e basth and th f r eo s ,a may be visible even in quite a ruinous broch, a fair number of examples are available. The figures are given in Table VI. This table shows that by far the largest number of stairs—thirty-one

peculiae 1Th r features note BaravaThoman y Cromorn db Du Du d t sa te an before 1890 wero en longer visible whe officere e B.C.A.Mnth th f o s . visited these site 1921n si t Thomas',bu s descriptions and sketches are too positive to be ignored entirely. Compare his descriptions as referred to above with E.C.A.M., Inventory Outer e oth f Hebrides, respectively1 etc.,7 d Nosan 8 3 . . B.C.A.M.e Se 2 , Inventory of Orkney Shetland,d an . 32 vol . p . .i 3 E.g. t Brounabana , sinc publishee eth d plaprepareds nwa . 5 VOL. LXXXI. 66 PROCEEDINGS OF THE SOCIETY, 1946-48.

TABLE VI.—DISPOSITIO P STAIR-BASENO COURTSN SI .

Position Stair-basesf so . (Entranc foroc o e"Vt n i s Ihi o'clock" position.) Region. Totals.

Left Opposite Right Left and VI-X. X-II. II-VI. Right.

. ShetlanI d 1 1 1 3

II. Orkney . 5 2 7 '

III. Northern Mainland. 14 1 2 1* 18

IV. West Coast and Inner Islands 7 2 9

V. Outer Islands 4 4

. CentraVI Easterd an l n Mainland 4 4

Totals . 31 8 5 1 45

three Th *e doubtful example somittede mentionear 5 6 . .p n do

totaa f forty-five—risf o o l t ou left-hane th n i e d segment,1 whil othee eth r two-thirds of the circuit contain less than half that number between them. And that thi dues si somn i , e degree deliberate th o t , e choicbuildere th f eo ; t simplno s i preponderance yresula d th f an o t f "left-handeo " cells that was noted above (p. 60), is shown by the fact that in fourteen of the thirty- one cases cells existed elsewhere than in the left-hand segment, from any of which the stairs could presumably have risen as easily as from a position o'clock X s possibl i betweed t I an . I e V ntha t this arrangemens wa t designed to expose the unshielded side of an intruder making his way to e staire furtheopenino th footh e N f th .o t t rg a inferences can, however, be drawn with safety from the table, as the rest of the figures are too scanty.

1 Between VI and X on the clock-face, the broch-entrance being at VI (cf. p. 60). SOME OBSERVATIONS ON THE BROCKS. &7 A final point of interest connected with the stairs is that, in seven cases foregoine th ou f o t g forty-five entrance th ,staie th eitheleves e ri o et th t lra scarcemene oth f f beloi , e scarcementwor tth , well abov e levee th eth f o l court. This is necessary at Gurness and Mid Howe, where the whole of the ground floor is occupied by a continuous gallery, and natural enough at Mousa, where the ground floor is a good deal taken up by large mural cellst BurriaA . n (Russland), again place staie o circuie s th , th s r i n di t that started ha t ii fd from ground-leve t mighi l t hav difficultd eha clearinn i y g the top of the right-hand guard-cell. At Clickhimin, Burray (E.), and Wester Broch, however, no reason for this .arrangement suggests itself. (v) Scarcements.——While it is naturally impossible to be certain that every broch originally possesse dscarcementa thay trus sa i t o t virtuallet i , y no broch can be proved never to have had one at all. Possible exceptions t Burriana e ar , , wher peculiaea t probablbu r y analogous arrangement exist . 68) t (p Yarrowssa ; , wher e secondarth e y reducing- wal facln i doe t t appeasno havo rt e covere scarcementa p du t Alla d t an ; Duin a Burnessd nan , where reducing-wall scarcementd san s see mhavo t e been confused in the published descriptions. Apart from these cases, scarcements have been recorded in something like forty brochs and, while a number of others in which no scarcements have been noted still stand to a sufficient height for the scarcements to have been preserved, when these apparent negative example analysee ar s s founi t di d tha evern i t y case some circumstance exists which upsets their value as evidence. For example blockeinterioe e b th , y drma with fallen masonr innee th r ryo face maskee b y secondary db ma y building, whil t Acharolea whole eth e record is suspect. Scarcements must consequentl givee yb importann na t place amon broch'e gth s constructional features. Some detail f twenty-fouso r scarcement forthcomine sar g from twenty- two brochs, two of the brochs in question, Mousa and Dun Telve, possessing two scarcements apiece.1 Gordon records an upper scarcement at Dun Troddan Pennand an , engraving-bus t hi show n i t si positioa n i t n whert ei certainly does not exist; it is possible that he may have depicted it at too lo elevationn wa oncs thad wa e t an ,carriei t wale para th lf n o dwhicto h has fallen since his time (p. 81), but it is also conceivable that Gordon confuse Telven e featureTroddan th dthad Du Du an ,f d t o sPennann an t followed him in transferring the latter's high-level scarcement to the former. The probability of an upper scarcement having existed at Dun an Ruigh Ruadh, perhaps some six feet above the existing one, is mentioned on p. 72. The commonest type of scarcement is probably an intake in the wall-face somewhat broadened by corbelling,2 but the published accounts

ledg1A e which appears abou fee2 1 t t abov groune eth circuie parn dth i f t o Clickhimi ta n (R.C.A.M., Inventory of Shetland, fig. t 558regardeno )s i trua s dea scarcement. 2 I am indebted for this piece of information to Mr 0. S. T. Calder. 68 PROCEEDINGS OF THE SOCIETY, 1946-48.

do not always make this clear, as they frequently allude to an intake without further particularisation. Another arrangement intakn a , e "with corbelt sse upon it at certain points, is found at Culswick, Mousa, Sallachaidh, and Ousedale Burn t Culswica ; Mousd k an corbele ath foue numbersar n i r d an , are equally spaced out around the circuit, while at Sallachaidh corbels are found on two of the three surviving portions of the scarcement. Consideratio e heightth f no f scarcement o s s somewhai s t complicated face byth Telve t n thaoccuo Du mentiones ttw t a ,Moust ra a d aan d above, even if Dun Troddan be ignored; and also by some uncertainty as to the original ground-level unexcavaten si d brochs. High-level scarcements like Telvn originally e Du seema e t on na e yth have existe n almosi d y an t broch except perhap Carlowayn Du s , disappearing wit e dilapidatiohth n of the upper parts in the manner already suggested in the possible case of Troddann Du ; comparisons will consequentl valie yb d onl betwees ya e nth single scarcement lowee th d r membersan pairse th f s.o The twenty-one scarcements that are available for this purpose range in height fro mfee4 inche6 t t Clickhimisa fee2 inche1 Know6 e t o nth t t sa ' eo Burristae largese th ; t numbe examplef o r 7-fooe sth (sixn i t e clas)d ar san the average height is 8 feet. It seems unlikely that this average would be raised by more than perhaps a foot if the original floor-level was laid bare in every case. The examples are too few in number for valid comparisons to be made between the various Regions. The two scarcements at Mousa are 7 feet and 12 feet 4 inches high respectively, and thus seem to be in a different class from those at Dun Telve, which are at 6 feet 6 inches and 29 feet 6 inches respectively. If an upper scarcement ever existed at Dun Troddan it must have been more than 25 feet above the ground—the greatest height of the surviving part of the structure—and would con- sequentl compareye neighbourine b fal th o t t l a e d Telv n on wit e gDu eh th rather than with the one at Mousa. The exceptional arrangement thas mentionewa t d abov s existina e g at Burrian, North Ronaldsay, has been described as follows: "At 3 feet abov flooe scarcemena eth s i r inche5 t depthn si , above-whic wale hth s i l gradually intake heigha r fee4 n fo inchesf 2 tto thed an ,n projects abruptly foinches4 r , thereafter rising vertically." 1 Thi s clearlyi se sam th t e>no e scarcementthinth s ga s described s functionaboveit t bu , , like theirs, havy ma e bee provido nt esuppora somr fo t e construction projecting from the wall; it is possible, for example, to imagine the recessed zone of the wall-fac servins e a anchorag n a s ga strutr efo triangular so r trusses, so f I . the height of the supported construction would have been rather more than 7 feet. Any discussio f whao n t structure e scarcementth s s supported must

B.C.A.M., Inventory of Orkney, No. 193. 1 SOME OBSERVATION9 6 E BROCKSTH N O ' S .

proceed in the light of Dr A. O. Curie's report on his excavation of Dun Troddan, where he found a ring of post-holes roughly concentric with the wall-fact somse fee6 ed tan e withi r Curin satisfies D it. ewa d thae th t

post-holes belonged1 to the broch's earliest period, and inferred that the. posts supporte corridoa rooe f dth o f r encirclin e courtg th e inneth , r edge of whic e scarcemenborns th hwa y e wall-faceeb th n o t . This conclusion, in fact, seems certain, and the corridor, which Dr Curie found reason for believing was divided from the court by a light partition, thus probably forme inhabitantse dth ' ordinary dwelling-space, while also perhaps serving to obstruct the approach of intruders to cells and stair. But two important questions still remai considerede b askefirse e o nt b th ty dn placI ma . t ei whether the roof of the corridor was a roof and nothing more, or whether t alsi o serve balcona s da gangwa r ysecone o th n traffic;i r d y d fo plac an e

whether the arrangement2 that obtained at Dun Troddan was necessarily the same in all brochs everywhere. firse Toth t questio Troddan nDu n itself provide direco sn t answert bu , notee b o dt s tha i scarcemene t i th t onls i t feey6 t abov grounde eth d an , that consequently a roof sloping down from it at even so gentle a pitch as woul3 n i d1 leave headroo littlf mo e moree thath e sid f th e o feen4 n o t corridor toward courte sth . This point tells somewhat agains idee f th tao a sloping veranda-roof though without disproving its existence; again, if hypothesie th broce th f hso being itself roofe favourabls di y regardede th , nee r roofinfo d e corridoth g r appears less urgent. Some more positive evidence is forthcoming, however, from nine brochs 3 in which doorways or large voids are found to open on to or just above the scarcements.

