1This paper takes up anew the subject of a paper presented at the 16^th Limes Congress in Kerkrade, Netherlands in 1995 on coin supply and the Roman army, looking at how coinage was actually supplied to soldiers from the Augustan to the Antonine periods1. Here the topic is revisited, looking at new evidence and a number of ideas proposed by other scholars in recent years, putting them into a broader perspective of imperial financial, administrative and strategic policy.

2In particular two periods will be considered in more detail: the reigns of Domitian and Septimius Severus: the latter because Severus undertook a number of interrelated reforms that directly involved military pay; Domitian since on the one hand his coinage is well dated thanks to the numerous consulates he held so that we can examine it in chronological detail, on the other because his reign saw a strategic reorientation of Rome’s military focus on the northern frontier. The central questions that I wish to consider are: to what extent can we observe a link between the army and the supply of coinage? And what were the mechanisms by which coin reached the troops?

3While we have a good understanding of military pay scales in the first and second centuries AD, we have almost no direct, written evidence of how the mechanisms of military pay functioned. We know practically nothing of how soldiers were actually supplied with coins: did they receive their entire pay – after deductions – on a regular basis in actual coins? Were they always paid in new coins, or did they receive it in old coins that were already in circulation? What sort of coin were they paid in? In silver or in bronze, or in a mixture? (Gold can surely be excluded from the discussion as the average soldier will rarely have had access to or use for it2).

4Given the almost complete lack of written evidence that might answer these questions, we have to turn to another source of information, and this means the coins themselves, or more exactly coin finds. They are a particularly potent source, for not only are they extremely numerous for the Roman period. Thanks to their rather unique character coin finds can also tell us a great deal about how and why coins were produced, supplied and used, for coins are quite unlike most other groups of archaeological material such as pottery or fibulae. Coin finds reflect mechanisms at different levels: on the one hand coins are an official product of the central authority, produced and distributed in order to meet the needs of the emperor and the imperial administration; they bore the emperor’s portrait (or that of a member of his family) and purveyed messages that were related to imperial policy and ideology. However, once they had been issued and entered circulation, coins passed out of the purely official into the public domain, becoming objects of everyday use much like pottery and fibulae. To be sure, there were official laws and regulations that could govern their use, for to some extent they remained an official medium, and they might return to the official domain in the form of taxes, etc., but generally speaking they were now subject to the dynamics of the public sphere.

5This double role means that coin finds are able to provide information at various levels in a way that other archaeological material cannot, and when analysing coin finds it is important to always be aware of this, to understand which level it is that a particular piece of evidence reflects. Another stage in the life of a coin must also be taken into account, one beyond production/issue and everyday use, and that is deposition. For a coin-find is essentially the result of the final moment in the life of a coin, the moment it enters the ground and is immobilised in the archaeological record. Thus, effectively, it is in fact from the moment of “death” that we try to extrapolate information about the life of a coin.

Coin production and supply

6So what do we know about the first phase of the life of a coin, its production, and of what relevance is this for the question addressed here? An important factor of production in the late-first and second centuries were the centralisation of minting of the imperial coinage in one mint in Rome. In the East there were a large number of municipal mints that produced bronze, what is generally known as Roman provincial coinage, but Rome was the sole supplier of coin to the West. During much of the first century AD a mint at Lyons had played a major role in supplying the Northwest with precious metal and bronze, but the mint had been closed by Vespasian in AD 78. Galba had also opened a mint in Spain when he rebelled against Nero in AD 68, but Vespasian struck no more bronze there after AD 70 so that after his death Rome was the only imperial mint until the reign of Septimius Severus. His rival for power, Pescennius Niger, having no access to coin from Rome, opened a mint in Antioch, and Severus continued to coin there after Niger’s downfall, heralding in an age of decentralisation that saw a proliferation of mints from the mid-third century.

