I've been reading some of Ian Morris' wide-ranging work, including his new book on war. It's got me thinking about a lot of things; one in particular is death in combat. If we wanted to know if Rome was doing better or worse at war, one ostensibly, simple way to do this would be to take a look at death rates among soldiers (let alone civilians) in combat over the course of Rome's documented history. So, if we're talking about using reliable accounts, that would mean, most likely, using everyone from Polybius to Theophylact (my cutoff for Rome being in the mid seventh century - and really my choice is just as arbitrary as anyone else's). Sure, there are narrative accounts for earlier periods, but given the questions pre-Polybius, not to mention issues with Dionysius and Diodorus. Best leave those out. Although that doesn't provide a complete list, it does give us some solid thick descriptions (here I'm borrowing terminology from Levithan's Roman Siege Warfare) of battle over the course of a considerable amount of time.
Ideally we'd like to use official records, but short of the occasional document like Hunt's Pridianum, or select Vindolanda tablets, we don't have what could be called official records of fatalities. And, that's not to say that those aforementioned documents are official themselves. And we're assuming that they had them to begin with. Given the general, and occasionally detailed, image we have of Roman military organization it seems unlikely to me that they didn't have some sort of record. Heck, even the list of names from the monument in Romania (not the Adamklissi one - name escapes me and this IS a blog post so I'm not going to check) points in that direction. But, as I say, we got nothing.
Archaeological data from sites of battle would be useful too, but we have almost nothing in this regard. The penchant for collecting and burying the dead, not to mention the pillaging of lost weapons by Romans and their foes alike has meant that we haven't found much of this, besides problems with tracking down the specific sites of individual battles. They did eventually find the site of Varus' last stand, and some remarkable stuff was recovered, but outside, say, the remains of Dura (and we're talking a siege here - and I have blurred the line so far), we haven't much to go on. In fact, even less of this than we have of the official record sort. What remains, then, are the accounts in the aforementioned historians themselves.
If we wanted some sort of control on our data, we might only consider those historians who fall within the classical or classicizing tradition, and so describe history - and for us war and battle - in a conscribed sort of way (speeches, digressions, etc.). This would mean using writers like Josephus (Jewish War), Herodian, and Ammianus, but not those like Josephus (Jewish Antiquities), Eusebius, and Count Marcellinus. We'd probably have to include Caesar too, even if he doesn't quite fit the mould, though he is similar. There are, unsurprisingly, problems with using a sample like this. Indeed, there are significant gaps during the period in question, not covered by these sorts of historians. Plus, even those that do survive describe battle in varying ways and in varying quantities. Their credentials too vary quite a lot.
Of course, it's clear that this attempt at control starts to falter quite soon too. There are too many gaps, many of which are filled by other historians working within different traditions. This does likely mean that they approach evidence and shape their narratives (if a narrative it is) in quite different ways - and hold different events and details of different worth than the classical and classicizing ones, though by the end of antiquity the lines had really started to blur. Still, it would probably be in our best interests to use as many different historians as possible, despite the glaring lack of consistency, and as along as we're sensitive to the vagaries of the various authors we should be ok...should be.
Now that we've established that we would use the various ancient historians (and anything literary that might be of value) and we feel "extremely" (read sarcastically) confident about that, we have to take a look at what sorts of figures they use. Some give precise figures ("612", some give vague figures ("no less than 1000", some use general phrases ("many died"), and some give widely fanciful ones ("70,000", or it could be read in some cases of "70 myriads" - which to me is less a real number than it is a vague generalization) - and some times the historians use all manner of figure. Of course, sometimes they give nothing, or only give it for one side only, and not the Roman side. If we look at chronicle (or chronograph - however you want to call them) accounts, sometimes we have little more than figures, and no sense of context. In fact, the closer you look the less ideal this evidence appears. None of this is really new, but to my mind it's still worth stressing.
So, what do we do? Do we discard this too? That would mean we're left with nothing - and there is no way to look at changing death rates amongst soldiers over the course of Rome's history. Sure, I could have mentioned inscriptions, but those are really only useful for a couple of centuries, and even then of the many dead soldiers recorded, few (so far as I know off the top of my head) come from combat itself. Rather, the epitaph was usually erected later in life. Even if it proved useful for at least some areas of the empire, it's no good to us for the republican period, or the late roman period. Ultimately, still nothing.
Maybe, however, this is the point. We don't have the sorts of data that we like, and the last thing we should try to do is to make it fit into any preconceived plans or theories. If we don't have the evidence, we don't have evidence, and we should leave it at that. Problem is, we do have some evidence, and as wretched as it might at a sweeping glance seem, I think it behoves us to use it. As long as we're all aware of what we're doing and what we're using, and as long as we don't make attempts to apply whatever results we have too widely, and we are careful and methodical as much as we can be when using it, then it is worth using. For, though it might not tell us give us the exact evidence we want, it's possible it might give us some sense of the rate of fatalities amongst Roman soldiers in combat over time. Then, if we used that in conjunction with other evidence (wins, losses, territorial gains, losses, etc.), we might go some way towards getting a better idea just how well the Roman military actually fared over time, particularly in late antiquity.