These openings must pretty certainl slopino t n o gt y roofhavno d sele but 4 ohorizontao nt l balconie stagingr o s whicn so inhabitante hth s could live, ot leasa r t move about. Again veranda-rooa , heighe uppee th th t f a fo rt scarcement at Dun Telve (29 feet 6 niches) would not have been effective as a shelter; while at Mousa, where two scarcements occur within a few anothere on fee t f leaslowea e o t e th , t on rmus t have supporte dbalcona y and not a roof. In this case, in fact, there would seem to have been two corridor galleriee above th otheon e e t eth ar muc sn se rs i a sam e y hth ewa withi e broch-wallsnth thid an s, analogy seem increaso t s likelihooe th e d e roofo th f othe o s r corridor s wela ss havina l g been available th o t e inhabitant movementr sfo . Finally t Clickhimina , Howe,d MousaMi ,d an ,

1 P.S.A.S., vol. lv. (1920-21), pp. 90 ff. 2 Tha structure tth e carrie low-levee th n do l scarcemen whole th roo a r s efo f court twa suggestes a , d by Sir W. Lindsay Scott, D.S.C., P.S.A. (Proceedings of the Prehistoric Society, New Series, vol. xiii. p. 10), seems quite unlikely, as this theory makes no provision for the structural features discussed in the next paragraph. 3 Burray (E.), Burroughston, Caisteal Grugaig, Clickhimin, Dun Telve, Gurness, Knowe o' Burristae, Mid Howe, Mousa. 4 At Mousa a doorway gives on to each of the scarcements. A third opens 4 feet 6 inches above uppee th r scarcement fro mlandin a stair e th n .go 70 PROCEEDING E SOCIETYTH F O S , 1946-48. perhapd an s elsewhere, ther bees eha nspaca e abov entrance-passage eth e accessible only throug voie hwhicth y db innes courthe openit d th en rn so ; if these spaces were put to any use they can only have been reached by ladders or from a balcony at scarcement-level, and of these two alternatives the latter seems to be preferable in view of the inconvenience of a ladder, as this woiild inevitably have obstructed the opening of the entrance- passage. Ther s thui e s good reaso r supposinfo n g thae structurth t e supported by the scarcement was often, if not always, something more tha nsimpla corridoe eth roor fo f r below. For the second question, as to whether a gallery of the Dun Troddan type necessarily existed everywhere a negativ, e answer suggests itselfs a , some evidence can be quoted against the idea that the structure resting on the scarcement was invariably of the veranda type, whether sloping or difficults i flatt I r example. fo , imagino t , e Telvn thaDu te containeda high balcony supporte 30-foon do t high-levef i poles d an ; l scarcemente sar supposed to have been commonly constructed, we should be faced with the further difficult e supplth f yo lonf y o g timber treelese th n si s regione th f so north. On the other side can be set the fact that a scarcement's normal function is to support the ends of beams and, while timbers long enough spao t nbroca h diametricall ylonge r lon s woulo a s ge a b r dthao t hav d n eha props for a high-level balcony, support for a balcony or partial floor could have been readily obtained by laving shorter timbers chord-wise round the circuit. Thus four timbers each abou fee1 inche2 t3 t lengtn si h would have made a square frame-work within a broch 30 feet in diameter, the radial distance of their centres from the face of the wall being about 4 feet 5 inches. A six-sided or eight-sided frame could have been constructed of corre- spondingly shorter timbers. And evidence that frames of this kind were used at Mousa and at Culswick may perhaps be seen in the corbels set in the scarcements of these two structures—six at Mousa and four at Culswick— t beini g supposed that these corbels were principal support whicn o s e hth cornerframe th f eo s rested face tTh .tha post-holeo n t s simila thoso t r e at Dun Troddan have been reported from any other brochs might be quoted as suggesting that supporting posts were n facti , , exceptionalo n t bu ; dependenc placee b sucn n do eca h negative evidence seeing tha suc) (1 t ha point would hardly have been notice oldee th rn d i excavations onle th y) (2 ; three brochs excavated since Dr Curie's discovery (Grurness, Mid Howe, d Kilmsteran ) were fulf secondaro l y structures e buildinth n i , f whicgo h l evidencal earlien a f eo r circl f posteo s might well have been obliterated. Moreover, if the function of the Dun Troddan posts had been performed in these case stony b s e supports s migha , t well have happene a regio n di n where trees were rare e masonrth , f theiyo r bases would almost certainly have been large enough to require demolition in order to clear the ground secondare foth r y structures. SOME OBSERVATION1 7 E BROCKSTH N O S . (vi) Voids.—Examples of the characteristic window-like voids, either single or forming vertical flights, survive in more than twenty brochs. They were probably a regular feature of broch-architecture in general, and it is no doubt onl accounn yo e dilapidatio th f o t e uppeth f nro e partth f so structures that many more examples are not on record. There seems to be no reason to question the usual theory that these voids were designed to reduce the weight on lintels; but in cases (e.g. Dun Telve, Dun Troddan, Carlowayn oDu r ) wher e flight t eth base f voidno o doorwaysn e o d ar s s leading into court or cells they may be supposed to have performed the function of lighting or giving access to the galleries—the latter, perhaps, b yladdea r woodeo r n stepslowese Th . t membe a simila f o r r flight a t Mous largs ai e enoug havo ht e havbeey nma edoora beed an , n approached by a ladder from a scarcement balcony 4 feet 6 inches below. This question possesses some interest as it bears on the function of the galleries; provision for lighting would indicate that some parts at least of the galleries were inhabitante th y b intende e werd t purelus s an r eno dfo y structural—as si also suggested by the distinction between rough and smooth finishing of the internal face . 63)(p s . Mousa alone show peculiae sth r featur fligha f dummf eo o t y voidsA . normal void occurs abov entrance e th staire t froth f bu ,emo this there rises a flight of small aumbry-like hollows which do "not penetrate the wall. merele b Thes y ye ma decorative difficuls i t i suggeso d t t an , practicay an t l purpose that they could have served unless, perhaps provido t , e pointf so support t varyin,a g heights timberr fo , s forming parrooa f to f (infra). (vii) Roofing and Wall-heads.—Brochs are commonly supposed to have been completely open at the top, but this is hard to believe in view of the obvious desirabilit f excludinyo g rain, wind d eneman , y missile—os t a r

least of reducing the size of any aperture by which1 these could enter. The positive evidenc e existenc th y kinr f an rooefo d, o howeverf is eo f , very meagre. One item is forthcoming at Mousa, where it is on record that "the

inner fac2 e of the wall of the lowest portion is vertical, while in the upper or galleried part the courses of masonry converge to some degree as they e summi towere th ris th o et f o t. The convergenc vers ei y markee th n di uppermos feet.0 1 t " This convergence hardly appear e publisheth n i s d

. HP .. Watsonsections,G r surveyeM o t wh , 4bu buildine dRoyath e 3 th r glfo Commission on Ancient Monuments, assures me that it is fully apparent to sixte levee th th f h o t l show s gallerya sketca s e ha e hi ey d nm n he i an , th t shouldI , however rememberee b , d that neither sling-stone r arrow-headno s s hav t beeye e n reporte1 d amon relice gth s foun excavaten di d brochs. * Boece's comparison, quote Arch.n di Scot., ancieno . vol192tw p f . o ,.v t building Boss-shiren si , which were presumably brochs, to bells, can hardly be stretched to mean that low-pitched roofs, corre- sponding wit crowne bellse hth th f ,o s wer e presen times hi n .i t Ubaldini, writing sixty years later

(1588 prettd an ) y clearly referring same th eo t structures , states that they wer(ibid.).p to ee opeth t na 3 B.C.A.M., Inventory of Shetland, . p.50 ' • Ibid., figs. 533, 534. 72 PROCEEDINGS OF THE SOCIETY, 1946-48.

field-notebook which brings thit quitsou e clearly. This convergence th f eo interior might be taken as suggesting that the tower ended in a false dome, forme encorbellationy b d t somethinbu , g wil e saib l d shortly aboue th t difficulties that attach to this theory. Evidence of similar convergence might also be held to exist at Dun an Ruigh Ruadh, where the inner face come markedln si y jxist abov e scarcementeth ; wale but th brokes i ls ,a n down on its inner face to a height of 5 feet 9 inches above the scarcement, it is impossible to say whether the encorbellation was intended to support a false vaul r merelo t foro yt uppen ma r scarcement, perhaps som feee6 t abov existine eth g latteonee Th .r alternative, however, seems preferable in view of the low elevation—the ruined wall-head standing only 7 feet 9 inches above ground-level on the outside.1 Another item conies from the Broch of Grurness, of which Mr J. S. Richardson, who conducted the excavations for H.M. Office of Works, has writte s followsa n d qualit:e ston ba d pooth ' ' f e Owinan eyo rth o t g construction, a downward and outward thrust, presumably from a heavy roof, forced the walls outwards, causing the passages to be crushed and the upper part of the building to collapse." He adds in a footnote that

similar displacement occurs to a greater or2 less extent in some other broch towers—the one that he has particularly in mind being Mid Howe—and that the most reasonable explanation for it is pressure from a roof-con- struction ;though he does not rule out the possible intervention of other

contributor3 y causes. It is probable that a roof of timbers covered with turf would have been heavy enough to produce the amount of thrust demande r Richardson'M y b d s suggestion, particularl e effecs th wa f yi t enhanced by wind-pressure, or by poor construction which permitted move- e thrusth ment t t bu develope; massiva y db e masonry dome, s suci s ha suggested by the signs of incipient contraction at Mousa, would naturally have been very much greater. In the absence of more definite evidence

the question must remain an open one; but it is to be hoped that, whe4 n next a ruinous broch is explored, care will be taken to examine the fallen material, for the sake of elucidating the circumstances of the structure's collaps estimatinf o d ean originae gth l natur uppes it f eo r portions. suppositione Onth , however, tha massiva t e false normaa dom s ewa l feature of all brochs, it is interesting to consider its probable size and features. That the dome, if it existed, was completely closed at the top is inherently most improbable, as this would imply that the whole of the interio permanenn i e towe s th f wa ro r t raty darknessan e t receivea r o , d

indebtem a I 1 thir dfo s informatio n. CalderEuign T a abou . n S . h. Du C tRuad r M o ht 2 B.C.A.M., Inventory of Orkney, p. 77. 3 Ibid., p. 79. objece Th t depicte Sueno'n do s Stone actuall,s evei t i nf i y brochbeea s ha ns suggeste,a . S . dJ br yD Richardson4 , throws no light on this point. Its top is flattish rather than domed or pointed, but the carving is too small and the stone too badly weathered to provide grounds for any inference. SOME OBSERVATIONS ON THE BROCHS. 73 no more dayligh e door—t th tha y b n a n i coulpropositio y dwa fins it d n which will hardly accord wit knowe hth n broch-builderse factth f so ' material culture even when allowance is made for the indifference of primitive people to the nuisance of smoke, and for the poor illumination tolerated, for example, in many mediseval buildings. On the other hand, a partial dome, which diminished the opening to a diameter of perhaps as little as 6 feet, would have allowed the inhabitants a reasonable amount of light—much

more, at any rate, than the same area of void distributed among, say1 , seven small vertical windows arranged in a kind of clearstorey—while providing a most difficult target for high-angle archery, if archery was in fact practised brochse byth ' inhabitants.2 The heigh curvaturd an t falsa f eo e e kindomth df eo suggeste d would depend on the average length and thickness of the stones used in its con- struction—the length of face of the stones determining the amount of over- han r projectioo g n obtainabl t eaca e h cours d theian e r thicknese th s corresponding rise. Professor Thomso kindls nha y undertake laborioue nth s calculations involved, and has found that, in theory, a false dome 15 feet in radius at the point of springing, and constructed of stones averaging 3 feet in length, could not be completely closed in less than about a hundred seved an n courses reducer o , aperturn a o dt feee 6 diamete n i t lesn ri s than a hundred and three.3 The foregoing dimensions were chosen as a kind of theoretical optimum, t Professobu r Thomso alss nha o mad esimila a calculationf o t se r meeo st t the special conditions obtaining at Mousa—a radius of 12 feet and an average stone onl feey 2 length. n i t thin I s cas t appearei s that theoreticalle yth

dome woul t contracdno 6-fooa o t t 4 t apertur lesn ei s than abou hundrea t d and fifty courses. The rate of contraction in each case is given in Table VII. If in the hypothetical case, where the diameter was 30 feet, the average ston supposes inche4 e wa e b thickness n si o d t heighe apee th ,th xo t woul d amount to 35 feet 8 inches, while the diameter would have decreased to 6 feet at a height of 34 feet 4 inches. At Mousa, however, the greater thickness of the average stone would give a height of at least 75 feet for the point of contraction to 6 feet. These heights, and particularly the one calculated for Mousa, are so great that they seem to rule out the possibility

' Mr G. P. H. Watson has pointed out to me that the Pantheon, which measures 142 feet 6 inches in internal diameter at ground-level and nearly 74 feet in height to the springing of the dome, is adequately lit by an aperture of this kind which is only 27 feet in diameter (Gwilt, An Encyclopedia of Archi- tecture (1899) . 916)p , aperturn A . f proportionateo e 30-foosiza n ei t broch woul aboue db fee5 t n i t diameter. 2 These considerations naturally apply to any type of roof. In the case of timber construction, the apex could simply be left without its covering. ' The amount by which the radius is reduced at each course is found by multiplying it by the cosine of halangle fth e subtende centree th t dthaa n ,i t courseaverage th y b , e stone. « E.C.A.M., Inventory f Shetland,o . 49 . p PROCEEDINGS OF THE SOCIETY, 1946-48.

of masonry domes, eve f contractioni n began well belo e wall-heawth s da s beeha n suggested above t beforbu ; e this inferenc s drawi e na furthe r possibility must be considered—namely that the rate of contraction may in fact have been more rapid than theoretical calculations suggest. It is assume purposee th r dfo thesf o s e calculations tha forware th t d cornerf so

TABLE VII.—CONTRACTIO FALSF NO E DOMES.

