An Assessment of the findings of The Ermine Street Guard.

INTRODUCTION

Over the past twenty-five years the Ermine Street Guard has reconstructed and utilised the arms, armour and equipment of the Roman soldier of the second half of the first century AD, and has displayed throughout Europe, winning "the respect and admiration of the public and academic world alike" (Zeinkiewicz, 1994, p36). However, such work is still primarily appreciated for its public entertainment value rather than the essential academic research that is involved.

Consequently, at this point, it is important to acknowledge and document some of the findings that are a by-product of the society's desire for authenticity. Although the Ermine Street Guard covers all aspects of the Roman army, of this period, such a large subject cannot adequately be covered here so the Roman legionary will be concentrated on. Also, before outlining these findings, it is essential to see them within the limitations of experimental archaeology on the whole and within the confines of the available representational, archaeological and literary evidence.

THE ERMINE STREET GUARD

The Ermine Street Guard (E.S.G.) was founded in 1972 after a village pageant, representing the history of the area, which aimed to raise money for the village hall of Bentham and Witcombe, Gloucestershire. Through an interest in what they had learnt about Roman arms, armour and equipment, the founding members decided to continue their research and improve upon what they had made by keeping in line with archaeological research, and to continue giving public displays for educational purposes. As a consequence of twenty-five years of research they have been described as the best re-enactment group in the world by Howard Giles, English Heritage's Special Events Organiser (unpublished, 1993) and give performances at Roman sites, museums and military events, for which fees are charged to finance further research and reconstructions. Since its founding the E.S.G. has become increasingly involved with academics from all fields of research; Dr. M. Hassall (artillery), Dr. M.C. Bishop (military equipment), Dr. LC.N. Coulston (military equipment), Professor P.Wild (textiles and dress), Dr. C. van Driel-Murray (leather),Dr. D.Baatz (artillery), etc. being but a few, as well as the major museums both in Britain and on the continent. Such museums also often ask the E. S.G. to provide replica equipment for museum displays as it is made to such high standards.

While the E.S.G.'s work is not directly seen as experimental archaeology the very nature of research into and reconstruction of arms, armour and associated equipment, from the artefactual, sculptural and literary evidence is in itself experimental and therefore could be considered experimental archaeology. While the equipment is not manufactured specifically for research into use, wear and breakage, there is no restriction on using it to test hypotheses, or to produce useful information.

It is important to remember, however, that the equipment made by the E.S.G. is generally made by the individual for his own use and that there may be a greater degree of care taken in its manufacture than perhaps would have occurred with the original artefact. Consequently, it could be suggested that to some extent equipment is 'over engineered'. Further more it is important to note that the materials used are of a more consistent quality than Roman materials.

EXPERIMENTAL ARCHAEOLOGY

Experimental archaeology has both its advantages and limitations Coles (1979, p. 1) that "by trying to make and use some of the weapons and tools of the past we can gain an insight into the importance of these objects to their original inventors and owners" By participating in these experiments archaeologists are able to improve their understanding not only of technical abilities but also the reason for choosing one course of action over another (Coles, 1979, p. 1). However, he also points out that it is only through the formulation of structured models for the examination and assessment of reconstructions that any conclusions arrived at can be of value. Most importantly, it must be remembered that such conclusions "can only show how something might have been done, not how it was done" (Bishop and Coulston, 1993, p.38). Results gained from reconstructions may overwhelmingly point to a certain practice or answer which is in line with the current archaeological evidence but it must be remembered that archaeology is not a static discipline and constantly new evidence comes to light which may conflict with present theories based on experimental archaeology results (Bishop and Coulston, 1993, p.38).

It is also essential to understand that while re-enactment groups like the E. S. G. do not pretend to know what it was like to be Roman legionaries, having not been brought up on the diet or discipline of the Roman soldier, it must be acknowledged that all such vital work has its limitations. Equipment may be incorrectly reconstructed no matter how much care has been taken with the replication. People such as Junkelmann (1986) have undertaken experimental marches, training with heavier loads until they have reached the approximate weight of full legionary kit; yet this overlooks the multitude of activities which the late fourth early fifth century writer Vegetius mentions in the Epitoma rei militaris which the legionary would have been trained for before he joined the legion. Essentially, without completely replicating the diet, standard of living and full activities of a Roman legionary no conclusions can be one hundred percent accurate especially as they are constantly affected by one's own inherent twentieth-century biases. Consequently, although experimental archaeology is often invaluable to research it is necessary to see each experiment within its own limited scope; within the bias of the present and to see information gained from it as constantly under review.

THE EVIDENCE

REPRESENTATIONAL EVIDENCE

For the first century AD, there is a wealth of representational evidence available in the form of propaganda sculptures and other monumental works. For example although constructed in the early second century AD, when considered alongside archaeological evidence, Trajan's column is of relevance to the latter half of the first century AD and the Arch at Orange which is probably of Tiberian date. Although these monumental works can produce some evidence, they are considered to be the work of urban sculptors trained in a Hellenistic manner, which is more concerned with the aesthetics of the sculpture than accurate representation. Also, what little the sculptors would have known of military attire and practice was almost certainly based upon the military units in Rome rather than on the frontier troops on the frontiers of the Empire, whom they were trying to portray, and who would have been subject to evolution in practice and attire as a result of changes in military policy on the frontier.

A second, perhaps more accurate and thus important source of representational evidence comes from funeral monuments erected by the deceased's colleagues to portray his status and profession as a soldier, to passers-by. These sculptures provide us with representations of military attire, which were carved locally where the deceased was stationed and, therefore, it could be suggested that the sculptors involved in these grave stelae would have a better understanding than the monumental sculptors of the military arms, armour equipment and practices they are portraying, which leads Bishop and Coulston (1993, p.25) to suggest that "military stelae provide an important foil to urban sculpture because so many details of sword fittings and horse-harness, for example, are verified by the artefacts."

ARCHAEOLOGICAL EVIDENCE

Whilst it is fair to say that there is a wealth of archaeological evidence for military equipment throughout the Empire, it can be argued that the frontier provinces, where the majority of the army was stationed, provides the greater part of the artefactual evidence for research into the equipment of the Roman army. This supposition is perhaps corroborated by the predominance of British and German and Dutch academics in the field of Roman military research, both countries having important 'military' frontiers of the Empire in them and consequently military activity. It is, however, important to consider why any artefacts thought of as military equipment entered the archaeological record because this very reason may influence their relevance to military equipment studies and the interpretation of use.

