KNOWLEDGE AS SOCIAL ORDER
This page intentionally left blank
Knowledge as Social Order Rethinking the Sociology of ...
115 downloads
1695 Views
786KB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
KNOWLEDGE AS SOCIAL ORDER
This page intentionally left blank
Knowledge as Social Order Rethinking the Sociology of Barry Barnes
Edited by MASSIMO MAZZOTTI University of Exeter, UK
© Massimo Mazzotti 2008 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise without the prior permission of the publisher. Massimo Mazzotti has asserted his moral right under the Copyright, Designs and Patents Act, 1988, to be identified as the editor of this work. Published by Ashgate Publishing Limited Gower House Croft Road Aldershot Hampshire GU11 3HR England
Ashgate Publishing Company Suite 420 101 Cherry Street Burlington, VT 05401-4405 USA
Ashgate website: http://www.ashgate.com British Library Cataloguing in Publication Data Knowledge as social order : rethinking the sociology of Barry Barnes 1. Barnes, Barry 2. Science - Social aspects 3. Relativity I. Mazzotti, Massimo 306.4'5 Library of Congress Cataloging-in-Publication Data Knowledge as social order : rethinking the sociology of Barry Barnes / edited by Massimo Mazzotti. p. cm. Includes index. 1. Knowledge, Sociology of. 2. Social structure. 3. Barnes, Barry. I. Mazzotti, Massimo. HM651.K563 2008 306.4'2--dc22 2007034124 ISBN 978-0-7546-4863-5
Printed and bound in Great Britain by TJ International Ltd, Padstow, Cornwall.
Contents List of Figures Notes on Contributors Acknowledgements Introduction Massimo Mazzotti 1
2
3
4
5
6
7
8
9
vii ix xiii
1
Relativism at 30,000 Feet David Bloor
13
Relativism: Is it Worth the Candle? Trevor Pinch
35
Who is the Industrial Scientist? Commentary from Academic Sociology and from the Shop-Floor in the United States, ca. 1900 – ca. l970 Steven Shapin
49
The Meaning of Hoaxes Harry Collins
77
Objectual Practice Karin Knorr Cetina
83
Producing Accounts: Finitism, Technology and Rule-Following Donald MacKenzie
99
Power and Legitimacy Mark Haugaard
119
Barnes on the Freedom of the Will Martin Kusch
131
Agency, Responsibility and Structure: Understanding the Migration of Asylum Seekers to Ireland Steven Loyal
147
vi
10
Index
Knowledge as Social Order
Against Maladaptationism: or What’s Wrong with Evolutionary Psychology? John Dupré
165 181
List of Figures Figure 6.1 Figure 6.2 Figure 6.3
Figure 8.1
Extensional semantics and finitism White to play and mate in a single move Examples of codes in the case-study firm’s chart of accounts
101 102
The main philosophical views on freedom of the will
135
104
This page intentionally left blank
Notes on Contributors David Bloor is Professor of Sociology and Director of the Science Studies Unit at Edinburgh University. He has written widely on the Kuhn/Popper debate, the cognitive functions of metaphor, and on the sociology of scientific knowledge and Wittgenstein’s philosophy. He is author of Knowledge and Social Imagery (Routledge 1976; 2nd edition Chicago Univ. Press 1991); Wittgenstein: A Social Theory of Knowledge (Macmillan and Columbia 1983); Wittgenstein: Rules and Institutions (Routledge 1997); and co-author (with Barry Barnes and John Henry) of Scientific Knowledge: A Sociological Analysis (Athlone and Chicago Univ. Press 1996). Harry Collins is Distinguished Research Professor of Sociology and Director of the Centre for the Study of Knowledge, Expertise and Science (KES) at Cardiff University. His has worked on the nature of artificial intelligence, publishing two book on the topic: Artificial Experts: Social Knowledge and Intelligent Machines (MIT Press 1990) and, with M. Kusch, The Shape of Actions: What Humans and Machines Can Do (MIT Press 1998). His main topic is the sociology of scientific knowledge publishing Changing Order: Replication and Induction in Scientific Practice (Sage Publications 1985; 2nd edition Univ. of Chicago Press 1992) and, with T. Pinch, a popular series about science and technology which so far has three volumes: The Golem: What you should know about science (1993; R. K. Merton book prize); The Golem at Large: What you should know about technology (1998) – both books are published by Cambridge Univ. Press and reprinted in their Canto paperback series – and Dr. Golem: How to Think About Medicine (University of Chicago Press 2005). In 1997 he was the recipient of the Society for Social Studies of Science J.D.Bernal prize. John Dupré is Professor of Philosophy of Science at the University of Exeter and Director of the ESRC Centre for Genomics in Society. He has previously taught at Stanford University and Birkbeck College. His research focuses on the philosophy of biology and on the cultural and methodological aspects of scientific research. His books include The Disorder of Things: Metaphysical Foundations of the Disunity of Science (Harvard University Press 1993); Human Nature and the Limits of Science (Oxford University Press 2001); Humans and Other Animals (Oxford University Press 2002); and Darwin’s Legacy: What Evolution Means Today (Oxford University Press 2003). Mark Haugaard lectures in Social Theory at the National University of Ireland, Galway, and is the author of numerous publications on power and social theory more generally. These include: The Constitution of Power (Manchester University
x
Knowledge as Social Order
Press 1997); ‘Power: Social and Political Theories of’ in Encyclopedia of Violence, Peace and Conflict, Vol. 3 (Academic Press 1999); Power: A Reader (Manchester University Press 2002); (co-ed) Making Sense of Collectivity: Ethnicity, Nationalism and Globalization (Pluto Press 2002); and (co-ed) Power in Contemporary Politics (Sage 2000). During the academic year 2001–2002, Haugaard was Jean Monnet Fellow at the Department of Politics and Sociology at the European University Institute, Florence. Karin Knorr Cetina is Professor of Sociology at the University of Kostanz and visiting professor at the University of Chicago. She has authored numerous books in German and English among which The Manufacture of Knowledge: An Essay on the Constructivist and Contextual Nature of Science (Oxford University Press 1981); Epistemic Cultures. How the Sciences Make Knowledge (Harvard University Press 1999) – winner of the L. Fleck Prize for the best book of the year in social studies of science; and, with A. Preda, The Sociology of Financial Markets (Oxford University Press 2004). Among her edited books are Science Observed: New Perspectives on the Social Study of Science (Sage 1983) with M. Mulkay; and The Practice Turn in Contemporary Theory (Routledge 2001) with T. Schatzki and E. von Savigny. Professor Knorr Cetina’s contribution to this book has also been published in Theodore Schatzki, Karin Knorr Cetina and Eike von Savigny (eds) (2001), The Practice Turn, London: Routledge, 175–188. Martin Kusch is Professor of History and Philosophy of Science at the University of Cambridge. He previously was Reader in the Science Studies Unit (of the University of Edinburgh) where he succeeded Barry Barnes in 1993. His main book publications are Psychologism (Cambridge University Press 1995); Psychological Knowledge (Cambridge University Press 1999); Knowledge by Agreement (Cambridge University Press 2002); A Sceptical Guide to Meaning and Rules: Defending Kripke’s Wittgenstein (Acumen 2006) and, with Harry Collins, The Shape of Actions (MIT Press 1998). Steven Loyal is a Lecturer in Sociology at University College, Dublin. In addition to his interest in sociological theory he has research interests in migration, and ethnic and racial studies. His most recent books have been The Sociology of Anthony Giddens (Pluto Press 2003) and The Sociology of Norbert Elias (Cambridge University Press 2004) edited with Stephen Quilley. Donald MacKenzie holds a personal chair in Sociology at Edinburgh University, where he has taught since 1975. His books include Statistics in Britain, 1865–1930: The Social Construction of Scientific Knowledge (Edinburgh University Press 1981); Inventing Accuracy: A Historical Sociology of Nuclear Missile Guidance (MIT Press 1990); Mechanizing Proof: Computing, Risk, and Trust (MIT Press 2001); and An Engine, Not a Camera: How Financial Models Shape Markets (MIT Press 2006). Trevor Pinch is Professor of Science and Technology Studies and Professor of Sociology at Cornell University. He has written extensively on the sociology of
Notes on Contributors
xi
technology, and his most recent publications include Analog Days: The Invention and Impact of the Moog Synthesizer (Harvard University Press 2002) with F. Trocco; The Golem at Large: What You Should Know About Technology (Cambridge University Press 1998) with H. Collins; The Hard Sell: The Language and Lessons of Street-Wise Marketing (HarperCollins 1995) with C. Clark; The Golem: What Everyone Should Know About Science (Cambridge Univ. Press 1993) with H. Collins; Dr. Golem: How to Think About Medicine (University of Chicago Press 2005) with H. Collins; and several edited books. Steven Shapin is Professor of the History of Science at Harvard University. He held previous appointments as Professor of Sociology at the University of California, San Diego, and at the Science Studies Unit, Edinburgh University. His books include Leviathan and the Air-Pump: Hobbes, Boyle, and the Experimental Life (Princeton University Press 1985) with S. Schaffer; A Social History of Truth: Civility and Science in Seventeenth-Century England (University of Chicago Press 1994); The Scientific Revolution (University of Chicago Press 1996; now translated into 14 languages), and several edited books. His awards include the J.D. Bernal Prize of the Society for Social Studies of Science, the Ludwik Fleck Prize for the best book of the year in social studies of science, the Robert K. Merton Prize of the American Sociological Association, and the Herbert Dingle Prize of the British Society for the History of Science. Professor Shapin’s chapter for this book was also published in Karl Grandin, Nina Wormbs, and Sven Widmalm (eds) (2004), The Science-Industry Nexus: History, Policy, Implications. Canton, MA: Science History Publications, 337–363.
This page intentionally left blank
Acknowledgements I thank Annalisa McNamara and Massimiliano Pagani for their aid in the preparation of this book. I’m grateful to numerous friends and colleagues in Exeter and Edinburgh for their comments and advice, and to the contributors, for their patience and availability.