Hypothetical Broch of Mousa : No. of case: average stone course average stone 2 feet long. above fee3 t long. springing. Radius (ft. ins.). Radius (ft. ins.)-

0 15 0 12 0 5 14 8 10 14 3 11 7 15 13 10 20 13 5 11 2 25 13 0 30 12 7 10 9 35 12 2 40 11 9 10 3 45 11 3 50 10 9 9 10 55 10 3 60 9 9 9 4 65 9 2 70 8 7 8 9 75 7 11 80 7 3 8 3 85 6 6 90 5 8 7 8 95 4 9 100 3 9 7 0 105 2 5 110 11 6 4 120 5 7 130 4 9 140 3 10 150 2 8

all stones are laid exactly flush with the faces of the stones in the course below, but in practice, if all the stones were properly locked so as to obviate any danger of movement, they might perhaps be laid with their corners slightly in advance of the stones in the underlying course, and this would naturally result in an accelerated rate of contraction. And it is a fact that some scalyeara o esag mode constructes . Strachanwa lA r M y db , formerly SOME OBSERVATIONS ON THE BROCHS. 75

employe foremaa s da n masopreservatioe th n no brochf no H.My sb . Office of Works, in which the theoretical height of the dome was very materially reduced. Unfortunately the particulars of this model perished in a fire, but both Mr G. P. H. Watson and the late Professor R. Gr. Collingwood, who examined the model, were left with the impression that this method of roofing a broch might well have been fully practicable. On the other hand, the contraction noted at Mousa is certainly not of the same order as that exemplified by Mr Strachan's model; and as a dome on the lines of this model, wit hpossibla e heigh f somo t fee 5 r lesse2 o t f superimposei , d on the original structure at Mousa, would have brought the total height of buildine th somethino gt g lik feet,5 e7 even this modified theor e forth f mo y

of the masonry vaul1 t is still difficult to accept. saie b d y dome Agaithath ma f ei tt n i wer e constructe f corbeldo t se s radially, not chord-wise, to the circumference of the tower, a considerably quicker rate of contraction could be obtained, though at the price of decreased stability and increased weight of material. But the surviving example falsf so e vaults—e.g . murae thosth f eo l cells—provid evidenco en e that the radial method of encorbellation was ever used in the brochs, and Mr Strachan is known to have adopted the " chord-wise" system in con- structing his scale model. It is probably now too late to look for a solution of this problem, e excepunlikelth n i t y even f roofino t g material being identified amon debrie gth broca f so h excavate somn do e future occasion; e meanwhilth an onln n di ca e obviouy e th resew n o t s conclusion thaa t timber roof would have been very much lighte d easiean r o construct r t tha fory nvaultf an m o timbeA . r roof, pitche degreesangln 5 a 4 t f dea o , would have risen t Mousaa , , onl fee5 y1 t abov e wall-headeth fee0 1 tr o , less tha Strachan'r nvaula M f o t s type. However the tower was roofed, and whether it was roofed or not, it seems necessary to assume that it did not prevent access to the wall-head, failins a provisioy gan wall-hear nfo d defenc mose eth t solidly constructed broch must ultimately have becom edeath-trapa absence th n I e. of positive evidence we are naturally reduced to guess-work, but in the case of a tower which carried accessible gallerie wall-heae higs th sa s ha actuao dn l difficulty of construction need be supposed to have existed. Thus if the tower were open at the top, or if a wooden roof rested on the inner part of the wall- head e staith , r could have been carrie t int dou oa parapet-wal k without speciay an l measures having been necessary; whiltowee th f ei r endea n di dom e waleth k could e havstais hauncit th ern d o carriebee han t nse d throug masonre hth y ascendinn masa n si g mural passage. This arrange- ment could hardly have been use n suci dn n TelvhDu caseDu r o es a s Carloway, wher e uppeeth r galleries becam narroo eto r passagewfo t bu , an alternative one has been suggested on p. 64. heighe originae th th n f o t1O l . structur80 . p e ese 76 PROCEEDING E SOCIETYTH F O S , 1946-48.

(viii) Wellsd Tanks.—an Arrangements e supplth r r storagfo o y f o e water have been reported in twenty-six places and, except as noted below, there is no reason to suspect that they do not belong to the primary period of habitation e structureth f O . questionn si e Norther, thirteeth n i e nnar Mainland Shetlandn i e muc,o ar eleve Orkneyn o To i h . e tw shoul nar d an , d not be made, however, of this apparent concentration in two regions, as a well or tank is unlikely to come to light unless a broch is cleared out to ground-level s jusi Caithnes n i t ti d Orkned an , an s y that most so-called "excavation" has been done. Again, much of the underlying rock in these regions is at once easier to quarry and more suitable for slab-construction , tha Wese nth t Coas r Shetlano t d formations. The provisions made include covered well-chambers and cistern-like cavities, cut in the rock or partly cut and partly constructed of masonry, and on the surface slab-built tanks. Some brochs contain more than one of these type constructionf so . Eightee welle th cisterns r f sno o 1 have steps leading e waterdowth o nt , whil e remaininth e g nine 2 have noned Mi . Howe and Stackrue possess cavities which may not actually have been wells at all, but dry storehouses or cellars. The finest of the well-chambers of whic hdescriptioa bees nha t Gurness na e publisheon e , th wher s di e eth botto chambere th f mo , itsel fee8 f t feehigh4 1 t s i belo, e surfacewth d an , the stair has nineteen steps. The largest cavity seems to be the one at Skirza Head, which measure -wello feet.0 1 Tw fee7 s y 3 fee b 0 ty 1 sb t (Burray (E. Netlaterd )an ) were approache shory db t underground passages roofed with slabs t Keist Elsaa A d .water-hol sa yan places ewa d within the thickness of the wall, at the foot of the stair, the one at Keiss being additional to a well close by in the court. The wells at Hillhead, Skirza Head, and Burray (E.) are all just outside the entrances, the steps leading down to the one at Hillhead opening within the outer door-checks of a pro- longatio entrance-passage th f no whicy suggesy wa hma a n connectioea ti n with external and perhaps secondary buildings. The method of access to the communicating underground passage at Burray (E.) is unknown, and consequently nothing can be said as to its primary or secondary character. Mention may also be made here of two shafts, some 5 feet apart and sunk respectivel fee1 y inche1 fee3 9 t inche 2 td san s int grounde ofarth e th m t a , of Oust, Caithness,4 the former containing a steep stair and the latter seven

In the folio-wing sixteen structures: Burray (E.), Burrian (N. Ronaldsay), Gurness, Hillhead, Hill of Works1 , , Keiss, Keiss Road, Kettleburn, Kintradwell, Loch of Ayre, Mamie Howe (an "uncertain example" not yet described), Ness, Netlater, Oxtro, Redland. In the following eight brochs: Burroughston, Cam Liath, Dunbeath, Elsay, Mousa, Nybster, Ousedal2 e Burn, Skirza Head. 3 The so-called "Roman Bath" at Burghead is comparable with a broch cistern, measuring about fee1 1 t feeb0 y1 inche8 tfee 4 inchey 4 t sb deptn si h (P.S.A.S., chambee th t f.)1 volbu ;r35 .. ivpp . larges i it n se rwhic s ani dt hi deeper than anything surroundinfoune brocha th t d a an , g ledgs ei

also without a parallel. 4 R.C.A.M., Inventory of Caithness, . 455No . SOME OBSERVATIONS ON THE BROCKS. 77

shelf-like steps projectin ga e sit f froth eo sidee n mon o . e b Thes y ema otheo brochn t r bu structur, e survives. The slab-built tanks describes ,a publishee th dn i d accounts, bear sufficient resemblanc o othet e r box-like constructions, foun siten o d s belongino gt various prehistoric periods, to suggest that they may in some cases be connected with secondary occupation e brochth f questionn o si s ; eight,

however, may be noted as perhaps primary, of whic1 h one at least (Mid springa y b n ever I d . Howefe y e founs brocb wa ) o t dwhicn hi h these slab-built tanks occur there is also a well or cistern—or at Mid Howe the suppose cellary ddr . (ix) Hearths.—One or more large open hearths was probably a feature of the court of every broch, but so few brochs have been excavated with a proper regar r stratigraphfo d y that primar d secondaran y y hearths cannot usuall e distinguisheyb e recordsth n di . JSTo useful discussiof o n this subjec thereforn ca t undertakee eb t presentna .

. SHAP4 DIMENSIONSD EAN . (i) Ground-plan.—Although broch tower e commonlar s ye b sai o t d circular, a study of large-scale plans and of carefully prepared descriptions will show that the courts frequently measure rather more on one diameter than another, and that the even curve of the wall is sometimes broken by minor irregularities. The external outline, again, is liable to diverge from circle greatea th o et r degree tha thicknesinnere e nth th walle s a ,th s f si so sometimes far from regular throughout the whole of the circumference.2 A question may consequently be raised as to whether the normal ground- plan should be regarded as a circle or not. This questio answeree b followinaffirmativee e y th th n ma r n di fo gd an , reasons. Internal diameter hundrede recorn totaa o on r e f d do fo lsar an thirty-two brochs, and it is only in six of these cases that the major and

minor axes diverge 3 interior cent.r e fro mucmeape e s Th ma 6 th 4 y s nhb a s greae oth f t majorit thue nearlyar o s y round tha divergenca t e froe mth true circle would hardly appear to the unaided eye. Again, it must be remembered that rough drystone walls, even whe t purposelnno y battered, will hardly ever presen trula t y vertical face—especially after sufferine gth vicissitude o thousantw f o s d years—and that measurements taket a n irregular heights above ground-level oftes i s a n, necessar sita n ey o encum - bere debrisy db wely ma l, diverge materially fro true mth e dimensions sa

In the following five brochs: Keiss Road, Mid Howe, Nybster, Ousedale Burn, Skirza Head. 1 2 E.g. t Berwicka , Burroughstou, Cama Duinn sa , Clickhimin, Kiutradwell, Lamn b HeadDu d an , Torcuill. 3 Region I, 14; 11,16; 111,69; IV, 21; V, 6; VI, 6. • Cama r cent.Duinn sa pe 3 ;1 , Dunrobin r Wood cent.Colbosn pe Du 0 Bdinshalld ;1 , an t r pe 8 , cent.nn a MaighDu ;cent.r pe 7 ,, possibl resula s y a reconstructionf to ; Dun Cromore centr pe .6 , 78 PROCEEDING E SOCIETYTH F O S , 1946-48.