Accidental loss

Artefacts may enter the archaeological record through accidental loss, and it is perhaps artefacts of this type that are of greatest value. The fact that they have been lost accidentally would suggest that they were in use at the time of deposition and therefore of serviceable quality or at least a required item of Roman army equipment. However, it tends to be the case that smaller items of low value or importance are lost and the artefacts which could be considered of more importance and greater value to the owner will be looked after more carefully or more diligently recovered. This argument is reinforced by papyri that show deductions were made from pay to cover the issue of equipment (Bishop and Coulston, 1993, p. 198), providing a financial incentive not to lose equipment.

The 'losing' of these artefacts may reduce their value for reconstruction purposes as the reason that they may have been lost was because they were of poor quality or badly made and subsequently broke and were deposited; the part which broke, not being found because of decomposition, for instance the leather apron straps on belts. It can be suggested that the reason these items have been lost is because they have been poorly or even incorrectly made and that to produce theories about construction and use of these artefacts and project them onto the army as a whole may be unwise.

Site Deposition

Deliberate deposition of items must be considered when examining the reasons why artefacts enter the archaeological record. The forts at Corbridge, Inchtuthil and Newstead for instance, relating to different periods of Rome's involvement with northern Britain, have produced 'hoards' of military equipment, which had been deliberately buried. It is suggested that this material was buried to stop valuable scrap metal falling into the hands of Rome's adversaries, when garrisons withdrew from forts which on withdrawal were subsequently raised to the ground (Bishop and Coulston, 1993, p.34). At Inchtuthil the items buried included nails and iron tyres for wheels, and, more importantly for Roman military equipment studies, the artefacts at Newstead and Corbridge included at Newstead swords and lorica segmentata and at Corbridge complete sections of lorica segmentata. The finds at Corbridge provided unique evidence as to the construction of lorica segmentata. Although finds of the copper-alloy fittings and small fragments of the iron plates involved in the construction of this type of armour had been found previously and there was sculptural evidence for its form, it was not until the 'Corbridge hoard' was found that it was possible to produce accepted reconstructions of this type of armour. Other examples of deposition are of soldiers found either in full or partial military attire. Such depositions occurred at Herculaneum where a soldier was buried by ash from the eruption of Vesuvius in AD79, and Velsen where a partially uniformed soldier was found down a well and is thought to have been murdered.

Votive offerings

Another way in which military equipment may be 'lost' is through its use as votive offerings. This practice is thought to have commonly occurred in rivers such as the Rhine where several helmets of good condition have been found, or in the river Thames where the Fulham sword was found as a consequence of dredging. However, it has been suggested that as so many helmets have been found in rivers that these items were the result of accidental loss; in the case of the Rhine through the loss of a barge. It is possible though, that they have been deposited as offerings after a prayer has been answered but the very fact that these items of equipment were the individual soldiers' responsibility and that he could be punished if he lost them would mitigate against such a theory. Despite the disagreement as to the reason for deposition a relatively large number of items have been recovered from water sites, including helmets, swords, scabbards, and shield bosses, and more are continually being found.

LITERARY EVIDENCE

Generally literary evidence tends to be indirect in its nature, but is provided by such contemporary writers as Julius Caesar, Josephus, Tacitus and, writing in the fourth century, Vegetius. These accounts obviously vary in their relevance to reconstruction, and to different topics. While Caesar's writings are literary in nature and therefore provide little direct evidence. Heron, however, wrote a technical treaty on torsion artillery, which gives dimensions, calculations and descriptions and in every aspect provides evidence. Vegetius' writings are different again as they hark back to a time when the army was at its best, using the since lost writings of contemporary writers to describe it. While it is possible to use the literary sources to different extents it is more satisfactory to utilise evidence from them when we have archaeological or representational evidence which corroborates it.

Finally, before any attempts at reconstruction can be made, it is important to realise why and by whom the original artefacts have been made. This is essential as it may be reflected in the standard to which the artefact was made and the materials used; ultimately having an effect on any reconstructions made. It is suggested by Bishop (1985, p. 17) that the Roman army, at least in the north and west of the Empire, used two mechanisms to provide arms, armour and equipment. Firstly there was the supply of these items by commercial concerns (factories) which were set up for the purpose of supply to the army and secondly, this method of procurement could be supplemented by the fabricae in forts. These fabricae would have been able to utilise the skills of trained legionaries, when necessary, to produce new equipment, repair damaged equipment and overhaul or refurbish worn equipment. It could thus be suggested that although no doubt there were specifications and quality controls upon the arms and armour produced by the commercial arms factories, there would be a continual struggle to maintain the required quality of items produced, as a result of the producers desire to maximise revenue. It could also be said but not proven, that where the production centre was distant from the point of use, the producer may feel less obligation to produce quality goods; the failure of such goods having little bearing upon his personal well-being. If these points were to be true, then it could be proposed that the armour and equipment produced in fabricae would be manufactured to a higher standard. Those producing the goods in these workshops would have a vested interest in ensuring that the soldiers fighting alongside them were equipped to the highest standard.

Indeed some examples of equipment found are poor if not crude in their execution and also utilise materials which are possibly inadequate for the job intended. This is particularly true of metal artefacts and a case in point are the copper-alloy hinges on lorica segmentata of the Corbridge type which appear very thin for their purpose. There is, however, a problem with this theory in that we do not know how worn the metal was before deposition, how much it has corroded during its deposition and in post excavation how much of it has been lost in the conservation process

ITEMS OF PERSONAL EQUIPMENT

While much is written about the individual pieces of Roman legionary equipment, most of this is from a theoretical, rather than a practical point of view. Even those who have reproduced pieces of equipment have tended to concentrate upon the function of the pieces rather than the functioning i.e. what it is like to use or wear them and how these items over time become worn and either need to be replaced before they break or break and as a consequence need repair, or replacement. When considering pieces of personal equipment i.e. arms and armour, it is also necessary to take into account the amount of protection given with regard to the amount of inconvenience experienced while wearing it. The greater the protection the armour gives, the heavier and less flexible it will become. In either case a sacrifice will be made, flexibility and manoeuvrability being lost to allow protection and the protection provided being limited by the desire for flexibility and lightness of armour. Therefore it is appropriate now to commit such observations to paper.

