Quantifying ARCHAEOLOGY STEPHEN SHENNAN EDINBURGH UNIVERSITY PRESS ACADEMIC PRESS, INC. Harcourt Brace Jovanovich, Publishers © Stephen Shennan 1988 Published in Great Britain by Edinburgh University Press 22 George Square, Edinburgh Printed in Great Britain by J. W. Arrowsmith Ltd, Bristol British Library Cataloguing in Publication Data Shennan, Stephen Quantifying archaeology 1. Mathematical statistics 2. Archaeology—statistical methods I. Title 519.5Ό2493 QA276 ISBN 0 85224 460 6 cloth 0 85224 473 8 paper Published in the United States of America by Academic Press, Inc. Harcourt Brace Jovanovich, Publishers San Diego, California ISBN 0-12-639860-7 Preface As chapter one of this book makes clear, quantitative methods have become increasingly important in archaeology in recent years, and indeed considerable progress has been made in terms of the degree of sophistication of the analyses and of the matching be- tween archaeological data and appropriate mathematical methods. The specialists in any sub-field of a discipline, particularly when that sub-field involves mathematics, are always going to be working at a level which non-specialists cannot match, but it is important for the general health of the discipline as a whole that its practitioners should have some awareness of the sub-field and its wider implica- tions. Archaeology now has a small group of impressively com- petent mathematical specialists but the bulk of the people practising the discipline do not even understand what they are talking about, never mind its implications. This is a dangerous state of affairs because it leads on the one hand to the outright dismissal of methods which may be useful, and on the other to an excessive gullibility with regard to claims backed by mathematical arguments and analyses. If you can't actually evaluate the arguments, what choice is there? This situation contrasts with what I believe is a much more healthy situation in geography, where the body of mathematical and statistical specialists is at the apex of a broadly based pyramid of practitioners with basic mathematical and statisti- cal knowledge. The difference lies in the educational process in the two discip- lines. Quantitative methods were quickly integrated into the educa- tion of geographers as a result of the quantitative geography revolu- tion of the 1950s and 1960s. Comparable developments in archae- ology were never incorporated into the education of archaeologists to anything like the same extent, for reasons which, I suspect, have more to do with the sociology of the discipline, especially within the universities, than with anything else. Correspondingly, textbooks on mathematics and statistics in geography have proliferated, while vu viii Preface it would be true to say that archaeology has had nothing compar- able. This book aims to fill that gap. It fits between Mathematics in Archaeology (Orton 1980) on the one hand, and Mathematics and Computers in Archaeology ( Doran and Hodson 1975 ) on the other. Orton's book is a very clear, indeed excellent, account of how quantitative methods can be of assistance to archaeologists, but in no sense is it a textbook. Doran and Hudson's book is basically more advanced than this one. Its brief coverage of elementary methods, while in many ways a model of elegant conciseness, is too compressed for people who don't know the material already. This book arises from a course in quantitative methods of data analysis which I have given over a number of years in the Depart- ment of Archaeology, University of Southampton. As time has gone on the course has changed considerably from the earliest version, in response to student feedback and developments in the field. (Statistics itself is a fast changing discipline, a fact which outsiders tend not to appreciate.) Indeed, some discussion and explanation is appropriate at this point, concerning what the book contains and what it doesn't. First, it is about data analysis in archaeology and is thus oriented more towards studies of objects, excavations and archaeological data from surveys than towards more laboratory based studies, such as soil properties, chemical analyses, etc. ; in general such studies have their own mathematical and quantitative tradition, derived from the discipline where the techniques originate. This orientation has undoubtedly influenced the selection of material to present here. And indeed my next point concerning the book's content is precisely that it is not comprehensive, in the sense of covering all the quantitative techniques which have been used or could be useful to archaeologists. To do that would have taken a book far longer than this one. The intention has been to give an impression of the kinds of things which can be done at different levels ; how the archaeological topic becomes translated into statis- tical terms, and the problems associated with this ; and to provide a technical grounding in some of the most important techniques. In a textbook such as this it is natural that the latter should take up most of the space. The aim of the first part is to show students how to carry out some of the most basic techniques for themselves ; the aim of the second part is to give an intuitive understanding of some of the more complex methods, based on a geometric approach and archaeo- Preface ix logical examples, as a basis for understanding the literature. Famili- arity with the material presented here in the context of archaeologi- cal examples will make it much easier for those who want to use other techniques not covered here to understand statistical litera- ture written for other audiences, such as geographers and socio- logists. At the basic statistics level the most obvious omission is the Mest, and more generally a lack of discussion of distributional theory, except rather briefly in chapters 8 and 14. This was not done without considerable thought. In fact, such material was covered extensively in the early versions of the course from which this book originates. It was dropped because the amount of complex technical detail which had to be covered at a fairly early point in the proceed- ings was proving to be a major obstacle both to understanding the quantitative methods themselves and to seeing any relevance of them to archaeology. Not including them proved to be a lot more satisfactory: the gain far