laid out on the ground by the builder. It will consequently be quite safe to regard the normal broch-plan as a circle, and to make use of mean diameter l caseal n si jusexcepx si t e mentionedth t . treatee Thesb y dema as abnormal, and left out of the present discussion. (ii) Diameter.—Although not always accurate for the reason that has just been given internae th , l diamete r morfa s ei r reliable tha e outenth r as an index of the size of a broch, as the inner wall-face is not battered. Material divergence fro e verticamth s onli l y shown wher e wall-facth e e recedes abov ea scarcement produco t , ea slightl y "cupped" effect when viewed in section, as in the cases of Mousa and Dun Telve.1 The external diameter, however, is considerably affected by the batter of the outer face, and decreases as the rising wall decreases in thickness. Consequently, in numeroue th s case whicn si base towea hth f eo obscures ri falley db n stones onld brokee an yth n stump seee b nn abovsca debrise eth outee ,th r diameter d wall-thicknesan n safel ca sassumee b y e les somy b b s o t de unknown amount tha nt ground-level a the e yar ; wherea e internath s l diameter, notwithstandin s shortcomingsgit takee b sufficientln s na ca , y accuratr efo practical purposes. The main facts concerning internal diameter, as observed in the approximately circular examples only, are shown in Table VIII. It must be noted tha exclusioe tth n from thiexamplex si s e tablth f eso (supra) which diverge' markedly fro e circulamth r plan causeo ignort s e largesu sth e t broc allf ho , namely Edinshall, which measure fee5 feet5 1 t 5 r fee 0 o ,y s6 b t on the average. t wil I seee b lt oncna e from this table tha diametere th t s recorded vary somewhat from one region to another, and a question consequently arises as to •whether these variations reflect real local differences in broch- constructio merele ar r no y accidental t somorden I ge .e e o t rligh th n o t mathematical aspect of this question I referred the detailed measurements Professoo t r Godfrey Thomsonkins wa d o enougwh , submio ht t themo t statistical analysis, and from his calculations the following facts emerge: (arespecn I e standar) th f o t d deviatio e measurementth f no s within each Region Regione t diffeth , no o r sd significantl y centr e 1-eve0th pe t ,n a point —tha saydifference o t th ,s i t e tha actualls i t y scattee foune th th n f di ro measurement eithen o s e rmea th sid f neo migh expectee b t occuo dt y b r chance sampling more often than once in a hundred times if no real difference existed, (b) In respect of their means, the six Regions do differ significantly, even at the 0-1 per cent, point—that is to say, on the hypothesis that there was really no difference between the Regions, the observed differences would occu chancy b r e sampling less often than oncthousana n ei d times. They should therefor regardee b t eno samples d a "population. e on f so Agai) (c "n

1 Mousfee9 diamete1 n i t s i a t ground-levea r t expand bu feel4 2 t o abovt s e cuppingeth e Th . corresponding figures for Dun Telve are 32 feet and 35 feet. SOME OBSERVATIONS ON THE BROCKS. 79 in respec theif o t r means IIId , Regionf groupean ,i , II , I sd togethere ar , homogeneoue founb o t d themselvesn i s , , whilVI d e an Region, V , IV s similarly grouped, are likewise homogeneous in themselves, i.e. the apparent differences between the subdivisions of these groups are not significant r centpe 5 , pointe eveth t n,a difference (d)Th mean i e n between these two groups of regions is significant at the 0-1 per cent, point, i.e. the northern distric a whol s a t e differs very significantly fro e res f mScotlandth o t . (e) Region II does not differ significantly from Region IV if compared with

TABLE VIII.—INTERNAL DIAMETERS.

Internal diameters (feet). Region. No. of Notes. examples. Max. Min. Average.

I. 14 35 19 29 Thi* s measurement, which II. 16 45* 27 32 is fully authenticated- ex , ceeds the next greatest in feet8 Regio y .b I nI III. 67 44 20 30 f This measuremen- re s i t IV. 20 52t 28 35| corded by Pennant and may be unreliable. The next V. 6 36 25 30 greatest in Region IV is 42 feet. VI. 4 40 30 34 | Or 34 feet if Pennant's figure is ignored. All Regions 127 52f 19 32

it alone, and if it is combined with Regions IV, V, and VI, the group so formed still differs significantly from I groupeRegionII d an d I s together. This suggests that, notwithstandin undoubtee gth d homogeneit Regiof yo I nI with occupRegiony IIId ma t ,an yi sI somethin intermediatn a f go e position between the groups having respectively large and small mean diameters. Statistical analysis thus appears to suggest that we may have to deal with two different races or strains in the species broch; and although no corroborative evidenc r plausibleo e explanatory broughtheore b n yca t forward, the possibility should perhaps be remembered in future study. Before passing on from the question of the mean diameter, it will be wel deao t l l wit hpoina t whic appeay hma r opee b criticismo ny t ma t I . 80 PROCEEDINGS OF THE SOCIETY, 1946-48.

said tha meae tth n diamete Regiof lowerero s i I nII d throug influence hth e of the nine brochs found within it that measure less than 25 feet, and that certain of these—actually seven in number—may not be fair samples seeing that they stan artificialln do y restricted sites suc s narroha w cliff-bound promontories, small islands tope hillocks.th f so r o , 1 Exit befor sucy ean h argument coul admittee db t dwouli necessare db y teso materialwhole e t t onlth th t no f y singla eo d an ,e diameter-clasn i s one district r sign fo ,f artificia o s l limitation d thian s; woul e quitb d e impossibl strengte th publishee n th eo f ho d descriptions alone. Moreover, in the light of some further calculations it is seen to be quite unnecessary, as if the seven small brochs in question are ignored the following results appear: (i) the six regions still differ significantly, though now at the 1 per cent, point and no longer at the 0-1 per cent, point; (ii) no change is e comparisomad thesth s a n e, i e IV stil f Regiond no l an diffe I II sr signi- ficantly at the 1 per cent, point as they did before; (iii) Region III, it is true, now no longer differs from the whole of the rest of the country even r centpe ,5 pointe th r cent t ,pe previousla o s wherea.5 d e di th t i st y a point and nearly at the 1 per cent, point; but in view of the comparatively low degree of significance this fact need not be accorded any great importance. Artificial limitation need therefore detai furthero n s nu . (iii) Height.—The only four brochs that still stand to any considerable height—in part at least of their circumference—are Mousa, Dun Troddan, Dun Telve, and Dun Carloway, and it is consequently to these structures tha muse w t tfirse looth t n ki instanc r evidencefo e regardin e originagth l height of the towers. And it will be convenient to consider the question wall-heaterme n i th f so d only, ignorin additionae gth l heigh allowee b o t d fo roof—supposina r g tharooy an tf existed (pp. 71 ff.). Broc e fee3 inche4 Th 6 tMousf hw o no s s ahighi musd an , t have carried at least one more gallery or a parapet-walk. Quite possibly there was more tha additionae non l store superstructuref yo t leas musa e t e on t b tbu , assumed. The tower can therefore hardly have been less than 50 feet high to the wall-head in its pristine state, and may well have been higher. It is worth noting here tha feefigure 0 5 th t t f woulo e d mak heighe e eth th f o t tower the same as its external diameter at the base. The surviving fragment of Dun Troddan shows two galleries intact and a third partially preserved. In its present state it is about 25 feet in height.2 1720n I , however, when Gordo it,w founsa r galleries were s intacta d an ;

additionae th l gallerprobabl y hala d ma f yan y have adde3 d som feee8 r o t 9 feet to to-day's figure, we arrive at 33 feet to 34 feet as the true height at that date. Gordon does factn i , cleas i , t giv i rheighe feett 3 eth 3 bu ,s a t

B.C.A.M.e Se 1 , Inventory of Caithness, , 203 Nos35 , , 518.33 ; Inventory of Sutherland, , Nos10025 , .4 2 P.S.A.S., vol. Iv. p. 85. ' 3 Itinerarium Septentrionale, p. 166. SOME OBSERVATION1 8 E BROCKSTH N O S .

fro allusios mhi fouo nt r internal doors leading int ogallerya s wela ,s a l frothicknese s figurth m hi e walr th t fo e whae a lf b so o tt seemem hi o dt ground level, that in his time the ground inside the tower stood 6 feet abov e truth e e e court leveth whed f e an necessaro l; nth y correctios ni made the resulting total height of 38 feet is seen to be impossibly great, as fourte th h missine gallerth d gyan thire parth df o t could never, between them, have accounte heigha r dthere s fo 1i f 4to r feetfactn ,i No ., anything at all impossible in the idea that Gordon should have exaggerated this measurement by as much as 6 feet, seeing that he was most unlikely to have been properly equippe measurinr dfo gbuildinga tha d base an ,th t e of a ruined fifth gallery—unnoticed and consequently unmentioned—might possibly have added a foot or two to the total. The discrepancy con- sequently need not be regarded as important, and we are left with the inference tha 172n i t Troddann 0Du , wit h heigha feetalreads 4 3 wa f , o t y reduced fro originas mit l heighunknown a y b t t possiblnbu y quite consider- able amount. That inneGordoe outed th ran w rnsa walls actually merging originan a t a inte l owall-heaon d seems quite improbable.1 On the question of the vanished portion the six feet of debris collected coure perhapn th ca n ti s throw some light. Theoretically this representsa volum somf eo e 3700 cubic fee pilef o tpurele dth stonesn o y arbitrard an , y assumptions (a) that laid masonry would occupy half the space of random debris, (6) that the debris inside the tower represented the ruins of the inner wall alone e outeth , r wall being suppose havo dt e fallen outwards, this volume would provide for an additional height of something under 7 feet. An unknown amount of stone-robbing must also no doubt be assumed. It would thus appear allowable to carry the tower up to an original heigh f ove o tfeet0 4 r , thoug s impossibli t hi e availabl th n o e e evidenc thay t sa weni t o higheeo t n t r proportioe thath nf i thisd f an no ; heigh basao t l diamete same th s e t Mousahera wa r s ea heigha , somf o t e 60 feet might well allowehave b o et d for. To Gordon's record Pennant adds littl value—thf eo e fourth gallerd yha disappeared by 1772, and he does not give any figure for the height ctf the thire th f d thirdo totas galler p feeHi 4 inchesto l2 .5 theigh s e yi th o ;t t and if this figure is corrected for the depth of the debris in the court, it accords fairly well wit e conclusionhth s reached above s engravinHi . g 2 cannoshowe h s relie e a high-levesa b t, don l scarcement crossin gpara f o t wall-face th e whic stils hi l preserve wherd dan scarcemen o en t exists. As regards Dun Telve, not only is Gordon's description regrettably short but his language raises the suspicion that he may have borrowed points somhi f eo s froTroddann mDu , then much better preserved thas nit neighbour, in the belief that the two structures were identical. In any 1 Cf. p. 63. 2 Tour in Scotland, vol. ii. p. 340 (4). VOL. LXXXI. 6 82 PROCEEDINGS OF THE SOCIETY, 1946-48.

