The site of Inverquharity is 4.5 km to the NNE of modern Kirriemuir and centres on NO 405581 (illus 1). It was discovered from the air in 1983 (Frere 1984, 274 and Maxwell & Wilson 1987, 15f), although it is possible that it was seen in the early 18th century by Sibbald (1707, 16. It lies on a marked, c. 1km long, promontory (illus 2) formed by the Prosen Water and the South Esk, to the north (whose confluence lies directly north of the site) and the much smaller Quharity Burn to the south. The top of the formation slopes gently from north to south, but is bounded on the north by a steep, eroded river cliff, c. 15m in height, and on the south, by a fresher, 3m - 5m high river cliff cut by the Burn. The eastern end of the promontory carries Inverquharity Castle and its grounds, but the rest has been merged into a single field that is regularly in cereals. The site commands reasonable views in almost all directions and provides a tactically strong position, close to water. It is in an excellent location to monitor movements along all three river valleys, including those entering or leaving Glen Prosen and Glen Clova.

Aerial and geophysical data
Nothing is now visible on the surface, but the site has shown repeatedly since its discovery (illus 3). It is thus possible to offer a rectified plan derived from the cumulative data of many years (illus 4). The accuracy of this plot has since been confirmed by a resistivity survey conducted by the Roman Gask Project in 2002, which also added significant extra detail (illus 5 & 6). Ideally, a magnetometer survey would have been conducted at the same time, but the technique has often proved unreliable in this area, probably thanks to a high iron content in the natural glacial drift. Resistivity was thus seen as the technique most likely to be effective, although further work may be possible in the future.

The Fortlet
Two main features have been detected from the air. The first appears to be a large Roman fortlet. The site is surrounded by a double ditch, with a third, outer ditch visible around its northwest and southwest sides. Only three of the four sides are visible from the air and only the southwest side is visible in its entirety. As a result, although it has been possible to say that the fortlet measured 79m from northwest to southeast, inside its inner ditch, it has not previously been possible to gauge its southwest to northeast length and thus its area. This was partly because the northern edge of the plateau has been eroded since Roman times so that some of the site's north-eastern face was known to have been destroyed. But the situation was made worse by the fact that the northernmost 10 - 20m of the plateau are thickly wooded and so invisible from the air. Here the geophysical survey was able to add vital information by detecting the beginnings of the turn for the northwest corner in all three of the western ditches, just before the top of the river cliff (illus 5 & 7). As a result, it is possible to extrapolate the entire circuit (illus 6) and estimate that the fortlet would have measured c. 68m from southwest to northeast (inside the ditches), giving it an area of c. 0.52 ha (1.28 acres) and confirming the aerial impression that the site's long axis ran from northwest to southeast. It is also possible to show that the most serious damage has occurred at the fortlet's northeast corner where approximately 16m of the site has been lost, or rather more if the outer ditch continued around this side (see below. Quite when the erosion took place remains unknown. The first 6" Ordnance Survey map of the area, dates to the 1860s and shows the Prosen Water and South Esk in essentially the courses they occupy today, which approach no closer than 130m from the base of the promontory. At the same time, there is a clear waterlogged line at the base of the cliff which appears to be the silted remains of an oxbow. This combines with the current steepness of the easily weathered glacial sand and gravel river cliff to suggest that the rivers must have washed against the promontory within the last few hundred years preceding the Ordnance Survey.

The resistivity survey also shed extra light on the rest of the ditch system. Firstly, it confirmed that the outer ditch does not make a full circuit of the fortlet for, although an area of low readings was revealed to the east of the site's northeast quadrant (illus 6, A) this is relatively short and appears too sinuous to represent the feature. The survey did, however, show a slight trace of the middle ditch along the eastern part of the site's south-western face, something that has never shown from the air. A trench dug across the south-eastern defences shortly after the site's discovery, but not yet fully published (Frere 1984, 274), found the two inner ditches to be 4.9m apart (from centre to centre) and the aerial data confirms this. The excavation also found the ditches to be V-shaped, but no width and depth data were published. The aerial rectification would suggest ditch widths of fractionally under 2m.