THE HELMET

Evidence of the form that Roman helmets took is amply demonstrated by the corpus of material in H. Robinson's "Arms and Armour of Imperial Rome". Although published in 1975 this book is the only work that attempts to discern a chronology for helmet development and although a number of new helmets have been found since publication it remains a 'bible' for the reconstruction of Roman legionary helmets. Finds of Roman legionary helmets in iron and copper alloy vary considerably in their thickness; the thickness also varying with where the measurement is taken upon the helmet bowl. The reason that the helmets vary in thickness across their bowls is because when the helmet is formed from a single piece of metal over a former, it has to be thinned or thickened in certain places to form the necessary curves. While thickening of the helmet bowl is beneficial, thinning of the metal is counter productive, and occurs most drastically at the brow of the helmet; resulting in the necessity for reinforcement in the form of a brow guard.

In the first century AD the Romans constructed their helmets, apart from the cheek-pieces and additional fittings, such as edging, rosettes, crest-holders and browbands, from single pieces of metal. Although Coolus type helmets, as described by Robinson (1975p.26-41) were sometimes made using the spinning method where a metal sheet is spun and formed over the mould and produces a round helmet, this is not ideally suited to the oval shape of the human head. Imperial Gallic and Italic helmets, however, are produced by forming the helmet without spinning and results in a greater degree of control over the form of the helmet allowing a greater range of helmet forms to be produced. The changes from first century BC spun helmets to first century AD beaten helmets may result from the change in the metal used for legionary helmet construction. Production methods had to change to suit the different properties of iron which is less malleable than copper-alloy.

Although the idea behind a reconstruction is to create an exact copy, this is not always possible due to either the manufacturing process used or the cost. As a result of the cost involved in producing single piece helmet bowls the E.S.G. purchases bowls that are composite in construction with the bowl made in two halves and depending on the metal welded or braised along a line from between the eyes to the middle of the nape of the neck with a neckguard added at the rear. Other people have produced helmets by spinning and squeezing to alter the shape, however, this produces a helmet with a steep neck-guard and is only suitable crude reproduction of type "H" Gallic helmets. Paradoxically the fact that cost causes the E.S.G. to purchase helmets of composite construction is perhaps reflected in the construction of later Roman helmets. Helmets of the later empire are produced in a composite manner, such as the Dominate helmets from Intercisa, Augst, Worms, Der el-Medineh, Berkasova and Budapest (Bishop and Coulston, 1993, p. 169-170). It could be suggested that the helmets of the Late Empire were produced in composite form for financial reasons, however, an alternative reason could be that the metal working skills needed to produce single piece helmets in quantity were for uncertain reasons, lost.

The average weight of a replica helmet with all fittings but excluding crest is 41/2-51bs (2.05-2.3kgs). A helmet of this weight upon the head, unless adequately secured and padded, would be particularly uncomfortable. Evidence for securing the helmet, by attaching a thong or string at the neck-guard loop and passing it through the cheek-piece loops, is apparent from archaeological finds such as the Imperial Italic type 'I' helmet from the river Rhine at Mainz (Robinson, 1975, p.53). Some form of padding or method of keeping the head out of contact with the helmet is necessary as most of the fittings on the helmets are riveted onto the helmet bowl; very few being soldered on. These rivets protrude into the inside of the bowl creating raised lumps, which could be uncomfortable if they came into contact with the head. There are a number of these lumps protruding into the bowl at the forehead for securing the brow-guard, brow-band, front crest securing loop; at the rear of the helmet for the rear crest securing loop, or hook and at the temples for the attachment again of brow-guard, brow-band, ear-guards, cheek-pieces and sometimes plume holding tubes.

As yet, no conclusive evidence has been found to suggest the true form of the lining used in helmets. However, based upon traces of fabric adhering to helmets from Hod Hill (a cheek-piece) and Newstead (a helmet bowl), Robinson (1 975:p. 144) suggests that the material used was not leather, as previously suggested, but some form of woven fabric such as linen or wool, which would have been glued to the helmet at the rim but secured at the crown with a tape so that adjustments could be made for fitting (fig.1). By securing the lining only at the rim, to some degree, air would be allowed to circulate and the wearer's head kept out of direct contact with the bowl. In effect, the helmet would be suspended from the head, using webbing. An alternative method of lining the helmet, for which there is no direct evidence, would be to stick a suitably thick material, such as felt, inside the bowl to provide padding. With this method, the helmet would rest on the head rather than being suspended however, this method would reduce air circulation within the helmet.

Experiments with webbing, based on suggestions by Robinson (1975:p. 144), do prove very effective, the helmet being reasonably comfortable. Nevertheless, due to their weight helmets are generally removed as soon as possible so comfort is a minor consideration in comparison to the need for protection. Even with webbing, which allows air circulation, the inside of the helmet can become stuffy. This suggests against simply applying padding to the inside of the helmet, which would allow less air circulation than webbing and could make the helmet uncomfortable to wear.

Another factor that needs to be considered is that the inside of the brow quickly becomes damp with perspiration and over several days of wear can become dirty and greasy (pity the poor souls who try on a helmet at displays). The problem of the lining accumulating dirt and grease, whether webbing or padding, could be solved by Robinson's findings (1 975:p. 144) that when hot water was applied to the glue adhering to the inside of the helmet found at Newstead it became tacky and was easy to remove. If this was the case with the original helmets, it would be relatively easy to remove a webbing lining which was only glued at the rim and replace it with a fresh lining. If padding was used, however, the process of replacement would be much more difficult, the whole bowl needing to be heated to remove the padding; the effects of all over heating possibly affecting the properties of the metal used in the helmet.

Although, either type of lining can be padded out at the rim to provide a good fit, it is necessary to start with a helmet bowl of reasonable fit. With too large a helmet, padding it out and securing the helmet below the chin with a thong through the cheek-piece loops, does not bring the cheek-pieces into close enough contact with the cheeks. This means that an element of lateral movement is introduced which can be uncomfortable, particularly when running, as the helmet will tend to 'bounce' on the head.

BODY ARMOUR

The Roman legionary had three basic types of armour available to him, plate armour: lorica segmentata (although the correct name is unknown) scale armour lorica squamata and mall armour: lorica hamata. The primary function of these types of armour is protection, mainly against downward sword blows. This is shown by the additional protection upon the shoulders, with the multiple layer construction. It is probable that the extra protection provided to the shoulders helped protect against blows which were deflected from the helmet onto the shoulders. However, the protection that it provides comes at the expense of manoeuvrability and extra weight to be carried.