case even partia mad le t onli b sens f assuming—whay n eo yb eca t no s i t in itself unlikely—that in 1720 the lower parts of the tower were deeply burie n debrisi d d thaan ,t Gordon entere a hol y eb d 1 whic intd le ho secone th d gallery secone Th .d lowes gallerye ratey th an s i , t a ,leve t a l whic coule hh d have made—a allegee sh s tha did—e h t a complete circuit buildine oth f g withi wall-spacee nth firse th ts a galler, barres yi t morda e tha poine thergallerd o groune non n an th t s ei n yo d t whilfloorBu e e.h give measurementso sn s drawinhi , g evidently represent e samsth e sector

of the wall as still survives2 , and it can be usefully compared with the section prepared by H.M. Office of Works and published by Dr Curie.3 This comparison shows that whereas in 1720 three voids of the flight rising from over the entrance existed above the high-level scarcement, in 1916 the lowest of these voids was intact but the second had lost its lintel; the total height of the wall at this point must thus have decreased since 1720 by the height of one void plus some half-dozen courses of masonry— probably some 5 feet at most. On this showing the true height of this part broce ofth 172n hi 0 would have been abou feet8 3 t . Pennant's illustration,4 though evidently taken from a sketch made on the spot, does not seem fully accurate in respect of these voids, but suggests none the less that very little height had been lost since 1720 in a vertical line above the door, though perhaps two or three feet from above the other long flight of voids (see fig. 3). This implies that the broch was about 35 feet high at its highest point 1772 n althougi d an ; figuree hth s statetexe Pennant'f th to dn i s description5 giv eheigh a fee 7 inches3 6 t f to lowee ,th r estimat probables i preferrede b yo t . In view of the fairly close correspondence of all these lines of reasoning, it is impossibl accepo et t Pennant's estimate tha heighe tbeeth d tha n reduced by as much as 10 feet 6 inches since the date of Gordon's visit, at least in this part of the circuit; though a robbery on the scale that he records may well have taken place elsewhere. The evidence obtainabl Telvn Du e t ethua s shows tha towea t r from which a great deal of debris had already fallen was probably just under 40 feet hig 1720n hi thad t an thaa t; t heigh intra-murae th t l spacd eha become so narrow that an inexpert observer considered the walls to have "joined together." This is not to say, however, that the wall-head had been reached or that it was even particularly near; Gordon's illustration, while showin walle facee th g th beins f sa o s g very close together, doet sno

1 A trace of what may have been this hole, now built up, can be clearly seen in one of Dr Curie's published photographs (P.8.A.S., vol . 251p . .1 , fig ) aboue entrance8 . th f fee9 to tp abov.to e eth Cf. also pp. 63, 65 on Olickhimin. 2 Op. cit., pi. 65. s P.8.A.S., vol. 245p . .1 , fig. 4 . 4 Tour in Scotland, pi. xli. (3). Ibid., pp. 337 f. He gives the height as 30 feet 6 inches, but his further figure for a wall-thickness 5 ofee7 f inche4 t s "takee distanc th feen t te na t f efroo e bottommth " shows e tha e leveth th tf o l ground insid towere eth beed ha n raised som fee e7 accumulatey b t d debris. SOME OBSERVATION E BROCKSTH N O S . 83

«1s3

Iw?g

§ -

I. I •o -d

o-- 1 1° " -«• 0:1

CO ">! 84 PROCEEDINGS OF THE SOCIETY, 1946-48.

by any means indicate that they had actually been bonded into one. Moreover, if the proportion of height to diameter was the same as that observe t Mousada towee th , r would have bee t leasna fee 0 heightn 6 ti t . e factTh s relatinCarlowan bettee Du b o n gt r yca obtaine d froe mth account writte Thomay b n n 186i s 1 than from that preparee th y b d

Ancient Monuments Commissio1 n exactly sixty years later, as during this interval something like 11 feet seems to have fallen from the upper part of the wall and a piece of modern walling has been erected in its place at the level of the vanished fourth gallery. As seen by Thomas, Dun Carloway

in its highest part still showed four gallerie2 s and the outer wall of a fifth, s comparea d wit e thresurvivingw hth eno r thi givee fo sh ; heigha s f o t 34 feet, which compares well enough wit e existinhth g heigh f abouo t t 22 feet for the top of the third gallery. He does not believe that the original heigh mors wa tthrer eo tha o e ntw fee excesn formee i t th f so r figuree h ; gives no reasons for this opinion, but may have been led to it by the close convergence of the walls—these are now only 12 inches apart at the floor thire oth f d gallerinche8 d regardyse an e h s topapar th it d l t ,an sal a t upper galleries as having been too narrow for access. This inference seems reasonable enough and, if it is accepted, we are left with a picture of a tower under rather than over 40 feet in height, and this on an external basal diameter of 47 feet. It would thus have been considerably lower than Mousa, and rather more squat in profile. While the evidence so far reviewed suggests that brochs may have varied in height, and that in particular some may have been a good deal lower than Mousa, there is nothing to show that any one of them was low enoug excludee b o ht d fro e clasm th f "tower o s respecn "i s genera it f o t l proportions. It is theoretically possible to argue 3 that the brochs now seen broken down to comparatively low heights never, in fact, attained to the proportions of towers, and that tower-h'ke brochs were rare and excep- tional structures. But against this suggestion both direct and indirect evidenc broughte b n passagca eA . n Georgi e e Buchanan's Rerum Scot- ticarum Historia4 compares Arthur's O'on with "numerous" (compluria) structures then standing "in a certain island," stating that they are "rather bigger and more roomy" (majuscula et laxiora) than the O'on.5 These structures can only have been brochs, and to give grounds for the comparison

1 Arch. Scot., vol. v. pp. 383 S. This modern walling is shown hatched in the section published by the Ancient Monuments Commissio2 n (Inventory of the Outer Hebrides, etc., fig. 57), though, without explanation in the text. But that this was the Commissioners' interpretation of the remains I am informed by Mr C. S. T. Calder, who prepared the published drawings. Wr Si . Lindsay b s A y Scott, D.S.C., F.S.A. P.P.8.,n ,i Seriesw Ne , vol. xiii. (1947). pp , '3 Ed. 1582, fol. 6. 5 Aikman, in his English version of 1827, p. 24, gives "in the islands," which is an obvious mis- translation; and "more loosely constructed" for laxiora, for which again there seems to be little justification. SOME OBSERVATIONS ON THE BROCHS. 85

they must have show a nsomewha t tower-like profile e havW e. thus direct evidence tha numbea t high-standinf o r g brochs, additiona thoso lt e now known,1 still existe sixteente th n di h century indirece Th . t evidence is furnished by the lower storeys of all the broch ruins alike, and is even more compelling immensele Th . y widd solian ed basal storey, combined with so ingeniou arrangemenn sa hollowe th s a t , galleried wal decreasinr fo l e gth volume stoneworke ofth , points clearl face th t ythao t wal intendee s tth lwa d high,e b tha d buildere o t 2an tth s consequentl neew reducyo dsa t e unneces- sary weighprovido t d stabilitr an t efo y unde laterae th r l pressur windf eo , as well as to save material. And in connection with this latter factor it is wel poino t l t tha broce ou t th t h metho constructiof do n wilt effecy no l an t saving unless the wall is of more than moderate height, as several galleries requiree ar compensato dt massive th r efo solie th siz df eo basa l portion. Thus an ordinary solid wall 20 feet high by 7 feet and 5 feet in thickness respectivelp to e th base d th yan e t woula perfectle db y stabl alsod ean , probably, sufficient as an obstacle to escalade; 3 but the section of this wall contains abou squar0 12 t e feet, wherea n lowermose sth Du e th fee 0 f 2 to t Carloway wall, if sectioned elsewhere than through the cell, gives a corre- sponding figure of about 140 square feet. It would therefore have been allowableCarlowayn case Du f th e o n infei o ,t , r tha e heigh th tbee d ha tn more tha t nknow somno facte d feed th 0 nan e2 e ;ha t thi b e eve o w st f ni the argument is even more compelling in cases where the solid basal portion is higher and thicker. Nor is it possible to explain away the thickness of the bases of broch walls by comparing them with other thick drystone walls, e.g. those of some of the great Irish forts. These latter may well have been intended givo t e spac thein o e r wall-head e deploymenth r fo s a larg f o te forcf o e defenders, with which tactic their easy, open stairs would have accorded very well indeeddefensiva t Bu . e pla thif no s kind would have been quite unsuitable to a broch, with its narrow, awkward stairs and tortuous galleries. The analogy woul rathee db r wit e outehth r defensive works founa t a d number of brochs (pp. 87 ff.); the one at Mid Howe, for example, being up to 19 feet in thickness at its highest surviving level, is already unnecessarily massive if regarded as a barrier or as cover, while the extraordinary con- struction at Kilmster, originally 22 feet broad and subsequently thickened to a maximum of 40 feet,4 strongly suggests a vantage-ground for hand-to- hand fighting at the lip of the ditch. The foregoing conclusions as to height, however, undoubtedly raise a addee whicb o vanishee T y * dth hma Alascaign dDu . <* That the principal purpose of the thick basal portion was not to accommodate cells is shown by hige th h proportio brochf no s which containe cele dlon only (supra, . 58)p . e sprehistori th Par f o t c parapet remains, intac unrestoredd an t fore f Locth o t t a h, Boon, near Ardara, Co. Donegal. The wall is about 17 feet high. * I am indebted for this piece of information to Mr C. S. T. Calder, who excavated the remains of this broch in 1940, before they were swept away in the construction of an air-field. 86 PROCEEDINGS OF THE SOCIETY, 1946-48.

difficult connection i y n wite verth h y large examples. Thus Edinshall, which is not round on plan, measures from 79 feet to 93 feet over all and feefro0 6 feet 1 m o 5 internallt t y acros quits i t courte i e sth d impossiblan ; e suggeso t t tha t evei t r showe e samdth e proportion profiln i s s Mousaa e , wit ha heighfee 5 r more9 same o t f Th o te . migh saie b f tAchaneas do , with an external diameter of about 80 feet; or of Dun Bhorairaic, to which Pennant's measurements woul feet eve6 r 7 o Torwoodleedf f ;no o giv e eon ,

which closely approaches this figure1 . In such cases one is virtually bound to suppose that the brochs were high enough for the owners' purposes, whatever these were, but that their profiles were squat and not tower-like—thus Edinshall fee0 4 t f i high, , would hav 'thd eha same profil a broc s a eh 26 feet high on a base 56 feet in external diameter. Apart, however, from these and other exceptional cases—^and no doubt exceptions occurred from a variety of causes—there seems to be no good reason to suppose that the broc nots essencehn wa i , tower-lika , e structure thad t normallan ,ti y stood 40 feet or more in height to the wall-head. t I remain consideo t s r whafo r t purpose structure f tower-liko s e pro- portions may have been built. One important object in the builders' minds was no doubt defence against escalade—a danger perhaps increased by the roughness of the outer face of a drystone wall.2 It is not at all clear, however, why the builders e brochoth f s should have taken t wouli s a , d thus appear muco s , h more seriou viesa thif wo s particular hazard tha buildere nth ordinarf so y forts, who were content with walls, again of drystone construction, of very much lower height—th hardln ca supposet e ei yb mors a o es d that their enemies were equipped with regular scaling-ladders. Whil generao en l comparisons properln ca made yb thesn eo e lines owininfluence th o gt varyinf eo g site- conditions on the practicability of scaling any individual structure, the difficulty remains preventd an , s this explanation from appearing completely adequate. However, if the broch was the residence of a single family and e strongholth t no a community f do s garrisoit , n would presumably have been smaller than a fortthad correspondingl f an o ,t y stronger material defences might have been desire thin do s account. s likewisi t I e commo o fint n d brochs regarde s watch-towersa d t bu , this explanation of their height is again far from satisfactory. The terrain Glen i n Beag (Grlenelg sucs i ) h tha onle tth y viewd an obtainabl p u e on s ei down the glen, and this would not have been materially improved by a ris f someo feet0 e4 ; whil t Mousea watchmaa e hilth l n behinno e dth broch could have observed any movement on the sea with perfect safety to himself and would have had ample time to give the alarm. It is possible, 1 On a recent visit Professor Childe found the diameter immeasurable on account of fallen debris. 2 Cf. the tradition of how Dun Carloway was scaled, with the help of two dirks used alternately as foot-holds (Arch. Scot., vol. v. pp. 387 f.). The lower, battered portion of a broch wall is not, in fact, difficult to climb. SOME OBSERVATIONS ON THE BROCKS. 87 however, that on a site among rocks, gullies, and scrub-wood a tower might have given greatly improved observation into pockets of dead ground, beside e opportunitsth f searchinyo g such areas wit hfira arrowf o e s from the wall-head—supposing always that the bow and arrow was in use.1 Indeed, were it not for the largely treeless condition of most of the broch country woule on , d certainlsuggeso t d le e ty b tha towere tth s were designed to give a view over tree-tops. A high firing-platform, again, would no doubt help to increase the range of missiles, though simple wall-head defence against an assaulting rush would not in itself seem to call for an exception- ally high wall-head. e ordinarNonth f o e y theories thus appea carro t r y conviction therefors i t i d ean , perhap t whollsno y fantasti whethek as o ct r heigh e broche th havth y f to sema been determined raty ean t influenceda r ,o , by other than material considerations. Magnificenc buildinn ei commoa gs i n foible of wealthy and aristocratic societies, and the leading men of this northern Iron-Age community may conceivably have built their strongholds to an unnecessarily imposing height as a means of asserting their prestige.

