The seminar on which this volume is based, and the chapters in this section of the volume provide considerable food for thought. I apologise if the following text is too self-referential, but my own engagement with Roman artefacts goes back a long way, to a period when I was first captivated by the archaeology of Roman Britain, and reflecting on my past work seems to serve a purpose in drawing out some points for discussion in the present context. In common with many others, my interest was stimulated first by excavating objects, and this curiosity developed into a desire to study and understand what I was finding. Only later did my work on finds and their excavated contexts lead me to wish to better understand the society that produced and used the objects, with the eventual product being an urge to make sense of Roman Britain and the Roman empire more widely. Hence, I return to the subject of artefacts now from the ‘bottom-up’ perspective of one whose archaeological career has always sought to engage with material evidence.
In the context of my own academic biography, it is worth noting that I focused originally on pottery, in part because no-one else in my local archaeological society in the early 1970s wanted to study it, instead preferring coins or small finds. But my interest was also increasingly caught because pottery as a finds class appealed to the then fashionable desire to have decent amounts of material to quantify in order to make apparently soundly-based statements that could be supported by numerical analysis. The first serious work I did used automatic seriation to establish the chronology of a pottery industry in an area where there was insufficient stratigraphic evidence to provide a time frame (Millett and Graham 1986, 63–93). Having sorted out the dating, I became interested in tracing the distribution of the pots produced by that local industry across southern Britain, hence thinking about the relationship of the industry to trade and the broader economy (Millett 1979b). I use this autobiographical example to make the point that although in the schema explored by the editors of this volume, my work at that time was partly using artefacts in a representational mode, this was the product of a progression in thinking based on the problems posed by first site, then material. I came to be influenced by approaches promoted at that time by progressive archaeologists like Ian Hodder, and by the development of computer-based methods of study. Using pots to say things about society seemed like a good idea at the time, especially when contrasted with the more conventional dry typological approaches that had treated such objects simply as materials for dating (Hodder 1979; Millett 1979a; Spriggs and Miller 1979). As this work progressed I gradually came to question the automatic association of pottery with ‘the economy’ and thought I could detect ways in which it represented social choices. Thus, for instance, lying at the margins of the distribution of both Oxfordshire and New Forest finewares, those living at the site at Neatham seem to have carefully selected particular products from each industry, to produce a complementary assemblage (Millett and Graham 1986, 90).
My career with finds then took a turn – which immediately afterwards appeared as a blind alley – but in retrospect I find to be of greater interest. As I worked on ceramic distributions across southern England, and as I was engaged in excavation, it became increasingly clear that variations in the composition of pottery assemblages (and incidentally, ‘the assemblage’ was then not that often the unit of study) were not simply a function of chronological differences, but reflected other factors. Hence, I came to argue, statistical analysis of variation between contemporaneous assemblages might provide a route for exploring social, economic and other variations within Romano-British society – provided one could control chronological variation (Millett 1979a). For my doctoral thesis, I thus examined all the known deposits supposedly associated with the Boudiccan destruction of AD 60/61 (Millett 1983a; 1987a). Partly as a result of the limitations of the computing methods available at the time, and partly because of a limiting empirical methodology, the exercise proved rather more barren than I had hoped and my career moved in other directions. It will be clear that this approach was also largely based on using pottery representationally, but perhaps in quite subtle ways, and significantly for me, seeking to access social information.
As my academic work has developed I have returned time and again to seek to use artefacts from both excavations and surveys in meaningful ways to try to tackle some big issues, like the relationship between survey data and ancient populations levels (Millett 1991), and in doing this I guess I am generally guilty of the ‘sin’ of using the material to represent other things without being sufficiently reflexive. In doing this I have at times moved away from pottery (e.g. Millett 2015), but I continually face the problem that it is more difficult to make general statements about patterns in the past with small data sets than with larger ones. My undergraduate training in the analysis of coinage clearly shows the influence of Richard Reece, who has done so much to encourage thinking about coin assemblages (e.g. Reece 1987). His influence can also be traced in some of the ways I have sought to approach other artefacts, including pottery, but most recently brooches and other small finds (Millett 2015). The tension between the study of finds individually or as groups/assemblages raises a series of questions to which I return below.
Now, my reason for opening this discussion in this autobiographical manner is not wholly self-indulgent: I hope that this brief contribution may help contextualise the fashion for thinking about artefacts as representational, even though I accept the editors’ critique of this trend. It is also worth highlighting that in reading the papers here I am reminded of some of the issues that I came across in my work on pottery, and the perspective that this brings may be helpful in current debates. Not least is the fact that the archaeological study of any artefacts for anything other than a simple descriptive catalogue is an extremely challenging business. Aside from the inherent complexity of the material – especially when studied contextually – there is also a major leap involved in linking the data with the sorts of big issues about past societies that most of us aspire to address. In the short space available to me here I would like to reflect on a few of these complexities stimulated by points in the papers in the section.
