Mapping for gender. Interpreting artefact distribution inside 1st- and 2nd-century A.D. forts in Roman Germany

Artefacts from military bases of the early Roman Empire potentially indicate not only the presence of women and children inside the walls, but also their movements, activities and impact on fort life. This paper explores dynamic approaches to categorizing and gendering artefacts for more holistic investigations of artefact assemblages. It uses GIS mapping techniques to analyse the distribution patterns of ‘gendered’ artefacts within three forts on the German frontier – Vetera I, Ellingen and Oberstimm. It investigates the social significance of these patterns within and between the forts to better understand women’s place in this sphere.


Introduction and background
The Roman world is epitomized by 'manliness' and 'masculine values and virtues, those of the military man and the administrator' (Lefebvre 1991, 249). To redress the balance Roman social and art historians have been foregrounding women in Roman public, private and commercial life (e.g. Pomeroy 1975;Treggiari 1976;Dixon 2001). However, Roman archaeology, and particularly Roman military studies, continue to be masculine and imperialist (Freeman 1997). As argued by Hingley (2000, esp. 15, 47, 58, 147-50), investigations of Roman military sites have traditionally been carried out within a late 19th-to early 20th-century world view that saw the Roman Empire as a model for contemporary imperial and military activities, resulting in approaches to Roman military sites as male domains, combat units at the edge of the civilized world.
Roman authors wrote about the inappropriateness of women and families in these combat zones. Wives were considered a hindrance to military discipline (Herodian, 3.8.4) and sharply criticized for involvement in political or military discussions (Juvenal,. Such views have supported the widely held perception among modern scholars, themselves male and often with military backgrounds (James 2002, esp. 10-11), that the Roman frontier was no place for women and families.
But textual and epigraphical sources also indicate that wives and families, as well as tradespersons, artisans and slaves, did indeed accompany the Roman under-floor heating, painted wall decoration and sometimes even mosaic floors. Whether or not centurions' marriages were legal, there is evidence that they existed and that the families were most probably accommodated inside the fort.
No such elaborate furnishings or structural variations are witnessed in the barracks of ordinary soldiers. However, Van Driel-Murray (1994; has investigated the size ranges of leather shoe remains from a number of 1st-and 2nd-century military sites. At Vindolanda she noted predominantly male footwear from the commanding officer's quarters during Period II (ca A.D. 90), an increased range of shoe sizes here in Period III (ending ca A.D. 104), signifying the presence of the commander's family in this later period, and concentrations of women's and children's shoes in ordinary soldier's barracks in Period IV (ca A.D. 104-20). These findings, possibly rubbish left behind by departing troops, have been used to argue for the existence, and perhaps habitation, of ordinary soldiers' families within the barracks during the early 2nd century. Van Driel-Murray wrote (1997,60) that once it is 'accepted that women did form a significant section of the camp population, we can begin to develop material correlates by means of which their social and economic roles can be investigated'. She also stressed that it is 'to whole packages of attributes that we must look' to understand the statuses and roles of the women and children in these domains, long considered male (Van Driel-Murray 1997, 55).
On the basis of house numbers and individual's names found on wooden tablets excavated from the rubbish dump of the legionary fortress of Vindonissa, in Switzerland, Speidel demonstrated (1996, 55, 80) that, opposite the main baths inside this 1st-century fortress, there had been a tavern, or perhaps even a brothel, where a female barmaid or innkeeper, Belica, worked. Next door there had been an inn, run by a female landlady. Speidel reported (1996, 186-87) that gaming stones, dice and kitchen utensils were found in the area, documenting entertainment and perhaps public eating and drinking in these establishments. This evidence does not verify the residency of these women within this fortress, but it suggests that they were employed within the fort walls, in establishments highly likely to have provided accommodation for their staff.

Engendering artefacts
Thus acknowledgement that women and families were an integral part of the community within a Roman military fort, before the lifting of the marriage ban at the end of the 2nd century, is taking root. While Phang argued that there is inadequate evidence for their domicile within the fort proper, Van Driel-Murray and Speidel found hints of their presence and at least some of their roles. This paper presents my approaches, first to identifying material traces of women and children within early imperial military forts, and then to analysing the distribution patterns of these traces for information they provide on the presence and roles of women and children in this sphere. I discuss frameworks for ascribing gender and gendered activities to Roman artefacts and demonstrate how the spatial distributions of these 'gendered' artefacts are analysed to see whether they reflect 'known' information about women discussion and children within particular parts of the fort. If they do, and thus give these gender-attributions some validity, then less well-'known' observations about the roles of women and families within forts can be suggested.
Essential to this study are critical, but constructive, approaches to the relationships between the different spaces within these forts; artefacts left behind in these spaces, the activities they document; and the people who carried out these activities. Recent consumption studies are concerned with relationships between material-cultural consumption, space and gender (e.g. Jackson and Thrift 2001;Delphy 2001), and Moore (1999, 156-57) has called for more emphasis on the 'doing' of gender rather than the 'being'. However, it is no simple matter to identify socio-spatial and gendered practices in the past through artefacts and artefact distribution.
In the first instance it is not always easy to ascribe a particular activity to a specific excavated artefact. For example, many pierced bone or glass discs have been found on Roman sites. Contextual evidence from Pompeii indicates that discs of these types had a number of functions (see Allison n.d.a). Some could be worn in necklaces, some used as spindle whorls, some used as furniture decoration and some used, or reused, as gaming counters. Similarly, while melon-shaped glass beads were worn in necklaces (Allison n.d.a), they also decorated horse harnesses (Bishop 1988, figures 1-4 and 8-10), as well as shields and axe sheaths (Hoffmann 2002, 230; axe sheath on display in the Bonn Museum).
One of the main obstacles to Roman archaeologists investigating consumption practices is that excavations of Roman sites are invariably carried out from production-oriented perspectives (see Allison 1997). In Roman archaeology, in particular, there has been a presumed precedence of production over consumption (see Green 2005), rather than a concept of reciprocity, with consumption as the logical outcome of production and not as an active agent. Therefore consumption studies are usually not an integral part of the original research design of an excavation, or of finds catalogues. Rather they have tended to be carried out at a later stage (e.g. Cool 2002;Cool and Baxter 2002;Gardner n.d.;Pitts 2005). This is a particular difficulty for spatial approaches to consumption analyses, which must address an artefact's deposition. For example, some artefacts, particularly in pits and wells, may have been 'ritually' deposited and their 'symbolic value need not reflect [their] practical use' (Clarke and Jones 1996, esp. 119). Only by considering deposition processes, examining artefact assemblages at sites with good depositional information (e.g. Pompeii -Allison 2004; n.d.a) and comparing associations and distribution patterns across a number of sites can one start to get a sense of the most probable functions and associations of artefacts and of their socio-spatial significance.
Gendering artefacts according to such probable functions is a further complicated and relatively subjective step. To ascribe a specific gender to a particular artefact, the artefact must either be a part of dress that is peculiar to the relevant sex, or be associated with an activity carried out only by a specific sex. While Tringham has argued (1991,94) that 'gender is an essential level of inquiry [in archaeology] for reconstructing past social behaviour', the premises that consumption, and activity, categories are culturally constituted (Kopytoff 2001,13), and that engendered relations and sexual identities (Voss 2005) are constituted in historically specific ways, mean that one cannot necessarily assume that particular categories of material always carried a particular status or gender value, without detailed consideration of the assumptions involved in that reasoning (see Rautman and Talalay 2000, 4).
Hill has noted (2001,51) that the 'quality of [archaeological] data to address gender issues is considerably greater [in the Roman period] than for any prehistoric periods, and as good, sometimes better, than much medieval evidence'. Nevertheless, gendered approaches to the broader range of Roman material culture, particularly from excavations rather than sculptural or epigraphical remains, are under-represented. As Allason-Jones has pointed out (1995), 'sexing Roman small finds' is no simple matter. Different forms of dress, for example, carried different gendered meanings in different contexts, and these contexts could often be overlaid in the various and diverse spheres of the Roman world. While the archetypal Roman male considered the wearing of jewellery effeminate, males from the eastern provinces and Africa could wear earrings and beads, become Roman soldiers and inhabit forts in the western provinces (Allason-Jones 1995, 25-26). Roman burial practices provide great insights into gendered material culture but do not necessarily reflect lived practices. Forms of dress worn by women in the western provinces in the early empire are discernible from sculptural evidence (Wild 1968;Böhme-Schönberger 1995). Brooches of certain types and their positions in female graves document this dress type and indicate that particular brooch types were worn by women (Böhme 1972;Böhme-Schönberger 2002;Martin-Kilcher 1993). However, such female brooch types have also been found, although infrequently, in German male graves (see Böhme-Schönberger 1994, 126;1995, 9).
Attributing artefacts other than dress to particular gendered activities is even more complex. As Moore has argued (1999), considering their historical specificity, the processes of gendered 'doing' are more problematic than those of gendered 'being'. This is particularly relevant for the Roman world, not least because of growing awareness of women's involvement in many spheres. In the civilian Roman world there is evidence that women were involved in a range of activities and professional and commercial occupations from prostitutes, vegetable sellers and physicians (see Evans 1991, appendices;Setälä and Savunen 1999) to wine merchants (bronze stamp in National Museum, Rome), to brick-factory owners (Anderson 1997, 158-59). These occupations were more likely to be distinguished along status lines (free, freed, slave) than sexual. There are few activities that both are exclusively female and leave material traces, although one potential possibility is cloth-working.
Essentially, the concept of gender in the Roman world cannot be expressed in a simple male/female dichotomy but is often determined by age, status and ethnicity (Monserratt 2000, esp. 162-65). Roman gender studies acknowledge that the basic categories of man and woman are problematic and so have moved on from simply foregrounding women to investigating the gendered relationships of differing statuses: child/adult; free/freed/slave; Italian/non-Italian (Monserratt 2000, esp. 164). However, for much of Roman archaeology and military studies the balance is far from having been redressed. In these spheres the roles of women are still marginalized, or even invisible, and need first to be foregrounded before more complex gender dynamics can be addressed (e.g. Cool 2002, 29-30;2004, 400 and 454).