5. EXTERNAL DEFENCES. While something mus e saib t d about this subjectn onle ca b y t i , approached in a tentative manner for lack of reliable data. Conclusions regarding the quality and strength of external defences can hardly be based on such summary mention as these works are frequently accorded in the published accounts, which tend to be chiefly concerned with the structural remain brochse th f so ; whil face eth t tha observatione tth s have been made numerouy b s observers wit uniforo hn f criteri o t mse a introduces further uncertainties. Again, artificial defence n hardlca s e considereb y d apart from the natural defensive features of the sites on which they are con- structed—no true comparison, for example, can be held to obtain between brocho tw s whic e alikhar e unprovided with artificialf o defence e on f i s them stands on open, accessible ground and the other on a sea-girt stack. The subject factn i , , require sfresa h approach, with appropriate comparative field-work failind an , g thiexistine sth g data mus treatee b t d wit hgreaa t deal of caution. In particular, mere lack of any mention of external defences in a given case should not be regarded as dependable negative evidence, seeing that so remarkable a system as the one at Gurness was only brought to light by the spade. Subject to all these provisos, external defences are recorded as shown in Table IX. On the figures in the table the following remarks may be made: (i) The totals (col. d) seem unexpectedly small t onl,no y when state percentages da s of the total number of "certain" brochs, broch sites, and "uncertain whicn O 1, not h. 71 ecf1 . p . 88 PROCEEDING E SOCIETYTH F O S , 1946-48. TABLE IX.—EXTERNAL DEFENCES.

Totals (d) Particu- Totals (d) larly Less Totals as percentage percentags a e Region. strong of Table I, strong defences. ) (6(c )+ (d)cols+ ) (b . of Tabl, eI defences. totalled. col. (e). (a)' (b) (c) (d) (e) (/)

I. Shetland . 15 6 21 r cent32pe . r cent22pe .

II. Orkney 7 10 17 20 „ ' 16

III. Northern Mainland 7 31 38 18 17

. WesIV t Coasd an t 19 19 39 39 Inner Islands .

V. Outer Islands 4 4

. CentraVI Eastd an l - 3 3 Not calculated: ern Mainland see below.

Totals 32 70 102

examples" t als(colbu o , .whe/) n state s percentagea d f "certaino s " brochs and "uncertain examples," without consideration of sites (col. e). While negative evidence—admittedly of doubtful validity—is exerting an influence here t weri implications y a ,e b figurw whilOrkner d lo efo ,an e eth y (col. d) must be suspect on the ground that it may be due in part at least to defective observations by early excavators as well, perhaps, as to extensive stone-robbery by farmers, it still seems clear that external defences are in fact less common, particularly in Shetland and Orkney, than the Ancient Monuments Commission suggest s summing-uit n si e subject.th f o po N 1 valid conclusion can be based on the very few examples recorded in the Outer Islands except that external defences mus vere b t y rar than ei t region; while the figure for Region VI should be ignored until more is know relationshie th f no p broche ofth question si worke th no t s among which they stand.2 (ii) Comparison of columns (6) and (c) shows a notable contrast 1 B.C.A.M., Inventory of Orkney and Shetland, vol. i. p. 33. 2 Infra, p. 90. SOME OBSERVATION9 8 E BROCKSTH N O S . proportione th n i f "particularlo s y strong "lesd "an s strong" defencen i s Regions I, II, III, and IV—the other two regions being again left out of consideration. While the figures have no doubt been influenced to some extent by the disturbing factors of which mention has been made above, there undoubtedly exist Shetlann si dseriea brochf so s heavil elaborateld yan y defended by ramparts of earth or walls, and wide, deep ditches (infra). That "less strong" defences predominat Northere th n ei n Mainlandn o d an , e Westh t Coas d Innean t r Islands alsy takee ob ma , s certainna ; their absence apparenr o , t absence, fro lattee mth r region being comparable with a similar blank in the Outer Islands, (iii) The figures for Region IV in cols. (e) and (/) are exaggerated in comparison with those for Regions I, II, reasoy face b anth tI f dnthaII o broco tn h site recordee sar Region di . nIV (ivf s compareTablI i ) X eI d wit e statisticahth l conclusions given above under Sectio (ii)n 4 marke o n , d correspondenc onln founds e i ca y b t I . said that "particularly strong" defences seem to be present in the northern group of Regions (I, II, and III) and absent from the southern group (IV, V, and VI), these groups also differing significantly in respect of their internal diameters (p. 79). However, in view of the comparatively small numbers of instances from Regions IV to VI that are shown in Table IX, this observation regardeshoule b t d no establishins da gfact rathea bu t s ra pointing to a possible subject for study. The main types of defensive outworks may be classified as follows: (i) Multiple Ramparts d Ditches.—an This typ s i magnificentle y exemplified by the complex laid bare at Gurness, where the defensive belt of three ramparts and three ditches must originally have measured at least 70 feet in width. With this system should no doubt be compared the Shetland examples just mentioned, though their details are still unexplored; in some of them the defences are drawn across a neck and do not encircle broche th thes a , y probabl t Gurnesa d ydi s befor site eth e became eroded. The most strikin e Shetlanth f go d example t Hoga e aar s Ness, Burland, Sna Broch (), and Underhoull. (ii) Walls and Ditches on Promontories.—With the foregoing works link up the single walls, usually accompanied by ditches, that are frequently drawn across promontorie narror so w necks massive Th . Howe d e walMi t ,la fee3 inche1 fee6 t9 thickness1 n i t o st witd outen ,an h a r ditc feeh9 t wide and a narrower inner one, is the finest example of this type that has yet been excavated; but analogous though slighter walls may be seen at Nybster and Ness, and a ditch at Skirza Head. (iii) Encircling Walls.—Clickhimin shows the best example of a simple encircling wall; this runs round the former shore-line of the islet on which the broch is built and averages 10 feet in thickness. At Jarlshof part of a thinner wall survives d unexcavatean , d trace f similao s r encircling walls occu t mana r y site varioun i s s localities greae Th t. wal f enceinto l t a e 90 PROCEEDING E SOCIETYTH F O S , 1946-48.

Edinshall cannot at present be placed in this class with certainty, as its connection with the broch, though quite probable, is unproved; and even superficialle saie th b f lesdo n sca y simila rLawe caseTorwoodleed Th sf an so . (iv) Foreworks.—A final word mus e saib t d abou e forth te wort a k Clickhimin n internaa , l gatehous t withise e e wal th f nenceinto l d an e spannin passage gth e tha broce t dooe leadth th f ho ro sthrougt hcomplea x of external buildings. No other such forework is known to be connected with a .broch, but it invites comparison with the gateway in the wall that surrounds an island in the Loch of (), and with the totally

detached building that stands within an outer defenc1 e of ditch and bank nea Nese poine th rth Burgi.f f so o t 2

. RECAPITULATION6 . chiee Th f points mad thin ei summarisese b pape y rma followss da : The facts of distribution and typology cannot safely be used as evidence broch'e foth r s plac origin f large o Th e . total number indicate formee sth r importance of the broch as a social factor, but there is no reason to suppose that all brochs were in use at once. Among the architectural features, entrances follow the "normal" plan except in a very few cases. The supposed "normal" arrangement of guard-cells (one on each side of the entrance) occurs in only 21 per cent, of the cases noted; a single cell occurs in 52 per cent., "right-hand" outnumbering "left-hand" cells by more than arrangemene r cent threpresens pe Th i .6 l one2 o eal t celo n t n i t;a l f o t three mural cells ("right," "left, "opposite"d "an ) made familia Mousy b r a s likewisi t normalno e s threa , e cell-entrance e founr ar s pe onln i d3 1 y cent, of the cases,- while one cell-entrance occurs in 47 per cent., two in cent.r 3 0morpe d ,an e tha centr npe thre.0 1 "Left-handn ei " cell-entrances preponderate. Complete basal galleries muce ar h less common than cells, and occur mainly in the Inner and Outer Islands. In respect of upper galleries and stairs Mousa, which is commonly regarded as the norm, differs from the three other least ruinous brochs in various ways which cannot be resumed summarily seconA . d stair exist foun i s re stair th cases f -O . entrances 69 per cent, are on the "left-hand" side of the court. Veranda- roofs resting on low-level scarcements probably served also as balconies; high-level scarcements (onl preservede yon havy )ma e supported some roof- structure. Voids, besides relieving somn lintelsi y e casema , s have given access to galleries from balconies. Theoretical considerations are against roofs towersf o presentf i , , having been masonry vaults, though some experi- mental evidence for this exists. Elaborate arrangements for water-supply made ar mann ei y cases. Statistical stud meanf yo internal diameters shows B.C.A.M., Inventory of Shetland, . 1316No . *1 Ibid., . 1154No . SOME OBSERVATION1 9 E BROCKSTH N O S .

tha northere th t n district (Shetland, Orkney, Caithness, Sutherland) differs significantly from the rest of the country in this respect. In height brochs probably varied, some being lower than Mous mord aan e squa profilen i t ; but there is no reason to believe that any were too low to be characterised as towers. Several explanations can be suggested for their height but none is completely convincing. Very strong external defences seem to be commonest in Shetland (15 out of total 32).

APPENDIX.

LIST OF BROCHS, BBOCH SITES, AND CERTAIN COMPARABLE STRUCTURES. The following list contains five hundred and fifty items, as follows: (i) Three hundre d fouan dr structures which have been positively identifie s brocha d s and are still in existence. For those printed in italics no dimensions or structural features are on record, (ii) Sixty-seven sites of vanished structures •which are recorde brochss da . Thes listee ear "brocs da casew h fe sites. s a description n I " s exist, (iii) One hundred and forty-one structures which probably are or possibly may be brochs, but which have not been positively identified as such. These are listed as "uncertain examples." (iv) Thirty-eight "comparable structures." These t brochsno t thee ar bu , y embody features which appea brocn ri h architecture. The material is arranged under the six regional headings adopted in the body of the paper, namely Shetland, Orkney, Northern Mainland,1 West Coast 2 and Inner Islands, Outer Islands Centrad an , Easterd an l n Mainlande (cf.Th fig. 1) . county, island districr o , whicn i t h each broch stand noteds referencsa i d an , s ei principagivee th no t l source from which information regardin bees ha n gt i obtained . e cas f brochth o e n I s surveye e Royath y lb d Commissio e Ancienth n d no an t Historical Monuments of Scotland, this reference takes the form of the structure's serial number in the appropriate County Inventory.