7This means that if military pay in the second century took the form of new coin, then it had to be supplied and shipped from Rome. Certainly there were good reasons why an emperor would have wanted to use new coin. It bore his portrait or those of his family, rather than those of predecessors, many of whom will have fallen into disfavour, and the reverses proclaimed important aspects of current imperial ideology. This will have been an important factor in retaining the loyalty of the army. On the other hand the logistics involved in the transport of coin from Rome might have been an incentive not to use new coin, but instead wherever possible to make use of coin already in circulation in the provinces that had come back to the treasury in the form of taxes, etc. Although given that taxes will have been paid predominantly in precious metal, this solution will only have been possible for silver. But it is worth noting that the logistics involved in the centralised supply of coin will not have been insurmountable for the imperial administration. For example, it has been suggested that the amount of new bronze coin entering circulation in Britain in the second century AD was some seven tonnes, a freight that could have been carried by just one ship, although due to the risk of loss involved it seems likely that the freight will have been spread between several loads3. For a power that was used to shipping grain to feed the inhabitants of the imperial capital across the Mediterranean in an enormous fleet this will not have posed an insoluble problem.

8A further important factor that impinges on coin supply is the fact that coin production was not regular, particularly in the early imperial period. For example, Augustus struck no bronze coinage between 3 BC and AD 9, while Claudius seems to have suspended the production of bronze soon after he became emperor, and it was only resumed by Nero in AD 62, nearly 20 years later. To be sure, from the Flavian period, in the final third of the first century AD, output became more regular, and this continued to be the case through the second century, but nevertheless output fluctuated, and in some years, for example in AD 83, could be suspended completely.

9It is also clear that coins were not supplied regularly to all provinces, so that a given province or region did not receive a regular amount of new coin each year. Instead the provinces seem to have been supplied intermittently, with the production of a particular year going out to only certain provinces. This was already recognised by Andrew Hobley in his work on the distribution of individual bronze coin types in the West in the second century AD4. Fleur Kemmers has investigated the distribution of reverse types on the bronze coinage in the second half of the first century AD in the Northwest and shown that there are regional differences in the representation of individual types, confirming Hobley’s findings. Kemmers concludes that these variations can only be accounted for by a conscious distribution policy on the part of the imperial administration that saw individual provinces supplied on a cyclical basis5. However, it is still not clear exactly what mechanisms were behind this rotation. Did the mint or imperial treasury take a central decision about where a year’s production was to be sent, or did the provincial treasuries put in a demand for new coin whenever their stock ran low?

Coin finds and the money supply

10While scholars such as Duncan-Jones have focused primarily on precious metal in their studies of the ancient economy6, the work of Hobley and Kemmers concentrated on the bronze because the vast majority of coin finds – apart from hoards, which are another matter – are bronze. Silver is much rarer, gold almost completely absent from finds, and as a result the precious metal does not provide enough material for a statistically valid study of the kind carried out by Hobley and Kemmers. It is for this reason that as long as we do not have sufficient numbers of precious metal coins finds, any study of coin supply based on coin finds will only be part of the full story.

11Hans-Markus von Kaenel has rightly drawn attention to the fact that this is not without its problems, arguing that concentrating on the bronze is to ignore the overwhelming proportion of money actually in circulation7. He points out that when the numbers of coins of the various metals are expressed in terms of value rather than as straight numbers, then the significance of bronze is dramatically reduced. For example, in Pompeii bronze is more common than precious metal (AV: 7666; AR: 10,168; AE: 21,797), but accounts for only 7 % of the actual face value of the coins8.

12Based on a comparison of the coins from the battlefield site at Kalkriese and the site finds from the contemporary fort at Haltern, von Kaenel also discusses the problem that site finds, which will consist to a great extent of casual losses, are themselves not a good reflection of the actual numbers of coins in circulation, due to differential loss of high and low value coins: higher value coins will have been looked for more intensively when lost, and so are rarer as finds, while bronze is correspondingly more common. Thus the numbers of the individual metals in the coin series at Haltern are: AV: 3; AR: 376; AE 16299, while at Kalkriese, where presumably there is no differential loss of high and low value coins, but during the disaster all metals were subject to the same loss mechanisms, bronze is much less common and the figures are: AV: 2; AR: 380; AE: 47310. von Kaenel concludes that the figures for Kalkriese will be a better reflection of the actual numbers of coins in circulation at the time than the numbers of coins found at a site such as Haltern.