The south-western side is penetrated by an entrance break (14m wide at the inner ditch) which faces the only easy gradient down to the Quharity Burn. Unlike the similar fortlet at Cargill, c. 32km to the southwest, the ditch ends appear to remain separate on both sides of the entrance, rather than joining in a so called "parrot beak" (in which the outer ditch turns inwards to join the inner. To the east of the entrance the inner and middle ditches terminate together but, to the west of the gate, the middle ditch ends c. 2.5m before the inner. This could have left room for the characteristic parrot beak inward swing but, if this was ever the intent, the connection does not seem to have been dug. As usual with sites of this size, no gates are visible on the northwest and southeast (short axis) sides and it has been suggested that none would have been present on the northeast side either (Maxwell & Wilson 1987, 16. Obviously this is a theory that cannot be tested since the relevant area has been eroded, but Cargill fortlet has entrances in both its long axis sides and it would not be surprising if Inverquharity closely mirrored this near neighbour. One assumes that the original theory stemmed from the steepness of the current river cliff, which would all but bar access to a northern gate. But we have no way of knowing how much more land has been eroded here in addition to the missing parts of the fortlet and it seems possible that the descent from the promontory in Roman times was at least gentle enough to justify an entrance on this side. The ground may still have been too steep to allow access by wheeled vehicles and may have been comparable to that approaching the gates of sites such as Milecastle 42 on Hadrian's Wall, but an entrance on this side would still have made pedestrian access to the main river valleys much easier.

There are noticeable differences between the fortlet's two inner ditches and the outer ditch, aside from the fact that the latter does not make a complete circuit. Both inner ditches appear to follow slightly convex courses away from the corners (especially on the south-western side), whilst the outer ditch runs straight except at its corners. The outer ditch also produces a heavier crop mark and seems to be rather wider, perhaps around 3m. Lastly, it follows a somewhat unusual course. At the southern entrance its centre line is roughly 5m from that of the middle ditch, but the bend for the southwest corner occurs too far to the west to maintain a parallel course and the centre line separation increases to c. 13m. The corner then turns through 92 degrees which causes the ditch to re-close with its neighbour and the two are only about 7.5m apart by the time they reach the cliff edge in the north. Indeed, if this ditch turned by a full 90 degrees at its northwest corner, its line would have converged with the middle ditch at some point along the now eroded northeast face, and it is tempting to wonder whether it was really designed to fit the existing fortlet.

No datable finds were reported from the excavations so the fortlet's exact date must remain uncertain. Its morphology alone makes the Roman identification seem certain and, as it seems to fit with the so called "glen blocker" series from Drumquhassle to Inchtuthil, a Flavian date would come as no surprise. This is doubly so given its distance from the nearest site with proven Antonine occupation, which lies some 44 km to the southwest at Bertha (Woolliscroft 2002, 40ff), and the similar, if slightly smaller, fortlet of Gatehouse of Fleet does have a proven Flavian date (St.Joseph 1983, 232. Nevertheless, it is worth noting that Antonine sites of this size do exist: for example Rough Castle on the Antonine Wall. No excavation has taken place in the fortlet's interior and neither remote sensing technique revealed internal details. However, the latter is hardly surprising if the site is Flavian, since any internal buildings can be expected to have been of timber. It is also noteworthy that no sign of roads or other metalled surfaces can be detected, either inside the fort or as a road stub emerging from the gate. It is possible that no such surfaces existed, but it is perhaps more likely that they have become too plough damaged to show, in the manner of the few patches of cobbling found inside the equally ploughed Gatehouse of Fleet fortlet (St.Joseph 1983, 227. Lastly, no trace of a bath house was detected either inside or outside the fortlet, although if any such a structure existed it may have been outside the survey beside the Quharity Burn on land that is unresponsive to air photography.