. SHETLANDI . All numbers refer to the Royal Commission's Inventory of Orkney and Shetland. BROCHS. Burland, 1247. Aithsetter, 1141. Burra Ness, 1716. Balta, 1596. Burraland, 1143. Brough Holm, 1548. Burraland, 1607. Burga Water, 1606. , 1114. Burgar Stack, 1544. Clevigarth, 1147.

Nort lina f heo joining Tai Gruinardd nan . 2 From Gruinar Kirkcudbrighto dt . 92 PROCEEDING E SOCIETYTH F O S , 1946-48.

. SHETLANI D (contd.). BKOCHS (contd.). •Brough, 1107. Brough, 1277, Clickhimin, 1246. Brough, 1343. Clumlie, 1145. Brough Lodge, 1238. Culswick, 1397. Brough Taing, 1580. Dalsetter, 1146. Firth, 1395. Burgan, 1386. Burland, 1535. Bastshore, 1148. Burra Voe, 1384. Feal, 1211. Burrastow, 1673. Footabrough, 1608. Burravoe, 1745. Fugla Ness, 1115. Burrian, 1308. Gossabrough, 1718. Greenbank, 1715. Burwick, 1528. Heogan, 1105. Hamnavoe, 1353. Hawk's Ness, 1500. Houllands, 1468. Head of Brough, 1721. Know HouUandf eo , 1188. Hoga Ness, 1545. Loch of Burraland, 1387. Holm of Copister, 1720. Loc Stavanessf ho , 1307. Houbie, 1212. Mail, 1187. Houlland, 1396. Musselburgh, 1582. Housabister, 1282. Orbister, 1385. Infield, 1116. Sand Wick, 1581. Islesburgh, 1354. Scousburgh, 1190. Jarlshof, 1149. Stoura Brough, 1674. Levenwick, 1144. , 1189. Loc Houllandf ho , 1352. , 1342. Loc Huxterf ho , 1605. , 1306. Loch of Kettlester, 1719. Loch of Watsness, 1609. Mousa, 1206. UNCERTAIN EXAMPLES. Noss Sound, 1085. Bousta, 1610. Nounsburgh, 1394. Brindister, 1399. Sna Broch, Fetlar, 1210. Burga Water, 1284. Brocha Sn , , 1546. Cullingsburgh, 1086. Southvoe, 1142. Fethaland, 1355. Stoal, 1717. Gord, 1150. Underhoull, 1547. Heglibister, 1501. Wadbister Ness, 1499. Loc Benstonf ho , 1283. Firth, 1393. Loch of Brindister, 1248. West Houlland, 1398. Loch of Brow, 1153. West Sandwich, 1722. Lunabister, 1152. Windhouse, 1723. Pinhoulland, 1611. Skelberry, 1151. Wester Skeld, 1400. BROCH SITES. , 1106. COMPARABLE STRUCTURES. Baliasta, 1579. Holm, 1529. Loch of Huxter, 1316. Brei Wick, 1744. Nes Burgif so , 1154. SOME OBSERVATION E BROCKSTH N O S . 93

. ORKNEYII . All number sRoyae refeth o t rl Commission's Inventor f Orkneyo Shetlandd yan . BROCHS. CoUi Ness, 473. Berstane, 405. Dennis Ness, 205. Berwick, 679. Harra, 852. Braebuster, 624. Harray Church, 138. Breckness, 920. Hoor Ness, 1071. Burgar, 261. Hunton, 980. Burness, 321. KnoU of Skulzie, 1072. Burray (East), 862. Loch of Westhill, 801. Burray (West), 861. Overbrough, 139. Burrian, Corrigal. 12 , Redland, 320 .descriptioA e th f o n Burrian, N. Ronaldsay, 193. structur thif eo s broc recordn o s hi . Burrian, Russland, 14. Scar, 182. Burroughston, 778. Scockness, 606. Burwick, 817. Smiddybanks, 850. Castl f Bothicaneo , 522. Stackrue, 677. A description of the Dingieshowe, 625. structur thif eo s broc recordn o s hi . Green Hill, 379. Stromness, 940. Gurness, 263. Tofts, 430. Heffiar Holm, 806. Westbrough, 183. Hillock of Breckna, 486. How Hoxaf eo , 815. UNCERTAIN EXAMPLES. Hunda, 863. Backaskaill, 159. Ingshowe, 322. Braebister, 380. Knowe of Burrian, 551. Burrian, Garth. 21 , Knowe of Burristae, 1034. Burrian, Loc Harrayf ho , 680. Knowe of Dishero, 265. Cantick, 1006. Knowe of Stenso, 262. Cummi Howe, 872. Lamb Head, 947. Deerness Church, 629. Lingro, 406. Finstown, 323.. Loc Ayref ho , 360. Green Hil Hestf o l i Geo, 1008. Loch of Clumly, 678. Green Hill, , 948. Mid Howe, 553. Green Hill, Walls, 1007. Ness f Ork,o 777. Hal Rendallf lo , 270. Ness of Woodwick, 264. Hillock of Baywest, 949. Netlater, 13. Hodgalee, 1035. Oxtro. 11 , How Farm, 158. Poin Buryanf o t , 437. Howan. 20 , Skogar. 16 , Howe of Langskaill, 627. Steiro, 779. Kir f Cleatonko . 23 , Taft, 15. Knowe of Gullow, 22. Yerron, 260. Knowe of Hunclett, 555. Westside, 552. Know f Ryoeo , 267. WasSo, 438. Loch of Hundland, 18. Loc Isbisterf ho . 17 , BROCH SITES. Mamie Howe; information fror M m Arion, 939. Richardson. S . J . Brough, 851. Mithouse, 19. Burrowstone, 1023. Nebister, 160. 94 PROCEEDINGS OF THE SOCIETY, 1946-48.

II. ORKNEY (contd.). UNCERTAIN EXAMPLES (contd.). Tankerness, 626. Ness of Boray, 313. The Howe, 921. Newark, 439. Skeoe Th , 1009. North Howe, 557. Tingwall, 268. Riggan of Kami, 628. Verron, 682. t Mary'S s Kirk. 24 , Viera Lodge, 556. t Tredwell'S s Chapel, 523. Vinquin, 266. Scarrataing, 681. Wass Wick, 269. Scockness, 554. Weems Castle, 816.

III. NORTHERN MAINLAND.

Number sBoyae refeth o t rl Commission's Inventorie Caithnesf so s (abbr an) f dC. .o Sutherland (abbr. S.).

BROCHS. Brounaban, C. 511. Achanarras,99. C. , 193.C. Achaneas,50. S. Burg Langwell, C. 201. Achaneas. 51 . S , Burg Euadh, C. 207. Acharole . 466C , . Burn of , C. 212. Achavar, 199.C. Camas an Duin, S. 157. Achbuiligan Tulloch, 350.C. Camster, 189.C. Achcoilleiiaborgie . 183S , . Canister, 522.C. Achies, C. 98. Carn Bran . 468S , . Achingale, 473.C. Cam Liath, S. 187. Achlochan Moss, C. 102. , S. 270. Achorn, 214.C. Carn Mor,53. S. Achow, 208.C. Car Mairga nn . 105C , . Achunabust, C. 351. Carrol, S. 27. Achvarasdal Lodge, C. 353. Castle. Cole25 . S , Achvarn, 112.C. Castlehill, C. 320. Allt a' Choire Mhoir, 312.S. Clachtoll, S. 7. AUt an Duin, S. 182. Cnoc Donn . 103C , . Allt Duin,an 313.S. Coghill . 469C , . Allt Breac . 395S , . Coich . Burn23 . S , A' Mheirle, S. 478. Coill' Ach a' Chuil, S. 176. Appnag Tulloch, C. 218. Croick, E. Ross; personal observation. Armadale Burn, S. 190. Crosskirk . 347C , . Backies, S. 272. Dail Langwell, S. 49. Balantrath, 213.C. Dalchork, 394.S. Ballachly, 192.C. Dale, C. 104. Bell Mount, 431.C. Doir a' Chatha, S. 52. Berriedale . 203C , . Duchary, S. 28. Berriedale, C. 205. Dun an Ruigh Ruadh, W. Ross; Borrowston, C. 510. information from Mr C. S. T. Calder. Brabstermire, C. 37. Carnachaidhn Du . 180S , . Brae . 107S , . Dun Chealamy, S. 179. Brinside Tulloch, C. 434. Creagachn Du . 175S , . SOME OBSERVATIONS ON THE BROCKS. 95

III. NORTHERN MAINLAND (contd.).

BROCHS (contd.). Loch Mor, S. 189. , S. 155. Loch Shin, 391.S. Dun Lagaidh, W. Boss; information Lynegar, C. 471. from Mr C. S. T. Calder. Mid , C. 195. Maigha n n ,Du S. 527. Minera, 197.C. Dun Riaskidh, S. 529. Murkle, C. 319. Dun Viden . 181S , . Murza,63. C. Dunbeath, C. 215. Mybster,96. C. Dunrobin Wood, S. 271. . Ness33 . C , East Kinnauld, S. 477. North Calder, 110.C. Eldrable, S. 309. Norwall . 508C , . Elsay . 521C , . Nybster . 518C , . Everley. 36 . C , Occumster, C. 198. Feranach, S. 314. Old Stirkoke . 499C , . Forsinain . 186S , . Ousedale Burn . 204C , . Framside, 111.C. Road Broch, Keiss . 517C , . Freswick Links. 34 . C , Roster, C. 191. Gansclet, 501.C. Rumster, 219.C. Gills, C. 53. Sallachadh . 392S , . Gledfield, E. Ross; personal observa- Sandy Dun. 184S , . tion. Scotscalder, C. 113. Golsary, 220.G. Scottag, 470.C. Green Tullochs . 348C , . Scrabster, C. 429. Greysteil Castle . 222C , . Shiness . 393S , . Grum More . 174S , . Sibmister, 321.C. Gunn's Hillock, C. 2. Skelbo Wood, S. 106. Gunn's Hillock, C. 194. Skinnet, C. 116. Gylable Burn . 311S , . Skirza Head, C. 35. Ha' of Duran, 436.C. , C. 209. Hempriggs, C. 504. Spital, 100.C. Hill of Works, C. 3. Spital, 101.C. Billhead, C. 520. Spital, C. 474. , 435.C. SuisgiU . 308S , . Inshlampie, S. 178. Tannach, 500.C. Keiss . 515C , . Thing' . 432C , . sVa Killin,26. S. Thrumster, C. 502. Killouran, S. 310. Thrumster Little . 503C , . Kilphedir . 307S , . Thurdistoft, C. 318. Kintradwell . 467S , . Tiantulloch, 196.C. Knock Urray, C. 349. Toflgun, 525.C. Knockglass, 117.C. Trantlemore, S. 188. Knockglass, 475.C. Tulach Bad a' Choilich, C. 202. Knockinnon, C. 216. Tulach Beag, C. 107. Kyle Tongue,of 530.S. Tulach Mor, C. 108. Kylesku . 168S , . TuUoch of , C. 346. Langdale Burn . 177S , . Tulloch of Shalmstry, C. 437. Latheronwheel, C. 211. Tulloch of Stemster, C. 344. Leosag, C. 109. Upper Borgue, 206.C. Loch Ardbhair, S. 4. Upper , C. 217. 96 PROCEEDING E SOCIETYTH F O S , 1946-48.