This page intentionally left blank
Introduction Massimo Mazzotti
For a long time, debates over the nature of knowledge have been informed by an influential general model, according to which knowledge and belief are separated by an insuperable epistemological boundary, belief being knowledge less something (warrant, for example). From this perspective knowledge is to be understood as the product of the encounter of the individual mind with reality – social and natural. Logic and evidence alone would determine the contents of knowledge, while the social dimension would mainly enter the picture as a distorting element, as when the analyst identifies the effects of ‘social pressures’, or the ‘weight of tradition’. The theoretical and empirical shortcomings of this approach to the study of knowledge have been fully exposed by decades of anthropological and sociological research. Within the sociology of knowledge an alternative view has taken shape, according to which knowledge/belief is to be understood as essentially related to social action and structure. The debate has thus shifted to questions about the nature of the relationship between belief, action and structure, and the analytic use of these notions; and, indeed, the opportunity to retain a knowledge/belief distinction. Clearly, the way in which these questions are answered is of great import not only for the sociology of knowledge but for social theory as well. The present volume aims to contribute to current debates in this area by honouring and reflecting upon the work of Barry Barnes. The volume intends to mark at once Barnes’s retirement from his current position at the University of Exeter and the fortieth anniversary of his affiliation to the Science Studies Unit at the University of Edinburgh. Barnes’s highly original and fruitful insights on the knowledge/power/ practice relationship are presented here through ten essays written by scholars from various fields who have been using, developing and challenging those insights. The diversity of these uses, which cut across disciplinary fields and concerns, is in itself a sign of the breadth of Barnes’s reflection and of its theoretical import. In fact, although Barnes is mainly known as one of the founders of what today is called science and technology studies, he has always conceived of his own research as dealing primarily with fundamental issues in the social sciences. The aim of this introduction is precisely that of rendering the profound sense of unity that pervades Barnes’ varied body of work, and illustrating the way his work bears on basic questions about the interplay between cognitive and social order in any given human collective. The relationship between structure and belief has been a traditional concern of the sociology of knowledge. The analytic tools used to explore this relationship derive primarily from two bodies of work. The first and most influential is the
2
Knowledge as Social Order
work of Karl Marx, in which the relationship between social structure and belief is conceptualised in terms of a materialistic theory of interests. It is the relation to the means of production that, according to Marx, defines the interests of social classes. Class interests in turn determine people’s beliefs – although the actual link between class interests and individual interests has proven to be quite difficult to analyse. This basic, mono-directional picture (where social structure influences belief) is complicated by the presence of ideology, which according to Marx distorts the true interests of the working classes, shapes their understanding of the world and favours the reproduction of class-relations. The idea that structure and belief are connected by a complex two-way relationship is therefore already present at the origin of the sociology of knowledge. A similar image emerged even more sharply in another major tradition of research, constructed around the work of Emile Durkheim. According to Durkheim, categories of thought should be understood as contingent products of a collective’s social experience. Scholars within this tradition, such as Mary Douglas, have highlighted the remarkable homologies between social structures and systems of beliefs as documented by anthropological research. However, as Durkheim was aware, the arrow of determination does not point simply from structure to belief. Those very beliefs that are produced through collective action are indeed functional to the sustenance of society and to the reproduction of its institutions. After a period of relative decline the study of the sociology of knowledge flourished anew in the 1960s, in correspondence with a more general process of reorientation of the field of social sciences, and especially a rethinking of the nature of social structure and of its relation to collective and individual action. In brief, resources from both the Marxist and Durkheimian traditions were mobilised to question monocausal, deterministic models for the production of knowledge, and particularly those strains of sociological and anthropological research that had reified – illegitimately, it was now believed – the notion of social structure. The work of such authors as Louis Althusser and Mary Douglas is emblematic of the revival and renewal of the two main traditions during the sixties. More generally, this period saw the strengthening of a more interpretive, interactionist theorisation, which emphasised human agency and the theoretical significance of practice, a shift emblematised by the raise and influence of ethnomethodology. Michel Foucault and Pierre Bourdieu, who were not associated directly with previous traditions in the sociology of knowledge, also played an important role in reshaping the theoretical landscape. Foucault’s analysis of power/knowledge proved to be especially suggestive, as it re-framed the debate by focusing on the concrete tools of social control and questioned, through the notion of discourse, the very analytic significance of tools such as structure and belief. Bourdieu’s notions of habitus and field also allowed for a rethinking of the relationship between structure and action. Barry Barnes’s early work should be situated precisely within the 1960s theoretical and methodological renewal of the sociology of knowledge. Drawing upon the work of Marx and Durkheim, but also on suggestions from Ludwig Wittgenstein, Mary Hesse, Mary Douglas, Thomas Kuhn and Harold Garfinkel, Barnes and his colleagues at the Science Studies Unit at Edinburgh University initiated a new and distinctive research tradition in the sociology of knowledge, which they called
Introduction
3
‘the Strong Programme in the sociology of scientific knowledge’ (SSK). When the publications of the Edinburgh School, as the group was soon labelled, began to appear in the early seventies, it was immediately clear that the scope of their work went far beyond the traditional concerns of the sociology of knowledge, and had implications for the very basis of social theory. In essence, they showed empirically that cognitive and social order cannot be understood in isolation from each other, even when one analyses the contents of the most esoteric forms of scientific and technical knowledge. During the 1970s the study of science and technology as social activities became, in Edinburgh and elsewhere, the raison d’être of a new, specialized field of research initially known as ‘science studies’. From the eighties onwards the field has been thriving within the Anglo-Saxon academic world, under the label of science and technology studies (STS), and its maturity has been signalled by a wave of specialized journals and undergraduate textbooks. The rise of STS within academia has been a striking success story. It was so successful that it is now difficult to recall the sense of outrage that the claim could provoke around 1970, even within the social sciences, that scientific knowledge should be studied just like any other kind of belief system. It might be useful then to start from there to sketch a map of Barnes’s own contributions to the sociology of knowledge and to social theory. In 1967, Stanley Barry Barnes joined the Science Studies Unit at the University of Edinburgh as a lecturer. He was employed as a sociologist to teach science and engineering students about their own work considered as social activity and about its wider social significance. David Bloor was similarly employed, and soon Steven Shapin and Donald MacKenzie would join the teaching staff of the unit. Needless to say, the necessity of teaching modules along these lines was far from obvious at the time. But then, the Science Studies Unit was a rather unobvious institution. The unit had begun its activities in 1966, having been set up by a group of prominent Edinburgh scientists led by the biologist Conrad Hal Waddington with the explicit aim of contributing to a bridging of the gap between what Charles Percy Snow had famously called the ‘two cultures’. The creation of the unit was inspired by the Scottish tradition of providing scientists with a broad education, and its main aim was to open science up to social responsibility, which at the time was understood primarily in terms of the scientist’s awareness of the social implications of their work. Waddington’s vision of science and its role in society was derived from a lively Marxist tradition that aimed to free western science from the control of capitalist priorities and bring it closer to the people’s real needs. A very similar agenda would soon find institutional expression in the British Society for Social Responsibility in Science (1971). It remained unclear, in this as in other traditions of a socialist sociology of science, if and how the political meaning of science should change our understanding of the notion of scientific objectivity (and hence of ideology) and our interpretations of the very contents of scientific knowledge. In any case, in 1966 a committee chaired by Waddington decided that science and engineering students should have the opportunity to take modules that would raise their awareness of the social and political implications of science and technology. David Edge was appointed as first director of the unit; he was a radio astronomer with a keen interest in science communication who had been working for the Science Unit at the BBC since 1959. Apparently Waddington told Edge: ‘we’ll teach them the science, you’ll
4
Knowledge as Social Order
teach them the rest’ (Bloor 2003). In a few years ‘the rest’ turned out to be a new and promising field of research: science studies. In constructing genealogies and myths of origins we tend to emphasise those aspects of the object in question that we deem essential. In my sketchy reconstruction of the origins of the Edinburgh School I have done exactly this. I would like to emphasise two points: first, its roots in the work of Marx and Durkheim, and secondly its prominent connection with the world of practicing scientists, especially biologists and physicists. The first point is relevant as it provides the context in which the long-lasting commitment to the materialist assumptions that characterise the Edinburgh School took shape, and especially Barnes’s ‘minimal realism’.1 At the same time these assumptions also have to do with the second point, i.e. the belief that science should be studied primarily through the careful observation of actual every-day scientific practice. The members of the unit began studying science as it is practiced, rather than as it ought to be. When contrasted with the ethereal worlds of 1960s philosophy of science or the institutional imperatives of Mertonian sociology, the early works of the Edinburgh School appear strikingly different, immersed as they are in the messy reality of experimental life. They certainly looked more like works in ethnography and historical anthropology than in philosophy. It is probably relevant that, unlike most philosophers and sociologists studying science at that time, Barnes and other members of the Edinburgh School were trained in the natural sciences. This might indeed explain their sharp perception of previous theories of scientific knowledge as irredeemably idealistic. Thus Barnes had read chemistry and was familiar with laboratory practice in this field (Barnes and Symons 1966; 1970a), while Bloor had done research in experimental psychology and Shapin, the historian of the group, had been trained in genetics, and collaborated closely with the Edinburgh Medical School while writing his groundbreaking articles on the history of phrenology (Shapin 1975). The enormous influence of this methodological revolution – looking at what scientists do rather than at what they, or philosophers, say – is testified to by the fact that this move would soon become obvious and unproblematic. This basic hands-on attitude shaped the methodological tools deployed in the works of the Edinburgh School. For them, the empirical exploration of scientific practices should itself be shaped by what natural scientists do and the way in which they orient themselves to what they study. Again, today it might be rather difficult to grasp the radical nature of this choice, its radical rupture with existing traditions in the study of scientific knowledge. In opposition to the then dominant rationalistic and idealistic perspectives, Barnes and Bloor defined their attitude as ‘naturalistic’. The conviction that the discipline of sociology would benefit from being informed by the methodology of the natural sciences was still relatively popular in the early 1960s, although on the wane. Interestingly, Barnes’s own postgraduate study in sociology had been possible thanks to a programme of the Social Science Research Council specifically designed to attract science-trained students to the social sciences. The limits of this form of scientific sociology, however, were becoming all 1 On Barnes’s minimal realism see Barnes 1992a, where this position is presented and contrasted to stronger and more familiar kinds of realism.
Introduction
5
too evident. Essentially, the idea that the methods of the natural sciences should also inform the social sciences was combined with misconceived notions of what those methods were. The sociological imagination of the 1960s was still very much permeated by a rationalistic account of science and scientific objectivity, shaped by the works of such philosophers as Karl Popper, who saw scientific knowledge as the product of the encounter of the individual mind with external reality, mediated by the rules of an alleged universal logic. While the rest of the social world was being boldly explored and deconstructed through new interpretive and phenomenological approaches, the world of science appeared thus frozen in an anachronistic, sui generis status that exempted it from any kind of social analysis – as if scientific practices were not social practices themselves. In Barnes’s opinion this special status – which paralleled that traditionally ascribed to theology – was the result of an unscientific attitude. In Edinburgh he thus began re-conceptualising the relationship between natural and social sciences in a way that went beyond the traditional, and at that point sterile, dualistic model. Barnes agreed that science should be studied scientifically. However, he also believed that social scientists were using a misleading model of how science actually works. This model needed to be freed from its idealistic and rationalistic components. What social scientists could certainly import from science was a fundamental attitude shared by natural scientists, i.e. naturalism. Rather than trying to offer some necessarily unsatisfactory definition of naturalism, Barnes suggested that the modus operandi of the natural scientists should be observed, alongside the basic assumptions that frame their interaction with the world, and the way in which they orient themselves to the world and to each other when they produce knowledge. Using a Kuhnian image, he suggested that exemplary scientific work incarnated the naturalistic attitude he was advocating. The first step towards a properly scientific sociology of knowledge consisted of importing to this field of inquiry the naturalistic attitude and realist presuppositions of practicing scientists. The adoption of such a perspective made it apparent that science should be analysed sociologically in just the same way as any other social activity. This meant challenging the dominant dualistic perspective on knowledge, and promoting a naturalistic and monistic sociology of knowledge instead. Barnes’s naturalistic commitment thus provided the basis for the development of that particular form of relativism that was soon to be become associated with the Edinburgh School (Barnes and Bloor 1982). After a few articles on the sociology of scientific beliefs and an empirical study of social dynamics of industrial research (Barnes 1969; 1971; Barnes and Dolby 1970), Barnes edited Sociology of Science (Barnes 1972), a groundbreaking reader that signalled the thriving status of the field. Barnes’s own contribution to this volume marked the start of his interest in the social construction of expert credibility, a line of research that was openly discouraged by policy-makers at the time (Barnes 1972a).2 The view of science as an essentially social activity was summarised in Scientific Knowledge and Sociological Theory (1974), Barnes’s first major contribution to social theory. Here Barnes mobilised a variety of theoretical and empirical materials in defence of this monistic position and engaged with dualist approaches of various 2 For Barnes’s more recent work on expert credibility, see Barnes 1999, 2002b, and 2005a.