The method by which these armours protect is to dissipate the force of a blow so that rather than it being lethal it becomes disabling bruising, or perhaps only annoying'. The method by which protection is achieved, however, varies with the type of armour. Lorica segmentata achieves this by bending or crumpling, when a severe blow strikes it; thus absorbing most of the force of the blow. With this method it is self evident that the more layers of metal the greater the protection achieved, which it can be suggested is why there are several thicknesses of metal on the shoulder. Being constructed of small plates or scales of metal Lorica squamata works on the same basic principle. At least three layers of scales overlap at any one point on the scale suit and in addition protection is increased by the use, in some cases, of shoulder doubling which effectively protects the shoulders and the back of the neck, with six layers of metal. However, with squamata the fact that rigid plates are not used means that the absorption of blows will be less effective, but the flexibility of the armour will be greater. This problem is counteracted to some extent, though, by the increased number of layers of metal. Lorica hamata is perhaps the least effective form of armour with only a single layer of inter-locking metal rings. Nevertheless, it

still reduces the impact of a blow, which could have caused a severe cut, to perhaps broken bones or severe bruising. Rather than absorbing the force of the blow, mail dissipates the force. Shoulder doubling is also often used to provide extra protection and absorb the force of a blow, but interestingly sculptural evidence for auxiliaries, such as that from the metopes of the Tropaeum at Adamklissi, indicate that they do not have shoulder doubling, perhaps to increase their manoeuvrability. Such manoeuvrability and flexibility is an overall advantage of mail, it being in effect a heavy T-shirt.

Lorica segmentata

Reconstructions of loricae segmentatae armour are primarily based upon the finds from the 'Corbridge Hoard'. Although this provides no full set, it contains three sets of girdle plates and five half collars with attached shoulder plates and three variations of the basic lorica type. All of the types of lorica segmentatae classified Corbridge type 'X, 'B' and 'C' by Robinson (l975:p.177) use an arrangement of internal leather straps to join together a series of metal plates of various shapes and sizes; when worn producing a flexible form of armour, which is able to concertina upon itself. The continual movement of the metal plates causes the leathers to be constantly under stress which eventually makes them liable to break. This most frequently occurs with the leathers connecting the shoulder plates. Due to their disturbance whenever the arms are moved, they are continually being put under differing stresses. After the E.S.G.'s trips abroad, to the south of France and Portugal, where the armour is worn in high temperatures, it has been noticed that these breakages increased and it can be suggested that the increased perspiration has been a direct cause of the perishing of internal leather straps. When these leathers break it is not possible to make temporary repairs. It is necessary to remove at least six rivets before putting in a replacement leather if a girdle leather breaks, and at least three rivets to repair a broken shoulder leather. Once a leather has broken it normally becomes advisable to replace all the leathers in that piece of the armour and to effectively strip it down and rebuild it. It could be suggested therefore, that before the Roman military undertook any form of campaign they would check every piece of leather and probably replace a great many of the leathers while in secure conditions, to minimise repairs in the field. Eventually, though, increased numbers of finds from campaign camps may clarify this.

Recent inspection of the Corbridge 'Hoard' lorica segmentata by Bishop (1996 unpublished) has revealed that they were manufactured using almost pure copper rivets and not, as previously thought, brass or bronze and iron rivets. Although it may seem that this is of minor significance, it can lead to some new hypotheses on the manufacture of lorica segmentata. It could be suggested that copper rivets would be too soft and malleable for use on protective armour, and that they may be adversely affected by wear. Even so it could be said that the leather straps are even less durable and would break before the copper rivets. The use of copper rivets may, however, be deliberate facilitating construction and perhaps more importantly, repair of lorica segmentata. If copper rivets are used it would mean that due to their malleability they are easier to remove using cold chisels (metal working chisels) and drifts (metal bars used to drive softer bars through holes) than copper alloy or iron rivets, speeding up the repair process,

In line with Robinson's research (1975) the E.S.G.'s reconstructions of lorica segmentata have utilised copper alloy and iron rivets. It has been noticed that unless incorrectly fixed, such rivets do not break, or fail, perhaps suggesting that using such durable metals does represent over-engineering and may even hinder repair. Whilst the E.S.G. has no experience of use of copper rivets in the construction of lorica segmentata, by a process of repair and renewal copper-alloy and iron rivets will be replaced with copper and perhaps some new evidence for armour construction will arise.

The first type of lorica, Robinson's Corbridge type 'A', uses an arrangement of buckles and leather straps to attach the shoulder pieces to the body girdle and amongst E.S.G. members is the most sought after type of lorica segmentata. This is because once assembled the parts do not become detached on their own when the armour is put down, as types 'B' and 'C' tend to, and so can be easily moved about and put down, but when necessary can be easily disassembled. The most serious problem. with this type of armour is that they are more likely to break as if one of the straps breaks it has a knock on effect upon the other straps.

Corbridge type 'B' consists of a series of meta. Hooks and loops to attach the shoulders to the girdles four at the rear (2 per shoulder) and two at the front (1 per breast). A problem arises with this type of attachment because when the hook is bent into position, the metal E weakened. This weakening is a result of bending and working the metal when cold, during construction and: can result in fracture lines developing at the point of bending if the metal has been over-worked. This can be avoided, however, by annealing (heating) the metal at certain temperatures at suitable intervals resulting in the (molecular structure of the metal returning to its pre-worked state, allowing it to be worked again. The process of assembling this type of armour can also result in the bending of these hooks and the resultant straightening which again weakens the metal. Although appearing perfectly satisfactory these hooks can break suddenly, due to fracturing caused by the stresses of bending am straightening; unlike type 'A' straps which can be seen to be wearing. Being of metal the hook and loops, being of metal also act against each other; wearing each other away, which can again weaken them. However, a visual inspection makes this evident.

Although as yet unknown to the writer, archaeological evidence of the breakage of these hooks may be available in the fabricae of Roman forts. Here the plates riveted to the girdle plates rather than the hook extension may be found. The basis of this supposition is that the hook will break and be lost whereas the plate will, to facilitate a repair, be removed and the piece recycled in the fabricae, some being lost in the process. It can be suggested that evidence of these broken pieces in the archaeological record could be used to support the accuracy of reconstructions.

While discussing the removal of hooks it is worth considering a point made by Bishop and Coulston (1993,p.85) that contact between iron and copper-alloy will produce electrolytic corrosion. When the broken hooks are removed this is not the case and in fact it can be suggested that the level of surface deterioration is lower behind the plate than upon the rest of the iron. It is possible however that modem metals with fewer impurities or perhaps modern cleaning products could be the reason for this, reducing the level of corrosion.