The Temporary Camp
To the east of the fortlet (80m at closest approach) lies a small Roman temporary camp. Only part of its north-western area was covered by the geophysical survey (Illus 6 & 8), but the rest has shown repeatedly from the air. As with the fortlet, the camp's northern side has been eroded away so that, at present, some 2.06 ha (5.09 acres) survives. But this time it is harder to estimate the site's original dimensions as there are signs that it may have been somewhat irregular in shape. 144m survives of the western side and there are faint aerial and geophysical signs that the ditch has started to turn into its northwest corner at the point where it reaches the cliff. On the other hand, some 159m of the camp's eastern ditch are visible, with no sign of a turn detected. If the northeast corner lay immediately north of the current erosion line, the camp would have covered around 2.2 ha (5.5 acres), but it may have been slightly larger. A section cut across the east ditch in 1984 (DES 1984, 35) showed it to be V shaped, 1.8m wide and 0.9m deep

The east and west sides represent the long axis, with the short axis southern side only 136m long, but again there are signs of irregularity. At the southwest corner the ditch turns through 98 degrees, whilst the southeast corner turns by just 86 degrees. This gives the southern ditch a slight inward turn, hinging at an angle of 6 degrees on the eastern side of the southern entrance. Three of the (presumably) four gates are visible. These seem to have been of the Stracathro type with a curved clavicula and straight, answering oblique traverse. However, none of the entrances have both features visible. The west gate shows a clavicula but no trace of a traverse, even though this sector was covered by the resistivity survey,. The other two entrances show the traverse only, but it is noticeable that in all three case the gate feature is 11m long and the clavicula (if originally present) would always have been in the normal position: on the left as one entered the camp. The entrances seem to vary somewhat in width (with the west gate c. 20.5m wide, the south gate c. 22.5m and the east gate 30m), although this could be an illusion caused by the vagaries of crop mark formation. None of the entrances lies at the centre of its ditch side and the eastern and western gates do not appear to face each other. There may, however, be a logic to their siting. The west gate lies c. 67m from the northwest corner and, if the northeast corner did begin just beyond the present cliff edge, the east gate would lie at a similar distance (and the east and west gates would, in fact, have faced each other. The south gate lies c. 65m from the southwest corner, which is suspiciously similar, and it may not be too fanciful to suggest an intention to site each gate 220 Roman feet (65.05m) from a corner. Finally, another slight anomaly can be seen in the camp's western ditch. For although the ditches on either side of the entrances on the other two sides line up with one another, the ditch to the south of the western entrance runs parallel to its northern counterpart, but noticeably (c. 2m) further to the west.

The supposed annexes
In addition to the two principle sites, a number of other features could be claimed as Roman. Firstly, a section of ditch (illus 4, B) has been traced from the air from a point just south of the Fortlet gate's eastern ditch termini to the southern boundary of the modern field. The known stretch is 28m long and leaves a c. 5m gap between itself and the fortlet ditches. A much shorter (c. 9m) length of ditch (illus 4, C) has also been seen just c. 7m to the east. The latter resembles a temporary camp titulus, although it lies to the south of the gap between ditch B and the fortlet defences, rather than across it, and no break in ditch B can be seen at this point. The two ditches appear to have roughly the same thickness as the fortlet's outer ditch and the features have been suggested as part of an annex (Maxwell and Wilson 1987, 16), although it is impossible to be certain of this from current evidence. Interestingly, although both Features B and C have shown strongly and repeatedly from the air, they were not detected at all by the resistivity survey (illus 7) and, given the strength with which the site's other principle ditched structures were detected, this is an anomaly that would benefit from further investigation.