6
Knowledge as Social Order
kinds, especially those defended by rationalist philosophers. In this context the work of Thomas Kuhn proved to be a powerful resource. Barnes’s fresh reading of Kuhn’s The Structure of Scientific Revolutions downplayed those aspects that were most discussed and celebrated, such as the idea that the history of science is characterised by a series of revolutionary discontinuities, or the notion of paradigm as virtually equivalent to that of scientific theory. Instead, Barnes brought to the reader’s attention Kuhn’s profound sociological intuitions, and developed them into an account of science as a set of practices based on exemplary achievements that are taken as authoritative in specific situated traditions of research. Through his reading, which focused on the dynamics of micro-social interaction within research collectives, Barnes emphasised the notion of paradigm as a concrete exemplar rather than a set of statements, and developed a vision of science as at once inseparably knowledge and action (Barnes 1982; 2002). Incidentally, it should be noted that studying science, for Barnes, meant to study instruments and technologies as well. Although the term technoscience had not been coined yet, it seemed natural to Barnes and his colleagues to take onboard the insights of an anthropological tradition that considered skills and artefacts as constitutive aspects of a culture (Barnes 2005). The fact that the vision of science outlined in Barnes’s 1974 book became a fundamental resource for the emerging field of STS makes it difficult to realise the formidable opposition elicited by the book at the time of its appearance. It was not only rationalist philosophers who would criticise Barnes’s monistic approach, but social scientists as well. Steven Lukes’s review is indicative of the kind of resistance elicited by Barnes’s project. Lukes applauded Barnes’s argument that all belief systems should be treated symmetrically for the purposes of sociological explanation, but he thought that this claim could and indeed should be detached from Barnes’s ‘radical epistemological relativism’ and its ‘vertiginous consequences’. While Lukes agreed that ‘the apparent truth or rationality of a belief or a set of beliefs does not preclude their sociological explanation’, he criticised Barnes’s relativistic stance, which he identified rather inaccurately with the claim that ‘there are no universally applicable criteria of truth and rationality’ (Lukes 1975: 503). Lukes’s review is remarkable: in its critical sections it outlines the major arguments that would be raised against the Strong Programme in the following two decades, and especially the claim that the social and the natural components of knowledge are mutually exclusive. Not unlike philosophers, leading social scientists were apparently uneasy with the idea that what is socially constructed can be constrained by the external world as well. Barnes’s early work focused on scientific knowledge, but mainly as a particularly interesting case through which to explore more general features of knowledge production and distribution. Barnes suggested that sociologists should study science not simply as interesting in itself, but as a way to tackle effectively the fundamental question of the relationship between knowledge – which to him has always meant collectively endorsed belief – and social order.3 His analysis of scientific practices 3 It is important to distinguish between two possible meanings of knowledge: knowingthat and knowing-how. Barnes’s definition of knowledge as collectively endorsed belief refers exclusively to the first meaning.
Introduction
7
was indeed intended as exemplary sociology, concerned with how best to understand and explain the basic features of human social life. Science, in other words, was a place to do sociology, and it was neither better nor worse than other places. What made the study of science particularly interesting though, was that it could be used to test and check certain general sociological conjectures, and to illustrate that individual reason and experience are invariably insufficient bases for an understanding of the provenance and perceived validity of a body of knowledge. The fruitfulness of looking at scientific practices to explore more general problems became clear in relation to the problem of reference. Based on his account of how scientists routinely apply concepts to make sense of empirical data, Barnes developed in a series of articles a distinctive finitist-sociological view of concept application, very much in opposition to the reified extensional view then widely accepted by English-speaking epistemologists (for a summary, see Barnes, Bloor and Henry 1996: 46–80). Developing insights from Wittgenstein, Kuhn and Hesse, Barnes argued that concept application is always the result of the contingent action of situated actors.4 Effective concept application cannot be the result of actions of independent individuals however, but only of members of a collective. Barnes’s key insight is that, in referring to things, actors respond to each other’s behaviour as well as to reality; and they could not do otherwise if reference has to be effective. Normativity, in other words, can only emerge and be sustained by the co-ordinated activity of concept application carried out by a collective as it engages with reality. Barnes was thus able to elaborate his sophisticated form of meaning-finitism into a more satisfactory way of expressing a relativist perspective, one that offered a constant reminder that concepts and hence knowledge are unintelligible as abstractions, and that what they involve has always to be understood as manifest in collective use. Through his elaboration and deployment of meaning-finitism Barnes related knowledge to collective practice, and more precisely to goal-oriented collective practice. Collective use of concepts cannot but be guided by a contingent set of priorities and goals, which Barnes referred to broadly as ‘interests’. His conceptualisation of the relationship between collective interests and bodies of knowledge was first articulated in Interests and the Growth of Knowledge (1977). The book’s key ideas began a research tradition of interest analysis that focused, especially in its early stages, on scientific controversies. Following pioneering casestudies such as MacKenzie’s on British statistics (1979), sociologists have been exploring contested bodies of knowledge in a search for the causes that led to their acceptance or rejection. As the indefinite richness of experience would allow for an indefinite number of accounts to make sense of a given state of affairs, the reasons for taking both sides of a scientific controversy are found in other areas of the actors’ social experience. Typically, in this kind of study, relevant collective practices and the varied interests that orient them in their engagement with the world are taken as key explanatory factors. This broad notion of interest and its systematic use in explaining belief production, sustenance and change, has since become pivotal to much empirical work in the field of science and technology studies. Barnes’s interest 4 The term ‘contingent choice’ could be also used here, but it would background the unthinking, blind, and matter-of-course nature of the greater part of this activity.
8
Knowledge as Social Order
analysis and the associated methodological relativism offered to sociologists a model of explanatory social constructivism that allowed for the empirical study of knowledge in a post-foundationalist context. They are still among the most empowering tools available to sociologists engaged in a critical analysis of technoscience rather than in the contemplation of its practices. The same array of conceptual tools could also be deployed to explore and assess a range of debated epistemological issues, such as, for example, the difference between the knowledge produced by the asocial operations of machines and knowledge made by humans (Kusch and Collins 1998). Barnes’s notion of interest is characterised sociologically rather than psychologically, as emerging from social interaction and organisation, thus emphasising the priority of social intercourse over instrumentally oriented action. Interests, including interest in prediction and control, are not understood as ready-made tools for monocausal explanations, but rather as produced and sustained by a whole culture.5 With the deconstruction of the content of scientific knowledge in sociological terms such as collective practice and goal-oriented action, Barnes brought the 1960s process of theoretical reorientation of the social sciences to its extreme consequences. Not only, as shown by Garfinkel, was there no ready-made society for the analyst to describe and explain, but there was not even any scientific knowledge or privileged scientific method that could be indefeasibly referred to as an ideal of rational argumentation. By arguing that knowledge, like any other social institution, is the product of goal-oriented collective action, Barnes had thus effectively shown that no attempt to provide absolute epistemological foundations for any given body of knowledge could be taken seriously any more. Barnes’s exploration of reference was not only at the origins of major developments in the sociology of scientific knowledge, however. Through his early work on this topic, he emphasised that self-reference is an essential characteristic of statements about social reality. This strain of his research began with an awkwardly titled paper, ‘Social life as bootstrapped induction’ (Barnes 1983), where the key idea was that references to some objects are reckoned to be correct on the basis of the examination of a state of affairs external to them, rather than to some intrinsic property. Declaring someone an ‘outlaw’, or a ‘leader’, or turning a piece of metal into ‘money’ are typical examples of the kind of collective performative acts that Barnes has in mind here. Since Barnes’s pioneering work, the presence of such self-validating feedback loops – or ‘Barnesian performativity’ – have been detected across social life in a series of groundbreaking empirical works.6 The realization of the essentially self-referential nature of social discourse brought Barnes to focus his attention on statuses as key exemplars. A status only exists in so far as it is believed to exist and the object possessing it is treated as possessing it. ‘To come to believe something about the status of an individual’, Barnes wrote, ‘is to do two things at once: it is to accept a claim about his status and at the same time to contribute to the constitution of his status’ (Barnes 1988: 49). Society, Barnes suggested, can be conceived as a system of statuses, and therefore 5 For the debate on interest as explanatory resource see, for example Woolgar 1981, Barnes 1981, and MacKenzie 1981. 6 See, for example, MacKenzie 2006.
Introduction
9
as a distribution of self-referring and self-validating knowledge: knowledge about the roles of individuals, the prevailing routines, norms and practices in that society. Barnes developed this intuition in The Nature of Power (1988): where he defines society as ‘a sublime, monumental self-fulfilling prophecy’, a term that has clearly lost Merton’s pathological connotation. This way of conceptualising social life, Barnes argued, allows a solution to the problems of the nature and basis of power in society. If one looks upon a society as a distribution of knowledge that manifest itself in a set of routines and actions, then social power is the added capacity of action that individuals obtain through constituting this particular distribution, that is through action coordination. It follows, among other things, that social power and knowledge ‘are the same thing under different forms of awareness, and a change in the one does not cause a change in the other but rather entails it’ (Barnes 1988, 169–70). To be powerful is to be known to have, and thus to have, discretion over added capacities deriving from a specific form of action coordination – that is, from a specific distribution of knowledge. Power and knowledge should not be understood as two separate and interacting entities, as they are in Foucault for example;7 they are rather two sides of the same coin. Since the book on power, the nature of status has occupied a central role in Barnes’s reflection. The problem of how is it possible that we agree in our collective attributions of statuses and in our treatment of those to whom they are assigned is indeed one of the central questions of sociology. According to Barnes, little has been done in order to clarify this point, and most theorists seem content with identifying status as a ‘social position’. In particular, he considers unenlightening the use of class and class interest as analytic tools in mainstream British sociology. Instead, he elaborated upon a notion of status group inspired by the work of Max Weber, Randall Collins, and by his own work on science and scientific expertise – as scientists tend to constitute themselves precisely as status groups. Schematically, Barnes suggests that class action – as with any other form of collective action – is better understood and explained as the action of status groups. That is, as instrumentally oriented collective action sustained by restricted social intercourse. This move allows him to bypass numerous problems associated with the Marxian notions of class action and interest, while bridging the alleged gap between micro and macro theory. In discontinuing the use of class as key analytic term in favour of status group Barnes is thus inviting the replacement of an explanatory model in which interests engender organisation and action, with one where close interaction engenders a degree of organisation, and organisation is prior to the specific objectives towards which it can be later directed (Barnes 1995: 172–92). Barnes’s status group theory elaborates Weber’s vital insight that independent individuals could never constitute status groups, and that only genuinely sociable human beings could do so. Barnes’s interactionist and collectivist understanding of status group, and of social reality in general, is grounded on a sociability intrinsic to the nature of human beings. But of what does this sociability consists? Barnes tackles this question in an important article on status groups and collective action (Barnes 1992), where he demonstrates that rational, self-interested individuals could 7
On this point, see Kusch 1991, pp. 150–151.
