Metal production was a major preoccupation in the Bronze Age but it was by no means the only one, nor for many people the most important one. Commoner materials were at hand for daily use. Far and away the commonest archaeological material from the period is pottery, though this reflects survival as well as original abundance. On the other hand, the material used in most areas for building construction was wood, though usually only the negative traces of its former existence survive, in the form of post-holes. In areas where stone was available, it was also used, usually in rough form, but for items such as querns and grinders, or for the production of high-quality display items such as mace-heads or battle-axes, it could be finely worked and highly polished to bring out the veining in the material. Leather, bone and antler working were important in most areas, though few detailed studies are available.

Other crafts were also important, notably textile manufacture and glass production. The production and distribution of salt is a crucial (though often forgotten) aspect of life in all societies, and it was no less important in the prehistoric world than in today's. These crafts and industries have left traces of varying extent in the archaeological record.

Potting

In spite of its abundance, little is known about the context of production of Bronze Age pottery. It is not known who made pots (male or female, higher or lower status individuals) or in what situations they were fired, transported around communities, or ‘commoditised’ (see p. 404).

Introduction

Reading almost any archaeological document, from a research proposal to a fieldwork report, it will not be long before one comes across the word ‘sample’ in one context or another. A project, whether at regional or site level, will be based on a sampling strategy, a research design may specify that features are to be sampled, and samples of various types will be taken for the delight or otherwise of specialists who wish, or who can be persuaded, to look at them. In fact, almost all archaeology involves sampling; indeed, one could say that there is a sense in which much of archaeology is sampling, echoing David Clarke's remark that ‘Archaeology … is the discipline with the theory and practice for the recovery of unobservable hominid behaviour patterns from indirect traces in bad samples’. (Clarke 1973, 17).

The word sample is all-pervasive, but one soon comes to realise that it does not have the same meaning each time it occurs. This is not surprising, as just one of the six definitions given by the Oxford English Dictionary is broad enough to encompass a wide range of meaning: ‘a relatively small quantity of material, or an individual object, from which the quality of the mass, group or species, etc. which it represents may be inferred’. At one extreme, one may encounter a ‘multi-stage probabilistic sampling strategy’, while at the other, one may encounter the casual use of the word sample to refer to collections of muddy objects in plastic bags.

Introduction

Well, you have made it. Here we are at the final chapter, ready to review what we have learnt from this survey of the role of sampling in archaeological theory and practice, and to see how it might be applied in the future, both in the discipline as a whole and in individual professional careers. I hope that you have read at least some of the chapters that are outside your own immediate practical concerns, and are beginning to see how a knowledge of work undertaken at other scales can broaden your vision and expand your ideas. At the same time, it should be recognised that the scales defined here are arbitrary, devised for the purposes of structuring this book. In reality, of course, there is a continuum, and work at no one scale should be considered in isolation from the others. In this chapter I shall try to highlight what I consider to be the main opportunities provided for archaeologists both by recent (and some not-sorecent) developments in statistical sampling theory and by rapid developments in electronic technology, as well as some of the problems in archaeological practice that remain to be overcome.

First, though, I must pause to re-emphasise the centrality of sampling to both archaeological theory and practice. We saw in chapter 3 how sampling theory can make a useful contribution to debates about archaeological inference in the face of distorting factors imposed by site formation processes.

Introduction

Having looked at sampling at a wide range of scales, from the regional to the microscopic, we now turn to our final topic – the sampling of objects from museum stores. This may seem to be an unusual topic for a book on archaeology, but there are several good reasons for including it:

1. (a) since the museum store is the ultimate destination of most archaeological finds, we as archaeologists have a legitimate interest in their care there;

2. (b) the museum store makes an excellent vehicle for demonstrating some of the approaches described in chapter 2;

3. (c) the application of some approaches is particularly clear here, since some of the problems typically associated with sampling in the field may be avoided;

4. (d) techniques have been developed for repeated sampling (p. 201), which have potential for use at other scales, but have not yet been applied there;

5. (e) I have had a long personal involvement and interest in this topic.

Historical background

In the UK, statistical interest in the care of museum collections was stimulated by the publication of a report from the National Audit Office (1987–8), although of course the care of collections has always been a recognised duty of museums. The first survey intended to establish the overall condition of an entire collection seems to have been one undertaken by the Horniman Museum (London) in the late 1980s (Walker and Bacon 1987).

Introduction

In this chapter we move down a scale, from the region to the site, and look at problems that arise in sampling at this level, and some of the solutions that have been suggested. The problems are not simply those of the regional scale writ small (see Cherry et al. 1978, 151–7), although there is overlap, and indeed there is a continuum of scale from the small regional (e.g. parish) survey to the large site (e.g. Gaffney and Tingle 1985).

Even site survey is not a homogeneous topic, because the term ‘site’ itself has many meanings. Here we need to distinguish primarily between the development site – an area of land subject to some form of proposed commercial, agricultural or infrastructural development – and the archaeological site, already discussed in detail in chapter 4 (pp. 67–111). For the former, the broad aim is to detect the presence and extent of any significant archaeological remains on the site, with a view either to recording them before damage or destruction (preservation by record), or to mitigating the damage by redesign of the proposed development (preservation in situ). For the latter, the aim may simply be to determine the extent and character of a site (perhaps newly discovered in a regional survey), or there may be a more site-specific research design. The aims and methods may vary according to whether the site is largely invisible on the ground surface, or whether it has extensive visible remains, and also whether it is likely to be shallow or deeply stratified.