III. NORTHERN MAINLAND (confd.). BROCHS (contd.). Canaster, C. 18. Cam a' Chladda, C. 467. Upper Sour, 114.C. Morm Ca , Baileuachdrach . RossE , ; Usshilly Tulloch, C. 221. personal observation. Warehouse , 190C , . Cam Mor, Birchfield, E. Ross; personal Watenan, C. 524. observation. Watten, C. 468. Creag Leathan, C. 352. Wester Broch . 513C , . Dun Mor, Doune, E. Ross; personal Wester Watten, C. 464. • observation. Westerdale, 106.C. East Kinnauld, S. 479. White Gate, Keiss, 0. 516. Gearsay . 472C , . Yarrows . 509C , . Geise, C. 430. Ha' of Greenland, C. 64. BROCH SITES. Halcro, C. 1. Hill of Stemster, C. 505. Achies . 180C , . Hollandmay, C. 39. Auckingill. . 52 . C , Housel Cairn, C. 115. Clerkhill, S. 265. Knockglass, C. 171. Cnoc Chaisteal . 386S , . Learable, S. 315. Alascaign Du . RossE , ; Arch. Scot., Lechanich, Uppe Let= ( r h Choinnich), vol. v. p. 192. E. Ross; information from Professor Buidhen Du . 544S , . Childe. Dun Phail, S. 387. Loch Watenan, C. 526. Ha f Bowermaddeno ' . 22 . C , Old Hal Dunnf o l . 461C , . Stationy Ho . 179C ,. Old Hall of Dunn, C. 463. Kettleburn, C. 588. Olrig Glebe . 322C , . Kilbrare. 24 . S , Olrig House, C. 323. KiJmster . 507C , . Oust, C. 455. Leckmelm . RossW , ; Ordnance Survey. Rattar, C. 83. Midgarty . 476S , . Scarfsferry. 62 . C , Rattar . Burn84 . C , Scoolary, C. 38. Stemster, C. 54. Scrabster, C. 433. Wilkhouse . 476S , . Smerral, C. 210. Stemster, C. 345. UNCERTAIN EXAMPLES. Thuster . 519C , . Torrisdail, S. 528. Achies. 97 . C , Tulloch Turnal, C. 200. An Dun, Drienach; information from , C. 523. M. S.. CalderC T r' . Banks of Watten, C. 465. Bilbster, C. 514. COMPARABLE STRUCTURES. Borgie Bridge, S. 185. Bowertower, C. 19. Dunan Diarmaid, W. Ross; P.S.A.S., Cairn of Dunn, C. 462. vol. xxix . 188p . . Cairn of Humster, C. 506. Sgarbach. 45 . C , SOME OBSERVATION E BROCKSTH N O S . 97

. WESIV T COAS INNED AN T R ISLANDS.

Numbers refer to the Royal Commission's Inventories.

BROCHS. Dun Choinnich, Skye, 605. Abhuinn Bhaile Mheadhonaich, Skye. Dun Feorlig, Skye, 516. 481. Dun Garsin, Skye, 482. Caisteal Grugaig, W. Ross; informa- DunHeanish, Tiree; Beveridge, op. cit., tion from Mr G. P. H. Watson. . f 7 8 . pp Dun Ard an t-Sabhail, Skye, 478. Dun Hiader, Tiree; Beveridge, op. cit., Dun Arkaig, Skye, 480. . ff 0 8 . pp Dun Beag, Skye, 479. Dun Ibrig, Tiree; Beveridge, op. cit., Dun Bhoreraic, Islay; information from pp. 112 ff. Professor Childe. Dun Liath, Skye, 655. Dun , Skye, 505. Dun Mhadaidh, Mull; information Borodaln Du e (Voradel), , 575. from Professor Childe. Borrafiachn Du , Skye, 510. Dun Mor a' Chaolais, Tiree; Beveridge, Dun , Skye, 506. op. . cit.,75 . p Dun Edinbain, Skye, 512. r VaulMo , n TireeDu ; Beveridge. op , Dun Fhiadhairt, Skye, 508. cit., pp. 76 ff. Dun Flashader, Skye, 513. n UrgadelDu , Mull; information from Dun Gearymore, Skye, 511. Professor Childe. Dun Greanan, Skye, 539. Mullach Dubh, Mid Argyll; personal HaUinn Du , Skye, 509. observation. Dun nan Gall, Mull; P.S.A.S., vol. Sean Chaisteal, Mull; information from Ixxvii. p. 40. Professor Childe. Dun Osdale, Skye, 507. Sean Dun, Lismore; information from Dun Raisaburgh, Skye, 540. Professor Childe. Dun Sleadale, Skye, 477. Stair haven, Wigtown, 310. Dun Suledale, Skye, 618. The information at present available Dun Telve, Glenelg; P.S.A.S.,. pp vol. 1 . does not warrant the inclusion of the 241 ff. following sites: Barchastallain, Castle Dun Troddan, Glenelg; P.S.A.S., vol. Chalamine, Castles, Duchoille, Dunan . ff 3 8 . Ivpp . Diarmaid, Kirkmichael Glebe, Lagan- Glen Heysdal, Skye, 514. druim, and' Tomaclare. (Arch. Scot., Kingsburgh, Skye, 619. vol. v. pp. 193 f.). Gordon's "Castle Sean Dun, Mull; P.S.A.S., vol. Ixxvii. Chonil" (Itin. Septent., p. 166) is p. 40. evidently the same as Dun Grugaig, Teroy, Wigtown, 28. Glenelg; cf. Anderson, Scotland in Tirefuar, Lismore; P.S.A.S., vol. xxiii. Pagan Times Irone Th : Age, . 183p . pp. 375 f. and 427 f. COMPARABLE STRUCTURES. UNCERTAIN EXAMPLES. Ardifuar, Mid ArgyU; P.S.A.S., vol. An Dun, Loch Fiart, Lismore; informa- xxxix. p. 260. tion from. Professor Childe. Castle Haven, Kirkcudbright. 64 , ArdweU, Wigtown, 433. Drui Duinn m Argylla d Mi , ; P.S.A.S., Dun Boraige Moire, Tiree; Beveridge, vol. xxxix . 286p . . . ff Coll8 d Tiree,7 an . pp Dun a' Choin Dhuibh, S. Knapdale; Bornaskitaign Du , Skye, 564. P.S.A.S., vol. Ixxvii. p. 41. Dun Borve, Skye, 515. n AisgainDu , Mull; information from Dun Borve, Skye, 620. Professor Childe. VOL. LXXXI. 98 PROCEEDING E SOCIETYTH F O S , 1946-48. . WESIV T COAS INNED AN T R ISLANDS (contd.). COMPARABLE STRUCTURES (contd.). Ringffln Du , Skye, 650. , Skye, 484. Dun Rudh 'an Dunain, Skye, 483. n BhuirgDu , Mull; information from Skudiburghn Du , Skye, 542. Professor Childe. Dun , Skye, 5ls. Dun Choinichean, Mull; information Dun Traigh Mhachir, Islay; P.S.A.S., from Professor Childe. vol. Ixxx. p. 00. Dun Chroisprig, May; P.S.A.S., vol Vallerainn . Du , Skye, 544. . Ixxx00 . p . Dun Voradel, Skye, 575. Dun Geilbt, Skye, 602. Dunan an Aisilidh, Skye, 576. Dun Grugaig, Glenelg; P.S.A.S., vol. Dunburgidale, Bute; P.S.A.S., vol. ff0 . xxix18 . pp . xxvii . 287p . . Grugaign Du , Skye, 651. Kildonan Bay, Kintyre; P.S.A.S., vol. Dun Kearstach, Skye, 649. Ixxiii. pp. 185 ff. Dun Liath, Skye, 541. . fortS , , Lome; P.S.A.S., vol. n MhuiligDu d ArgyllMi , ; personal xxiii. p. 406; xxv. p. 476; xxvii. observation. p. 375. Dun na Mhuirgheidh, Mull; informa- Peinduin, Skye, 630. tion from Professor Childe.

V. OUTER ISLANDS. Numbers refer to the Eoyal Commission's Inventory of the Outer Hebrides, Skye, and the Small Isles. BROCHS. Dun Loch an Duin, Barra, 445. an ' ChaolaisDu , , 442. Dun na Buaile Uachdraich, S. , Dun an Sticir, N. Uist, 171. 374. Dun Borve, Lewis. 11 , Dun Sandray, Sandray, 444. Carlowayn Du , Lewis. 68 , Dun Sleibhe,1 Lewis, 30; cf. also Arch. n CromoreDu . , alscf Lewiso ; 38 , Scot., vol. v. p. 392. Arch. Scot., . 380volp . v .. Dun Smirvig,1 Lewis; Arch. Scot., Dun Cuier, Barra, 441. . 372volp . .v . Torcuilln Du . UistN , , 172. Dun Stuigh, Gt. Bernera, 70. Loch an Duna, Lewis, 10. Dun Traigh na Berie, Lewis, 69. , S. Uist, 375. UNCERTAIN EXAMPLES. Dunan Ruadh, , 443. Dun Airnistean,1 Lewis, 33; cf. also Dunan Ruadh, Pabbay, 447. Arch. Scot., vol. v. p. 373. Loch Baravat, Lewis, 36; cf. also Arch. Dun Aligarry, S. Uist, 427. Scot., . 373volp . v .. Dun Ban, Barra, 446. Dun Baravat . BerneraGt , . 71 , COMPARABLE STRUCTURES. Dun Borranish, Lewis, 74; cf. also , , 450. Arch. Scot., vol. v. p. 393. Bann Du , , 299 . alscf ; o Arch. Dun Borve, Berneray;1 Arch. Scot., vol. Scot., vol. v. p. 399. . 399p . .v Bilascleitern Du , Lewis. 34 , Borven Du , Harris, 125 . als;cf o Arch. Dun Loch an Duin, Lewis, 51; cf. also Scot., vol. 396p . .v . Arch. Scot., vol. v. p. 378. Dun Buidhe . , , 373. Scurrivaln Du , Barra, 449. Chlifn Du , Barra, 448. Loch Hunder, N. Uist, 173. structuro 1N e remains; description from record only. NOTE SCOTTISN O S H INCISED SLABS (III). 99

VI. CENTRAL AND EASTERN MAINLAND. Numbers refer to the Royal Commission's Inventories. BEOCHS. Torwoodlee, Selkirkshire; P.S.A.S., vol. Bow Castle, Midlothian, 233. xxvi. pp. 71 f. and Ixvi. p. 341. Coldoch, Perthshire; P.S.A.S., vol. ix. ^ p. 38. Edinshall, Berwickshire, 115. UNCERTAIN EXAMPLES. Struy, Inverness-shire; Arch. Scot., vol. Hurley Hawkin, Angus; P.S.A.S., vol. . v194 p informatio.d an , n fro . 210mp . Pro.vi - e LawsTh , Angus; fessoP.S.A.S.,r Childe vol. . iii. Tor Wood, Stirlingshire; P.S.A.S., . 440volp .. . 29 . p . ix