10
Knowledge as Social Order
never produce status groups and collective action. Instead, Barnes argues that it is the existence of mutual symbolic sanctioning as an aspect of communicative interaction, and our susceptibility to this sanctioning, that make collective action intelligible.8 As with his earlier invitation to pursue a monistic understanding of science, Barnes’s work on status groups and on the necessity of genuine collective action to their functioning runs against assumptions that are deeply routed in mainstream philosophical and social theory. Barnes’s book on human agency (Barnes 2000) was conceived precisely as a critique of the facile moralistic mode of speech that pervades the social sciences at a time when individualism is running rampant and human agency is increasingly being spoken of as our independent power of choice. Based on his symbolic-interactionist understanding of groups and action, in this book Barnes identifies the putative rationality of human individuals as accountability, and their putative free will as their vulnerability and susceptibility to others. The discourse of agency and choice is itself understood as a medium of social control, and the communications that constitute it are treated by Barnes as speech acts with causal power. It is a discourse through which we affect each other and thereby mutually regulate our conduct. This understanding of human agency is currently being put at work at the empirical level, and Barnes himself is exploring if and how it allows a reformulation of some of the putative accounts of human action offered by geneticists and biologists (Barnes 1999; 2002a). The biological sciences, and especially genomics, are indeed the terrain in which Barnes has been exploring ideas about status groups and agency after he took up the co-directorship of the ESRC Centre for Genomics in Society at the University of Exeter. Barnes’s recent work on agency and the institution of responsible action should be seen as exploring another aspect of the same phenomena studied in his previous work on knowledge and power, and on scientific knowledge, rather than as a change of subject as some reviewers have implied. Overall, Barnes’s reflections on structure, belief and action provide an insightful and unitary perspective that allows original solutions to problems such as the relationship of structure and agency or that between the micro and macro level of analysis (Barnes 2001). Garfinkel famously claimed that old structural-functionalist theories were unsatisfactory as they turned individuals into ‘dopes’. As noted above, they have been largely replaced by accounts that emphasise the way in which active agents draw upon sets of rules to achieve their goals. However, as shown by Bruno Latour’s historical reconstructions, these accounts can easily turn into celebrative narratives of omnipotent agents. Or, as in Anthony Giddens’s individualist micro-causal account, agency can become an obscure hidden power, a metaphysical postulate. In contrast, Barnes’s seamless micro/macro sociology of interacting sociable agents producing and sustaining status groups, collective action, and extended status orderings – i.e. societies – offers powerful tools for a genuinely empirical and sociological analysis of collective action and belief. In his monist and collectivist account, human beings are certainly independent from and creative about rules and normative systems. However, it is an essential component of their humanity that they are not, and cannot be, independent from each other.
8 Barnes has used especially the empirical findings of Erving Goffman and Thomas Scheff to make his point about human sociability and interdependence.
Introduction
11
References Barnes, B. (1969), ‘Paradigms: Scientific and Social’, Man, 4, 94–102. ––– (1971), ‘Making Out in Industrial Research’, Science Studies, 1(2), 157–175. ––– (ed.) (1972), Sociology of Science (London: Penguin Books). ––– (1972a), ‘On the Reception of Scientific Beliefs’, in Barnes (ed.), 269–291. ––– (1974), Scientific Knowledge and Sociological Theory (London: Routledge). ––– (1977), Interests and the Growth of Knowledge (London: Routledge). ––– (1981), ‘On the “Hows” and “Whys” of Cultural Change (Response to Woolgar)’, Social Studies of Science, 11, 481–498. ––– (1982), T. S. Kuhn and Social Science (New York: Columbia University Press). ––– (1983), ‘Social Life as Bootstrapped Induction’, Sociology, 17(4), 524–545. ––– (1988), The Nature of Power (Urbana and Chicago: University of Illinois Press). ––– (1992), ‘Status Groups and Collective Action’, Sociology, 26(2), 259–270. ––– (1992a), ‘Realism, Relativism and Finitism’, in Raven et al. (eds), 131–148. ––– (1995), The Elements of Social Theory (London: UCL Press). ––– (1999), ‘Biotechnology as Expertise’, in O’ Mahoney et al. (eds), 52–66. ––– (2000), Understanding Agency: Social Theory and Responsible Action (London: Sage). ––– (2001), ‘The Micro-macro Problem and the Problem of Structure and Agency’, in Ritzer and Smart (eds), 339–352. ––– (2002), ‘Thomas Kuhn and the Problem of Social Order in Science’, in Nickles (ed.), 122–141. ––– (2002a), ‘Genes, Agents and the Institution of Responsible Action’, New Genetics and Society, 21(3), 291–302. ––– (2002b), ‘The Public Evaluation of Science and Technology’, in Bryant et al. (eds), 19–36. ––– (2005), ‘Elusive Memories of Technoscience’, Perspectives on Science, 13(3), 142–165. ––– (2005a), ‘The Credibility of Scientific Expertise in a Culture of Suspicion’, Interdisciplinary Science Review, 30(1), 11–18. Barnes, B. and Symons, M.C.R. (1966), ‘Electron Spin Resonance Spectra of Irradiated Potassium Persulphate’, Journal of the Chemical Society A, 66–77. Barnes, B. and Dolby, R.G.A., (1970), ‘The Scientific Ethos: a Deviant Viewpoint’, European Journal of Sociology, 11, 3–25. Barnes, B. and Symons, M.C.R. (1970a), ‘Radiation Damage in Persulphate Crystals’, Journal of the Chemical Society A, 2000–2002. Barnes, B. and Bloor, D. (1982), ‘Relativism, Rationalism and the Sociology of Knowledge’, in Hollis and Lukes (eds), 21–47. Barnes, B., Bloor, D. and Henry, J. (1996), Scientific Knowledge: A Sociological Analysis (London: Athlone). Bloor, D. (2003), ‘David Owen Edge: Obituary’, Social Studies of Science, 33, 171–176. Bryant, J., Baggott la Velle, L. and Searle, J. (eds) (2002), Bioethics for Scientists (Hoboken, NJ: John Wiley & Son).
12
Knowledge as Social Order
Hollis, M. and Lukes, S. (eds) (1982), Rationality and Relativism (Oxford: Blackwell). Kusch, M. (1991), Foucault’s Strata and Fields. An Investigation into Archeological and Genealogical Science Studies (Dordrecht: Kluwer). Kusch, M. and Collins, H. (1998), What Humans and Machines Can Do (Cambridge, MA: MIT Press). Lukes, S. (1975), ‘Review of Barry Barnes’s “Scientific Knowledge and Sociological Theory”’, Social Studies of Science, 5, 501–505. MacKenzie, D. (1981), ‘Interest, Positivism, and History’, Social Studies of Science, 11, 498–504. MacKenzie, D. (1979), Statistics in Britain, 1865–1930. The Social Construction of Scientific Knowledge (Edinburgh: Edinburgh University Press). MacKenzie, D. (2006), An Engine, Not a Camera. How Financial Models Shape Markets (Cambridge, MA: MIT Press). Nickles, T. (ed.) (2002), Thomas Kuhn (Cambridge: Cambridge University Press). O’ Mahoney, P. et al. (eds) (1999), Nature, Risk and Responsibility: Discourses of Biotechnology (London: Macmillan). Raven, D., van Vucht Tijssen, L. and de Wolf, J. (eds) (1992), Cognitive Relativism and Social Science (New Brunswick, NJ: Transaction Publishers). Ritzer, G. and Smart, B. (eds) (2001), Handbook of Social Theory (London: Sage). Shapin, S. (1975), ‘Phrenological Knowledge and the Social Structure of Early Nineteenth-Century Edinburgh’, Annals of Science, 32, 219–243. Woolgar, S. (1981), ‘Interest and Explanation in the Social Study of Science’, Social Studies of Science, 11, 365–394.
Chapter 1
Relativism at 30,000 Feet David Bloor
In his book River Out of Eden the well-known biologist Richard Dawkins takes a swipe at something he calls ‘cultural relativism’. It is, he says, a ‘fashionable salon philosophy’ and its adherents clearly irritate him. These people, in Dawkins’s view, need to confront a few simple and undeniable facts: facts such as, ‘Airplanes built according to scientific principles work [...] Airplanes built to tribal or mythological specifications [...] don’t’. He throws down a challenge: ‘Show me a cultural relativist at thirty thousand feet’, he says, ‘and I’ll show you a hypocrite’. Who are the targets of Dawkins’s challenge? He is quite precise on the matter. They are people who think it clever to say that ‘science is no more than our modem origin myth’. They assert that ‘science has no more claim to truth than tribal myth’ and that ‘Neither way is more true than the other’ (1995).1 I share Dawkins’s impatience with these formulations, but the question I want to ask is what happens when his challenge is directed at other forms of relativism, for example those found in the history and sociology of science. Advocates of what is sometimes called the ‘strong programme’ certainly don’t say that science is no more true than myth, indeed, this is explicitly rejected (Barnes and Bloor 1982: 1–47).2 Their position is that the truth of both is, in a sense, equally problematic. Both bodies of belief require the local causes of their credibility to be investigated. Credibility is always to be viewed as an empirical problem and sociological variables will always be involved. Is this subject to Dawkins’s knockdown argument? I don’t know what answer Dawkins would proffer to this question. I do know that others think that such relativists can’t make sense of a technological feat like flying at thirty thousand feet. A case in point is Christopher Norris, in his book Against Relativism (1997). This includes a sustained attack on ‘the strong sociologist’s programme’ whose claims, says Norris, tempt him to an exasperated endorsement of Dawkins’s challenge – which is duly quoted (on p. 314). This is the line of thought I want to address. I think it is wrong. There is no inconsistency or hypocrisy involved in a strong programme, relativist account of the technology of flight. It is those who 1 Dawkins prudently adds that what he calls his ‘knock-down argument’ is aimed strictly at such persons. There are others, he says, who call themselves ‘cultural relativists’ but have perfectly sensible views, for example the view that ‘you cannot understand a culture if you try to interpret its beliefs in terms of your own culture’ (1995: 31–32). 2 The explicit repudiation of the position Dawkins is attacking is to be found on pp. 22– 23.
14
Knowledge as Social Order
think they can generalise Dawkins’s accusation to cover this sort of relativism who need to confront reality and examine their conscience. The Nub of the Problem The nub of the problem is how we should analyse practical success in dealing with our environment. That we do have a degree of success is obvious and the aeroplane is a good symbol of it and of all the complexities involved. A widely accepted starting point for the analysis of practical action invokes the idea of trial and error. This seems applicable to the discovery of the secret of flight.3 First there was the desire to emulate the wonderful ability of birds. People tried to build bird-like, flapping wings and failed. All manner of contraptions were assembled and discarded. Fixed-wing gliders, what we would call ‘hang-gliders’, represented the real break-through. Otto Lilienthal built and flew these with some success in the 1890s though he eventually killed himself experimenting with a new wing. He brought to the task not only courage but an engineer’s systematic, empirical curiosity. Similarly with the Wright brothers who took good care to inform themselves of Lilienthal’s work. Their goal was powered flight. They too did a lot of experimenting with models of different wings. They built a small wind-tunnel to test different cross-sections, though there was a lot they did not understand about wind-tunnels. Nevertheless their work still represented a further stage in the process of moving carefully back-and-forth over the terrain, accumulating and recording data and becoming ever more familiar with its intimate details. On 17 December 1903, they achieved their goal.4 I shall pick up the historical story again in a moment, but let us ask what we, as analysts of the process, should be making of all this? There are methodological choices to be made, and these will have consequences for the question of relativism. The big choice is this: should we see the trial and error of Lilienthal and the Wrights as itself a natural phenomenon? Or are we going to allow it to be glossed in nonnaturalistic terms? I should explain what I mean by the word ‘naturalistic’. The best way is to start with the work of experimental psychologists. They try to build causal models of learning and adaptation, hence all those experiments on rats in mazes and all the work on human perception, memory and learning. We could say their aim is to build
3 For a thorough, clear and technically sophisticated account of the history of flight and aerodynamic theory see Anderson (1997). A good illustration of the range of themes and approaches demanded by the history of flight is provided by Peter Galison and Alex Roland (2000). 4 Why did the Wright brothers succeed when others failed? The interesting suggestion has been made that they had a more efficient search strategy: they explored a ‘function space’ rather than a ‘design space’. They built fewer complete experimental aircraft than their rivals who seemed to keep trying out different numbers of wings, different positions for the wings, different power-plants etc. The Wrights identified functional problems such as how to achieve stability, lift etc. (see Bradshaw 1992: 239–250).