Perhaps the reason that type 'A' lorica are preferred is not that it seems to break less frequently but that the process of repair is easier. The removal of two rivets the cutting of a piece of leather and its replacement is all that is necessary whereas type "B" lorica requires the removal of two rivets, the cutting shaping and bending of a piece of copper-alloy and then its replacement, which takes considerably more time.

There is a type 'C' Corbridge lorica segmentata. This however differs very little in construction to type 'B' and consequently needs no discussion; it differs only in the manner by which the girdle sections are attached to the shoulders at the rear.

It is perhaps worth mentioning a different type of lorica segmentata, the 'Newstead' type. This form, found at Newstead north of Hadrian's Wall; is dated to the Antonine period by Bishop and Coulston (1 993,p. 119) and considerable differences to the Corbridge types of lorica. Firstly the girdle plates have been reduced in number but made broader and the shoulder-pieces have been very much simplified. This disposes of the unnecessary hinges between plates and substitutes them with rivets, reducing the number of plates. It is perhaps surprising that the method of attaching the shoulders to the girdles is the metal hook and loop method, the method most disliked by E.S.G. members. The loops that allow the two halves of the girdle to be tied together are no longer riveted to the girdle plates but are effectively split pins where the tails have been opened out. This method would certainly make it much easier to replace them, by simply folding the tails back together and withdrawing it. The Newstead lorica is also plainer in its appearance, doing away with the purely decorative rosettes that backed many of the round-headed rivets on the Corbridge type lorica.

Lorica Squamata

The sculptural evidence for scale armour appears to indicate that there is a distinction between the design worn by legionaries and those worn by centurions and standard bearers. The scale shirts worn by legionaries have no shaped scallops at their bottom edge or at the sleeves, they have a curved bottom edge and cut-off sleeves. Whereas the scale shirts worn by standard bearers exhibit a single scalloped bottom edge and scalloped sleeve edge. However, the shirts of centurions, exhibit a double row of scallop's at the bottom edge as well as at the sleeves. These scallops would serve no function and must therefore be decorative and possibly a method of distinguishing rank (fig.2).

The construction of scale armour involves the linking together of the individual scales with a twist of wire, so that each one overlaps the next, producing a fairly rigid length of scales. These rows of joined scales are then attached to an undergarment so that the rows overlap the ones below them by about one third. This produces an armoured garment, that relies upon the backing garment for support, unlike the segmentata and hamata which are armour in their own right. One problem with this is that the weight of the scales drags on the material used and can stretch it. If this occurs too greatly, the overlapping of the scales vertically can be reduced, possibly even exposing the backing garment, thus reducing the effectiveness of this form of armour. It is therefore necessary to take account of this when the backing garment is being constructed; unless the properties of the material are such that it is unlikely to stretch vertically. It could be suggested that the material chosen for the backing garment may be of a particular type of weave that would be unlikely to stretch, or was used in the direction in which it was least likely to stretch.

A point of particular importance with squamata is the construction of the sleeves where they come part way over the shoulders and down the upper arm. The sculptural evidence shows the scales running in curves across the torso as though they are radiating outward in concentric circles from a fixed point. This means that the scales form circles around the neck opening. However, if these circles radiate outwards and include the partial sleeves there is no flexibility and the arms cannot be raised freely. This freedom of movement can be achieved if the sleeves are formed in scallops which divide them up. However, as noted before scallops are only shown on suits of squamata worn by officers and standard bearers, so it must therefore be the case that the sleeves on legionary suits were formed in some other manner, as yet unknown. There seems less known about the form that lorica squamata took than about any of the standard legionary armours, perhaps because the basic principles of scale armour are so simple that nobody has thought, to any great extent, about how the actual garments are constructed.

Because the rows of scales are staggered to maximise the effects of the overlapping, with squamata the edges of the armour appear rather ragged. This raggedness could become a nuisance, by catching. It is therefore necessary to provide some form of binding to the vertical edges. There is a precedent for this in the representation of squamata in sculpture, where the bottom edges of the armour and the edges of sleeves and shoulder doubling are always depicted with a plain narrow strip (fig.4). This could be seen as an edge binding, possibly of leather, which stops the bottom edge of the armour chafing upon the tunic worn beneath. It would be sensible to continue this binding around the vertical edges, particularly where they join together, to secure the armour when worn and where this fastening halts for the arm apertures; to prevent the chaffing mentioned.

Lorica Hamata

Lorica Hamata or 'mall', (not chain-mail), is the most flexible armour of the three available. The form of mail is very simple being in form like a T-shirt except with extra protection at the shoulders. When worn the mall tends to cling to the upper torso and hangs below this point. It is, however, possible to reduce some of this weight by the use of the military belt, which if clinched tight enough will support some of the weight at the waist, rather than just on the upper torso. This is one of the advantages of mail, the weight can be supported at two points; unlike lorica segmentata which has to be supported completely by the shoulders. Although if a padded garment is worn under the armour it can be fitted tightly to the torso, supporting weight more evenly.

Padding

Bishop (l995: p.1-2) suggests that there is evidence, which may point to a padded garment being worn below all forms of armour, called a thoromachus. If this garment was worn in whatever form it took; it would certainly make the wearing of all these armours more comfortable but perhaps would have reduced the circulation of air about the body. This would represent, then, a compromise to accommodate the additional protection, as the padding would also reduce the effects, on the wearer, of a blow to the armour.

SHIELDS

Evidence for the reconstruction of the traditional playing card shape Legionary shield comes from collapsed remains of tower nineteen of the Roman defensive wall at Dura Europos. In a sap, or tunnel, below this tower were found the remains of several combatants from the armies of Rome and its enemies the Sassanid Persians. Although the finds from this tower date to the third century AD, there is evidence in the form of shield bosses from the first century AD and on Trajan's column that there was little alteration in the size of these shields and in the absence of more contemporary evidence it is therefore suggested that there would have been little change in the methods of construction between the first and third century.

The shield from Dura is constructed from thin laths of wood which are glued together over a mould, in three layers. The grain of the inner and outer layer run vertically and sandwiching between them the layer runs horizontally. When the glue has set, these laths can be removed from the mould and will retain the curve they have been formed over. A covering of linen or leather is then applied to the front and back, and reinforcing wood and metal strips are attached to the back of these. A painted design, individual to each legion according to literary accounts of the battle of Cremona, is then applied along with a copper-alloy edge binding and an iron or occasionally copper-alloy shield boss.