13But does this mean that studies of the mechanisms of coin supply that rely on an analysis of bronze are necessarily invalid? Leaving aside gold, which will most likely not have played any part in the supply of coins to the average soldier, it is by no means certain that the ratio of the metals at Kalkriese represents the usual ratio actually present in the forts of the military zone. First and foremost, we cannot exclude the possibility that the soldiers who perished in the battle, and whose coins have been found, deliberately took more of their money with them in the form of higher value silver coins than was usual, as the latter were less bulky and the soldiers would have had less use for bronze while on active service than when stationed at their home base11. Bronze may actually have been appreciably more important in circulation than the numbers in Kalkriese suggest. But as there is no way of proving this, it remains a matter of speculation. However, there are other reasons for thinking that the role of bronze was greater than the monetary value it represents in coin series from catastrophe sites such as Pompeii or Kalkriese, for it is an over-simplification to take only the numbers of coins into account and to ignore the dynamics of coin use and coin finds. Coins do not end up in the ground simply because they were present in the ancient world. As a general rule, coins only get lost and enter the archaeological record when they are used, or else the container which they are held in gets lost12. People mostly lose coins when they take them out of their purse to pay for something: if they are sitting in a container in a cupboard they don’t enter the archaeological record (unless, of course, that cupboard was in Pompeii in November AD 79). This means that since low value coins get used more frequently – just as today 10 cent coins change hands more frequently than 10 euro notes13 – bronze coins will have had more of a chance of being lost and later recovered as coin finds. At first sight this would seem to add weight to the suggestion that bronze is over-represented in site finds; not only were bronze coins not looked for so intensively when dropped, they were more likely to be dropped. But the fact that bronze was more intensively used adds an additional factor to the equation, that of dynamics. Bronze coins will have been used more often than silver, silver more often than gold, and this reflects the fact that the role a coin plays in circulation is defined by more than just its value: how often it changed hands is just as important. Today money supply is defined as value of coinage in circulation x velocity of circulation. And since bronze will have had a higher velocity of circulation, so the part it played in the money supply will have been greater than pure numbers and face value suggest. As a result we must ask whether the numbers provided by a site like Haltern are not after all a more accurate reflection of the role of bronze coin than Kalkriese or Pompeii. The latter two sites are like snapshots, static reflections of a moment; while the finds from Haltern are more a result of the dynamics of everyday circulation, which must also be taken into account. To some extent this is reminiscent of Heisenberg’s principle, which states that you either know where an arrow is, or how fast it is travelling, but not both14.

14In the absence of figures for the velocity of coin circulation in the ancient word, we will never be able to quantify exactly the role bronze played. Even if this article has to rely on an analysis of bronze coins primarily because they are the best source and the numbers of silver coins is too low to allow statistically significant results, nevertheless there are good reasons to believe that bronze played a more significant part in coin circulation in the ancient world than is sometimes suggested, and that such an approach has an important contribution to make.

Bronze coin and the Roman army

15There has been much discussion in recent years over the role of bronze in military pay in the early imperial period. Reinhard Wolters has proposed that pay took the form of silver, whereas the current author draws attention to the much quicker turnover of bronze compared to silver in the military camps of the Rhine in the Augustan period, and takes this as evidence that bronze was used to pay the troops15. Johan van Heesch suggests that both metals were used, with bronze playing an important role alongside silver16.

16In a new study on finds of Roman coinage in Britain, Philippa Walton has noted that denarii are more common in the frontier zone on North England than in the civilian Southeast, and interprets this as confirming that the army was mainly supplied with silver and not with bronze17. However, this interpretation would require that the proportions of coins found today correspond closely to the proportions of coins as supplied to the military zone. Yet as was shown above, there were a number of mechanisms in play which make this unlikely: coin finds are a result not just of supply, but also of circulation dynamics, coin use and deposition. The question must be asked whether the higher proportion of silver in northern Britain is not in fact due to the fact that in the North, with its reliance on a pastoral economy and “native” settlement structures, there will have been less demand for low denomination coinage than in the more romanised and monetised Southeast with its urban centres and rich villa landscape. Did the North have any great need for small denomination coinage outside, or even in the forts? Is the low proportion of bronze there not the result of less intensive monetisation resulting in bronze coin being used less and so having less of a chance of being lost and entering the archaeological record? It is therefore debatable whether the figures presented by Walton are in fact evidence for military pay everywhere being supplied in silver.