Relativism at 30,000 Feet
15
a ‘learning machine’, for example, one that would model our tendency to extrapolate from experience and generate inductive inferences.5 Of course, the naturalistic stance doesn’t end with psychologists. It is shared by many practitioners of anthropology, sociology and history who adopt a matter-of-fact approach to knowledge. They bring into the story a concern for interaction, convention and shared cultural resources. The ultimate learning machine is a social institution. Two points need underlining. First: a naturalistic, causal analysis of knowledge does not challenge the reality of knowledge. Scepticism is not at issue. To explain is not to explain away. Psychologists studying vision do not conclude that we cannot really see. What they challenge are alternative accounts of knowledge that are naive, metaphysical or non-naturalistic.6 Secondly, I accept that we may be a long way from an adequate analysis of cognition, but the issue here is the goal of our efforts, rather than their current success. The aim is to see our own cognitive achievements as part of nature: as causal, law-like, material processes. Make no mistake about it. If we opt for the naturalistic approach to knowledge, relativism will come out as the winner. I realise that the argument has moved rather quickly here, linking relativism and naturalism in such an immediate way. In order to consolidate this link let me go back to the definition of relativism and offer you some further resources for thinking about its relation to what I have called ‘naturalism’. Then we can get back to aeroplanes. Relativism versus Absolutism So far I have characterised relativism very simply. It is the idea that both true and false beliefs are, with regard to the causes of their credibility, equally problematic. I now want to make two farther suggestions about the definition of relativism. One will concern a ‘formal’ requirement on any adequate definition; the other will concern the metaphors and comparisons that give substance to the formal specification. Formally relativism is the opposite of absolutism. If the word ‘relativism’ is to mean anything it must stand in clear opposition to something else that can qualify as ‘absolutism’. To put the point in another way: if you are not a relativist about knowledge, you are an absolutist. If you reject cognitive relativism then be prepared to embrace and defend cognitive absolutism. I am assuming here that to qualify as a relativist it is necessary to deny that there are any instances of absolute knowledge. An absolutist, by contrast, asserts that there are at least some instances of knowledge with an absolute character. 5 One of the pioneers of an explicitly mechanistic approach (though, of course, not the first) was the influential, Edinburgh trained, Cambridge psychologist Kenneth Craik (see Craik 1943 and 1966). 6 It is unfortunate that the point needs underlining but the contrary , sceptical, view is frequently attributed to relativist sociologists of knowledge. Norris’s Against Relativism (1997) proceeds on the assumption that the strong programme is an exercise in scepticism (‘strong sociologists and other sceptics’, 1997: 303). Another such imputation, directed against the present writer, is to be found in Franklin (2002: 621). For very good reasons, these imputations are not supported by evidence or quotation.
16
Knowledge as Social Order
Relativism and absolutism thus stand as jointly exhaustive and mutually exclusive, but there is still the problem of giving substance and content to the contrast. We have got to get this right otherwise the boundary will be drawn in the wrong place. We must not let some insipid, ersatz candidate pass muster as authentically absolute when it is no such thing. To anchor the distinction in the right place we need more than a few slippery, verbal formulae: we need to grasp the basic ideas that are at stake. We need a model of what it is for something to deserve the label ‘absolute’. To provide such a model we must go to the area of our culture where our intuitions about the word ‘absolute’ are at their firmest and where the category seems most at home. This, of course, is theology and our talk of God. Our best and most enduring exemplar of something absolute is God. God does not come into existence and then go out of existence. He is eternal. God does not change and decay. He is perfect and unchanging. God’s properties do not depend on this or that precondition or contingency. He is necessary and the uncaused cause of all else. He transcends causality. Again, God’s commands are not made by reference to anything external that validates, justifies or constrains them. They are infinite in scope, ultimate and self-sufficient. The attributes of God are absolutes, or they are nothing, and our attitude towards them must be the appropriate one: they are not useful fictions or conjecture but the objects of faith. This is the point I want to emphasise. When we accord a knowledge claim the status of ‘absolute’ it must display an appropriate analogy with properties we should want to attribute to God. It may not share all of these, but it surely must share some. If it doesn’t then we can be quite sure that our attribution is suspect and that our standards, as it were, are slipping. Now you can see the strategy of my argument. It is to force the critics of relativism to acknowledge their absolutism and then to impress upon them the demanding character of the position they have wittingly or unwittingly chosen to occupy. The relativist is in a relatively comfortable position. There are no absolutes in the contingent realm of nature where everything is bound up in causal relations or arises meaninglessly by chance. For the relativist our knowledge is just one more phenomenon within this realm. The relativist is at home with conjecture, inconsistency, expediency, partiality, a mixture of success and failure and all the shortcomings of this vale of tears. For non-relativists things are not so comfortable. They can acknowledge the existence of things with a merely relative status, but only as stations on the road to the absolute. The familiar pragmatics of scientific work (theories that are approximate, that have limited scope, that are inconsistent, that get this bit right and that bit wrong, or that are accepted as useful fictions) cannot themselves belong to the realm of absolute truth. To qualify as absolute, a knowledge claim must transcend these limitations. It may sound strange to put it in this way, but the critics of relativism have something like a theological problem on their hands. It is a form of what theologians call the problem of incarnation. To ask how absolutes are manifest in the natural world is like asking: how was God made flesh? How did an infinite spiritual being come to be the finite human person that was Christ? Remember that orthodox Christian theology stipulates the identity of God and Christ while insisting that this in no way compromises either the reality of the Godhead or the true humanity of
Relativism at 30,000 Feet
17
Christ. This is acknowledged as a mystery that cannot be penetrated by the human intellect.7 I fear a similar impasse awaits the anti-relativist. If this sounds like a fanciful comparison recall the situation in the philosophy of mathematics. Those who adopt what is called a ‘Platonizing’ stance think of mathematics as a set of truths about a realm of abstract objects. They then confront the problem of explaining how the human brain can grasp such truths and connect itself to this realm. This was Gotlob Frege’s problem in his Foundations of Arithmetic (1959). He poured scorn on the relativizing tendencies of his contemporaries who appealed to ideas from psychology or evolution. Numbers, for Frege, are selfsubsistent objects – selbständige Gegenstände (p. 73). They are non-physical and non-sensory and outside space and time (‘raum-und zeitlos’ – p. 15 – ‘weder räumlich und physikalisch [...] noch auch subjective [...] sondern unsinnlich und objectiv’ – p. 38). So how do we know them? How is the Platonic heaven of number brought down to earth? Frege talked of an intellectual or spiritual activity (‘eine geistige Thätigkeit’) which can ‘fasten upon’ concepts (‘anknüpfen’ was the verb he used – p. 32). These obscure metaphors were all he had to offer. This is the quasitheological impasse of which I spoke. Not all anti-relativists are explicit Platonists like Frege, but what he did blatantly and on a large scale others, sooner or later, must do surreptitiously and on a small scale.8 We can now anticipate the strategy the anti-relativists will adopt if they cannot confront this truth about themselves and explicitly embrace the absolutism to which they are committed. They will deploy their rhetorical skills to disguise their resort to the absolute. They will seek to break down the distinction between the sacred and the profane and to undermine the ‘otherness’ of the absolute. They will try to ‘domesticate’ the absolute in order to convince us that it is something familiar we meet with every day. Before I examine this response, however, there is a confusion that I need to clear up. It concerns the relation of relativism and idealism. Relativism and Idealism Critics of cognitive relativism frequently confuse relativism with idealism, as if they were the same thing or as if a relativist were logically committed to idealism. Idealism has a long and complex history, but for our purposes it is sufficient to see it as comprising two main claims. First, that there is no independent object of knowledge distinct from the knowing subject and, second, that the knowing subject has the
7 The orthodox position was defined at the Council of Chalcedon in AD 451 (see Hall 1991, ch. 21). 8 An example of this impasse is to be found in the reflections of an eminent physicist on the question of why mathematics applies so successfully to the material world. E.P. Wigner writes, ‘the enormous usefulness of mathematics in the natural sciences is something bordering on the mysterious and there is no rational explanation for it’ (1967: 223). In fact he says it is a ‘miracle’ (1967: 237).
18
Knowledge as Social Order
character of a mind, soul, spirit or a centre of consciousness. For the idealist, the material object of knowledge is collapsed into the mind of the knowing subject.9 One can see at once how one might be tempted to criticise a simple-minded idealist by appeal to our ability to fly high into the air. The ground looks a frighteningly long way down. We don’t want to fall and this may serve as a reminder of our instinctive commitment to the reality of the material environment. This is a version of Dr. Johnson’s response to Bishop Berkeley. Berkeley said that ‘to be was to be perceived’, so Johnson kicked a rock as a knock-down proof he was wrong.10 Those who raise a cheer at Dawkins challenge may well be thinking along these lines. If so, then their target should be idealism not relativism. We must not confuse them. The ideal-material dichotomy is not the same as the relativeabsolute dichotomy. One is ontological the other is epistemological. The opposite of ‘idealism’ is ‘materialism’, not ‘absolutism’. Time after time critics of relativism run these logically distinct oppositions together and collapsed them into one confused hybrid.11 A sense of cultural history should have been enough to warn of the danger. Historically, cognitive relativism has typically been associated with materialism, not idealism.12 It is the believers in absolutes who have typically embraced idealism. 9 The complexities that I am putting aside here concern the historical dimension of many idealist theories. Historically the idealist story is simply a transfigured form of the Christian story: some form of alienating distance between matter and spirit is progressively overcome in a final, redemptive unification – when, as Hegel put it, ‘consciousness itself grasps ... its own essence’ and Absolute Spirit finally reaches its Calvary (Hegel 1977: 57, 493). 10 Was Berkeley really refuted by this? However much one may sympathise with the unity of theory and practice implicit in Johnson’s gesture, a proper refutation would surely have to take into account the role accorded to God in Berkeley’s story. 11 To give just one example, the slide is evident in the article on epistemological relativism in Honderich (1995: 757). The writer informs the reader that ‘full-blooded’ relativism involves idealism. The implication is that when relativism is revealed in its true colors the idealism implicit in it will emerge. It might be thought that there is some justification for this assumption given the terminology adopted by Berkeley himself. In the Principle of Human Knowledge we read, ‘For as to what is said of the absolute existence of unthinking things without any relation to their being perceived, that is to me perfectly unintelligible’ (see Campbell 1911: 34). This describes independent, unperceived matter (which of course Berkeley rejected) as having an ‘absolute’ character and so appears to put his idealism on the side of (some form of) ‘relativism’. Terminologically this deviates from my use of these labels but it in no way justifies the confusion of the ontological with the epistemological issue. The doctrine that there are no knowledge claims that can qualify as having an absolute status is and remains different from the doctrine that there exists no independent material world. One can, however, see why Berkeley conjoined the concepts of the ‘absolute’ and the ‘material’ in the way he did. It was because he wanted to argue that those who believed in matter would end up worshipping it. Matter would displace God and thus usurp the role of the absolute. ‘Matter once allow’d. I defy any man to prove that God is not matter’ (see Luce 1945: 72). 12 Historically, materialism has been the ideology of those who have struggled against ‘priestcraft’. Such relations, however, are always contingent rather than necessary. For a fine study which explores these connections in the context of the history of science, see Desmond (1989).