While the shields produced by the E.S.G. differ in their construction; three sheets of 1.5min birch plywood being substituted for the laths and modern glues used; some useful Information can be learnt both about their construction and use. Unless the plywood is firmly held in place and the glue is applied to all the areas of contact the plywood tends to 'bubble' apart which can, if the bubbling occurs at an edge, produce problems when the metal edging is attached at a later phase of construction. Staying with the gluing of the shields, it is necessary to ask what glue the Romans were using? If, as is probable, a natural (animal based) glue was used, what happens to glue of this type when it is exposed to rain or water, should be considered. It is probable that they would have lost their adhesive properties when wet and so if left long enough, would have resulted in the shield disintegrating. If this is the case it may help explain why Julius Caesar emphasises in his 'Gallic Wars' that in one encounter the Legionaries were caught by surprise and did not have time to remove their shield covers. The reason for their shield covers being on, it could be suggested, may have been to protect the shields from the wet as well as for protection against accidental damage. This argument may be supported by the evidence for water-proofed shield covers from Caerleon and Vindonissa (Bishop and Coulston, 1993,p. 47)

A construction peculiarity of shields is the metal binding which would have protected the shield edge from day to day wear and would have also helped to prevent damage to the shield during combat. Examples of these copper-alloy edging pieces indicate that this edging was attached to the shields by means of nails or rivets through lugs protruding from the edge of the binding strips. Although this binding edge is perfectly functional it raises the question as to why they went to the extra effort involved in shaping and cutting these binding strips and lugs in one piece, rather than producing binding strips and using additional loops of metal to retain the binding in place (fig.5). The only plausible reason possible for this is, perhaps, that this method was thought to be more satisfactory than loops holding the binding in place as it is possible that the loops could have been chopped through in combat. It could be that as with modem professional armies items of equipment tend to be over-engineered.

The writer's personal experience of using Legionary shields has shown that although it may, due to its curved shape, appear difficult to carry and use, this is not the case and in fact once one is experienced in its use the benefits of this type of shield for a unit operating in formation become evident. To carry the shield an overhand grip is used and the shield is held vertically at arm's length, as if one was carrying a shopping bag. Although this type of grip may appear awkward it is better than a underhand grip where the weight of the shield is continually trying to straighten the arm against the joint's natural articulation.

Using a overhand grip also has the benefit of making the shield more manoeuvrable, meaning that you can hold the shield away from the body; close in to the body; at arms length at shoulder height; above the head; in front of the body or braced against the shoulder with the top edge of the shield at eye level. Because of its curvature the shield wraps around the user providing a large degree of protection which increases when the carrier is part of a formation. Although as said, the shield is easy to carry and manoeuvre, after prolonged periods of carrying it, the arm begins to feel the weight, despite its lightweight construction.

If it was necessary to carry the shield for prolonged periods of time, such as on route marches, it is possible that some form of strap over the shoulder, would have had to be used. If this was the case this strap, or straps, may have been an integral part of the aforementioned leather shield cover. This probably would have resulted in the shield being carried relatively high upon the back so that the bottom edge of the shield did not interfere with the free movement of the legs. There may in fact be evidence for this from contemporary sculptures. On Trajan's column the legionaries crossing the bridge are shown carrying the marching pack high over the left shoulder, possibly because the marching pack pole was resting on the shield.

CINGULUM MILITARE

Although commonly referred to as a cingulum militare or cingulum militiae it is believed that the military belt of the first century AD was called a balteus, the term cingulum not being used until the third century AD (Bishop and Coulston, 1993, p.96). This belt consisted of a belt of leather with applied metal plates, (either cast or pressed); an 'apron' and two frogs for the hanging of a pugio (discussed below). Although finds of complete belts are rare, Velsen and Herculaneum providing an example each, there are a multitude of the individual metal components that made a balteus as evidenced by Grew and Griffith's 1991 article, 'The pre-Flavian military belt : the evidence from Britain' which only covers a limited period and region.

Sculptural evidence in the form of tombstones indicates that there was a gradual change in the first century AD from two belts, worn 'cowboy' fashion, each belt supporting a sidearm; to one belt supporting only the pugio, the gladius being worn on a baldric over the shoulder. The belt seems to have had three functions of varying significance. The belt not only provided suspension for the pugio but also (as will be discussed later) provided a means of securing the gladius scabbard at the side and supported the apron. The purpose of this apron, which it is thought developed from the loose strap end of the belt, is unknown, although several purposes have been proposed. Two of the more plausible suggestions are that it was designed to protect the lower stomach and groin region of the legionary, or that it was designed as a modesty device due to the relative shortness of the legionary tunic, when compared to civilian tunics which would tend to ride up when moving or sitting down Practical use of balteus has shown that although the apron could provide protection, this can be outweighed by the annoyance of the apron catching on the legs and other parts of the male anatomy, when running. The possibility that the apron did work as a modesty device cannot be ruled out as the weight of the metal fittings on the apron does tend to weigh the tunic down particularly when sitting. Alternatively the apron could be designed so that the metal fittings jangle as the wearer moved or walked and either made civilians aware of his presence, or acted to bolster his confidence.

CALIGAE

There are ample archaeological examples of Roman caligae from waterlogged sites such as Mainz and Valkenberg (Bishop and Coulston, 1993, p. 100). These examples are consistently rugged in their construction and provide ample information as to their component parts and methods of construction; including the stitching used and the arrangements used for hob-nail layout. With the help of Dr. C.van Driel-Murray a Dutch leather expert it has been possible to produce serviceable reconstruction's of Roman caligae based upon finds from Mainz and Valkenburg; where complete examples have been found and partial remains from water-logged sites such as Chesterholm and Nijmegen. A noticeable problem with these reconstructions is that they have a tendency to stretch at the heel; in some cases causing the stitching to either break, or become abraded. Interestingly the literary evidence states that legionaries were issued three pairs of sandals per year and although E. S.G. sandals are not used as intensively as the real examples, the above evidence seems to corroborate the necessity for frequent replacement. Alternatively this necessity could be due to the manner in which the Roman hob-nails are clasped over meaning that once worn away, which would be rapidly on stony surfaces, they cannot be replaced.