Because of the uncertainties of gendered associations of artefacts, it has been relatively easy for scholars to explain away evidence for potentially female-and child-related artefacts, especially in military contexts, by considering that such artefacts more probably carried the (often quite rare) male or animal associations. When Böhme first identified certain brooches from the fort at Saalburg as being of female types (1970; 1972), scholarly opinion questioned her identifications. Similarly, melon beads and bronze pendants found inside military forts are usually considered to be associated with cavalry horses. However, both artefacts are found in domestic contexts (Allison 1997, 80;Allison n.d.a), and melon beads are common finds at military sites that were unlikely to have had a resident cavalry (see Schönberger 1978, 148-50, Böhme in Schönberger 1978. Hoffmann found (n.d.) that melon beads were as numerous in the supposed civilian settlement outside the fort, the vicus, as in the fort proper, if not more so. If it is not possible to argue away the female associations of certain artefacts found within military forts, lack of careful stratigraphical excavation often renders it conceivable that they are intrusive. For example, spindle whorls recorded at the auxiliary fort at Ellingen were considered possibly pre-Roman (Zanier 1992, and 4.31% of the stray finds recorded at this site are potentially female-or child-related, which is a very high percentage given that only 0.48% of all recorded artefacts at this fort are potentially female-or child-related. It intimates that scholars prefer to consider such artefacts stray finds rather than assign them to a military context. In summary, approaches to the interpretation of material culture from Roman military forts have tended to encompass perspectives that perpetuate the view that these were male-only communities. Because traces of women are relatively invisible, it is widely assumed that women were not present. As Van Driel-Murray pointed out, 1 it is equally difficult to trace definitive male presence within such communities. If women and children were indeed present inside these forts (Van Driel-Murray 1997, 55), then we can assume that some of the artefacts within the forts belonged to them. If so, then we should be able trace the areas that they frequented and perhaps even their roles within the forts, through the distribution of items mostly likely to have been associated with them. This paper is, therefore, based on the premise that certain items found within military forts were more probably associated with the activities of women and children than with those of particular male groups, such as ethnic soldiers or cross-dressers. It is by no means certain, but the latter, at least, were more likely than women to have been anomalous in this sphere. Thus the paper assumes that there is a tendency for certain items to carry a specific gendered meaning and that patterns of association can reinforce the gendering of certain items.

Categorizing engendered artefacts
Rather than ascribe a specific activity or gender to each artefact, the artefacts in this study are ascribed a series of categories that first offer a potential range of activities with which the artefact is likely to have been associated, and then the potential range of people who might have been involved in these activities -as a range of maleness to femaleness, childhood to adulthood. I have focused on artefacts associated with dress and on tasks whose gendering is, except in exceptional circumstances (e.g. Boudicca and Plancia Magnussee Monserratt 2000, 165), relatively 'safe' (e.g. spinning for women and combat for men). These categories do not define the activity or a gendered ascription but rather explore the range of activities and genders with which a particular artefact might be associated.
For example, as mentioned above, pierced bone and glass discs had a number of possible functions. From the associations of various types of these discs, found in the Insula of the Menander in Pompeii (figure 1), it is possible to construct a simple typology of the types and size ranges most likely to have been used as either jewellery, spindle whorls, furniture decoration or gaming counters (Table 1). Thus a bone disc with a diameter of less than 50 mm and a central hole of less than 10 mm diameter was probably either a bead or a spindle whorl, whose activities can be classified as 'dress?/cloth-production?'. Given that such beads were more usually worn by women and that spinning was usually carried out by women, they are gendered as possibly 'female?'. discussion a b c d Figure 1 Pompeii: a. spindles found in room 2, Casa del Fabbro (Pompeii inv. no. 5234A-B); b. furniture legs from room 36, Casa del Menandro (Pompeii inv. no. 4915); c. bone beads from a necklace from above room 7, Casa del Fabbro (Pompeii inv. no. 5333C-I); d. bone and glass discs from upper levels in House I 10, 2-3 (Pompeii inv. no. 5077A-D) (photographs by J. Agee).

Analysing Roman military sites
Using this process I have ascribed a range of possible activities and gender associations (Table 2) to artefacts recorded at 1st-and 2nd-century forts.
There is obviously a certain amount of subjectivity in these ascriptions. The aim is not to substantiate them but to explore their spatial distribution, using geographical information software (GIS), first to test the validity of these ascriptions and then to investigate the information they provide on the movements and activities of women and children within these forts. For Raetia G e r m a n ia S u p e ri o r G e r m a n ia example, the categorization of melon beads as 'female?/child?' is not intended to indicate that those found in military forts were, of necessity, worn by women and children. Instead it allows for investigating the overall distribution patterns of all potentially female-and child-related items to analyse any spatial correlation between these melon beads and other so-gendered items. I have investigated artefact distribution patterns at three military sites on the German frontier -Vetera I, Oberstimm and Ellingen (figure 2).
Vetera I The double legionary fortress of Vetera I, in the Lower Rhine, covered an area of about 600 m × 900 m. It had three earlier periods of construction but the main stone-built fortress dated to the Claudian-Neronian period (ca A.D. 40-70). The central area was excavated in the early 20th century so the stratigraphy is not always clear. The finds and their provenances have recently been comprehensively published by Hanel (1995). The distribution patterns of items ascribed gendered activities (e.g. combat equipment, metalworking, weaving, toilet, gaming etc.), 2 or gendered dress, indicate, as one might expect, that the overwhelming majority can be gendered 'male' (Ma), or possibly 'male?' (ZMa) (figure 3). 3 However, there is a scattering of items that can be gendered 'female' (Fe), or possibly 'female?' (ZFe), and a number potentially associated with children (ZCh or ZFe_Ch). Removing all the items gendered 'male', 'male?' or 'male?/female?' reveals more clearly that material assoiated with women and children is predominantly found in the gateways, the main street and central market area, and in the officers' residences, Buildings K, J, M, H and P (figure 4). The items categorized as possibly female-or child-related tend to be associated with definite female-related items, and to be found in clusters with each other, irrespective of type. This implies that there is some validity in my gender ascriptions. The distribution pattern also conforms to the view that women and children within this 1st-century legionary fortress were either members of officers' households or traders who frequented the main street and market areas. The only apparent anomalies are a handful of possibly women's and children's items scattered across the central administrative buildings. However, multivariate analyses indicate that the main distribution pattern in the GIS plots is fairly robust, with women's and children's items clustering with officer's quarters and street areas, and male items clustering with administrative buildings and barracks (Allison et al. 2005, figures 28-29). This suggests that women within this fortress are most probably officers' families and tradespersons, although the ordinary soldiers' barracks have not been extensively excavated.
Of the items gendered according to activity (supplementary figure 8), the most prolific material was combat equipment, and stone-and metalworking equipment. The former was found mainly in Building 7, the praetorium or commander's residence, and barracks 6. The latter was found in Building 1 and the area of Building 3. Schönberger (1978, 45-46) identified Building 1 as an industrial area during Period 1.
Items possibly associated with the women and children (figure 5) are scattered across the excavated areas. Definite female-and child-related material is concentrated in the area of Building 3, identified as accommodation for craftsmen or immunes (Schönberger 1978, 68-73), and on the west side of Building 1, particularly in the north-west corner. It was also found in Building 7, the commander's residence, and between Buildings 12 and 14, which have been identified as taverns and soldiers' barracks, respectively. As at Vetera I, there seems to be a relatively close association between definitive and less certain female-related items. At least six of the potentially femaleand child-related items in the area of Building 3 and one in Building 12 can be dated to Period 1b-c, the main occupation period of the fort. Most of the others found in these areas are likely to belong to Period 1. The female-and child-related items in Building 7 belong mainly to the transitional Period 1d, when the fort was probably being rebuilt (Schönberger 1978, 143-44). The items in the area of Building 1 belong mainly to Period 2, when the fort was reoccupied and when this area appears to have been a relatively open space, with seemingly commercial activity in the north-west section (Schönberger 1978, 143). The female-and child-related items at Oberstimm include 28 melon beads. Because of their small size, Böhme (in Schönberger 1978, 288-89) felt that these beads were unlikely to have been worn by horses. They are evident in the area of Building 12, which Schönberger argued was not a barracks but possibly coach houses, tabernae and storerooms (Schönberger 1978, 118), and where some harness equipment was found. However, these beads were more prominent in other areas of the fort, notably in the eastern part of barracks 6, which may have been officers' quarters, and in the area of Building 3.
This distribution pattern suggests that women and families were prominent in a number of areas of the fort and probably in a number of aspects of fort life, in all its occupation periods. If Schönberger's identifications of the various buildings and areas are correct, then it is possible to surmise that the immunes and conceivably also the troops, or the centurion, in barracks 6 resided with their families within the fort, and that either these or other women were actively involved in commercial and possibly industrial activities in the area of Buildings 1 and 12. Rather surprisingly, the evidence for female presence in the commander's residence is mainly dated to the period when the fort was occupied by troops involved in its reconstruction. Given that this fort is identified as a supply station and was unlikely to have housed an active garrison, it is perhaps unsurprising to find that women seem to be as integrated into this community as they might have been in a civilian community. There is no rationale, beyond the perspectives of elite male authors on ideal womanhood (Phang 2001, 368;Fraschetti 2001, 2) and long-standing prejudices, for assuming that women did not play an active role in the functioning of a unit of this type.

Ellingen
The first fort at Ellingen was built ca A.D. 120 and replaced in A.D. 182 with a stone-built structure covering 0.7 hectares (Zanier 1992). Zanier has suggested (1992, 165) that it housed a garrison involved in the construction of the Limes. The date for the fort's demise is unknown but it was probably at the beginning of the 3rd century.