Relativism at 30,000 Feet
19
Mind and spirit are the traditional homes of absolute truth. Historically, relativism and idealism have been on opposite sides of long-standing ideological and political divides and yet, oblivious of the barricades, the critics of relativism slide from one to the other as if they can take the association for granted. This is what Norris does in Against Relativism.13 It may help to cut through some of the verbal confusions if we imagine a twoby-two matrix with the relativist-absolutist dichotomy located along one side and the materialist-idealist dichotomy along the other side, set at right angles. There are then four possible combinations. The position I am defending is the combination of relativism and materialism. Perhaps the most puzzling of the four possibilities is the combination of materialism and absolutism. What on earth is materialist absolutism? Actually, this mysterious combination is sometimes graced by the name ‘Realism’, a label that only obscures the difficulties. Sometimes the word ‘realism’ just means ‘materialism’ but, as here, it often indicates a hybrid combination of materialism and the correspondence theory of truth. Absolutism can then be smuggled in by glossing correspondence in an appropriately absolutist way.14 Whatever we call it, this position requires the overcoming of a dualism. The difficulty is the one I alluded to earlier when, using the example of Frege, I spoke of the absolutist facing a version of the theological problem of the incarnation. The absolutes must be made incarnate in material form.15 Let us now look at a strategy for trying to overcome this problem. It is the one I mentioned earlier, namely, trying to convince us that absolute knowledge is really 13 Thus, ‘[…] relativists routinely deny […] the existence of a real-world (mind-and belief-independent) physical domain […]’ (Norris 1997: 314). Norris’s discussion throughout chapters 9 and 11 of Against Relativism is ostensibly directed at the strong programme but systematically misrepresents the position. Not only is it presented as idealist rather than materialist it is understood as saying that only socio-cultural factors play any role, as if people’s capacity to observe the world, and act on it and in it in an intelligent way, were of no account or were somehow outwith the causal perspective. On p. 313 the argument also seems to assume that psychological processes are excluded or denied in the strong programme (see also p. 314–316, and numerous other places in these chapters). This is a pity because one of the examples used in these chapters concerns the history of aeronautics and it would have been useful to have a proper confrontation rather than one in which the position of one side is not properly expressed. It is indicative of the misunderstanding that the sub-title of chapter 9 is ‘Aerodynamics as Test Case for Anti-Realism’, but see the comments on anti-realism below. 14 Notice this renders ‘anti-realism’ ambiguous. Does it mean anti-materialism or anticorrespondence theory? The strong programme is only anti-realist in so far as it embodies opposition to an absolutistic form of correspondence. 15 There is one area where a circumscribed version of relativist idealism is appropriate and to be recommended (see Barnes 1983). The late Prof. Anscombe introduced the term ‘linguistic idealism’ for those cases, discussed independently by Barnes, where there is no independent object of discourse and where the users of the discourse create the object through their references to it. Anscombe made it clear that this form of analysis is appropriate for social and moral entities like rules and obligations. It should be clear, however, that ‘linguistic idealism’ actually presupposes a background of commonsense materialism (see Bloor 1996).
20
Knowledge as Social Order
familiar and commonplace. Given that it is so familiar, the argument goes, there can’t really be a problem about making the absolute incarnate, can there? A Quotidian Absolute? One way to make the idea of absolute truth appear non-mysterious would be to argue that any attribution of truth is, automatically and by its very nature, an attribution of absolute truth. If a proposition is true at all, then it is absolutely true. In support of this it could be argued that our concept of truth has no place in it for relativity. Relativists may talk of something being ‘true for’ some group or ‘true in’ one culture but, the argument goes, this is confused talk. To believe a proposition is to believe it is simply (and absolutely) true, not that it is merely ‘true for’ us. There is something right in this. Talk of ‘true for’ is suspect. However many people believe something, we can always ask: but is it really true? This is so central to the concept of truth that we can say it is part of its definition. But this definition can’t generate the existence of absolutes. It doesn’t follow that we have a hold on knowledge that is absolute in any substantial way. It no more proves this reality claim than putting existence into the definition of God proves that anything exists which satisfies the definition. If we cannot conjure absolutes out of the general definition of truth perhaps we can do better if we look at specific cases where we impute truth. Alfred Tarski codified some of the common-sense conditions of adequacy that must be satisfied and, significantly, his approach is sometimes called an absolute theory of truth.16 We want to domesticate the absolute so take a homely case. The sentence ‘The cat sat on the mat’ is true if and only if the cat in question indeed sat on the mat in question. We would normally say: the words are true if they fit the facts. If the fit seems perfect why not say the truth is absolute? Perhaps we are, even now, in the presence of absolute truth. This is the position adopted by John Searle in the otherwise valuable discussion of the correspondence theory in his book The Construction of Social Reality (1995, ch. 9). Suppose we have in some appropriate way defined the word ‘cat’. Searle says, ‘But once we have fixed the meaning of such terms in our vocabulary by arbitrary definitions, it is no longer a matter of any kind of relativism or arbitrariness whether representation-independent features of the world satisfy those definitions’ (p. 166). Notice that if it is not a matter of any kind of relativism we must have achieved a state of affairs that can be called ‘absolute’. Searle’s position is an example of how the correspondence theory can be given an absolutist gloss.17 For a naturalistic thinker, however, the ‘fit’ between words and the world should never be treated as absolute. It is always complex, partial, problematic and potentially changing. How does this apply to the sentence about the cat and the mat? Well, first of 16 This is what Karl Popper calls it (see Popper 1965: 223). 17 Searle offers as his reason why no kind of relativism is involved that the features which do or do not fit the words ‘exist independently’ (p. 166). I do not see how this supports the conclusion. It looks to me as if Searle is conflating the categories of truth and reality, i.e. confusing relativism and idealism.
Relativism at 30,000 Feet
21
all, those who use and evaluate this sentence, and accept or reject its truth, must have learned the relevant words and concepts. They must have acquired the words ‘cat’, ‘mat’, ‘sat’ and ‘on’ in relation to a specific and finite set of examples. The words must have been introduced by ostensive definition. This makes their subsequent extrapolation and their application to further instances a matter of judgement. It involves a judgement of relevant similarity. We are guided by analogy with the past applications that have been collectively endorsed and conventionalised. Judgements of similarity and analogy are by their very nature imperfect and fluctuating things. They are influenced by past history, by background knowledge, by the particular cases already encountered, by the purposes informing the classification, by precedents set (and decisions taken) elsewhere on related matters, and so on.18 Searle’s line of argument (and it is a very common one) is based on a false picture. Contrary to his assumptions, meanings are not the sort of things that can be ‘fixed’. Here is the point at which the theorist jumps out of the realm of the natural world and manipulates ideas in a way that only make sense in non-naturalistic terms. Frege’s Platonizing tendency may have been thrown out of the front door but a disguised version of it has crept in at Searle’s back door. Rather than allowing that meanings can be ‘fixed’, the relativist will follow Wittgenstein and insist that meanings are created as usage proceeds. Meaning follows from use, not use from meaning. Meaning has none of the hall-marks of the superlative or metaphysical. There is nothing absolute here. We are in the realm of contingency, conventionality, causality and relativity.19 If the particular example that I have used – the sentence about the cat on the mat – looks as if it could fit the world in some surpassingly immediate way, then this is an illusion. It is produced by the stereotyped character of the example, and perhaps by the hypnotic effect of repeating the same words when formulating the sentence and the fact that ‘fits’ it. The endlessly problematic character of the ‘fit’ between words and the world becomes more evident if one moves away from stereotyped examples to the more complex cases found in science and technology. Here the analyst (the historian or sociologist of science) becomes acutely aware of how concepts change and develop, of how their application is subject to constant revision and adjustment. In other words, the ‘conventional’ character of the concepts becomes more visible with scientific examples.20 To show this I want to take up the historical narrative again and go back to aeroplanes. 18 For a fine statement of this account of meaning see Barnes (1982, ch. 2). 19 It is important to notice that we define for ourselves what counts as fitting or corresponding as we go along. What is more, this is a case where Anscombe’s category of linguistic idealism can be put to work. There is a parallel here with an important feature of scientific thought first identified by Thomas Kuhn. In his paper ‘The Function of Measurement in Modem Physical Science’ (1961), Kuhn reflected on the meaning of numerical tables which set out, side-by-side the predictions of a theory and the measurements from experiments. Their function, he said, was not so much to confirm the theory as to define what was to be counted as confirmation. In our terms it was to exhibit what counts as ‘correspondence’ or ‘fit’. The tables did not report a correspondence; they were constitutive of what correspondence was to be (see Barnes 1982: 21). 20 This position might be called ‘meaning finitism’ (see Hesse 1974). For a discussion of theories of truth in the light of meaning finitism see Kusch (2002, ch. 16).