17Recently Kemmers has looked in depth at the question of how military pay was organised in the late first century AD in her extensive study of the coins from the Flavian canabae legionis in Nijmegen18. In this work she identifies a remarkable example of a consignment of unusual bronze coins supplied to the garrison of the legionary fortress. The excavations at the site have produced no less than 304 quadrantes of the type RIC² 123 among the 825 coins of Domitian, that is some 37%, an otherwise unparalleled high level in the Northwest. No other site has anything like as many quadrantes, indeed in the late first century AD quadrantes were quite rare outside central Italy. Although the coins themselves are undated, the second edition of RIC has suggested a date of AD 81/82, and Kemmers links the arrival of the coins in Nijmegen to Domitian’s Chattan campaign of AD 83. She interprets the unusual consignment of the smallest Roman denomination, one that is otherwise hardly found on the Rhine, as a deliberate reaction to a widespread shortage of small denomination coins in the region19. However, given that the quadrans was normally a central Italian phenomenon, another possibility is that the coins arrived on the northwest frontier in Domitian’s coffers, and were perhaps originally intended to be paid out to Praetorian guardsmen who could have accompanied the emperor from Rome and who would have been well acquainted with them. But whatever the precise reason behind this injection of quadrantes into the Lower Rhine coin pool, whether or not it was the result of the presence of troops from Rome (and we should not forget that the campaigns against the Chatti were launched from Mainz, and we have no evidence that Domitian was ever actually in Nijmegen). The distribution of quadrantes of the type in North Gaul and Germany confirms that the coins entered circulation through a military site, the fortress of the legio X Gemina at Nijmegen/Noviomagus. At no other site in the region are more than 19 specimens known, and they are found almost exclusively at military sites along or near the Rhine20.

18In an article published in 1996, Markus Peter had presented other evidence for a clear link between the presence of Roman troops and the occurrence of bronze coins. He demonstrated that the withdrawal of the legio XI Claudia from Vindonissa in AD 101 led to a dramatic reduction on the amount of bronze coin being supplied to the entire region (fig. 1)21. Clearly the legion was the major factor in the supply of bronze to the area. What is more, whatever the mechanisms involved were, new coin must have played an important part, for otherwise the drop off in the coin series in AD 101 would not have been so dramatic. But what exactly were the mechanisms behind this? Does it mean that troops were actually receiving their pay, or part of it, directly in the form of bronze coin? Or was bronze being supplied to the area by other means, for example via money-changers? As Kemmers points out, the restricted distribution of the Domitianic quadrantes makes it unlikely that nummularii were involved, for if they were, then we would expect to find the quadrantes more evenly spread over a wider area22. All in all, the two examples by Peter and Kemmers would seem to confirm that, at least on occasions, new bronze coin was being supplied directly to the army as part of the soldiers’ pay.

Fig. 1. Proportions of pre- and post-AD 100/101 Trajanic bronze coins in Vindonissa and along the northern frontier (Peter 1996, Abb. 7).

19Can we perhaps find other examples for this? For example, if Peter demonstrated for Vindonissa in the early second century AD a connection between the withdrawal of troops and a reduction in the amount of new bronze coin reaching an area, can the phenomenon be observed elsewhere? Following the revolt of Saturninus in Germany in AD 89, Domitian reduced the garrison of the double legionary fortress in Mainz/Moguntiacum to just one legion. The legio XXI Rapax was transferred soon after the revolt to the Danube, where it apparently met its end in AD 9223, while the legio XIIII Gemina moved to Pannonia, possibly to Mursella, in AD 92/9324, to be replaced in Mainz by legio XXII Primigenia25. Is this reduction in the garrison visible in the coin finds from Mainz? At first sight it would seem so, for the bronze coin series in Mainz records a significant drop after 90/91 (fig. 2a)26. However, a comparison with other military sights on the Rhine where there was no such reduction in the garrison that we know of reveals a similar or even sharper drop at these sites (fig. 2b-d). What is visible at Mainz is not a local but a widespread phenomenon to be found elsewhere along the Rhine and Upper Dannube frontier: at the legionary fortresses at Nijmegen/Noviomagus, Neuss/Novaesium, Vindonissa and Carnuntum. In the case of Mainz the withdrawal of the legion did not have as significant an effect as is to be observed at Vindonissa ten years later. However, this is not surprising given that not only was Mainz the capital of the province of Upper Germany, but also even after the reduction in the legionary garrison there was still a very significant concentration of auxiliary troops on the area, along the limes beyond the Rhine, so that the demand for coin will have remained high. Vindonissa, on the other hand, lay further back from the frontier, there was no sizeable concentration of other military units in the nearer vicinity, and the garrison was withdrawn almost completely.