The distribution of items associated with gendered activities (supplementary figure 9) indicates large quantities of stone-and metalworking equipment across the fort, with the next most prolific activity being cutting and sharpening. There seems to be little combat equipment, and only one item that was positively identifiable as combat dress. This pattern contrasts with that at Oberstimm and Vetera I (see Allison et al. 2005, section 8.6.2a), where Vetera I had good evidence of combat equipment but limited stoneand metalworking equipment and Oberstimm had these only in specific areas.
At Ellingen, Building C has the greatest concentration of artefacts of all types. Zanier has noted (1992, 69-70) that material had been redeposited here for a new floor in either Period 1b or 2 (i.e. ca A.D. 150-82). He surmised that this material must have been brought in from outside the fort because it contained perinatal human skeletal remains and tubular tiles, of the type used for hypocaust heating. However, tubular tiles and perinatal skeletal remains (figure 6) occur elsewhere within the fort, and many of the perinatal remains were found in pits and as partial skeletons, suggesting that they were in situ infant burials (Schröter in Zanier 1992, 305-6). Such fragile partial skeletal remains are most unlikely to have survived redeposition. It is more probable that infant burials were dug into the redeposited material in Building C and, therefore, that they are associated with the activities of this fort, and especially of this building. The distribution of other female-and child-related material at Ellingen seems to bear this out (figure 7), particularly definite female material in Well 4 and in the main street area outside Building C. It is difficult to assess how much material within Building C had been redeposited, but some of this material certainly belongs to the later occupation of this building, notably that in Well 4. Other areas which had the main densities of women's items are again discussion the streets and gateways but also barracks B. Areas with concentrations of workshop activities (e.g. Building D and Area G), and administrative buildings (A and E) had a dearth of female-related material. Zanier identified Building F as the commander's house in Period 2 (1992, 86). Only two artefacts, found in the upper levels of Shaft 6 associated with Building F, might support this identification. A greater density of femaleand child-related material was found in barracks B. The prominence of female-related material in Well 4 and the likelihood that infant burials were cut into the redeposited material under the floor in Building C suggest that women lived in this building, at least during Period 2 (i.e. after A.D. 182).
Thus the distribution of female-or child-related artefacts at Ellingen indicates that areas identified by Zanier as administrative buildings and workshops had little or no such material but those identified as soldier's barracks (Buildings B and C) had considerable quantities, including infant burials. The evidence in Building C could be taken to indicate that Building C, and not Building F, was the commander's house. Considered together with the evidence from barracks B, though, the remains in Building C imply that women were bearing children in the ordinary soldiers' barracks and were undoubtedly resident there, very probably before Septimius Severus' marriage reforms of A.D. 197, and irrespective of the legitimacy of such families. Evidence for spinning items in the vicinity of these barracks suggests that these women continued to carry out their 'ordinary female' tasks within the fort walls. Unlike Vetera I and Oberstimm, there is little in the way of structural evidence for commercial activities at Ellingen but the main street between the north and south gateways may well have formed a market area. Possible female-related material from the earliest period of the fort's occupation was found in Well 1. It is conceivable that the material from this well had been ritually deposited at the end of this well's life (see Clarke and Jones 1996, 121-22) and that the female items were associated with a female deity rather than living women (see Allason-Jones and McKay 1985). In any event, such items in this area are minimal compared with those in barracks B and C, suggesting that women were residents rather than visitors coming into the fort from outside to sell their produce.
The need for Zanier to argue away the evidence for infant burial inside the fort at Ellingen is undoubtedly based on preconceived ideas about who should have inhabited the fort rather than on the actual depositional evidence. Given the unclear stratigraphy, and apparent post-fort reoccupation of Building C (Zanier 1992, 67) it is easy to assume that the material found here was intrusive. However, the spatial analysis indicates that material found in Building C differs from that found in other parts of the fort only in quantity. It seems unreasonable to assume that this material was redeposited from outside the fort. Given its relatively high artefact content, any redeposition was very probably of rubbish from within the fort itself. No coins dating later than A.D. 186 were found in the material from Building C (Zanier 1992, 66) and Zanier has not suggested that the artefacts from this building could have been from a later reoccupation (for further discussion on Ellingen see Allison, n.d.c).

Presence, habitation and roles of women in these forts
Material remains, mostly dress-related, document the presence and movements of women and children within these three forts. Much of the female-related material from the gateways, streets and more public areas of these forts conceivably belonged to itinerant traders and service-people coming into the fort from a settlement outside. However, while material at Vetera I might fit into this category, this is less apparent for Oberstimm discussion and Ellingen. But even at Vetera I this material did not necessarily belong to workers coming into the community by day, not least because no such outside settlement associated with this fortress has been discovered. The evidence at Vindonissa for female innkeepers and barmaids within the fortress supports the argument that similar personnel might well have been resident at Vetera I. At both Vetera I and Vindonissa this evidence is associated with material related to eating, drinking and gaming (Speidel 1996, 80;Allison n.d.b, figures 5-6). It therefore seems that commercial and leisure activities, possibly including the provision of sustenance and perhaps also prostitution, traditionally believed to have taken place in settlements outside the fort (Allason-Jones and McKay 1985, 64-65), were part of the established activities within these fortresses, performed by women who were resident in these commercial and leisure establishments. The evidence is sparse at Oberstimm but it seems that here, too, women were actively involved in commercial activities, possibly related to the preparation and supply of food and drink.
At Vetera I the distribution pattern seems to conform to the traditional view that women were resident within the officers' quarters, but this may well be due to limited excavation. At Oberstimm and Ellingen such a pattern is less clear for officers' quarters. At these two forts there seems to be a more prominent pattern for female residency within the barracks of ordinary soldiers and craftsmen. Only during the rebuilding period at Oberstimm is there evidence for female presence within the commander's house. In this fort women seem to have frequented all areas and were therefore very possibly significant players in the fort activities, which involved the supply of necessities to other military establishments on the frontier. At Ellingen their movements seem to have been restricted to the soldiers' residential areas, with other parts of the fort given over to what might be considered the more masculine activities, especially metal-and stoneworking for the construction and maintenance of the frontier, and possibly also woodworking and meat production (see Zanier 1992, 171). Whatever their other activities at Ellingen, these women certainly appear to have been producing soldiers' families within the fort precinct. The limited and unprovenanced 3rd-century material at Ellingen suggests an end date close to A.D. 200 (Allison n.d.c). This dating, and women's apparent presence in early occupations of the fort, imply that the existence of these families was unrelated to Septimius Severus' lifting of the marriage ban at the end of the 2nd century. Even if this material is datable after A.D. 193, these families were certainly not housed outside the fort walls (cf. Phang 2001, esp. 35 and 127-29).

Implications of the presence of women within these forts
Issues of space, of the necessary efficiency of a military unit and of the systematic layout of military forts (see Polybius VI,[27][28][29][30][31][32][33][34][35][36][37][38][39][40][41][42] are frequently given as reasons why women could not have been accommodated within these communities. Calculations of the space required to accommodate a single soldier (Petrikovits 1975, 36) are based on very proscriptive approaches to military life and to human behaviour. In many excavated forts the systematic layout is hypothetical, as few forts have been comprehensively excavated. An assumed disruptive, as opposed to supportive, role of women in military life seems to stem from age-old misogynistic perspectives on female behaviour.
With the exception of marching camps, most Roman military bases were relatively long-lived communities. Those discussed here have been selected because of their comparatively short life. Given that a soldier usually served for 20 to 25 years, much of his adult life was spent in active service. It is difficult to make a direct analogy between this life and that of soldiers who have served in armies in modern times. While the Roman administrative system would have considered it an expensive option to allow these soldiers to have legitimate Roman marriages, this does not render ordinary soldiers' families non-existent (see Phang 2001). To assume that married men and their families occupied separate quarters outside the fort proper (Birley 1977, 46-48;Phang 2001, 35) is to construct a solution to their evident existence that is more dependent on Roman elite prejudices, traditions of Roman military history and modern analogical inference than on anything inherent in the material evidence. That there is a need to foreground the presence of women within military forts is a tribute to the strength of such prejudices.
If families of ordinary soldiers and tradeswomen providing food, entertainment and other services were accommodated within the fort, then the proscriptive approaches to spaces available to each soldier, and calculations of the strength of a specific troop based on space, are rendered suspect. Views that soldiers cooked and ate together in their units of eight also need to be revisited. Likewise, perceptions that entertainment and married quarters were only to be found in the accompanying vicus or canabae are misleading, and relationships between the fort proper and settlements outside need to be reconsidered. In general, our perspectives on the community life of a Roman soldier in the 1st and 2nd centuries need to be rethought (see Goldsworthy and Haynes 1999). Fortresses, even in the 1st century A.D., might be viewed more as towns than as segregated communities. More comprehensive analyses and gendered explorations of artefact assemblages within military sites are needed, to develop more holistic approaches to them as lived communities. The body is not a thing, it is a situation: it is our grasp on the world and a sketch of our projects.

Notes
(Simone de Beauvoir 1949, 34) As a Palaeolithic archaeologist I am often surprised by the paucity of research on issues of social relations in more recent time periods. The wealth of archaeological evidence and written texts seems so plentiful from where I stand. Thus when I look into Greek and Roman archaeology for examples of research on gendered being and for a discussion of difference in the past I am reminded that archaeological evidence is the consequence of research questions as much as their starting point. We only see that which we are able to comprehend, and archaeological finds and their interpretations become visible through the questions we ask. Issues of gender in the archaeological record are a very useful reminder of the firm grip that disciplinary traditions have on our imagination. In archaeology material determinism and scientific positivism, coupled with the straw man of subjective distortion of objective reality, continue to drive our imagination at a comfortable speed that does not threaten to overturn the theoretical cart. Furthermore, firmly held disciplinary boundaries, such as those that separate history and the study of texts and representations from the practice of archaeological fieldwork, can result in a limited conversation between the social and material traces of the past. In this sense classical archaeology can appear suddenly familiar to a prehistoric archaeologist, if the refusal of history in one context effectively parallels its absence in another. As someone both used to prehistoric contexts featuring minimal physical evidence, and interested in issues of gender as understood by contemporary feminist scholars, I wonder why this should be so.