22
Knowledge as Social Order
From Practice to Theory The crowds in both America and Europe who came to see the Wright brothers demonstrate their skill (they were soon staying aloft for half an hour) could no longer doubt the reality of powered flight. But how did the wings keep the machine in the air? What about the fundamental aerodynamics of flight? This, too, must be given a relativist analysis. Basic Newtonian principles told scientists and engineers to think in terms of a resultant force, directed upwards, generated by the flow of air over the wings. The resources for understanding the flow are twofold: theoretical and experimental. Theoretically it is necessary to construct a model of the flow in the imagination, to show by reference to the basic principles of fluid mechanics that the flow could generate the requisite forces and then, by empirical study, convince oneself that the supposed flow is really present. Things need not be done in that order: the experiments might come first and suggest the model, but somehow these components must be woven together. A major problem was that the governing equations of fluid dynamics, the formidable Navier-Stokes equations, were mathematically intractable. They were non-linear, partial differential equations which took account of both viscosity and compressibility. The standard response was to ignore these variables in order to simplify the equations. Unfortunately, they then no longer dealt with a real fluid but only with an ‘ideal fluid’. The cost was high. A wing-like shape set in uniform motion through a continuous, stationery body of ideal fluid would indeed experience a calculable lift and drag, but the calculated value was precisely zero in each case. The possibility of understanding how one night gets to 30, never mind 30,000, feet looked remote. A way round this difficulty came from two applied mathematicians: one a Russian, Nikolai Joukowsky; the other a German, Wilhelm Kutta (see Joukowsky 1910; 1912; and Kutta 1910). In the early 1900s, stimulated by Lilienthal’s flights, they independently suggested superimposing two flows: one a uniform flow over the wing, the other a circulating flow around the wing. The idea is that the actual flow is such that it can be analysed into these two components. This combination, even in an ideal fluid with no viscosity, could generate a lift. This is because the circulating flow works against the main flow on the lower surface of the wing and works with the flow on the upper surface of the wing. The resulting speed of the air is therefore lower beneath the wing than above it. Since the pressure of the air is inversely related to the square of its speed there is an imbalance of force with a resultant pushing upwards, providing the lift. The formula for calculating the lift even turned out to be very simple. The lift, L, is simply the product of three quantities: the density of the air symbolised by ρ , the velocity of the main flow approaching the wing symbolised by V, and a quantity, gamma, Γ, called the ‘circulation’. The Kutta-Joukowsky formula is; L=ρVΓ
Relativism at 30,000 Feet
23
A number of things about this formula and the theory behind it deserve notice. First, if this theory could, in principle, explain the presence of a lift force, it still predicted a zero value for the drag. This is, of course, a strikingly false prediction. Secondly, the theory does not explain where the circulation comes from. Its causes remain obscure and, within ideal fluid theory, there are no resources for ever providing an explanation.21 Thirdly, the theory does not predict the magnitude of the circulation. Kutta and Joukowsky did, however, notice that one and only one value for the circulation would ensure that the flow came away smoothly at the trailing edge of a wing and that, for a given angle of attack, the flow did, indeed, settle down in this form. This became known as the Kutta-Joukowsky condition. Fourthly, and last, the theory did not apply to a real aircraft wing of finite span. It dealt with an abstraction, albeit a very useful one, called an ‘infinite’ wing. The analysis proceeds on the assumption that the flow across every cross-section is identical when, in reality, it must be different. Regardless of these simplifications the practical question is whether the theory generated good, quantitative predictions about the amount of lift. It is necessary to say here that, while two of the quantities in the formula, the density and the speed of the wing through the air, can be measured easily, the so-called ‘circulation’ is much more difficult to get at. Mathematically the circulation is defined as the value of a certain line-integral. You imagine, say, a large circle, that goes round the cross section of a wing. You then add up all the velocity components that are tangential to the circle. That gives the magnitude of the quantity gamma in the formula. It was not until about 1924 that anybody managed to put a wing in a wind tunnel, measure its lift, take the measurements needed to compute the circulation, and then see if the two were related as the formula said. When they eventually did, there was reasonable agreement – though the measured circulation turned out to be sensitive to the choice of contour along which it was determined when, theoretically, it should have been independent (Bryant and Williams 1924).22 After Kutta and Joukowsky the next big step was taken by the brilliant engineer Ludwig Prandtl at the University of Göttingen.23 Ask an aerodynamicist when the workings of a wing were first understood and the chances are he or she will refer to Prandtl’s classic paper Tragflügeltheorie of 1918. If the Wright’s gave us the practical know-how, Prandtl gave us the theoretical knowledge to make sense of it. Prandtl was able to shed light on the origin of the circulation by proposing that as a wing 21 The fact that circulation and vorticity cannot be created or destroyed but must be present or absent for all time has encouraged theological reflections. ‘A vortex in a perfect fluid is therefore permanent and indestructible; and the enunciation of these properties by Helmholtz suggested to Lord Kelvin that vortex rings are the only true atoms, inasmuch as the generation or destruction of vortex motion in a perfect fluid can only be an act of creative power, a theory long since abandoned.’ (Ramsey 1935: 216). 22 This paper also played a significant role in overcoming the scepticism felt by British aerodynamicists towards the circulation theory – but that is another story. 23 Prandtl would have been famous even if he had never turned to aerodynamics: he had discovered the boundary layer and the boundary-layer equations in 1904 (this was a vital step in making the abstract equations of ideal fluid theory applicable to real fluids – see Busemann 1959).
24
Knowledge as Social Order
begins to move the air curls round the trailing edge, in the way predicted by ideal fluid theory, but then detaches itself in the form of a vortex which floats off down stream. This starting-vortex, as it is called, will itself have a circulation, but since the original circulation was zero, and once zero always zero, another, contrary circulation of the same magnitude must be created to offset it. This is the circulation round the wing. Prandtl called it a ‘bound-vortex’ to distinguish it from the starting vortex and signalise its novel character. It is novel because it disobeys the fundamental laws of vortex behaviour previously introduced into fluid mechanics by Helmholtz. It does not, throughout its existence, always consist of the same fluid particles. For this reason Prandtl insisted that it had no physical reality. It was an ‘abstraction’, a mathematical fiction.24 Prandtl also took the great step of generalising the circulation theory from the idealised, two-dimensional, or ‘infinite’ wing, to the more realistic, threedimensional, finite wing. One should not imagine, however, that as the theory is made more realistic in this respect, the idealisations and the theoretical fictions were progressively stripped away. There was no move towards some ideal of pure description in which conventionality was transcended by an immediate and direct correspondence with reality. On the contrary, Prandtl introduced his so-called ‘lifting-line’ theory. The wing, as it exists in reality, and as we can picture it in our mind, is replaced for the purposes of calculation by a simple, straight, geometrical line. The line is taken to be the centre of a vortex motion equivalent in strength, per unit length, to the circulation introduced by Kutta and Joukowsky. This gave Prandtl a way to think about the unsolved problem of drag which, in many respects, was even more difficult to explain than lift. He imagined the circulation around the wing continued at the wing tips in the form of two trailing vortices leading back from the wing tips. These absorbed energy and generated drag. But a single lifting line was too crude an approximation. To do greater justice to the contribution of different parts of the wing to the overall lift, in particular to account for the fall-off of lift at the tips, Prandtl then replaced the wing by a whole set of such bound-vortex lines of different length. The point is this: Prandtl was substituting a mathematically tractable simplification for the reality that was too complex to work with. If he used a mixture of ideas, some drawn from ideal-fluid theory and some from the behaviour of real fluids with viscosity, then so be it. Logically this rendered the theory inconsistent, but it was expedient because it allowed the work to go forward.25 From Theory to Fact In drawing attention to these facts about the work of Kutta, Joukowsky and Prandtl, I am not making any criticism of them. On the contrary, what stands out is their admirable pragmatism and creative ingenuity in exploiting and extending the 24 ‘Der Begriff des tragenden Wirbels hat keine physikalische Realität, sondern bedeutet eine Abstraktion [...]’. This formulation comes from Tietjens (1931: 177–178). 25 This inconsistency is noted in Moritz Epple, ‘Präzision versus Exaktheit’ (2002). I should like to thank Michael Eckert for drawing my attention to this paper.
Relativism at 30,000 Feet
25
intellectual tools at their disposal. But what must be stressed is the impossibility of glossing all this in absolutist terms. An absolute truth cannot be a useful fiction or inconsistent within itself; it cannot be a partial or pragmatic approximation or a mixture of right and wrong. Absolute truths are not tailored to the demands of expediency. Could the absolutist take refuge in the idea of progress, for example, blaming all these features of knowledge on its incomplete and undeveloped status? It is true that certain contradictions are removed as theories get tidied-up, but as old contradictions are removed new ones are frequently generated (see the section called ‘Paradoxes of Airfoil Theory’ in Birkhoff 1950: 15–19). As old fictions, approximations and expedients outlive their usefulness, new ones take their place at the front line. And some problems linger on unresolved. Long after the circulation theory was accepted as the true account of lift, the reasons why the circulation obeyed the KuttaJoukowsky condition remained obscure.26 It is important to learn to live with these facts about knowledge, and to treat them as facts about real knowledge, not about pre-knowledge or sub-knowledge or second-rate knowledge. For the naturalist, materialist and relativist, Prandtl and his colleagues knew around 1918 why a wing generated lift. It would be fatal to allow knowledge to be defined in absolute terms. Follow that route and knowledge will always be over the horizon. I wonder if those who cheer Dawkins see the danger? It is the critics of relativism, those who embrace absolutism, who are leading us towards sceptical conclusions. They are pitching the standard so high that no science, and certainly not Prandtl’s aeronautics, will ever satisfy it. Any absolutist who doubts this should read the perceptive account of aerodynamics and engineering in Vincenti (1990). What about other appeals to progress? Surely we can say that the empirical reality of the circulatory component in the air flowing over a wing had been made evident, not perhaps by 1918, but certainly, on my own admission, by 1924? Theoretical obscurities may have remained but, at a certain point, surely, the idea of circulation stopped being speculation, or a useful mathematical fiction, and became a clear fact. For example, the circulation around the wing expressed itself in trailing vortices which come off the wing-tips and under certain circumstances can actually 26 Writing as late as 1967 an eminent Cambridge authority on fluid dynamics G. K. Batchelor said of this smooth flow from the rear edge: ‘It was used as an empirical rule in the early development of aerofoil theory, but current knowledge of boundary layers enables us to account, at any rate in qualitative terms, for the establishment of the circulation with a specific value’ (1967: 437). The specific value in question is the one at which the circulation puts the rear stagnation point of the flow precisely at the trailing edge thus canceling the infinite velocities that would arise. ‘It is a remarkable fact’, says Bachelor, ‘that in practice a circulation is generated around an aerofoil, owing to the convection of a non-zero amount of vorticity from the rear edge of the aerofoil at an initial stage of the motion, and that when the aerofoil is in steady motion the circulation is established with just this special value’ (1967: 437). That the effect of viscosity in the boundary layer at the inception of motion should produce just the amount of circulation ‘that enables effects of viscosity to be ignored […] in the subsequent steady motion’ is called a ‘fortunate circumstance’ (p. 437). These comments may be contrasted with the rather uncritical discussion of this topic provided by Norris (1997: 315).
26
Knowledge as Social Order
be rendered fully visible to the observer. They can even be photographed. From being what empiricist philosophers call a theoretical term ‘circulation’ became an observation term and described a clearly known fact of the matter. This is right, as long as we remember that there are no such things as ‘pure’ observation terms.27 As more work was done with wind tunnels, and more and more detailed measurements were made, an ever-more fine-grained picture of the flow of air over a wing emerged. Engineers came to know their way round the terrain in the way that we get to know our way round a new town, or in the way a laboratory rat gets to know its way round a maze, by roaming and exploring and building up a cognitive map of it. It might even be true to say that the phenomenon of circulation became as evident as any of the commonplace facts of everyday life, for example, the fact that the cat is sitting on the mat. But to grant this conclusion is not to make a concession to absolutism, unless, that is, we have already conceded that the cat is on the mat is an absolute truth. This is precisely the question I discussed earlier, and precisely what I did not concede. Are things different if we confine attention to general laws of nature or general principles rather than particular matters of fact? Perhaps a better case can be made for the claim that these have an absolute character than can be made for sentences about cats on mats. Let us focus on the Kutta-Joukowsky theorem that lift is the product of density, velocity and circulation. Here, surely, is one of the “scientific principles” of aircraft design that Dawkins had in mind. It is a principle that is as true in Berlin as in Boston or Baghdad. Does it not possess the sort of universality that the relativist appears to deny and that could justify our calling it an absolute truth? We need to look carefully at the universality imputed to this principle. From the perspective developed here scientific inference moves from particular to particular.28 Inferences pick out the analogy between the instances that are said to fall under the generalisation. Like all analogies they go so far but no further. There will also be disanalogies and differences between the cases that fall under the principle. For example, the flow pattern of the air over an insect wing is very different from the steady flow over a classical aeroplane wing. Insect flight depends on exploiting the rapid production and reproduction of leading-edge vortices. These momentarily increase circulation and hence lift. Such vortices do sometimes occur on a classical aircraft wing, though as the herald of disaster. They involve what is called the ‘separation’ of the flow from the wing which indicates the onset of a stall and hence the breakdown of lift (see for example Ellington et al., 1996; Dickinson 2001). Nevertheless, in the late 1950s one of Prandtl’s pupils, Dietrich Küchemann, envisaged a new kind of aircraft wing that involved and exploited just such a separation under all flight 27 The most cogent proof of this that I know is in Hesse (1974). The appeal to ‘pure’ observation terms is the point at which the advocates of empiricism forget their robust, naturalistic ancestry and embrace absolutism. Theologically this form of empiricism is like radical, individualistic Protestantism: observation becomes revelation. 28 This is J.S. Mill’s way of describing inductive inferences. Notice that nobody has ever justified induction, so inductive knowledge is not a plausible candidate for an absolute knowledge claim. What about deductive knowledge? This has always been a favored candidate. In fact it is in no better position than induction. Attempts to justify it inductively fall short of the goal, while attempts to justify it deductively beg the question. So no secure candidates for absolute knowledge are to be found here either (see Haack 1976).