20But if we cannot identify a close correlation at a local level between the incidence of coins and the presence of troops, what the example of Mainz does demonstrate is that on a regional level there is a clear pattern in the bronze coins of the period along the entire Rhine frontier. Is there perhaps a more general, supra-regional explanation for what we can see? The reign of Domitian presents an excellent opportunity to look at the correlation between military activity and coinage in a broader context, for in the middle of his reign the northern frontier of the Roman Empire underwent a significant strategic realignment.

21The early years saw intense military activity in the Northwest. In Britain Agricola conducted a series of campaigns from AD 79 to 84 that saw the frontier of the province moved forward into southern and central Scotland. On the Rhine front, in AD 82 Domitian himself travelled to Gaul, ostensibly to conduct a census, but in fact to campaign in Germany, where in AD 83 he crossed the Middle Rhine and attacked the Chatti. He collected together three legions plus detachments in Mainz, as well as raising a new legion, legio I Minervia27.

Fig. 2a-j. Proportions of bronze coins of individual years struck during the reign of Domitian.

22In AD 85 the strategic focus then shifted. While there was a renewed campaign against the Chatti in this year, further east Rome faced a new challenge. On the Danube, which had been relatively peaceful, there now came the first dramatic signs of the intensive struggles that were to follow in the next 20 years. The governor of Moesia, Caius Oppius Sabinus, suffered a decisive defeat at the hands of the Dacians who had crossed the river and were plundering the province. Domitian reacted by going to the Lower Danube himself, accompanied by the Praetorian Guard, and ordering the transfer of several legions to the area. After initial success on the Danube in AD 85, Domitian returned to Rome, but the following year a Roman expedition into Dacia under the command of Cornelius Fuscus ended in disaster and resulted in the renewed presence of the emperor on the frontier. There followed several years of campaigns before Domitian and the Dacian king Decebalus concluded a peace treaty in AD 89, after which the emperor again returned to Rome. A further campaign, this time against the Iazyges, saw Domitian on the Danube once more in AD 92. Rome’s strategic focus in the North had shifted, culminating in Trajan’s conquest of Dacia, while the Rhine frontier remained quiet for over half a century.

23These events stretched Rome’s military resources to the limit. The army on the Danube had to be strengthened by the withdrawal of units from elsewhere, and in AD 86/87 one of the four legions stationed in Britain, legio II Adiutrix, was transferred to the Danube, presumably together with auxiliary units. At the same time the greater part of the newly annexed territory in Scotland was abandoned28. This is reflected directly in the coin series at Richborough and Bath in Britain, which peak in AD 86, before falling off after AD 87 (fig. 2e-f, fig. 3)29. With the reduction in the garrison of the province, and the shift in the strategic focus, bronze coins virtually stopped being supplied to the island.

24The Rhineland sites have a very different coin profile to Britain, but all share the same tendencies, if to differing degrees; a first peak in AD 82, followed by further peaks in AD 86-87 and 90-91. Later coin, as noted above, is rare. The peaks all coincide with military activity on the Rhine or its immediate aftermath: in AD 82 Domitian prepared for his first campaign against the Chatti (the actual campaign took place in AD 83 but Domitian struck no coins in that year); the second peak follows immediately on the second campaign against the Chatti in 85 when it is likely that forts were being constructed and troop dispositions reorganised along the new frontier on the right bank of the Rhine; and the third peak follows the revolt of Saturninus in AD 89. It is interesting that as was the case with Britain this last peak occurs at the point at which troops are being transferred from the province and not during the hostilities which preceded the transfers: in Britain after Agricola’s campaigns at the time of the withdrawal from Scotland; in Germany during the period of reorganisation after the revolt of Saturninus that culminated in the reduction of the legionary garrison in Mainz. As a result of the strategic reorientation which marked the end of 10 years of military activity on the Rhine, and saw Rome turn her attention to the Danube, the supply of new bronze coin to the Rhineland was significantly reduced.

Fig. 3. Bronze coinage of Domitian in Britain – distribution by date (Hobley 1989, fig. 2).