25m to the east of Feature C, lies another length of ditch (illus 4, D). This has again been detected only from the southern boundary of the modern field, but from here it makes a straight north-easterly run of c. 53m (D1) at the northern end of which it crosses the fortlet's middle ditch and appears to touch the inner ditch at its southeast corner. It then makes a sharp, angular, 82 degree turn to head southeast (D2. From the air, the feature has only been traced as far as the temporary camp ditch (illus 4, D2) and it has been suggested that it too might have formed an annex which utilised pre-existing camp defences at this point (Maxwell and Wilson 1987, 16. But the resistivity survey traced it running through the camp ditch to a total length of 99m, before making an angular 90 degree turn to the northeast and continuing for a further 19.5m to the cliff edge (illus 6, D3 & illus 8. It might be tempting to relate all these structures by envisaging the fortlet's outermost ditch running straight past its northeast side to join up with ditch D3, whilst ditch D1 turned to join ditch B, to the south of the modern field boundary. This would produce a double annex but, although parts of this model are beyond testing, thanks to the erosion of the northern face of the promontory, it does seem unlikely for a number of reasons. Firstly, ditch D appears narrower than both ditch B and the fortlet's outer ditch. Secondly, ditch D's sharp angular turns contrast markedly with the normal Roman curved ditch corners. Thirdly, the resistivity survey was extended south of the field boundary (as far as the low river cliff cut by the Quharity Burn) opposite ditches B and D1 and no sign of a linking ditch was detected (illus 7). Finally, although it is not unknown for annexes to be added to existing Roman military installations and, thus, for their ditches to cross those of the parent site, the fact that ditch D crosses the fortlet's middle ditch twice without breaking on its berm does suggest that the two are far from contemporary. Only excavation can answer this question with certainty, but D has the look of a relatively modern ditch, perhaps cut after the fortlet and camp had ceased to be significant surface features. It seems rather large to be agricultural and it may be connected with the castle. There is a possibility that Inverquharity was involved in Montrose's campaigns of 1644-6 (Williams 2001, 263) and one could speculate as to whether the feature may be part of a fortification dating to that time.

Prehistoric features
The site is also replete with probable prehistoric features and it is worth pointing out that the aerial and resistivity surveys produced very different collections, with remarkably little overlap. In all, twelve have been detected. Air photography has detected an 11m diameter ring feature (illus 4, E) cut by the fortlet's western outer ditch, and a similar sized penannular ring just inside the fortlet's entrance (F. The latter contains two small maculae, which may be either postholes or cist burials. Two more penannular rings are faintly visible towards the western end of the promontory, of which Feature G is again 11m in diameter, whilst Feature H is 1m larger. Further east, a tiny penannular ring feature (I), just 5m in diameter, is visible inside the temporary camp, whilst a larger, c. 15.5m ring (J) has been photographed outside the camp's eastern entrance. The resistivity survey failed to detect any of the above, but did add three more penannular features. In the area between the fortlet and camp, two rings were found close together, of which Feature K (illus 6 & 8) is c. 10m in diameter and Feature L, c. 8.5m, whilst to the west of the fortlet (illus 6 & 7), a single feature (M) measured 13m across. It is dangerous to draw too many conclusions from penannular features detected by any remote sensing technique, as there may be various reasons why only part of what may really be a complete ring has been detected. Nevertheless it is worth noting that the Inverquharity ring breaks show no consistency in orientation. This is especially relevant since only Feature F has produced internal maculae and so the remaining structures seem more likely to be domestic rather than funerary. A certain dogma has grown up that roundhouse entrances are generally oriented to the east or southeast but, although Features G and H do fit this pattern, the site also show breaks facing north, northeast, northwest, and southwest.

This concentration already has the makings of a substantial settlement, but it may well be only a beginning. For the single ditch section dug in 1983 located what might be a roundhouse construction trench on the berm inside the fortlet's inner ditch, a point where nothing has been recorded by remote sensing (DES 1983, 33. As a result there seems every possibility that larger scale excavations might locate more. More importantly, air photography has detected a 25m long curved feature inside the temporary camp (illus 4, N) which has long been suspected of being a souterrain. Here, the resistivity survey was able to add materially to existing information. Firstly, this was the only feature on the site to show as a high, rather than low, resistance anomaly, which would be consistent with the stone linings found at many souterrains and, secondly, a smaller (7.5m diameter) penannular feature (illus 6, N & illus 8) was detected attached to the south side of the known site's east end. This too showed as high resistance readings and might represent a stone lined side chamber like that at Shanzie, 15 km to the southwest (Coleman & Hunter, 2002. Alternatively it might be a stone floored roundhouse attached to the souterrain, similar to a free-standing, Roman period example excavated by the writers at East Coldoch, Stirling (Woolliscroft & Hoffmann forthcoming.