One of the concerns I frequently return to in thinking about archaeological material is the false security created by the familiarity of Roman material culture. This issue is one that I contemplated again in reading Hella Eckardt’s contribution to the volume. At a very superficial level, this is manifested by our apparent ease in identifying and hence labelling the objects we study, for instance cooking pots or drinking vessels. A lack of awareness that this is a potentially problematic area can weaken our arguments, as can ignoring the much more varied real-life uses of objects made for one purpose but used for another. I am reminded here of an odd group of burials from Roman Gaul in which some of those interred wore upturned mortaria as hats (Schneider 1965; de Boüard 1966). Such occasional aberrant (?) uses can be identified as such because they stand out from the general pattern of use that can be seen from significantly sized samples of material. The careful contextual study of large samples may equally reveal unexpected and widespread patterns of changed use – such as Jerry Evans’ demonstration that many mortaria in northern England were apparently used for cooking on an open fire (Evans 2016, 518). Naturally, any recurrent pattern of alternate use will call into question the use of artefact evidence in a representational manner – for instance linking the distribution of mortaria to the adoption of Roman-style cuisine. The identification of such patterns of use is much more difficult for less widely distributed types, like the inkwells discussed by Eckardt – especially in the absence of obvious evidence of use.
So I am cautious about questions like ‘when is an object an ink-well?’ or ‘how far can objects of very specific design-function be used to identify specialised knowledge in the ancient world?’ This seems to me to be a problem that is especially difficult when studying thinly spread distributions of scarce objects – like metal inkwells, whose inherent qualities may have made them desirable for a variety of reasons quite distinct from showing knowledge of the technology of writing, unless there are contextual reasons for believing that they were widely used for their designed function (always assuming that they were all made as ink-wells). In a similar way, when one deals with more common object types – like the belt sets and crossbow brooches discussed by Rob Collins, I worry that in the absence of contextual information at the margins of a distribution, the specificities of meanings conveyed by objects may have been both more mutable and more subject to ambiguity of understanding. In the specific case of the late Roman world discussed by Rob Collins the distinction between the military and civilian officials was arguably far less distinct than might be assumed. In this context objects like the crossbow brooch may have become more generalised symbols of power or authority, so it is very easy to see how their use may have spread beyond just those representing the Roman state. Although such objects thus arguably represent social power, it is not easy to relate them to very specific historical manifestations of it – like the army. Equally, their scarcity leaves them prone to becoming special objects, used for instance as heirlooms, representative of the ancestor rather than the office.
Against this background, it is interesting to note how both Rob Collins and Hella Eckardt turn to the funerary evidence in seeking to understand their material. There is no doubt that the very deliberate construction of funerary assemblages provides an accessible way into appreciating the use of objects in the Roman past, even allowing for the fact that their layout and content are not as straightforwardly representational as has traditionally been assumed. In these circumstances, it is possible to gain valuable contextual information but this is evidence that it is rather difficult to generalise from. This leads to the paradox that there may be excellent evidence for a single case, but little clarity about how this relates to others. In the context of contemporary trends in archaeology this may be a good outcome, but for me it leaves a certain level of dissatisfaction because of my wish to think about bigger issues – the lack of representativeness of the individual example seems a problem for me.
This seems to be recognised as a problem by other scholars too, hence the trend towards broader contextual work in finds studies over the last few years – a period during which there has been something of a renaissance in the sub-discipline (e.g. Hingley and Willis 2007; Eckardt 2014). Such work has commonly adopted two complementary approaches, both of which seek to overcome the issues of low-density distributions and often low grade contextual data: first, the classification of finds into broad ‘functional’ categories as initially suggested by Nina Crummy (1983, 5–6) – so, inkwells as ‘objects used for or associated with written communication’, crossbow brooches as ‘items of personal adornment or dress’), and second, looking at occurrence by site type (town, fort, villa etc.). These methods have certainly led to exciting new information, and the utility of this kind of comparative analysis of aggregated data is well illustrated in Martin Pitts’ paper in this section, which discusses how certain ceramic types have different patterns of deposition in various regions and on certain types of site.
Despite the success of such studies I have some concerns about them, both at a theoretical level and because of the way in which methods are sometimes applied. At the theoretical level, I am worried by the unproblematised use of simple exclusive categories for classifying both finds and sites. In thinking about the categories used for objects, it is obvious to anyone who has attempted to apply Nina Crummy’s categories that attribution is problematic in at least two ways. First, do we categorise by supposed designed-use, or can we use broader contextual information about actual function? Second, many objects have multiple functions even in designed-use. So, in the case of the late Roman soldier’s belt buckle from a grave, the object might legitimately be categorised as a ‘personal adornment or dress’, or an item of ‘military equipment’ (using Crummy’s categories), whilst its context shows that it was also a grave-good – and if it has an image of a deity on it, it also has a religious aspect. Given that most analyses of objects using this classificatory system label objects as representing single functions, they are evidently over-simplifying the evidence. This is not to say that Crummy’s approach is wrong – it is immensely better than previous systems that classified finds in publications by material (the ‘Objects of Iron’, etc.). It does however suggest that more sophisticated – multivariate – approaches to artefact classification might be more useful. Similarly, the system used for the classification of sites is also an over simplification, both in terms of categorisation and scale. There is immense value in distinguishing site types, as Eckardt’s study of lamps (2002) or Pitts’ work on pottery (2014) both demonstrate, but there are enormous pitfalls if we are not very careful. Richborough starts out as a fort, but develops into a town (? small town) with a port, before resuming a parallel function as fort – but its finds will almost invariably be dealt with as coming from a military site. Furthermore, it is interesting how such studies tend to privilege the more Roman site types (towns, forts, etc.) whilst categories like ‘rural settlement’ are left undifferentiated, thus effectively undervaluing the lives of perhaps 90% of the population. Such categorisations also tend towards positing materials as representative of certain site-types, without allowing sufficiently for individual human agency.