outweighed the loss. Another obvious omission is spatial analysis. Hodder and Orton (1976) still provide a good introduction for archaeologists here, backed up by the many geographical texts. In this light it was felt that the desirability of including something rather more elementary than Hodder and Orton on spatial analysis was outweighed by the significant increase in size of this book which it would have in- volved. The introductory intention of the book precludes examination of the more advanced techniques now being used in archaeological research, such as those based on computer simulation, but there is a brief discussion of these in the final chapter. One or two sections of the book, however, are more advanced than the rest and many people will want to skip them on the first time through. This is especially the case with the last section of chapter 7 and, to a lesser extent, parts of chapter 10 and 11. They are included to show what detailed data analysis can involve and how it may be done. They are examples, as are others in this book, of the realisation of a sceptical approach to data and data patterning. Indeed, if I have one hope for this book, it is not that readers will retain a memory of the details of the statistical techniques it covers, but that they will have acquired an informed, sceptical and questioning attitude to the quantitative analyses of themselves and others. In this way things can only improve. Over the years, in teaching quantitative methods and preparing this book I have incurred a number of debts which must be acknow- x Preface ledged. First, and most important, I have to thank the students who have taken my courses, especially Todd Whitelaw, Hans-Peter Wotzka and Nick Winder. Their critical questioning has meant that I could never get away with anything and I've benefited enormously as a result ! I'm also grateful to Professor Colin Renfrew for his early encouragement and support for the teaching of quantitative methods at Southampton, and to Archie Turnbull of Edinburgh University Press for encouraging me in the writing of this book and offering many constructively critical comments on an earlier draft. Finally, I owe a special debt of gratitude to Dr Nick Fieller, Depart- ment of Probability and Statistics, University of Sheffield, for his invaluable assistance in reading the manuscript with an expert eye ; and to Professor R.Barry Lewis, Department of Anthropology, University of Illinois at Urbana-Champaign, who made many help- fully blunt comments and suggestions. Neither they nor anyone else except myself is responsible for the inadequacies that remain. One Introduction The aim of this text is to make students familiar with some of the basic quantitative methods currently used in archaeology. Of course, these techniques are not specific to archaeology, being used in a great variety of fields, but experience has shown that archae- ology students do not gain a great deal from attending statistics classes for sociologists or biologists because, although the statistical theory and methods are the same, the examples used are alien. To the student of archaeology such examples are boring and often incomprehensible. To most people quantitative methods tend to be sufficiently forbidding to need no handicaps of this kind. Teaching in such an alien framework is particularly unfortunate because many non-mathematically inclined people find that they can best get an initial grasp of a topic not by learning about the theory behind it, but by following through a worked example. For these reasons a specifically archaeological introductory text seemed worthwhile. It is hoped that by the end of the book students will themselves be able to use the simple techniques described, will have some feel for the way archaeological questions can be translated into quanti- tative terms, and will have a basis for talking to statisticians, in their own terms, if problems are more complex. This last point is of some importance. If you turn to a statistician for help and neither of you knows what the other is talking about, you will probably end up with the wrong answer to the wrong question. The text assumes very little in the way of prior knowledge. Only the most basic mathematical operations of addition, subtraction, multiplication and division, together with roots, powers and loga- rithms are needed. Calculus and matrix algebra are not required. WHY QUANTITATIVE METHODS? This question must be considered before we go further. In fact, it is possible to divide it into two rather different questions : why should 1 2 Quantifying Archaeology students concern themselves with quantitative methods in archae- ology, and why should archaeology, as such, concern itself with quantitative methods. One answer to the first of these questions is that the archaeologi- cal literature is increasingly given over to papers whose arguments depend on the application of such methods. A knowledge of them is therefore essential if their arguments are to be understood and evaluated. This is certainly true, but it does not answer the second and larger question. The most cynical view would be that archae- ology has become involved with quantitative methods purely as a result of disciplinary fashion. The last thirty years have seen the biological sciences, geography and many of the social sciences become increasingly quantitative ; it is now a matter of prestige for a discipline to appear to be 'scientific', an endeavour in which quantitative methods have a key role. Archaeology has simply followed this trend, adopting the 'archaeologists in white coats' image, and in the process has carried out a successful piece of disciplinary imperialism, expanding its influence, its manpower and the resources allocated to it generally. In due course, such an argument might go, such approaches will become less fashionable, indeed perhaps are already becoming so, and will gradually fade from significance. I think it would be foolish to try to deny al- together this aspect of the 'quantitative revolution' as it has been called in geography, but such arguments from the sociology of science are certainly only part of the story. One key factor has been the rise of the computer. As we are well aware, computers now have a great variety of roles in archaeology. In the last decade they have become increasingly widely used as