As mentioned these sandals are rugged and hardwearing and even though they are of an open work construction they provide ample protection for the foot while being free-draining. The hob-nails in the soles of the boots are also surprisingly good at providing a firm grip except on areas of flagstones, or marble where it is easy to slip over. This problem is recorded by Flavius Josephus, in his accounts of the Jewish Wars of 66-70AD with the demise of a centurion by the name of Julian of Bithynia who upon entering the fortress of Antonia slips over on its flagstone surface due to wearing the wrong sandals and is slain by the Jews (Josephus, Book VI, p.81). On the other hand, whilst partaking in some educational film-work portraying scenes of Romans fighting Celts, the E.S.G. found that on grass, the re-enactors portraying 'Celts' could easily be pushed back and were unable to gain a grip, with their boots lacking hob-nails. Although, reconstructed boots are similar to boots found, they do not correspond to those found in quantity upon Hadrian's Wall, at Chesterholm (Vindolanda). The boots found here and dating to the early second century are more similar to modem boots with enclosed sides and in some cases toes. As such these Chesterholm boots are different to boots found elsewhere in the Empire. It could be suggested that the creation of a frontier upon Hadrian's Wall in harsh surroundings led to the development of a new type of boot more suitable for keeping the foot dry and warm rather than a boot that was hard-wearing, 'free-draining', allowed free air-circulation and thus quick drying, which was perhaps more suitable as a 'campaign' boot. This supposition could perhaps be supported if these enclosed boots occurred across the less hospitable areas of the Empire. Even so, as these boots occur at the beginning of the second century they fall outside the period of study for the E.S.G. which is the later half of the first century AD.

OFFENSIVE EQUIPMENT

Whilst the items of equipment that we have discussed so far can be considered defensive or protective in nature, the offensive items of the Legionaries equipment, such as the pila, gladius and pugio must also be considered as they were a vital part of the Legionaries' function: the enforcement of the edicts of Rome.

Pilum

It seems sensible to start any consideration of the offensive weapons the legionary carried in the order used, or disposed of in battle; starting with the pilum. There is abundant archaeological and representational evidence for the forms that pila or throwing spears, took in the first century AD, as outlined by Bishop and Coulston (1993, p.65-68).

There are two main components to the pilum; the wooden shaft and the iron shank and head. The wooden shaft has a top pyramidal section into which the iron shank is fitted and a lower circular section of shaft by means of which the weapon is carried and thrown. As the wooden shaft and top section are made in one, it would be necessary to work the main body of the shaft down from the dimensions of the pyramidal section. The easiest way to achieve this would be by using coppiced wooden poles; possibly of ash, hazel or birch, which would provide straight poles of regular diameter that could be turned, or bodged, on a pole lathe, to provide the correct diameter. The metal component of the weapon consists of the pyramidal, hardened steel point; the shank, which is of tempered steel and the tang or plate that fits into or onto the wooden shaft. The method by which the metal section is fixed to the wooden shaft either involves the metal plate being placed into a slot in the shaft with two rivets used to fix it in place, or the tang being driven into the wooden shaft. In addition to these parts the pilum also had a metal butt on the bottom end of the wooden shaft to protect against wear.

The pilum is designed so that on impact the hardened steel head pierces what it hits and the metal shank then bends due to the weight of wooden shaft. Although primarily designed to kill, if the pilum penetrated flesh, a shield or armour; the shape of the head would make removal difficult and due to the weight make a shield so cumbersome that it would be abandoned. If, however the pilum struck a hard obstacle such as stone or the ground, the shank of the head would bend on impact and mean that the weapon would not be as effective if thrown back by the enemy. It has also been suggested that due to the shaft being of soft steel the pilum could be collected after use and the shank re-straightened by the Legion's blacksmiths, making it a reusable weapon.

Practical use of the weapon by the E. S.G. indicates that the best method of carrying it is back-handed with it resting against the shoulder as portrayed on sculptures such, as the Adamklissi monument and the Cancelleria Relief in Rome. The use of pilum, thrown for television productions has also indicated that the weapon does bend upon impact; the soft metal shank producing some most peculiar shapes.

Gladius

The short sword or gladius as commonly known (although not technically correct (Bishop and Coulston, 1993, p.69) has many variation of shape in the first century AD, for which there is ample archaeological evidence. The leaf shape Mainz and parallel sided Pompei type swords, are represented at Mainz, Bremen, Fulham (Mainz type) and Herculaneum, Whaddon Hill, Hod Hill (Pompei type) (Bishop and Coulston,1993,p.71). Little that is new can be said about an item about which so much is known, or about which there can be so little doubt apart from to say that it is a relatively light weapon which, it can be suggested, could be more easily handled for long periods of time when compared with the longer Celtic and Germanic swords of northern Europe. If, as suggested, the weapon was primarily for thrusting or stabbing while still of use for slashing when the opportunity arose, economy of movement and energy are achieved when compared to the longer swords mentioned which could realistically only be used for slashing.

While the notion that the gladius could not be drawn when worn upon the right side has been disproved, it is perhaps essential to note that for a gladius worn on a baldric (a strap over the shoulder) to be drawn it is necessary that the sword scabbard is held at the side; otherwise a snug fitting the scabbard lifts up with the gladius. However, while no evidence for the method by which the Romans achieved this exists, the E. S.G. adopts a method where the military belt passes through the baldric and thus holds the scabbard in place at the side. This problem does not arise though if the scabbard is supported upon the belt, rather than a baldric, as appears to have been the case in the early first century AD; the belt can also be made to support the gladius and scabbard in such a way that they slope forward, not only producing a better appearance but also making the gladius easier to draw and more importantly replace; a problem if the scabbard hangs vertically rather than at an angle.

The gladius is relatively broad in relation to its length, making it relatively heavy for its size and it is perhaps, for this reason that such a large pommel was used to counter-balance the weight of the blade and improve the balance of the weapon. Even with the counter-balance the sword naturally adopts a point down position in the relaxed hand, although it is no uncomfortable to use. Another reason for the size of the pommel is to prevent the sword being twisted or pushed from the hand; when it is thrust with.

Pugio

I.R. Scott in his article 'First-century military daggers and the manufacture and supply of weapons for the Roman army: Decorated dagger scabbards found in Britain' (1985) excellently demonstrates the evidence for pugiones and their decorated sheaths for the first century AD but does not cover the issue of sheaths that are of open metal work, such as the example from Exeter, and undecorated sheaths. Even when these omissions are taken into account, his work amply demonstrates the considerable evidence available for reproduction and reconstruction of first century daggers and their scabbards whether related or not.