The chronology of the likely prehistoric features and their relationships both inter-se and with the Roman sites are hard to disentangle from remote sensing data alone. Obviously Features E and F cannot be contemporary with the fortlet and neither can the feature found on its berm in 1983. Likewise, Features I, N and probably J seem at least unlikely to have been occupied at the same time as the camp. Features G, H, K, L and M are another matter, however, and it would be interesting to see what excavation might show of their histories. Moreover, the camp occupation is likely to have been brief, so it would still have been perfectly possible for Features I, J and N to have coexisted with the fortlet. Armit (1999) has suggested that souterrains may have been a specific native response to Roman military supply needs in the Flavian and, more particularly, Antonine periods, and that they were then abandoned en-mass in the late second and early third centuries. Certain aspects of this model have since been challenged (Coleman & Hunter, 2002, 95ff), but again excavation on a site so close to a Roman installation may well add to our understanding. To say more would be speculation, however, except to point out that similar native concentrations are known around other Flavian sites in the area, notably the forts of Cargill (RCAHMS 1990, 84f) and Cardean (Hoffmann, forthcoming.

Probable Medieval and modern features
Aside from Feature D, a number of other features have been revealed of indeterminate or relatively modern date. A series of lines of low resistance were detected between the fortlet and camp which seem likely to represent rigg and furrow (illus 6), dotted & illus 8), whilst air photographs show a straight, c. 45m long, crop mark (illus 6, O) to the west of the fortlet which may be a modern land drain. The resistivity survey also detected a 22m long curved feature (P) passing across ditch D1, and a more sinuous 27m feature (Q) inside the fortlet. Both could be taken for further souterrains on morphological grounds alone but, as both showed as low resistance anomalies, such an interpretation might seem unlikely, albeit timber lined souterrains are known (DES 1997, 13. They may be connected with drainage, especially Feature Q, which passes through the fortlet's entrance, but they might also represent geological features or paleo-channels in what was once a glaciated landscape.

Acknowledgements
The writers wish to thank Miss R.Dundas, Miss S.Moore, Miss G.Williams and Mr D.Hodgson for their help during the resistivity survey, Cambridge Air Photography and the RCAHMS for providing air photographs and Mr J.Stormont Darling, Mr W.Taylor and the Kinnordy Estate for allowing access to the land. We also thank Geoscan Ltd for their very prompt help in repairing a damaged resistivity probe array and SUAT Ltd for the loan of a replacement. The Roman Gask Project is sponsored by the Perth and Kinross Heritage Trust.

Bibliography
Armit, I 1999 'The Abandonment of Souterrains: Evolution, Catastrophe or Dislocation?', Proc Soc Antiq Scot 129, 577-96.

Coleman, R & Hunter, F 2002 'The Excavation of a Souterrain at Shanzie Farm, Alyth, Perthshire' Tayside & Fife Archaeological Journal, 8, 77-102.

Frere, S S 1984, "Roman Britain in 1983", Britannia, XV, 265 - 332.

Hoffmann, B Forthcoming, Excavations at the Roman Fort of Cardean.

Maxwell, G S and Wilson, D R 1987, 'Air Reconnaissance in Roman Britain 1977-1984, Britannia, 18, 1-48.

RCAHMS 1990 North-east Perth an Archaeological Landscape, London.

St.Joseph, J K 1983, 'The Roman Fortlet at Gatehouse of Fleet, Kirkcudbright' in Hartley, B & Wacher, J (eds) Rome and her Northern Provinces, Gloucester.

Sibbald, R 1707, Historical Inquiries, Edinburgh.

Williams, R 2001 Montrose, Cavalier in Mourning, Edinburgh.

Woolliscroft, DJ 2002 The Roman Frontier on the Gask Ridge, Perth & Kinross, BAR (British Series), 335.

Woolliscroft, D J and Hoffmann, B forthcoming "Palisaded Enclosures, Cist Burials and a Roman Period Iron Age Site at East Coldoch, Stirling".