data management tools for such tasks as the recording of excavation data and the building up of regional data banks of archaeological information, the former development in particular greatly en- hanced by the advent of microcomputers ( Richards and Ryan 1985, Gaines 1981 ). The use of computers in archaeological model-build- ing has also become important: computer programs have been written to simulate processes as diverse as the collapse of Maya civilisation ( Hosier et al. 1977 ) and tool manufacture and discard in Australian aboriginal subsistence-settlement systems ( Aldenderfer 1981); numerous examples are to be found in books edited by Hodder (1978), Renfrew and Cooke (1979) and Sabloff (1981). Here I want to consider only their use as tools for carrying out data analysis, which is in fact the purpose for which computers were first introduced into archaeology, and indeed into many other sub- Introduction 3 jects. Prior to the development and first application of computers in the 1950s and early 1960s uses of mathematics and statistics were largely restricted to the 'hard sciences'. This was at least partly because the solutions to many problems of interest could be ob- tained by means of elegant methods of mathematical analysis which did not require enormous numbers of calculations. Similarly, the statistical techniques which for the same reason were practically possible, proved very useful in many scientific, technological and industrial applications, but less so with the more intractable data of geography, archaeology or the social sciences. Only with the de- velopment of a means of carrying out enormous numbers of calcula- tions at very high speeds did it become possible to apply methods appropriate to the kind of problems which the data from such disciplines presented. It might be said that such an exciting new toy as the computer was eventually bound to be tried by archaeologists, and that the involve- ment of archaeology in quantitative methods simply stems from an attempt by archaeologists who like such toys to find a use for them. This cannot altogether be excluded. However, it still does not take us to the heart of the matter, which lies not in fashion, nor in the availability of computers, but in the fact that quantitative reasoning is central to archaeology, and that a better grasp of this fact might well improve our work as archae- ologists. Clive Orton's book Mathematics in Archaeology (1980) provides an excellent demonstration of why this is the case, by taking some of the standard questions which archaeologists ask, such as 'What is it?', 'How old is it?', 'Where does it come from?' and 'What was it for?', and showing how a quantitative approach can help to provide the answers. It follows, therefore, that quantita- tive methods should be seen, not as a distinct scientific specialism within archaeology, like pollen analysis, for example, or the various techniques of artefact characterisation, but as part of every archae- ologist's mental tool-kit. Statistical, mathematical and computer specialists may often be required to cope with particular problems, but archaeologists must have sufficient quantitative awareness to recognise when problems arise which can be helpfully tackled in a quantitative fashion. No one else can do this for them. Given that this is the case, it remains to be specified exactly where the mathematics and the archaeology come together. Part of the answer is in the simple description of the archaeological record : numbers of potsherds of different types, sizes of pits, and so on. Such quantitative information is an essential part of all modern 4 Quantifying Archaeology archaeological reports, and simple quantitative description is the first topic we will consider. Much more important than this, however, is the link described by Orton (1980). The archaeologist makes his inferences about the past on the basis of patterning and relationships in the archaeologi- cal record. Mathematics is an abstract system of relationships. The possibility then exists that mathematics may help us to recognise patterning in the archaeological record and to specify its nature. The area where mathematics meets the messier parts of the real world is usually statistics. It is precisely this fact that makes statistics in many ways a tricky subject, because mathematical and factual considerations are both involved, and because the relationships which we look at are almost never perfect ones. Orton shows very clearly that all interpretation of the archaeo- logical record is concerned with identifying patterning and is cap- able of benefiting from a quantitative approach. It is nevertheless a historical fact that the main impetus for the introduction of quanti- tative methods of archaeological data analysis came from the North American 'New Archaeology' tradition of the 1960s, and as a result of this 'New Archaeology' and the use of quantitative methods became inextricably associated in the general archaeological con- sciousness, both being labelled as 'anti-humanistic' (Hawkes 1968). Doran and Hodson (1975) were at pains to point out, correctly, that^there was no necessary connection between the two, and that quantitative approaches could be used to tackle traditional archaeo- logical problems. However, it is still the 'New Archaeology' tradi- tion, now known, twenty-five years on, as the 'processual' school, that has made the greatest use of such techniques and it is worth asking why quantitative analysis has been, and remains, one of its distinguishing features, despite the fact that some applications of such methods have since been shown to be classic examples of misuse and misinterpretation (Thomas 1978). I believe there are several reasons for this. First, least praise- worthy and probably least important: quantitative methods are regarded as 'scientific' and the New Archaeology specifically set out to adopt a scientific approach to the subject, making the use of quantitative methods ideologically necessary. Second, New Archae- ology emphasised explicitness and objectivity, both of which are considerably aided by the rigour of quantitative analysis, which has a vital role to play in removing at least some potential sources of self-deception. Third, it advocated a hypothetico-deductive approach to the study of the past, in which hypotheses are gener-
Description: