Collection online


  • Object type

  • Museum number


  • Description

    The upper part of an iron sword, 557 mm long, including the metal fittings of its handle. The blade, 435 mm long to the end of the hilt end and 52 mm wide, is flat in section and there is no discernible taper. It is corroded and its edges are damaged, especially in the lower part. It ends at a fairly straight, and probably deliberate, break, and it seems that a sharp bend has been straightened 143 mm from the present end. There is a copper-alloy crown-shaped hilt end (66 mm wide) with a ribbed band below (57 mm wide). Both sides and both ends are decorated with pairs of stippled comma motifs. The tang is rectangular in section and tapers towards the top, which is burred over a four-clawed copper-alloy pommel cover (55 x 25 mm) that is decorated with a grooved border and cross-grooves over the top. In the centre of the tang is a thick, rounded copper-alloy washer that is circular in plan and 29 mm diameter. It is decorated with diagonal grooves and is bordered top and bottom by two bands, the inner notched and the outer plain. There are two other thick washers that would have separated the pommel from the grip and the grip from the guard. Each is oval, 22 x 17 mm, with narrow vertical ribs separating alternate broad and shallow grooves. They seem to be composite, with the ribbed and grooved element bordered by separate washers. There is another thin washer towards the top of the grip. The type of hilt-guard here represented belongs to the first century AD.


  • Culture/period

  • Date

    • 1stC
  • Findspot

  • Materials

  • Technique

  • Dimensions

    • Length: 557 millimetres
    • Length: 435 millimetres (blade)
    • Width: 52 millimetres (blade)
    • Width: 66 millimetres (hilt end)
  • Curator's comments

    Stead 2006
    From the Durden Collection. C.R. Smith 1868: 5, pl. 1; Piggott 1950: 21 and 27, pl. ii:I (Group IV); Brailsford 1962: 1 and A1, pl. ii:a; Spratling 1972: no. 289; Manning 1985:149 andV3; Jope 2000: 128 and 280, pl. 211;a-b.Manning 1985
    In the past this sword has usually been identified as an Iron Age piece, dating from the years immediately before the conquest (Piggott 1950, 21; Macgregor 1976, 80) but, as is argued below, it is more probably a Roman sword to which Celtic fittings have been added. The fragment of the blade would accord with Roman types; indeed with a width of c. 5 cm it approaches the extreme for such blades. The hilt, in so far as it can be reconstructed would have had a ribbed grip, with the moulded bronze ring forming the central rib, and a large pommel, all of which conform to the normal Roman bone hilt, only the metal elements being abnormal. Later swords usually have a spherical pommel, but a gladius of the Mainz type from Rheingönheim (a fort founded c. AD 43) (Schönberger 1969, 153) has a discoidal pommel similar to that from Hod Hill (Ulbert 1969a, 120, Abb. 3, 1).

    In archaeological terminology Roman swords are usually divided into two types, the gladius regarded as the traditional weapon of the legionary, and the spatha (which was longer than the gladius), seen as the weapon of the auxiliary.¹ The spatha is usually assumed to have been derived from the long Celtic sword, a type intended for cutting rather than thrusting. Roman tacticians, however, strongly disapproved of slashing with the sword when a thrust could be used; slashing rarely killed the opponent, while the swinging action exposed the attacker in a way which stabbing did not. That auxiliaries as well as legionaries were trained to stab is confirmed by a passage in Tacitus' Agricola (36.1) where the stabbing action of the Batavian and Tungrian auxiliaries at the battle of Mons Graupius is contrasted with the actions of the Britons who were fighting with swords lacking points. Ultimately the spatha is thought to have replaced the gladius throughout the Roman army, although the change was not completed until the third century.
    The situation in Britain in the first century AD as revealed by archaeology, is less clear cut than the traditional view would suggest. The type of gladius in use at the end of the first century AD had a relatively short blade, around 50 cm long and 5 cm wide, with parallel edges which converge some 7.5 cm from the end to form a point. Probably the best example of the type found in Britain comes from the fort of Newstead, a site which has produced several other swords. Of these two were identified by Curle (1911, 183, pl. XXXIV, 6 & 7) as spathae, another he accepted as Roman on the strength of the bone hilt (1911, pl. XXXIV, 13), but the remaining pair, which had hilt-guards of simplified Celtic form, he regarded as 'native' (1911, pl. XXXIV, 8 & 10). The dimensions of these swords are given in the chart from which it can be seen that the so-called Celtic sword has a blade of the same width as, and only slightly shorter than, the spathae. Thus the only reason for rejecting it (and by association its companion) as another auxiliary sword is the form of the hilt-guard.
    Hilt-guards of this type occur in some numbers in the north of Britain with a smaller group in Dorset. They form part of Piggott's Group IV (Brigantian) scabbards, although not all of the scabbards included in that class had these hilt-guards (Piggott 1950, 17 ff.). Piggott divided guards of this general form into two types, the 'cocked hat' (IVA) and the 'crown' (IVB). In her discussion of Early Celtic Art in North Britain Morna Macgregor lists nine crown hilt-guards from the north of Britain (1976, 79 ff.). Two were those on the swords from Newstead, another came from the Roman fort at Manchester, three were in some way associated with Roman material, one came from a hoard of small bronzes, probably of early Roman date, from Middlebie, Dumfriesshire, and another formed part of a sword hilt from Cotterdale, North Yorkshire, which is of a type clearly modelled on the standard Roman form. The cocked-hat guards were less common, but of the four recorded by Macgregor, two are from Newstead, and another is on the sword from Fendoch. None of these is likely to date from before the Roman conquest of the area, and none from long, if at all, after the end of the Flavian period.
    The second and smaller group comes from Dorset. Hod Hill itself has produced four crown hilt-guards (Brailsford 1962, A1-A4), and a scabbard mount of related type (Brailsford 1962, A5); two crown-guards come from the Roman fort at Waddon Hill (Stoke Abbott) (Webster 1960, 91 no. 212, fig. 7; Webster 1981, 67, fig. 30, 57); and a cocked-hat guard comes from the late Iron Age hoard from Bulbury (Cunliffe 1972, 298, no. 7, fig. 3).
    In the present context the key group is that from the Roman forts. As we have seen only one of the swords with such a guard from Newstead is sufficiently well preserved for its dimensions to be useful, but on those grounds it could be Roman. The sword from Fendoch with the cocked-hat guard was found in a foundation trench of the principia (Richmond 1939, 142, 147), and as Richmond made clear in his report on the site, there can be no doubt that this was an auxiliary's sword. As it is closely comparable with that from Newstead, it largely confirms that the Newstead swords, too, were Roman. This was a conclusion which Richmond accepted (1939, 147), and it is the only reasonable explanation for the discovery of such swords in what are entirely Roman contexts (Manning 1981, 57).
    Metal fittings of Celtic design are quite common in the Roman forts of northern Britain as Macgregor's distribution maps make clear, and the explanation is obvious. Roman auxiliaries, largely drawn from the Celtic provinces including Britain, were attracted by such decorative pieces, which were firmly within their own artistic traditions, and as they did not materially affect the functioning of their equipment, they were allowed to use them. If, as seems almost certain, the Roman soldier of the first century AD had to buy his own equipment (Watson 1969, 102; Breeze et al. 1976, 93), some degree of personal preference in the inessentials might well be allowed. Nor need the legionaries have been immune to the attraction of Celtic design; one must have found a British mirror sufficiently attractive to have carried it to Nijmegen.
    If these arguments are accepted for the northern swords of the late Flavian period, they must be equally acceptable for the fittings from Hod Hill and Waddon Hill which are half a century earlier, and therefore date from a period when standardisation had proceeded less far. The Hod Hill sword is essentially a Roman sword with an unusually decorative hilt in which bronze replaced part of the normal bone grip and guard, and the other hilt-guards in the Durden Collection no doubt came from similar swords. That this type of guard first appeared late in the pre-Roman Iron Age is probable; that they were made by native craftsmen is almost certain, but that they were made for Roman swords at the behest of Roman soldiers must be considered undeniable. Whether those soldiers were auxiliaries or legionaries is an open question. Both appear to have been in garrison at Hod Hill, and while auxiliaries might be thought the more obvious choice, the width of the existing sword and the size of the other guards would certainly allow them to be identified as being for gladii; a point which was made in passing by Graham Webster (1979, 130).
    In addition to the swords so far discussed there are three others of mid-first century date from Camulodunum (Hawkes and Hull 1947, 112, 340, pl. CIV, 3, 4, 5). Two of them were identified as gladii, but although they have blades of similar length to the Newstead gladius they are markedly narrower, being 3.8 cm wide as against 5.0 cm. The third is both narrower and shorter (3 cm wide and c. 39 cm long), and thus falls outside all the obvious groups. Nor does the sword from Fendoch, a Flavian auxiliary fort, fall neatly into the same group as the Newstead spathae, for its blade is of gladius length (52.1 cm), with a comparable width at the top (4.9 cm) although it then tapers to c. 3 cm just above the tip. Such variations in the proportions of blades are not confined to Britain. Of the three gladii from Pompeii illustrated by Ulbert (1969a, Taf. 17), two have blades of about 50 cm in length, while the third is some 8 cm shorter.
    All these swords must date from the second half of the first century and they make it clear that although the basic form was standardised there was a considerable degree of variation in both the length and width of Roman sword blades at that time. In fact the major change in this field which took place during the first century was probably the replacement of the older form of gladius, with its slightly waisted blade and long point, by the new form of equal length but with parallel edges and a short tip. Once this had been accomplished both gladius and spatha had achieved a basic similarity of form which inevitably led to a blurring of the distinction between them.
    The date at which this change took place is uncertain, and in any event it must have taken a period of some years to reach completion. The Fulham sword (registration no. 1883,0407.1) is unlikely to date from before AD 43, which indicates that some of the troops entering Britain still carried such swords. However, a scabbard tip from Hod Hill, which must date from before the end of Claudius' reign, is for a gladius of Pompeian type.² (Brailsford 1962, A14, fig. 1. Compare Ulbert 1969a, Abb. 4, and Taf. 17, 19, 26, 27.) Thus we may conclude that the decade AD 40-50 was one in which the new form was replacing the old.

    1. The ancient authority most often quoted in support of this distinction is the passage in Tacitus' Annals in which he describes the defeat of Caratacus at the hands of Ostorius Scapula: “si auxiliaribus resisterent, gladiis et pills legionariorum, si, huc veterent, spathis et hastis auxiliarium sternebantur” (Ann. XII, 35). Unfortunately, it would seem more likely that Tacitus is using the terms rhetorically to gain contrast and variety of language, than that he is reflecting a technical distinction. Less often quoted is the passage in the Agricola (36) in which he uses gladius for the swords of both the native British and the Roman auxiliaries in the battle of Mons Graupius, even when he is specifically (and rhetorically?) commenting on their excessive length. Vegetius, probably reflecting a later situation, refers to the legionaries using larger swords (which he calls spathae not gladii) and smaller ones, semispathae (Vegetius De Re Mil. II, 15).

    2 What is most probably a gladius of the Pompeian type may well be associated with the Claudian fort at Waddon Hill, unfortunately its findspot is recorded as Long Windsor, Dorset, a place-name which does not appear to exist in that county. However, Waddon Hill lies but one and a half kilometres from Broad Windsor which suggests an error in recording the findspot and that the sword is actually a stray find from that site: its early date is confirmed by an association with a coin of Claudius (Britannia 10 (1979), 259-60). The blade has a length of 48.4 cm and a width of 4.4 cm at the shoulders; whether it tapered before the reduction to the point is made uncertain by extensive damage to its edges.


  • Bibliography

    • Lang 1988 bibliographic details
    • Stead 2006 234 bibliographic details
    • Brailsford 1962 A1 bibliographic details
    • Manning 1985 V3 bibliographic details
  • Location

    On display: G49/dc8

  • Exhibition history


    2015-2016 24 Sep-31 Jan, London, BM, G30, 'Celts: Art and Identity'

  • Acquisition name

  • Department

    Britain, Europe and Prehistory

  • Registration number



If you’ve noticed a mistake or have any further information about this object, please email: 

View open data for this object with SPARQL endpoint

Object reference number: BCB9078

British Museum collection data is also available in the W3C open data standard, RDF, allowing it to join and relate to a growing body of linked data published by organisations around the world.

View this object

Support the Museum:
donate online

The Museum makes its collection database available to be used by scholars around the world. Donations will help support curatorial, documentation and digitisation projects.

About the database

The British Museum collection database is a work in progress. New records, updates and images are added every week.

More about the database 


Work on this database is supported by a range of sponsors, donors and volunteers.

More about supporters and how you
can help