These types of analysis raise the further problem of smaller-scale variation. The process of aggregating data to provide assemblages for study is clearly useful in allowing us to make more robust statements. So, the careful analysis by Martin Pitts allows reasonably strong statements to be made about the kinds of sites and communities that used different types of pots in the early years after the Roman conquest of Britain (Pitts 2014). But this evidence is not without its problems – to take two examples, evidence of absence and contextual variation. Aggregation of data works well where there are finds, but the evaluation of absence is problematic, and gaps are prone to be overlooked. For instance, in the introduction to his chapter in this volume Pitts suggests that over much of southern England terra sigillata was initially generally related to areas of military supply, whilst away from these, people were ‘far more likely to encounter standardised vessels in the guise of so-called Gallo-Belgic wares’. I am happy to support this as a general statement based on the aggregated data, but this fails to acknowledge large gaps in the distribution which may or may not be significant. My own work long ago on sites in northeast Hampshire in the environs of Silchester (Millett 1983b; 1986; Millett and Russell 1984) showed that Gallo-Belgic wares were largely absent from most rural sites (as was early terra sigillata) implying some mechanism that led to its rejection prior to the mid-late first century AD when forms of this style began to be produced in local fabrics, albeit platters and bowls, not beakers. I do not have an explanation for this absence, although it does seem real, but I think it makes a more general point – aggregating data allows for sound (statistical) generalisations but tends to suppress patterns of absence. This is important because very interesting information often lies in the patterning of deviations from any norm as much as in the norm itself – and more generally in archaeology, as elsewhere, statistical generalisations, although robust, suppress information about (potentially interesting) variation between samples.
This brings me to another point raised by Pitts’ paper, returning to my own D.Phil. thesis work and concerning intra-site variations. It is interesting to see how Pitts’ work on Gallo-Belgic wares has picked out some of the same variations between the Colchester assemblages (at Sheepen and the colonia) as emerged in my own work. But I had also noted some other subtle variations for instance between the Balkerne Lane sites just outside the colonia and those within (Millett 1983a). This highlights the fact that identifying intra-site patterns depends to some extent on the size and definition of the sub-groups studied – and it also brings into question any simple categorisation into ‘civilian’, ‘military’, and ‘official’. This is further underlined by the pattern I noted at Verulamium where there are significant variations between contemporaneous assemblages. Within the settlement, terra sigillata was in regular use although Gallo-Belgic wares were comparatively rare, but by contrast Gallo-Belgic vessels were common in the King Harry Lane cemetery whilst terra sigillata was generally excluded from the graves (Millett 1993, 270–5). We might debate the significance of these variations, but their occurrence underlines the problem of using aggregated data. I think it may be time to start further work exploring intra-site patterning for two reasons. First, to see how finds analysis might contribute more to the understanding of settlements. Second, because unless we better understand the extent of normal variations from regional patterns of distribution we will not be able to evaluate whether patterning can be linked to factors like the character of the population rather than being a result of stochastic processes.
Finally, and returning to gaps in distribution patterns, I note Pitts’ comments about the supply of Gallo-Belgic wares in civilian areas. This raises for me another issue that I have wondered about on a number of occasions. I think that many of our presumptions about archaeological distributions in Roman provinces like Britain are unduly influenced by modern western experience where whatever we require is readily available if we have the resources to obtain it. In much of the Roman world, this was perhaps not the situation, and objects were probably available less frequently and less evenly: a situation that brings to mind my experience of East Berlin in the 1970s where commodities were available in very unpredictable ways. I suggest that we need to be ready to rethink some of our assumptions and conclusions in terms of the very imperfect market. To take a final example, returning again to mortaria, I would draw attention to a small but well-excavated site at Thorpe Thewles in the Tees Valley of northern England. Here the Roman pottery assemblage was very small, but largely dominated by mortaria – suggesting to me that their presence was perhaps partly a product of periodic availability, or a glut in supply, without denying the agency of those who chose to acquire them (Millett 1987b). Whatever the explanation, this surely implies that we cannot think of the region as one where the pattern of distribution is subject to any neat categorisations, and this underlines an important general point. The investigation of material culture patterning in the Roman world remains a fascinating challenge. In addressing this challenge, we need to be aware of the sorts of complexity touched upon here as revealed by earlier research, but we should equally aspire to think about the problems and materials in new ways, like those explored by Van Oyen and Pitts in the introduction to this volume. Only by thinking creatively in different ways will the subject continue to flourish.
* Faculty of Classics, University of Cambridge.