25Carnuntum on the Austrian Danube shares many similarities with the Rhineland, the coin series also thins out after AD 91 (fig. 2g). However, there are two significant differences: the peak in AD 82 that is connected with the first Chattan campaign on the Rhine is almost invisible, and instead there is a new peak in AD 85, one that is also apparent at two sites further East on the Danube, Solva and Brigetio (fig. 2h-i). It was in this year that the forces under the command of the governor of Moesia, Caius Oppius Sabinus, were annihilated by the Dacians, requiring the presence of the emperor himself on the Danube. Hostilities on the Middle and Lower Danube continued on and off until AD 92, when Domitian personally led a campaign against the Iazyges. Subsequently the frontier garrison was reinforced, possibly with a view to a renewed attack on the Dacians, a task which was to be left to Trajan. Once again the Danube coin series reflect military activity and a strategic focus. The peak in AD 85, which is not found anywhere in the Northwest, marks the beginning of two decades of almost incessant fighting; and in contrast to the Rhine frontier, which was quiet in the AD 90s, there is no drop in the frequency of bronze coins after AD 91. On the Danube the strategic situation was far from secure in the final years of Domitian’s reign.

26A series of coin finds from Rome (fig. 2j) provides an illuminating comparison and would seem to be a good indication of the actual output of the mint at Rome, a background against which the individual areas can be understood, for it reflects most of the peaks found in the various provinces: the first peak of (81/)82 found in Britain and on the Rhine; the high level of coin of 85 found on the Danube, and the peak of 90/91 from the Rhine30. Only the small peak in the years 86/87 on the Rhine is not visible.

27This comparison of the frequency of bronze coinage of the reign of Domitian from the three areas, Britain, the Rhine and the Danube, has demonstrated that there is a clear connection between military activity and the supply of bronze coin. Quite clearly the army and military pay were an extremely significant factor in the provision of bronze to the provinces. However, the correlation between military activity and the presence of coin is often a relatively general one. Peaks do not always coincide exactly with campaigns, for example the campaigns of Agricola in Britain are not reflected in the coin finds. Instead we often find peaks in periods of reorganisation following campaigns, for example in Britain in AD 86/87, or in Germany in AD 90/91. On the other hand, mid- or long-term focuses of military activity are clearly visible in the coin record, and the overall strategic shift from Britain and the Rhine to the Danube had a clear effect on coin supply.

28A final example of the important role of military supply in the overall mechanisms of coin supply and circulation was demonstrated by Holger Komnick in his study of the distribution of coins of Titus in North Gaul and Germany31. He showed that sestertii are more common in the military zone than in the civilian hinterland. While sestertii account for nearly 25% of Titus’ bronze coin along the limes in Germany, the figure for Luxemburg is less than 5%. Here asses are correspondingly more common contributing 80% of all bronze coins of Titus against just 55% on the limes. Two explanations are possible for this. Either a different spectrum of denominations was supplied to military and civilian areas. Or else the difference in the relative proportions of individual denominations is the result of preferential use of bronze coins after they had entered circulation; in other words the percentages found at military sites on the frontier reflect the percentages of the coinage as supplied, but when coin moved out of the military zone into the civilian milieu, coin users preferred the smaller denominations, in particular the as. In either scenario it is clear that bronze coin was entering circulation via the army.

29However, this is not to say that the military was the only or even the primary factor in the distribution of bronze coin. But it was clearly important. As we have seen, there will have been other factors which governed when or whether coin was supplied to a particular region or province. There were years when no bronze was struck at all, for example in AD 83, so that the first Chattan campaign is visible in coins of AD 82 instead. On other occasions the relevant provincial treasury may already have had enough coinage to cover its needs – whether in the form of recent supplies from the mint or in the form of older coin that had flowed back in tax to the treasury – so that a supply of new coinage will not have been necessary and no peak will be visible for the year in spite of high expenditure. Perhaps sometimes the mint or treasury may even have stockpiled older coin from over production and only sent it out several years later32.

30Given that there is good reason to believe that much of the coin reaching the army consisted of new bronze coin, the question remains as to whether the soldiers received it directly as pay. There are other models for how bronze may have been supplied, in particular the possibility that money-changers, nummularii, were involved. Under this scenario soldiers will have received their pay in silver, but changed this to bronze in the market in order to have the small change they needed for everyday expenditure. The bronze will have been transported from the mint or the treasury to the military camps by the nummularii. One attraction of this model is that it saves the imperial administration the expense and effort of transferring bulky bronze, instead transporting the same value of coinage to the army in the form of lighter silver.

31However, it is questionable whether this mechanism would produce the correlation between military activity and the frequency of new coin observed here. Kemmers drew the same conclusion on a smaller scale from the distribution of the Domitianic quadrantes found at Nijmegen, interpreting them as a deliberate delivery made directly to the troops. On a larger scale, the money-changer scenario would require a situation whereby the nummularii reacted extremely quickly to the changing strategic situation; that they were effectively attached to the garrisons they served, moving with them. Whether money-changers could be this flexible is open to question, given that they will have been subject to a number of other market forces other than the demands of military activity. It is also debateable whether the amount of new coin coming into the military zone would have corresponded so closely to military activity if it was the nummularii who were responsible for obtaining the new coin. They will normally have had regular access to older coin, receiving it as payment for the precious metal needed for paying tax, leading to a less variable demand for coin than peaks such as that in AD 86 in Britain or AD 85 in Carnuntum.

32On balance, it seems reasonable to assume that the central administration was responsible for the transport of coin to the frontier provinces, and that this coin generally included new bronze.

Silver under Septimius Severus

33The late-second century provides more incontrovertible evidence for the central role of bronze in military pay. In the second century AD bronze typically accounts for more that two thirds of the coins of the second century found along the northern frontier – in some cases, such as Mainz even more than 90 %. However, with the accession of Septimius Severus the situation was suddenly reversed, with silver making up 80 % or more of the coin (fig. 4). This sudden reduction in the supply of bronze and concentration on silver was just one of a whole range of fiscal measures introduced by Severus, measures which taken together suggest that the imperial finances were coming under increasing strain and that there was a shortage of metal, both copper and silver33: Severus increased military pay in AD 197, although we do not know by how much34; in AD 194 he significantly debased the silver denarius from 68% silver, as it had been under Commodus, to just 46 %35; and he ceased the payment of subsidies in silver to the Germans of the north European Barbaricum36. All of these measures revolved around one theme: military pay and securing the resources to guarantee it. By raising military pay, Severus was forced to debase the denarius and to restrict the official flow of silver out of the Empire. At the same time the debasement of the denarius meant more copper was required to mint the silver coinage, requiring the diversion of copper resources from the bronze to the silver production. As a result the bronze production had to be reduced and the supply to the Northwest was suspended, being replaced by silver instead.

34This was clearly a decision taken by the central imperial administration. It was not a decision on the part of the money-changes to stop supplying bronze to the soldiers. The demand for small change will still have been there, and further south in Central and Southern Gaul, the Iberian peninsular and Italy bronze was still entering circulation in significant amounts (fig. 4c: Conimbriga)37.

35The salient point is that if military pay was already in the form of only silver in the second century AD, then there is no reason why the amount of silver reaching the military zone in the Northwest should have risen so dramatically. For example, at Mainz there are only seven denarii from the 31 years 161-192, but 98 (!) from the 29 years 193-22238. This increase in the number of finds is significantly higher than the increase required to finance any increase in military pay, and the conclusion must be that in the second century a very significant proportion of military pay had been handed out in bronze. Silver may well have been involved too, indeed very probably was, but it will have played a subordinate role39.

Fig. 4. The proportions of silver and bronze coins at Mainz, Carnuntum and Conimbriga.

Conclusion

36The evidence of the coin finds provides several clear indications that there was a close link between military financing and the supply of bronze coinage in the late first and the second centuries AD. Not only can we identify cases where this link was immediate: in one situation bronze was being supplied directly to a military milieu, in another the cessation of the supply of bronze to a legionary fortress led to a dramatic drop in the amount of bronze arriving in the entire region. A more general correlation can also be observed between the supply of bronze to individual areas and the level of military activity there.

37Furthermore, there are also a number of indications that the supply of bronze to the military involved new rather than old coin.

38Even if this study neglected the precious metal coinage on the grounds that finds are not yet numerous enough to provide statistically reliable data, there are a number of reasons for asserting that bronze played a, if not the major role in the supply of coin to the army, and that silver was less important. In particular the massive increase in the amount of silver being supplied to the Northwest from the reign of Septimius Severus onwards is evidence for a thorough reorganisation of military finance as part of a larger package of changes; bronze was no longer supplied to the army there, and pay now clearly took the form of silver.