Introduction

The regional archaeological survey is an instrument whose nature and status have changed throughout this century, and particularly since about 1970. Four main reasons have been put forward for undertaking such surveys:

1. (a) discovery of sites, or prospection,

2. (b) estimation (e.g. of site densities or numbers),

3. (c) characterisation or specification (Wobst 1983, 37) (of the archaeological distribution within a region), and

4. (d) hypothesis testing (Wobst 1983, 40).

The emphases both between and within these aims have shifted over the years, following changes in academic and legal climates.

As a preliminary, we need to establish what is meant by both a region and a site. Although the term region may have a definite administrative meaning in some countries (or even several conflicting ones, as in the UK), it is used very broadly in archaeology to denote a spatially defined unit of territory, ranging from a few sq km in the British ‘Parish Survey’ (e.g. Aston and Gerrard 1995, 5) to 10,000 sq km or more (Read 1986, 479). The smaller examples can be thought of as ‘samples of size 1’ within larger regions, while the regions per se should exhibit some sort of topographical or cultural coherence (for example, an island (e.g. Cherry 1982, 13), a drainage basin (e.g. Read 1986, 477; Barker 1995), or a distinct topographical zone such as the English Fens (Hall 1987) or the Agro Pontino in Italy (Voorrips et al. 1991)).

Introduction

In this chapter we consider the smallest of the scales at which sampling is undertaken in archaeology – that of sampling from artefacts or ecofacts for ‘scientific’ analysis, such as chemical, physical or textural analysis (p. 184). This process usually divides into two stages:

1. (a) the selection of objects (whether artefacts or ecofacts) from an assemblage,

2. (b) the selection of samples from the chosen objects themselves.

The aim is usually to estimate parameters, such as chemical compositions, isotopic ratios or grain size distributions, which will help to characterise the assemblage in some way, for example in determining its provenance(s), in dating it, or in providing information about manufacturing techniques.

The first stage is needed because assemblages can be very large, and the scientific techniques can be very expensive, or very time-consuming. A few examples may therefore have to represent a very large collection, for example in the characterisation of the products of a pottery production site, but the issues involved in ensuring that the sample is truly representative are rarely discussed, and often taken for granted. Even the question of sample size seems often to be determined by financial or time constraints rather than by any archaeological or statistical consideration. Frequently, it seems that inferences have to be drawn from very small samples about very large populations, which must make one very concerned about the cavalier or unthinking attitude that is sometimes taken towards the selection of such samples.

A brief history of statistical sampling theory

The origins of the fundamental idea of sampling, that a small part can in some way be taken to represent, or stand in for, the whole, are lost in the origins of civilisation itself. There are many everyday, and especially commercial, situations in which the need for sampling, however informal, would have asserted itself. For example, merchants buying grain would wish to test the quality of their purchases, or a community would wish to maintain the standards of the weights and measures of goods (e.g. bread, ale) being sold within it. Such checking is at best time consuming and at worst destructive of the material being checked. For purely practical reasons, a small part had to stand in for the whole. Of course, such a procedure must be seen to be ‘fair’, to both the producer and the consumer, and this need must have been a frequent cause of disputes. In the absence of any appropriate theory, trust in whoever was doing the checking must have been important, and procedures similar to what we would today call random selection (see below) may have grown up.

Later, in the eighteenth and especially the nineteenth century, the need for sampling was felt in other areas, like scientific research and economic and social policy, and more theoretical questions began to be asked.

Introduction

Before moving on to look at the applications of statistical sampling theory and practice at a range of archaeological scales (chapters 4–8), we first make a diversion into archaeological theory. In this chapter we shall consider what insights the statistical theory of chapter 2 might be able to give us into the broader question of archaeological inference, that is, of inferring past activities from present-day archaeological remains. Archaeological inference is of course an enormous topic (see Sullivan 1978 for an overview; for a more recent view see Adams 1991), deserving of a book in itself, and it is not my intention to cover its full breadth here. But an approach grounded in statistical theory may at least help to clarify some issues, and to suggest appropriate methods of recording and analysis in certain situations.

It is a truism that archaeological remains, whether in the form of features or of artefacts, are rarely the totality originally created by, or used in, the activities that they now represent. In a sense, they are a ‘sample’ from some original but unknown ‘population’. For example, coins found on the site of a Roman town may be regarded as a sort of sample of those in circulation at that location at a certain period, or the sherds from a waster dump at a pottery production site may be regarded as another sort of sample of the pottery produced there. The big question is: what sort of a sample?

Introduction

We now move a step further down in scale, to sampling from archaeological deposits, done with the intention of recovering artefactual or ecofactual material, or geoarchaeological material for the characterisation of the deposit itself. The necessity for sampling is immediately apparent; every deposit potentially contains very many ‘objects’ (embracing all the above categories), some of which may be visible to the eye of the excavator, but a high proportion of which are invisible, either because of their size or because of other characteristics, such as colour or texture (see, for example, Keighley 1973), which make them indistinguishable from their soil matrix. It is simply not feasible to retrieve them all, not only because of the resources needed for retrieval itself, but also because of the resource implications for the subsequent stages of sorting, identification, recording and analysis, and the problem of storage. Nor indeed is total retrieval necessary, since the amount of data that could potentially be extracted is likely to be far in excess of what would be required for any conceivable analysis.

Given that sampling is a necessity, what particular issues are involved at this scale, and how do they differ from those at other scales? First, the issue of bias is ever present, both in the sampling process and in the associated literature, because the material exists across a wide spectrum of sizes, and there is no practical way of recovering it ‘all’ (whatever that might mean).