If archaeology seeks to invoke the particularity, complexity and contingency of past lives beyond general, faceless processes, then surely we would make use of any and all available evidence that might better render people in the archaeological record as individuals and groups that actually lived in their own terms in the past. So long as we limit our research imagination to analogies of our expectations and the material evidence they make visible, we will barely glimpse prehistoric lives. Surely one of our constant and continuing tasks is to refine, and complicate, our very questions.
Although writing about women was once itself an innovation in most historical disciplines, it is now both a well-established genre of research in many areas and a line of enquiry with an elaborated critical legacy. Scholars even moderately interested in theoretical issues would shirk from writing about homogeneous 'people', in Ruth Tringham's words 'faceless blobs' (Tringham 1991), recognizing that lives are experienced through sexed, raced, differentiated bodies. Feminist scholarship is one of the major forces in literature, art, religion and law and has made significant contributions to geography and cultural anthropology. In an effort to counter universal claims of biological determination of sexual inequality, feminist scholars have moved on from 'discovering women' to researching gender as a culturally constructed, changeable category that is historically moveable. The existence of women is no longer the focus of such scholarship; rather the making of women is seen as far more interesting and particular to cultures, classes and historical contexts (e.g. Butler 1993;Grosz 1994;Joyce 2000). In this shift from sex to gender, men join women as research subjects, in recognition that masculinity, just as much as femininity, is a category to be questioned rather than perfunctorily accepted. Indeed, some have suggested that it is a sign of mature, self-confident scholarship that women, the historical focus of studies politically rooted in the women's movement, were suddenly not enough (Frantzen 1993). The biology of sex concealed the relationship between a sexed body and power directly marked by gender identity. This shift allowed a move away from debates of biologically defined identity to discussions of power and inequality. Focusing on gender opened new areas of research, allowing scholars to differentiate between classes, factions and age and ethnic groups, to ask why some men have more power than others, why some women have greater access to high status than some men, and to consider how specifically gendered behaviour intersects with class, age or social context. Furthermore, gender destabilizes identity as a fixed category; even more than ethnic affiliation or race, gender provides an opportunity to examine change on a personal and social level simultaneously. Individuals are seen as gendered in a dynamic way, travelling through differently gendered landscapes depending on age, situational and cultural context, and historical circumstance. Bodies are real but also changing and changeable, dependent on the material and social settings simultaneously, linking spaces and objects with gendered existence. Gender is lived and needs to be carefully examined in its production rather than accepted as known, understood or predictable. The revelation of the dynamic nature of identity also held the additional benefit of revealing analytical categories that framed and constrained the picture we were trying to see, and consequently encouraging us to call them into question. One result of such questioning has been greater interdisciplinary communication across and between strands of historical discussion thought. Anthropology in particular brought attention to the questionable assumptions of universal human behaviour with countless ethnographic and archaeological accounts. Studies of everyday life in diverse historic and cultural contexts have highlighted the need to see spaces, objects and acts as both gendering and gendered themselves.
Sherry Ortner (1996) suggests that, to understand how gender operates in a culture, we may wish to consider the 'making' of gender, as it is not a stable category but rather a changing, working concept, yet one that does have cultural boundaries and is not entirely free-form. Explorations of how gender is made through particular practices follow numerous theoretical frameworks, but Michel de Certeau's Practice of everyday life (1984) gives one of the more compelling arguments that it is through the mundane, routinized activities that social life comes into existence. De Certeau also leaves some room for an awareness of the subject's position that may not necessarily have an effect on the structure or world that the subject occupies. This conception of agency that does not simply reduce to individual, voluntary action affecting structure would seem particularly suited for studies of gender. The focus on everyday life would seem most appealing to archaeologists, especially those with an interest in the household and in spatial research. Yet despite a number of creative and interesting archaeological projects (e.g. Gilchrist 1999; Morris 1999), we still have a long way to go to deal with gender and the everyday existence in the past in an imaginative way that would make our contributions to a discussion of human experiences worthy of the attention of other disciplines.
Prehistoric archaeologists have grappled with some of these issues for over 20 years, first finding women in the archaeological record, then gendering people in an effort to refute the universality of the biological determination of sexual differences (among numerous studies see e.g. Claasen 1992; Claasen and Joyce 1997;Gero and Conkey 1991;Gilchrist 1999;Joyce 2000;Kent 1998;Nelson and Rosen-Ayalon 2002;Schmidt and Voss 2000;Scott 1994;Seifert 1991;Wright 1996). Yet despite this larger record, some studies of the past, particularly in fields where feminist thought has not had a major impact, are still only searching for women, even if it is under the rubric of gender. While such a task remains a necessary stepping stone for drawing attention to the omission of categories of people such as women or children in general historical narratives, it constitutes only the beginning of a gendered analysis, not its full realization. From this perspective, articles that announce a study of gender but only deliver tentative identifications of women in the archaeological record are bound to disappoint readers expecting more. Such was my experience with the article under discussion.
Classical archaeology has the advantage of investigating a context containing far more written evidence than most other time periods, other than medieval or recent historical archaeology. Yet it seems that disciplinary boundaries between history and archaeology and tensions within classical studies are so firmly entrenched that they prevent the realization of what appears to outsiders like a mutually beneficial relationship (Sauer 2004). The divide in the classical world appears to resemble the prehistoric and historic divide in the North American context, where disciplinary and political histories have constrained the parameters of research. In that context Kent Lightfoot cogently argues for a need to include written documents in archaeology: The question we should be asking is not whether North American archaeologists should be using ethnohistorical and ethnographic documents, but rather how they should be employed most effectively in archaeological research. If critically read, there is a wealth of information in written documents that can be employed by archaeologists . . . (Lightfoot 1995, 204; added emphasis).
A similar call actually has been made by a few scholars in classics (e.g. Sauer 2004), who suggest that ancient history, and particularly those aspects of it that have addressed social issues including questions of gender, should constitute one of the primary sources for classical archaeologists, encouraging greater dialogue across disciplinary boundaries.
Ian Morris, exploring gender ideologies in archaic Greece in an analysis of the spatial organization of Greek houses, convincingly argues that the archaeological record has to provide the context for literary texts: 'the only way to put the literary evidence into a longer historical and a broader sociological context is by combining it with the material record' (1999,306). Morris approaches the change in gender ideology from an archaeological perspective through a detailed analysis of the use of domestic space, one that is not clearly expressed in literature, and he creatively takes advantage of the tension between the two sources. Moreover, Morris calls for a greater exploration of practices common among historical archaeologists, in his case classicists, through both theoretical debates and detailed empirical studies, relying to a much greater degree on the knowledge of ancient history and archaeology: 'Ancient historians and classical archaeologists have grown used to being consumers of theories and methods developed for other parts of the world: as the new millennium opens, we are about to become producers in our own right ' (1999, 312). Such an optimistic note urges classical archaeologists to boldly tack back and forth between written words and material evidence, in order to divorce themselves from the 19th-century methodological divisions between classical studies and material objects collected for museum display.
In the spirit of this suggestion I will propose a few avenues that might prove productive for exploring gender at Roman military bases. From the discussion and background included in the study it would appear that Roman historians have traditionally considered military bases to be uniquely male spaces. Such an assumption offers a fertile ground for a gendered analysis, once we recognize that an all-male population (whether or not historically valid as an interpretation) would not simply translate into a space of 'unbound', predetermined masculinity. Allison reminds us that this was a lived community where some members spent over 20 years. Do we know whether all the spatial arrangements of military camps were uniform and unchanging or universally 'military' in derivation? The mapping of gender in such a camp setting would surely extend beyond raising the possibility of literal female presence (however likely, given lengthy traditions of discussion camp-followers in other contexts, that might prove to be). What sorts of male relationship existed within such 'masculine' spaces, and how might they have varied between individuals and subgroups and changed over time? In a fascinating case study that should serve as an example for any archaeologists interested in material culture and gender, Carol van Driel-Murray offers an analysis of Roman footwear from military bases and nearby settlements in the provinces of north-western Europe (Van Driel-Murray 2001). While the main focus of the research is seriation and dating of the sites with the help of time-sensitive footwear, the article creatively combines recovered material everyday objects with notions of fashion, status and gender (Van Driel-Murray 2001, 194). Noting changing styles of sandals, Van Driel-Murray records also sizes in relation to changing styles, and offers an interpretation of different women's and men's fashion, as well as of children's 'masculine styles', small sandals with missing toe indents, suggesting that boys' gender identity was shaping already at the age of four or five (Van Driel-Murray 2001, 194). Thus the analysis suggests that even a modest, everyday object could allow for the expression of personal choice within a delimited range of possibilities, effectively engendering the person while indicating quite clearly that ancient Rome might not map easily onto our modern dualist categories of male and female.