Despite Scott's evidence (1985, p. 165) that there are two types of sheath: 'A' where a metal sheath with lining has a design applied to the front and 'B' where a leather covered wooden sheath has decorated metal plates applied to it, it can be suggested that there is another type which can be included as demonstrated by the aforementioned Exeter find. This sheath appears to have been constructed using guttering joined at the point of the sheath with a ferrule and at the top and approximately one third of the way down the sheath with bands of metal. Although no evidence exists for a lining to this type of sheath it is probable that a construction similar that used in gladius construction is used. Here the metal binding encloses a wooden lining that is covered with leather: this contrast between metal and presumably dyed leather provides a decorative effect.

Although a failing of the E. S.G. is that it at present only utilises the evidence for type 'A' sheaths, some evidence for use of these sheaths exists. The pugio and sheath which is worn upon the left side is suspended from the belt by means of two 'frogs'. The E.S.G. hangs the sheaths from these frogs using pieces of leather that are figure of eight in shape. One end of the piece of leather is placed over the frog, the other end being passed through a suspension loop on the sheath and then secured over the frog. This process is carried out on both sides of the sheath. It is, however, useful if the rear of the two leathers is shorter so that the sheath hangs in such a way that the handle of the pugio points forwards. A problem with some of the pugio and sheaths made by the E.S.G. is that the pugio on occasion falls out. Although this information could be used to suggest that Roman infantry had this problem, it is perhaps more likely a failing of interpretation of the evidence, or lack of care in the manufacture of replicas.

While the issue of whether the pugio is a second cutting weapon, or a general purpose tool, for the legionary is still unresolved, it is worth noting that the sculptural evidence for cavalrymen, which is of most use for their study, indicates that they do not wear the pugio. This may be as a result of the tendency to depict cavalrymen facing the right and consequently hiding their left side where the pugio would be worn, but if it was the case that cavalrymen did not wear pugio we must ask why they did not while legionaries and auxiliaries did. A possible reason, as suggested by the E.S.G.'s reconstruction of the equipment of two cavalrymen and their associated equipment, is that due to the cavalrymen already wearing a gladius scabbard on the right side the addition of a pugio sheath upon the left would make mounting and dismounting difficult. This could especially be the case if we consider what use a pugio would be to a cavalryman as a weapon. As is well known, the cavalry used a sword or spatha that was longer than the infantry gladius to facilitate the extra reach they needed, due to being mounted; with the pugio being such a short weapon, having a blade of between 143 and 280 min as against a blade in excess of 620min for the spatha it would be almost useless to a mounted person and therefore unnecessary. Nevertheless, epigraphic evidence does record that Caecilius Secundus, a cavalryman used an inlaid dagger sheath as collateral for a loan. With this in mind, then, it could be argued that it is quite plausible that both the pugio and gladius, being not too dissimilar in size (between 143min and 280min, and 358min and 493 min respectively) were both of use to the legionary and auxiliary in combat.

Thus it can be seen that the construction and use of reconstructions can give a valuable insight into the problems which a Roman Legionary may have had to contend with and the practicalities of equipment manufacture. Also, such reconstructive experiments can produce possible solutions to questions left unanswered by representational, archaeological and literary sources as well as posing questions of its own. Perhaps the greatest benefit of the work of re-enactment groups such as the Ermine Street Guard, is that the pieces of equipment which they produce are not seen in isolation but as part of the larger picture of a complete legionary kit and that therefore the interaction of pieces of equipment, arms and armour, can be studied and analysed.

CONCLUSION

In conclusion, there is still a great deal which is unknown about the armour worn and equipment used by the Roman legionary. However re-enactment groups such as the Ermine Street Guard have gone a long way to increasing our understanding of this subject and their work should be appreciated for its academic and educational merits as well as its entertainment value. It has been shown that reconstructions can provide valid information on the use of the equipment and armour; how pieces interact to form a complete kit and the processes of wear and breakage. Nevertheless, it is essential to see such reconstructions within the wider context of not only their own limitations as experimental archaeology but also within the limitations of the evidence upon which they are based and ultimately, the biases caused by our own twentieth century prejudices, preconceptions and technology. As a result of an increase in the available primary evidence upon which reconstructions are based (representational, archaeological and literary) and the study of long-term problems, which arise with armour and equipment, the future should bring a refinement in the reconstructions made and thus the information gained from them. Furthermore, a closer working relationship between re-enactment groups and academics will result in greater understanding of the artefacts and of the reconstructions based on them. It is only through constructive criticism of reconstructions and the acceptance of this criticism by individuals and groups, that attempts at reconstruction can result in anything near a correct solution. It could be said then, that reconstructive archaeology is beginning to play an ever greater and valid role within the field of Roman military studies; representing the public face of the academic research.

ACKNOWLEDGEMIENTS

1 would like to thank all those who have helped in the production of this dissertation. Gail Renwick for taking photographs; Andrea Cox, a most patient girlfriend for advising on and assembling the figures and proof reading the finished product; Janet Cox for printing numerous rough drafts and most importantly my father Chris Haines for clarifying innumerable points of detail.

BIBLIOGRAPHY

Bishop, M.C. 1995, Aketon. Thoromachus. and

Lorica Segmentata in Exercitus : The Bulletin of the Ermine Street Guard (Ed) Mayes, B. Haines, C., Shaw, T.

Bishop, M.C.& Coulston, J.C.N. 1993, Roman Military E9A12ment London : Batsford

Coles, J. 1979, Experimental Archaeology London: Academic Press

Coles, J. 1973, Archaeology by Experiment London : Hutchinson University Library

Grew, F. and Griffiths, N. 199 1, The Pre-Flavian military belt: the evidence from Britain Archaeologia 109, p.47-84

Thackeray, H. St. J. 1927 (Translated), Josephus , The Jewish War: Books IV-VII Cambridge, Massachusetts : Harvard University Press

Junkelmann, M. 1986, Die Legionen des Augustus Mainz

Robinson, H. 1975, The Armour of Imperial Rome London: Arms and Armour Press

Scott, 1. 1985, First Century Military Daggers and the Manufacture and Supply of Weapons for the Roman Army in The Production and Distribution of Roman Military Equipment. (Ed.)M.C. Bishop BAR International Series 275

Zeinkiewicz,1). 1994, Roman Legion Wales: National Museum of Wales.