Similar strategies of engendering artefacts also featured in earlier studies when identifying women was the main goal. Allison follows this path but adopts, in my view, an overly cautious approach that ultimately leaves the basic assumptions framing her research unexamined. Noting that many artefacts cannot be ascribed to specific activities, she reminds us that our interpretations related to gender issues remain circumscribed by the larger complexity of social life: 'Attributing artefacts other than dress to particular gendered activities is even more complex' (p. 5). Differences in ancient Rome ran along multiple lines, most importantly status -free citizen, freed person, slave -that refracted sexual divisions in numerous ways. While her point is certainly well worth taking in the sense that gender is not simply a longignored label, neatly attached to static artefacts, I would argue that for precisely this reason we need both a wider frame of comparative reference and an ever-evolving set of categories to frame our research questions. While the range of social possibilities may inevitably exceed the certainty of our knowledge -even in contexts rich with historical evidence -it is not thereby simply infinite or unthinkable; when carefully foregrounded and interrogated, different social attributes can be fruitfully explored. Historical evidence suggests that the concept of gender in the Roman world was not one of a simple binary opposition of male to female. That fact, however, does not reduce us to wild guesses. Rather, by exploring gender relations carefully and extensively, disclosing assumptions along the way, we might recognize more general problems of interpretation and analogy applicable across all archaeological endeavours. The notion that gender is inherently more subjective than other lenses through which we view the past has been addressed numerous times (e.g. Gilchrist 1999;Wylie 2000), yet it remains a widely held belief among many archaeological practitioners, particularly those who cling so tightly to their artefacts that they do not recognize the shaky philosophical ground beneath them. Unfortunately, studies that attempt to remedy the absence of women in archaeological narratives while leaving larger gender assumptions unchallenged inadvertently reinforce this perception. The study presented here does usefully refute the notion that Roman military camps were populated by men only. However, when placing women into a now shared space the author too quickly and easily interprets the archaeological pattern as evidence of wives and families. This may indeed be a plausible interpretation of the evidence. However, it fails to acknowledge other possibilities, or explain why the simple presence of a woman would automatically equate with the social role of a wife. As Beryl Rawson has suggested, a concubinage, or cohabitation, appeared to have been a choice for a segment of the Roman population, including the military, thus broadening our understanding of partnerships, and the concept of family (Rawson 1974;. In addition, this was a society in which slavery, prostitution and an extended range of sexual practices were far from unknown. When disrupting the traditional view of military camps as genderless (in the sense of unmarked) male space, it would be fruitful to remind readers and fellow scholars of the wider range of ways in which different denizens of the Roman world could be both 'female' and 'male', whether or not all of those possibilities seem equally relevant to the particular context. I will add a final comment about gender amid our general disciplinary problem of method. Archaeologists are always and understandably eager to amplify the amount of information we can squeeze from material artefacts through technical means. Recently GIS has received a great deal of attention among archaeologists by opening a promising new horizon of techniques. As a number of feminist geographers have noted, however, we may wish to consider whether this is the best tool for all the questions we may wish to ask (Kwan 2002a;2002b). Mapping gender is not an easy task, requiring not only a broader range of theoretical tools but also a methodological approach that is open to greater scrutiny and questioning. Yet gender, just as any aspect of past societies, is within our interpretative reach if we search less for certainty than for multiple plausible scenarios, bracketed by an explicit description of assumptions along with methodological and interpretative steps. Roman military forts in the provinces would appear to be fertile sites in which to investigate issues of gender in ancient Rome, suggesting that the margin may be far more interesting. First and foremost, however, we need to approach the topic with a willingness to cross disciplinary boundaries and to expand the scope of interpretation through which we seek to recognize and analyse material remains.

'Safe' genders? Eleanor Conlin Casella
As historical archaeology has expanded from its origins in the 'settler nations' of the post-colonial world, an increasing number of European scholars have argued for a radical reorientation of the subfield. But if historical archaeology is to be accepted as the 'archaeology of literate societies', the profound discussion interdisciplinarity of the subject must be embraced. While the presence of written documents can indeed be a luxury, providing for richer or more personalized interpretations of the past, it can also be a curse, demanding careful attention to the interplay between material and textual sources of evidence. Allison's example of 'gendering' the Roman military landscape provides a crucial first step towards a more socially oriented archaeology of the classical world. It also demonstrates both the opportunities and limitations faced by scholars of the literate past.
Citing Henrietta Moore's distinction between 'doing' and 'being' gender, Allison observes that identity associations of such activities are constituted in historically specific ways. Thus not only to link artefact patterns to certain activities, but also further to associate those activities with specific identity categories (such as age, status, sex or indeed gender), require an explicit presentation of the reasoning and justifications that forge those crucial links. Drawing from the available corpus of both primary textual sources and secondary studies of the Roman frontier provinces, how did the specific activities mentioned in this study become associated with specific identity categories?
If we do accept a range of 'safe' (or perhaps 'normative') categories of activity, such as 'spinning for women and combat for men', can textual sources be integrated to support or challenge these interpretative assertions? What about the wider range of activities incorporated into this study? Why were gaming, metalworking, stoneworking and weaving indicators of male gender? And is weaving not a component of 'cloth working', an activity previously identified as 'safely' feminine? Could this productive activity actually be interpreted as a material relationship between the genders, the finished cloth ultimately representing a combination of both masculine and feminine crafts?
Particularly given the compelling results of Allison's spatial distribution models, it might be interesting to consider whether moments of dissonance occur between her artefact patterns and contemporary written accounts of 'doing' gender on the Roman military frontier. Studies of predominantly male-occupied settlements from the 18th to early 20th centuries -such as mining camps, labour gangs, whaling stations, pastoral properties and colonial military encampments -have frequently considered the process by which particular subgroups of men assumed responsibility for essential, yet feminine-gendered, tasks. The provision of sustenance may have been 'performed by women' at the Roman-era sites of Vetera I and Vindonissa, but military establishments since the 18th century typically had teams of young male recruits assigned to staff the kitchens and mess halls. Studies of mining, railroad construction and pastoral camps in 19th-century Australasia and the American West have combined documentary and artefactual evidence to demonstrate that male Chinese immigrants frequently served as the cooks and market gardeners for these frontier settlements. In the case of monastic settlements, older men often assumed responsibility for the caregiving activities typically associated with women. Do written accounts justify Allison's exclusion of 'ethnic soldiers' or even 'cross-dressers' from her analytical model? Could other non-normative actors have assumed responsibility for the feminine-inscribed activities? Perhaps it is the moments of dissonance that actually expose the 'doing' of gender. If historic accounts claim that no women were present, yet material evidence demonstrates that female-gendered activities occurred, then obviously someone at the site was 'doing' femininity, regardless of their biological sex.
Perhaps gender identity held a limited relevance within these frontier communities. In mentioning the range of professional and commercial occupations women maintained within Roman civilian society, Allison herself observes that activities 'were more likely to be distinguished along status lines' (p. 5) than by either gendered or sexual aspects of identity. Closer interrogation of documentary sources might reveal those specific social encounters during which gender categories emerge as the primary node of social belonging. Additional concepts of 'status' or 'age' could then be mapped against gender to illuminate the plural and overlapping subjectivities that structured these social worlds.
Finally, are children 'doing' gender? Or are they 'doing' age? Presumably the 'child-related material' represents a constellation of activities different from those of adults. Do we therefore assume that they had no gender, or 'did' a somehow different form of gender? Perhaps their very presence was itself gendered, simply because childcare is typically a feminine-associated activity. Would textual accounts help illuminate the historically specific process by which Roman children learned to 'do' gender through their material world?
Ultimately, Allison's study suggests some tantalizing new directions for Roman-era archaeology. By explicitly adopting a social approach she has recast Roman military sites as living communities -as places shaped by the materiality of commercial relations and domestic encounters as much as by imperial warfare. In searching for women and children within the material remains of these forts, Allison offers some intriguing possibilities for understanding the dynamics of family life within these frontier settlements. The 'romanization' of gender archaeology Marie Louise Stig Sørensen Allison's paper shows how the complex world of real-life social relations, dependencies and needs may be extracted from the small finds and the mundane domestic aspects of life associated with a site -even when that site is a Roman fort. In her approach Allison does not presume that society in its totality is mirrored in these finds; rather her apparently modest, but in fact potentially significant, point is that concrete aspects of how people organized their relationships, and especially the spatial aspects of these relationships, are revealed by such finds. In this emphasis Allison makes important contributions to two core areas of concern: one is the character discussion of domestic life within Roman forts during the 1st and 2nd centuries A.D., and the other, more theoretical concern, is how we make interpretative links between objects and social roles and identities. These two areas of concern are not, however, given equal standing in her arguments. The reflection upon the interpretative project of engendering is largely nested within the discussion of how one may find evidence of women and children within the Roman fort. There appears to be little acknowledgement of how some aspects of this concern need to be discussed at a general level separate from the specific questions about gender as part of social relations within this particular type of settlement. This omission is probably due to the limited space, but it brings up important points about how we reach interpretation.
That the paper primarily sees itself as contributing to the first area, the discussion of women's presence and roles within Roman forts, is indicated by the detailed manner in which the need for further research within this area is argued. The paper exposes how the view of the military fort as a male-only environment has been based on a mixture of 19th-and 20th-century ideals of the military man coupled with an apparent lack of material unambiguously related to women. The paper shows us how this skewed interpretation can be rectified, and using an engendering approach to the small finds it gives us a glimpse into the greater social complexity of life in the forts. This is an important accomplishment. Furthermore, by demonstrating that women and children are present in the deposits found at various places within the forts, the paper provokes further questions about the relationships between the different groups within the sites. In turn this calls for scrutiny of other aspects of current interpretations of life in the Roman forts. We also get an enticing sense of the temporal dimension to this involvement, and we are thus given hints about how gender relations are interwoven with other structures and relationships, although Allison does not begin to explain why this is the case and some of the historicity of gender gets a gloss-over.
The approach Allison uses to reach her conclusions may well come to provide a model for the use of small finds in more detailed social interpretation, and it is therefore of interest to consider the structure of the argument in greater detail in order to identify both its strengths and its more tricky aspects. The essential structure of the argument has been used effectively in other gender studies. The engendering of small finds and their spatial distribution has in particular been used as a means of engendering the archaeology of contact, to insert gender dynamics into production sites, or to engender historic urban landscapes (e.g. Devonshire and Wood 1996;Trocolli 1992;Wall 1994). This approach developed in order to engender contexts in which gender cannot be established through direct reference to human bodies, and where therefore a number of inferences have to be made to insert gender as a significant parameter. This approach was explored early on by Conkey, who coined the term 'context of action' in reference to the need to place objects within their context of use so that we can trace the interconnection between different objects and the social relationships and identities invoked through their use (Conkey 1991). Such analysis begins by making a number of assertions about association between artefacts and activities of certain types and it then argues for links between specific groups of people and these artefacts; in effect this means that artefacts of different types are associated with gendered activities. On this basis, identifying the distribution and clustering of these objects will provide us with the means of linking gender to the places in which such objects are found. Allison uses the example of pierced discs. Discs of a certain size were probably spindle whorls, these artefacts are associated with spinning, and women usually carried out this activity. Therefore, while these types are used as markers for women, this association, rather than assigning gender to artefacts, is in principle based on the assumption that certain activities are gendered. When engendering takes such a form it becomes an interpretative project rather than a fixed result as in itself it will encourage further investigation of the relationships exposed, particularly exploration of how and in what ways the various activities affected people's lives (Sørensen 2000, 39f). In the case studies discussed by Allison the questions, as acknowledged in her conclusion, will change from being concerned with demonstrating the presence of women in the forts to the task of characterizing their roles within it. The distinction between ascribing a gender to each artefact and ascribing gender to an activity is, however, potentially subtle and there is an obvious risk of gender-stereotypical assumptions (such as spinning being a female activity) affecting these associations. Allison's paper at times seems to put this distinction at risk. Thus it can appear as if her argument, that the spatial distribution of some of the objects shows the presence of women, is pre-given, due to the direct link she makes between activities and gender. This problem may partly be caused by her discussion not making a distinction between spatial patterns based on objects associated with gendered dress (such as beads) and those indicating gendered activities (such as weaving). This means that the interpretation of the presence of women is based on inferences of two different types. The ambiguity of whether (and through what argument) the data directly suggest that women were present or whether it is female activities that are documented would have been avoided if the discussion of the spatial organization of different practices had been disentangled from the gendering of these practices.
Pursuing gender at greater depth than that found in presence-absence discourses is, however, clearly within Allison's agenda, and in these terms the paper is an exemplary case study while nonetheless still somewhat limited in its reflection upon gender. This, I believe, is because the paper tends to frame the second concern, the question of the links between objects and gender, solely within the debate about Roman society and Roman archaeology. As a result general issues that are about epistemology, and more specifically about archaeology's classic concern with solving how we understand the social through material objects, are discussed as if they were unique to Roman archaeology. They are not. Gender archaeology from its very beginning has been debating this question and various solutions and resolutions have been formulated regarding whether the person genders the object or vice versa. It is neither more difficult nor easier to sex Roman or Neolithic or Byzantine small finds. It is in general difficult to 'sex small finds', and that point should not be lost. Moreover, it is not very clear what 'sexing small finds' actually means and whether it is necessarily always the best method for the engendering of discussion a particular kind of space. Lacking other evidence, we do, however, often resort to using objects in this manner as a means of reaching engendered (as well as other) interpretations of the past -and the results may indeed tell us about gendered activities and segregation as demonstrated by the case studies here. It is, however, important to reflect upon whether the coherence of the interpretation in and of itself validates the gender assertions that it was initially based upon. The format of Allison's paper does not force us to return to the initial assertions, although she rightly points out that when different types, all with the same gender ascription, show the same spatial distribution then this provides some validation for the gender ascription. In this paper the validation of the assertions is not, however, significant per se, rather the overshadowing aim is the interpretation of women's presence in the Roman forts. Our confidence in the assertions about associations between gender and activities, and in particular in the reflexivity with which we reach them, is, however, important for the further development of engendering approaches within archaeology and especially for our continued concern with the difference between sexing people and sexing things. The tendency to root the epistemological issues solely within Roman archaeology limits the curiosity and degree of reflection that the paper brings to the question of how we investigate gender. The documentation of women's presence in the forts is the prime aim of this paper, while the means of reaching the interpretation, although accounted for and argued, are less central. At times this is clearly the right strategy in terms of the needs of the discipline, but the fallout resulting from the way in which the argument is structured needs to be recognized and revisited or our engendering of the past will become cluttered with a new range of stereotypes and assumptions. This paper is therefore primarily about the Roman forts, and only secondarily about gender. This prioritizing can be seen in several of the comments about gender, as when it is stated that 'Roman gender studies acknowledge that the basic categories of man and woman are problematic' (p. 6). This realization is not unique to Roman archaeology, nor has it been formulated in isolation from the debate carried on within gender studies; her discussion of gender needs to be simultaneously aware of gender as an analytical concept, as referring to a particular set of social practices and as part of the performance of historically constituted views about difference within a community. It must have been with this tension in mind that Allison introduced the distinction between being and doing gender (as discussed in Moore 1999, 154). This distinction can help to separate more clearly the generalized discussion of the nature of gender, which is not unique to Roman archaeology, from the investigation of gender as constituted by the demands of life in the Roman forts. It is at the latter stage of analysis that gender first becomes about doing.
Ironically, gender archaeology from its beginning has found it the hardest to develop succinct methods for the engendering of the domestic context, despite this being the contexts which women and children were routinely associated with. For prehistory the engendering of the domestic sphere has remained a substantial challenge, while more progress has been seen in the archaeology of later periods. Allison's sustained interest in this area is therefore of great importance as her work may provide methodological guidelines for how contexts that lack bodies can be engendered. It is, however, for the same reasons that it is so interesting to scrutinize the structure of her arguments and the inferences she makes.
Archaeological Dialogues 13 (1)  Engendering change in our understanding of the structure of Roman military communities Simon James One of the most important developments in the study of the northern frontier provinces in recent years has been growing realization of the complexity of the populations living within, as well as around, imperial Roman military bases. It is clear that, even within the walls, such bases were not the exclusive preserve of soldiers, or even males. (It has been especially delicious to see the outraged reaction of some conservative practitioners of Limesforschung to the idea that there could possibly have been women living inside Roman forts.) Women associated with the soldiers were by no means confined to the stereotypical prostitutes lurking outside the gates. The Vindolanda tablets, supported by female-sized, feminine-styled and child-sized footwear from the praetorium (Van Driel-Murray 1995, 8), make it undeniable that, by A.D. 100, commanders' wives were present at least in some forts some of the time. Centurions, too, apparently cohabited with their wives inside the walls. However, Van Driel-Murray's startling conclusion that women and children were also present in some numbers in Vindolanda's barracks -and, by implication, that families even of ordinary soldiers may have been common in garrisons, despite an official ban on marriage in service -remains persuasive, but controversial (Van Driel-Murray 1995, 8-20; see below). Women, then, are certainly attested in bases, but which women, where they were, and how routinely they were present, remain unclear. Allison's work on the issue is therefore welcome, especially using data from outside Britain.
Allison's main focus here is to try to identify the presence in intramural spaces of women, through mapping deposition of artefacts interpreted as gendered, or (it seems) through the presence of small children as a proxy indicator (e.g. intramural burial of neonates at Ellingen making intramural residence of women highly likely).
In my view this paper raises, but does not fully develop, two distinct issues. The first is the place(s) of women as part of each Roman 'military community' during the principate. By this term I mean a standing formation of soldiersusually a whole named unit (legio, cohors, ala) -combined with individuals of various other attested categories who in diverse senses, including literally, belonged to it. These latter groups are often referred to simply as 'civilians'. discussion However, some, such as regimental servants, were entirely part of the military, receiving training and rations, and subject to military discipline (Roth 1999, 91-116). Others, including soldiers' families, directly relied on the military economically, and will have taken much of their identity from the association. All these groups are best thought of as 'non-soldier' components of the military community. If there was any well-defined boundary between the civilian and military worlds, it was where each such extended community encountered the surrounding local population.
The second issue is about how early imperial military communities were spatially deployed. The conventional answer has always been that, insofar as non-soldiers were present at all, they lived outside the walls of 'forts and fortresses' (better termed military bases, since their conception was essentially offensive, not defensive; they were thought of as places to operate from, not to skulk within). Space precludes much comment on this aspect, beyond underlining Allison's conclusion that some women and children did reside inside the gates -and to add that it is likely that some soldiers routinely worked and lived outside, wherever extramural settlements developed.
With regard to the place of women within military communities, it seems to me that the interpretations offered by Allison -some tradeswomen, and especially the emphasis on soldiers' wives per se -could be considerably refined. Even in the current state of knowledge, it is possible to go further in drawing informative inferences from textual sources against which to interpret archaeological data such as the potential 'gendering' of artefacts and their spatial distributions. Not least, there is more to be said on demographics using data from the military epigraphic record. We may combine Phang's quoted evidence that ordinary soldiers tended to marry relatively latetypically in their mid-30s, i.e. well into the second half of the period of service -with Scheidel's work on the demography of Roman military units. The latter underlines how many soldiers died, or otherwise left military service (due to incapacity, desertion etc.), during their 25-year engagement; it seems that rather less than 50% of recruits would become veterans (Scheidel 1996, 117-29). Based on Scheidel's figures, I calculate that only around 30% of serving soldiers will have been 35 or over (Table 1). The implication of these studies is that only a minority of serving soldiers were married at any given moment. Even if most eventually did marry, some surely failed to, or chose not to, while serving. Others will have lost their wives early, e.g. in childbirth; still others will have divorced while still serving (Welles, Fink and Gilliam 1959: P. Dura 32). All this suggests that, even after the marriage ban was lifted, the proportion of married serving soldiers in most units at any one time was probably of the order of 20%. So, while we may accept that soldiers' wives were a normal component of military communities, we need to be careful not to exaggerate their numbers or potential social significance.
On the other hand, Allison's focus on wives tends to obscure the possible significance of other categories of women which might serve to increase the numbers of females present within military communities. For example, some soldiers probably owned female slaves, who were not wives (Varon 1994). Further, Saller's (1987) demographic projections suggest it was common for serving soldiers to find themselves heads of families, financially responsible for Table 1 Idealized age profile of serving soldiers in an early imperial legion if it was maintained at a strength of 4,800 men, based on analysis of epigraphic data and interpretation offered by Scheidel 1996, 117-29, and Table 3.14). At any given moment, only about 30% of serving soldiers will have been aged 35 or over -and therefore likely to be married according to Phang's work (2001) Nevertheless, even accepting the routine presence of significant numbers of women, by the standards of the Roman era military communities remained exceptionally male-dominated, both in composition and in purpose. It is likely that other groups of non-soldiers, notably various categories of male servants discussion and (in modern terms) support personnel, were at least as numerous as the females in military communities, and some perhaps had higher standing. It seems reasonable to suggest that gender was not the most important dimension or axis for internal differentiation of identity or function. Within a community overwhelmingly masculine in terms of both gender balance and distribution of power, axes such as legal status (free/freed/servile) and age may have been far more significant than gender.
In military communities, then, I suggest that while routinely present, women were normally relatively few in number. In such circumstances we might anticipate that even gender distinctions common elsewhere in the Roman world might be 'bent' or transgressed, e.g. by assignment according to other axes of identity. For example, was spinning, a task Allison plausibly argues is strongly associated with females elsewhere in the Roman world, really only ever conducted by girls and women in Roman military communities too? This is obviously crucial to her choice of spindle-whorls as indicators for female presence. But if there were not enough resident females to meet community requirements for yarn, were arrangements always made to acquire it from women elsewhere? Or was this task sometimes assigned to low-status males -e.g. young slaves? 'Gender-bending' of textile-related tasks is attested in more recent British military and naval contexts, as is reflected in the name of servicemen's sewing kits for clothing maintenance: 'housewives'.
A central inspiration for the current interest in women and children in military contexts was Van Driel-Murray's study of the footwear from Vindolanda mentioned above, especially the early 2nd-century material from the Period IV barracks (Van Driel-Murray 1995, 9-20). In some of the rooms a large proportion of the shoes were of sizes too small for adult males. Some clearly attested small children, others were of intermediate sizes. Van Driel-Murray interpreted these in terms of soldiers cohabiting with de facto wives and children, but was also quite clear that the data actually gave no direct indication of the biological sex of those individuals too small plausibly to be adult males (neither the styles represented nor wear-patterns can currently be 'sexed' or 'gendered'). She notes the possibility that the footwear evidence might be explicable solely (pardon my pun) in terms of males.
Van Driel-Murray was candid that it was her preference to interpret the Vindolanda footwear data in terms of women and children, rather than just boys and youths; part of this choice she attributed to her familiarity with Dutch colonial experience, where 'informal' families in barracks were common, but part, she also acknowledged, derived from distaste for the perceived implications of the all-male alternative, which, she felt, made the Vindolanda barrack rooms seem 'more like a male brothel than anything else. I am culturally averse to the idea of children abused in the barracks at Vindolanda, but', she continued, 'given a different society, and especially a slave society, it might not have been seen that way' (Van Driel-Murray 1995, 19-20). This admirable self-awareness makes it easier to consider the implications of her arguments, and of the alternative which she does not prefer.
If we try to look beyond what is, to us, the distressing vision of barracks swarming with young catamites, we can consider the possibility that military communities actually did incorporate large numbers of what we would regard as pre-adult males, primarily employed for purposes other than sexual gratification, for a case can be made that the 'non-soldier' groups of personal and regimental servants mentioned above may have been substantially composed of youths or older boys.
There is little doubt that many such servants, of whatever age, were slaves. Many of these are likely to have been young: when Clodius Super, centurion of cohors viii Batavorum, wrote of his (?six) pueri, he may have been using the word in its attested infantilizing sense of 'slaves', but they could also literally have been boys (Bowman and Thomas 1994, Tab. Vindol. II, 255). However, there is also this question: what did sons of soldiersthought usually to have been resident at their fathers' bases -do after boyhood during the decade or so before they were old enough to enlist (unless they left for another career)? Such youths in their mid-teens were probably regarded as young adults, but apparently could not become soldiers until they were 18 or 20, perhaps because only then were they fully developed in skeleton and muscle, and able to withstand the full rigours of military training. I suggest that many teenage soldiers-to-be, camp-born or from other preferred backgrounds for recruitment, worked in a free servant capacity for the regiment, earning their livings and learning the rudiments of soldiering, such as equipment maintenance and, in cavalry units, care of horses and how to ride, i.e. in roles roughly analogous to that of the medieval knight's page. Demographically, then, military communities may not only have been unusually male-biased, but also exceptional in age profile with, even by Roman standards, an abnormally high proportion of males in their teens and 20s.
The purpose of these arguments is not to seek to drive women back out of the camp gates, so to speak. It seems to me that Allison's present work, and Van Driel-Murray's before her, have clearly been successful in establishing the residency of numbers of women within the walls. (That said, paradoxically, our best material evidence for the presence of women in and around Roman military bases may not come through attempts to identify feminine-gendered artefacts. It may be quite indirect: the presence of small children, already attested through infant burials and children's footwear.) More generally, they have been successful in making us think harder about the nature and workings of Roman military communities. Their work is helping revivify a stale and introverted subject -the study of Roman 'forts'.
However, I argue that, while women were usually a constituent of Roman military communities, even against the wider background of a distinctly 'phallocratic' world these entities were exceptional, 'supermasculine' both demographically and functionally, and the implications of this, too, require further consideration. We need now to develop more nuanced approaches to the internal structure and operation of military communities, which incorporate gender as one, but only one, of a number of significant axes of identity and functional differentiation at work within them. In this way we will best be able to engender -in the primary meaning of the word, to bring into being -change in our understandings of the military as a component discussion of Roman society, and not just as a 'war machine' comprised solely of male soldier-cogs. Women in Roman forts -lack of knowledge or a social claim?

Thomas Becker
The classic idea of the Roman army, especially of the legions, is that of a man's world, where discipline and military drill dominate, and where there is no room for women, whatever their social status or function. This idea has been fostered by the picture painted by the antique authors, in which fighting by women is reserved to goddesses (Athena/Minerva) and exceptional personages. The normal female is described as a mother or wife, whose chief occupations were confined to the organization of the household, the up-bringing of the children, spinning and weaving (Marquardt 1975, 58). This role model fits in excellently with the social structure of 19th-century Europe, where women were also absent from military camps. This, in turn, can be traced back to the Prussian view of military virtues, which would be diminished by the presence of women. Many concepts of Roman military archaeology have their origin in this period. In many ways these traditions still influence our views on Roman life, as analyses of the roles of women and children in archaeological illustrations have shown (Röder 2002;Becker and Hölschen in press). German archaeological research, especially, concentrates on questions of building-structures, military units or dating, whilst social aspects of life in the camps or on the frontier are normally neglected.
In recent research an attempt has been made to overcome this influence. Thus the picture of the Upper German-Raetian Limes has in recent years changed from that of an invincible borderline into one of a traffic-controlling and toll-collecting system. A similar change can also be observed in the idea of the 'female-free' fort, resulting from the work of Allason-Jones (1999), Van Driel-Murray (1994; and Speidel (1998). This is also true of the work presented here by Penelope Allison. The change in ideas is certainly justified; however, a qualified discussion of the statements presented is necessary.
The question central to this analysis is: which types of small find indicate the gender or age of the owner? Specific parts of female dress, as designated by Allison as possible proof of gender, are relatively few. Foremost of these are the fibulae, which have only a decorative function, as there is little space beneath the bow for a large amount of cloth. They are associated with female dress (Gechter 2003, 207). Such fibulae give only a hint as to the normal user. Finds from forts may be explained by exceptional circumstances, such as a soldier wearing such an ornament as a keepsake from a woman. Likewise hairpins may have other functions, for example as a stilus (Becker and Schallmayer 1996, 145).
In a civil context, small finds connected with certain occupations can be more easily gender-linked. Distaffs found in late Roman women's graves provide an impressive example of a role model as described by the antique authors and feature in P. Allison's analysis of the making of clothes. A similar evaluation is given by Marquardt (1975, 58). Beginning with the premise that military units and life in the forts was strictly male, the question is whether 'female small finds' are really an indication of the presence of women. Is it not more likely that men were producing a small number of objects for their own use? Unfortunately it is not possible to identify further small finds from fortsother than the fibulae and spinning whorls -which P. Allison associates with the presence of women in forts. The publication of a find-list is necessary. The reference to an Internet journal at this point is not helpful .
For this kind of analysis it is necessary to focus on small finds which have definite gender associations and which exclude any exceptional usage. These are objects which have been made in different sizes, because of the difference in size and form between the female (slender) and male (robust) physique. Such objects are jewellery worn close to the body, like finger-, arm-or neckrings, or items of dress, like shoes. As yet there has been no analysis of the sizes of finger-rings. Finds of finger-rings are relatively rare and often they cannot be distinguished from other types of ring. In her analysis of shoe assemblages C. van Driel-Murray has demonstrated the presence of women in forts (1994; 1997). However, in some cases a closer look at the find-context is necessary. The shoes from the eastern fort at Welzheim were excavated from three wells within the fort (Van Driel-Murray 1994, 347) which seem to have been backfilled at the end of the camp's use. The fill of the wells was probably taken from the civilian settlement (Becker and Hölschen in press).
The concentration of gendered finds in the quarters of high-ranking officers inside the legionary fortress of Vetera I seems to point to the presence of women in this area. This can be simply explained by the career of the staff officers (tribuni laticlavii; tribuni angusticlavii). As part of their senatorial or equestrian career (cursus honorum) they served with the legions. The earliest age for marriage being 14, they were mostly married at the ages of 17 to 20 (Becker 2002, 161). Entering military service, they took their family and household into the provinces with them. Such an example is Germanicus, whose wife Agrippina and son Gaius joined him during the campaign to Germania Magna in A.D. 13-17. He went together with his friends and advisers, who also brought their families to the Rhine (Speidel 1998, 54). It is also possible that some of the higher-ranking centurions housed their wives within barracks. This could be indicated by the often luxurious accommodation at the top end of the barracks. New finds from the legionary fortress of Windisch/Vindonissa substantiate this (Trumm 2004, 49). Three neonatal skeletons were found within the quarters of a centurion, indicating births and the presence of women within this barracks. Unfortunately this is the only burial of its sort to be observed in a legionary fortress. It can possibly be due to a gap in our knowledge. The analysis of animal bones from the legionary camps might produce some surprises. Simply an anthropological determination of the few human bones from the camp of Vetera might have produced evidence for the presence of women (Hanel 1995, 285).
GIS-based mapping of small finds always omits the third dimension. The Vetera fort shows an interesting distribution pattern of finds associated with discussion women in the officers' quarters and along the viae principales. However, it is not clear whether these finds were actually lost and deposited at this particular spot, e.g. in a rubbish pit, or whether they were brought from other parts of the fort, e.g. in earth for levelling an area. For legal reasons, it seems improbable that earth would be taken into the fort from the civilian canabae. A transfer from within the fort is easier to imagine. In cases of known find-contexts, an analysis of the features could verify the distribution pattern.
A special situation pertains at the fort of Ellingen. The ground plan of the fort indicates that a different type of unit was garrisoned there. It was not the pedites singulares as mentioned in a building inscription found there, marking the renovation of the fort in A.D. 182. This unit, due to their higher rank and normal location at the governor's headquarters, cannot have formed the garrison (Becker 2004, 62). The ground plan indicates the presence of a non-independent unit (numerus), which is usually assumed to be the case in forts of this size. The presence of the neonatal burials should be viewed in this context. A translocation of the neonatal skeletons with earth from the vicus can definitely be excluded, because of their completeness (Zanier 1992, 70-72). A translocation might only be possible in the case of single long bones, which are robust enough to withstand it. The burials were discovered in pits and wells, and not in levelling layers. Thus a direct connection can be drawn between the burials and women living inside the fort. The analysis of the datable small finds indicates that the neonates had been born not only in the latest phase of the fort, but also in earlier ones (Becker and Hölschen in press). Only the cooperation between archaeologists, archaeozoologists and anthropologists produced results in the case of Ellingen. The neonates were recognized during the examination of the animal bone and subsequently analysed by an anthropologist. Unfortunately this type of cooperation is unusual. In archaeozoological reports of finds from Roman forts human bones are only described generally, or not even mentioned (Becker and Hölschen in press). This also pertains to Oberstimm, where no human bones are mentioned (Boessneck 1978). The analysis concentrates on bones from datable contexts, so it is possible that information on neonate skeletons from non-datable contexts has been lost.
In summary, the rethinking of the presence of women in Roman forts has definite justification. However, a careful and well-grounded assessment of the methods used has to take place in order to ensure that statements are sound, and that finds have not been misinterpreted. Just as the militarism and imperialism of the 19th century influenced our ideas on the Roman army, so too can our modern social outlook colour our views.

Response. Historical complexity Penelope M. Allison
The concerns of these five commentators are too diverse to be dealt with comprehensively in this short response, but I emphasize that this paper introduces a project which aims to 'redress the balance' with regard to Roman military bases as domestic spaces and whose theoretical and methodological underpinnings are wide-ranging and complex. I will briefly address a few basic points.
The need for a critically 'interdisciplinary' approach to textual and material evidence is an ongoing concern in the archaeologies of the Greek and Roman worlds and is by no means straightforward (see Rotroff 2005). The agenda for many of these archaeologies was set centuries ago, with archaeology initially considered the 'handmaiden' of history (Finley 1985, 7, 18-26). Many simplistic assumptions and unsophisticated approaches to relationships between the different types of evidence have become the baggage of Greek and Roman scholars. We must first unpack the biases and misinterpretations of relationships between textual and material-cultural remains (which have been woven into extant interpretations of material-cultural remains) and critically assess how and in what contexts remains of different types are useful for comprehending meaning in other remains (Allison 2001;Foxhall 2004;Hoffmann 2004). The disappointment experienced by many prehistorians and post-medieval archaeologists reading the slow and laboured attempts by Roman archaeology to investigate social behaviour often stems from a lack of exposure to the complex and intricate histories of this subdiscipline and from a lack of comprehension of the huge amount of material that has been continuously investigated and interpreted since antiquity itself (see Trigger 1988, esp. 27-72). Scholars who fuse textual and material-cultural evidence by analysing the interpretations of previous scholars, without critical approaches to the formation of these interpretations and to the sociocultural biases with which they are imbued, are only repackaging past misreadings of the original data, be they textual, material or both (e.g. Laurence 1994).
Documentary evidence pertinent to Roman archaeological remains is wideranging in type, content and chronological and regional specificity, but is sparse compared with that for the post-medieval archaeology. For example, while we can learn much about families in Rome from the ancient authors, be they a biased source, and while some ancient authors did write about the army in the west (esp. Tacitus; see Hoffmann 2004), references to the place of women in the latter arena, as can be seen in the introduction to my paper, are limited and derogatory, and showed little concern for women below the status of commanding officers' wives. Studies like Phang's show that epigraphical evidence provides more information for investigating life in the provinces and the place of women in the military sphere. Her detailed analyses of specific classes of written evidence have moved the debate forward from discussing prostitutes and concubines running brothels and drinking houses in settlements outside the camp to arguing that the term 'wife' (Phang's term, not mine) might be appropriate for many of the women who accompanied ordinary soldiers while on active service. It is now recognized that concubinage was a widespread marital-type relationship for those ineligible for formal Roman marriage. It is thus appropriate to include them in discussions of 'wives' and families. The Roman concept of familia can include all women and children under the power of the paterfamilias, discussion and embraces widowed mothers, slaves (male and female) and natural and adopted children (Gardner and Wiedemann 1991, 3-4). Because Phang was unable to find written evidence that such 'wives', or any other women except those in officers' families, actually inhabited military bases, she assumed that they inhabited external settlements in the 1st and 2nd centuries.
Socio-spatial information is lacking in the written sources but it can be gathered through investigations of the archaeological evidence, particularly through the distribution patterns of artefacts that have indeed been interrogated through the textual and other material-cultural evidence to imbue them with potential social values concerning task and gender. In archaeology generally it has long been acknowledged that GIS and spatial analyses are, to date, the best tools for exploring the 'fuzziness' that is the archaeological record (Kirkinen 1999, 255). Such 'fuzziness' is characteristic of investigating gender through material culture, with its 'multiple plausible scenarios '. Contra Tomášková, Kwan (2002a, esp. 272-73) advocated the usefulness of GIS for both 'quantitative/empiricist' and 'critical/qualitative' analyses and has called for 'more diverse and nuanced reading of complex relationships' through GIS. 1 The lack of 'explicit presentation of the reasoning' (Casella, p. 26) behind the artefact 'engendering' processes that are fundamental to the spatial mapping of artefacts is a consequence of the lack of space in a paper such as this one. Discussion of these processes, as presented in the original conference paper, has been put to one side for more theoretical and interpretative discussion here. The procedures have required critical analyses of previous textual and material-cultural approaches to the uses of particular artefact types, to the people involved in these activities and to the clothes that men, women and children wore in the western provinces in the 1st and 2nd centuries A.D. I, and other scholars, such as Allason-Jones, Böhme, Dixon, Van Driel-Murray, Martin-Kilcher and Treggiari, have long been investigating such questions and Furger has explored the question of finger-ring size (see Allison et al. 2005, section 8.2.2c). For example, the studies of these scholars, integrating textual and materialcultural evidence, have shown that, as discussed in my paper, weaving is carried out by men and women, that usually only women were spinners in the Roman world, but that soldiers mended their own clothes. Such studies also highlight that perceptions of military forts as all-male, 'supermasculine', communities have led to presumptions that certain brooch types within forts must have belonged to ethnic soldiers, and necklaces to cross-dressers, because women were assumed to have not entered this context. Examples of my own processes can be found in my listed previous and forthcoming publications. Full discussion of all the relevant and current debates will be published in the final presentation of this project (Allison n.d.d). And more is still required, through the development of this approach to Roman artefacts, to investigate less 'safely' gendered activities like pottery-making, cooking, mending or indeed leatherwork, that may well have been carried out by male non-soldiers in this context. Indeed, far from 'clinging tightly to their artefacts', and despite the evident centrality of artefacts to materialcultural studies, Roman archaeologists have, to date, paid little attention to the wealth of questions concerning social behaviour that can be addressed through artefact analyses and through more critical approaches to artefactual evidence and textual information. This paper is principally concerned with Roman artefacts and their functions in a 'male-dominated' space but can potentially contribute to dialogues across archaeology and social history. My own previous contribution to such dialogues has been recognized (e.g. Rawson 2003: 124).
The reliance of past Roman military scholarship on the authority of ancient authors, particularly the views of such authors that effeminacy or homosexual practices by Roman soldiers were severely punished and, like the presence of women, were considered disruptive to military discipline (see Phang 2001: 262-95), has ostensibly removed all but the most masculine of males from this arena. But Roman distaste for homosexual practices did not include pederastic relationships with male inferiors, slaves or male prostitutes (Parker 1997), who may well have cohabited with soldiers, as Van Driel-Murray, Phang and James have suggested, and may also have carried out the more 'feminine' tasks. Focusing on a concept of 'the feminine', whether biological or social and expressed as 'women and families', within these masculine domains is indeed the first, and only the first, step to breaking down this prescriptive view. The artefactual and particularly the skeletal evidence for the presence of small children and babies may not be directly concerned with the social concept of gender, but provides a stronger argument for the presence of biological women than for the presence of male slaves and male prostitutes, transsexuals or cross-dressers. This model does not exclude 'ethnic soldiers' and 'cross-dressers' from these domains but rather suggests there is a higher likelihood that women were a prominent part of the community than were ethnic, or any other, soldiers wearing a type of dress normally associated with women. Exceptions can no doubt be found, as discussed, but the views of the ancient authors of Roman military life and standards need no longer be used to explain away the archaeological evidence for the presence of biological women. As argued in a forthcoming paper (Allison n.d.), these artefact distribution studies indicate that women probably constituted some 5-15% of the fort population, a figure that would accord with James's. The next step is indeed to address the more complex issues of the performers of other traditional female activities within these forts, and the identifications and roles of the 'pueri'.
Despite the wealth of excavation of Roman forts in the western provinces, the nature of past excavations has meant investigators asking socio-spatial questions are frustrated by the lack of precise contextual and temporal information. Again limitations of space in this paper prohibited detailed discussion on these issues. Essentially, the three forts discussed here were chosen because they have better contextual documentation than any others, and all three experienced relatively rapid abandonment. The advantage of spatial mapping of artefact distribution patterns is that it can potentially overcome the specificity of 'redeposited' material. It is the consistent patterns that are important, not the precise locations of individual artefacts.
Finally, I thank the commentators for their useful and insightful comments and critical debate. Over the last decade the term 'engendering' has taken on a