Relativism at 30,000 Feet
27
conditions. He saw how to make leading edge vortices steady by the use of a slender delta shape. At the time it was deemed ‘a heresy’, but Küchemann’s heresy has become familiar today through his design of Concorde’s remarkable wing.29 Küchemann also pointed to other ways to generate lift. The shock waves which build up at supersonic speeds interfere with the circulation of the air round the wing, but it might be possible to exploit them to generate lift. Küchemann envisaged hypersonic aircraft that would, in effect, ride on the shock waves. As a creative engineer he had a lively sense of diversity and variability rather than a tendency to think that final essences might be within our grasp. In his book The Aerodynamic Design of Aircraft (1978), in which he summed up both his technical thinking and his general design philosophy, Küchemann said: The most drastic simplifying assumptions must be made before we can even think about the flow of gases and arrive at equations which are amenable to treatment. Our whole science lives on highly-idealised concepts and ingenious abstractions and approximations. We should remember this in all modesty at all times, especially when somebody claims to have obtained “the right answer” or “the exact solution” (Küchemann 1978: 23).
For our purposes the reference to the right answer and the exact solution can be understood as claims to absolute knowledge, the very thing that aerodynamics cannot have. Relativism, by contrast, embodies the modesty for which Küchemann was calling. We should remember this and not allow ourselves to be dazzled by words like ‘universal’. The generalisations used by engineers to organise their practices have a utility that is contingent and inductive – and applicable within a limited range of circumstances.30 Perhaps it would be better for us to talk, not about ‘universality’, but of operating successfully within a limited ‘ecological niche’. Let me try to take that metaphor a little further. 29 The heresy designation (which was almost certainly meant in a good-natured way) is attributed to Sir Morian Morgan, FRS, then the deputy director of the Royal Aircraft Establishment at Farnborough (see Owen and Maskell 1980: 312; also Küchemann 1970. This article was first published in the Aeronautical Journal of the Royal Aeronautical Society, Feb. 1968). 30 Two points may be developed in more detail to reinforce this conclusion. First, in 1878 a version of the Kutta-Joukowsky law was developed by Lord Rayleigh and applied to ballistic problems e.g. the deflection of a tennis ball or bullet given spin (Rayleigh identified a force on a moving and spinning body normal to its direction of translation through an ideal fluid. The magnitude of the force was the product of velocity and circulation, density being assumed to be unitary). Empirically, for low rates of spin the predicted force is in the wrong direction (Birkhoff 1950: 36). Second, consider Newton’s calculation for the force on a flat plate of area A set at an angle a to a flowing fluid of density ρ and speed V. For Newton a fluid was taken to be a collection of independent particles not a continuum. By considering the momentum of the impact of the particles he predicted a force of ρAV2sin2a. If this is taken as the basis for understanding a classic aircraft wing the prediction is wide of the mark and differs considerably from the prediction of the Kutta-Joukowsky law. However, for predicting the pressure distribution of blunt-nosed, hypersonic vehicles operating under circumstances in which the continuum character of the air breaks down the Newtonian law comes into its own (Anderson 1991, ch. 14).
28
Knowledge as Social Order
Adaptation: Natural and Super-natural One popular way to develop a naturalistic perspective is to draw a parallel with biological evolution. We can follow Popper’s invitation to liken trial and error in science to trial and error in the evolutionary process.31 Technical success then becomes an analogue of biological adaptation. To say our technology ‘works’ is to say we have achieved a degree of adaptation to our environment. Natural processes can and do generate successful adaptation, though we must never forget how contingent, variable and precarious these successes are. We must never be tempted to gloss them in the language of absolutes. For example: adaptation is never ‘perfect’. There is no such thing as perfect adaptation and evolution is not to be understood as a progressive move towards such a goal. In the 18th and early 19th century, when the idea of ‘adaptation’ first began to play a significant role in the thinking of biologists, all the talk was indeed of ‘perfect adaptation’ (Ospovat 1978). Each species was thought of as perfectly adapted to its habitat. Of course it was God who provided this admirable arrangement. Here we see a blend of naturalism and super-naturalism. There may be nothing logically contradictory about it, but scientists would not be comfortable with it today in biology. Isn’t this scientifically reactionary mixture exactly what the anti-relativist, the absolutist, is still offering us today? The success of our technology provides no justification for it. Birds can fly successfully and we don’t think they need supernatural assistance; it suffices to say it is a form of adaptation explicable by evolution. If the birds can do it this way, we can too. The Myth of the Third Way I have now set out my reasons for saying that the relativism of the strong programme is in no way inconsistent with technical success, like flying at 30,000 feet. I have concentrated on the instrumental, pragmatic and opportunistic character of the knowledge involved, but the sociological dimension of my case should be obvious. The cognitive instruments were shaped by local, contingent and shared purposes. The pragmatism was a socially interested pragmatism. Aviation clearly had military potential. Soldiers need to know what is over the next hill and a scouting aeroplane could tell them. The task of Prandtl and his colleagues was to connect the existing body of idealised, and largely useless, mathematics about ideal fluids to this new technology. This is why the classical, hydrodynamic preoccupation with flow over straight, circular or elliptical surfaces gave way to a focus on the peculiar geometry of an aircraft wing and why classical results were selectively used and ignored. So there is the social interest and there is its creative and transforming effect on the existing body of conventions and paradigms. 31 Popper gives a systematic account of his understanding of the evolutionary nature of scientific knowledge in chapter 7 of Popper (1972). My use of the comparison emphasises its naturalistic implications. Popper’s use of the theory of evolution seems to me to combine it with non-naturalistic ideas, in particular, his theory of the so-called ‘third world of objective knowledge’. For a naturalistic application of the evolutionary analogy see Vincenti (1990).
Relativism at 30,000 Feet
29
As well as meeting a Dawkins-style challenge I have also issued one of my own. How are anti-relativists going to explain and justify their absolutism? I do not think that I have proven relativism true or proven absolutism false, but I do believe that I have shown that absolutism is a deeply unattractive stance. Anti-relativists no doubt see themselves as champions of science, but anti-relativism implies absolutism, and absolutism is a betrayal of what modem science stands for. When I sought to make relativism seem obvious by grounding it in an overarching naturalism I could just as well have said that it is the product of the scientific attitude. It isn’t the careful relativist who should be worried by the successes and failures of technology, it is the absolutist. The anti-relativist may be tempted to dismiss this challenge by claiming that my portrayal of their position is a caricature. I don’t accept this and insist that it exposes something close to the heart of their position. Nevertheless I understand why they might want to disown what they see in the theological mirror that I am holding up before them. This is why I explored the various attempts to make absolutism seem innocuous and familiar, e.g. to treat the very idea of truth as an absolutist preserve. I have argued that these expedients carry neither compulsion nor conviction. The critics of relativism must not think they can evade the problem by claiming to be both non-relativists and, simultaneously, non-absolutists. Obviously someone might reject some (particular) version of relativism whilst also rejecting some other (particular) version of absolutism, but this doesn’t meet the point. There is no ‘third way’ that allows the critic to escape the ultimate choice between the two positions. For example, no-one can reject both relativism and absolutism simply by expressing their commitments in terms of some other category, say, by invoking ‘objectivity’. This merely pushes the problem back, for what are we to understand by objectivity? Is it something absolute or something relative? Objectivity is not a middle, neutral ground; it is merely an idea that itself calls for analysis and requires positioning in terms of the dichotomy. A relativist could give an account of objectivity by following the lead given by Durkheim and Wittgenstein. Objectivity is the opposite of subjectivity and subjectivity is characterised by its dependence on the psychological resources of the individual. Thus the objectivity of a belief, judgement or standard can be accounted for by reference to the social, that is, to shared conventions and institutions. These can sustain standards which transcend the subjectivity of the individual without stepping outside the natural world. A similar argument would apply to the appeal to rationality. This, too, fails to offer a third way between absolutism and relativism because it merely raises all the same questions in a different form. You might indeed be an anti-relativist because you believe in ‘rationality’ but that can only mean you subscribe to an absolutist account of rationality. If you do not embrace such a view then a naturalistic understanding of rationality should suffice. This will appeal to our natural, psychological dispositions to make inferences, along with the shared norms, standards and conventions that
30
Knowledge as Social Order
enter into their evaluation. In other words, a naturalistic account will make rationality relative to biological and social contingencies.32 What status, according to my own argument, can be accorded to the dichotomy between the absolute and the relative? I am insisting that it is exhaustive and exclusive. I am saying there is nowhere else to go. But is that claim advanced as an absolute or a relative truth? If it is taken as absolute then it represents a counter-example to my relativism and I have refuted myself. If it is merely relative to some set of conventional categories, or the rules of some language-game, then how compelling is the inference from the rejection of relativism to the embrace of absolutism? Clearly I cannot, as a relativist, treat the dichotomy between the relative and the absolute as itself absolute. But I can appeal to its prima facie plausibility and its force as a current convention of discourse. I can still use it to pose my challenge. If the anti-relativists really think there is a third way they have a job of work to do. They need to explain it both to their relativist critics and to themselves. They cannot simply make abstract and unsubstantiated claims. For my part I cannot see how to deny relativism without embracing absolutism and vice versa. I shall believe this remarkable conceptual feat when I see it, but not before. Until then the antirelativists’ problem is the one that I identified; namely, making their commitment to absolute knowledge naturalistically intelligible, i.e. it is the problem of incarnation in cognitive form. How will they explain it without resorting to dogmatism on the one hand or obscurantism on the other? I don’t envy them their task.33 32 Rational persons proportion their belief to the evidence. The proportionality, however, is always mediated by conventions. Inference is not inductive rather than social, it is necessarily both at once. This may be seen in abstract terms from Carnap’s discovery that, formally, there are an infinite number of inductive strategies. He shows that the strategy of choice is an expression of pragmatic considerations such as willingness to take certain sorts of risk (Carnap 1952). For a concrete discussion which exhibits the conventionally structured character of inductive inference in science see Barnes (1984), see also Barnes (1976). 33 Here is an egregious example of how not to do it. It is taken from a paper mentioned earlier (Franklin 2002: 621). The author is addressing the character of arithmetical knowledge and how our ability to do elementary arithmetic might be explained. He begins with the question of why an electronic calculator produces the output 4 when 2+2 is punched in. The approach adopted is dualistic. On the one hand we have the causal workings of the machine, how it is wired up; on the other we have the abstract fact or truth that 2+2 really is 4. ‘The causal apparatus is designed specifically to be in tune with or track the world of abstract truths’ (p. 621). Unfortunately the nature of this abstract ‘world’ is never explained nor its postulation justified. It is simply taken for granted and the reader is expected to accept it as obvious. How we could ever design anything to ‘track’ it or to be ‘in tune’ with it is left as a mystery. All we have are a couple of metaphors. Then the picture is taken further. Calculating machines have humans to design them, but what about the human designers and their knowledge of 2+2 = 4? ‘It is the same with brains that do science’ we are assured. Thus, ‘a brain that draws the correct conclusions is performing well, while a brain that does not do well needs its mistakes explained’ (p. 621). Notice the implication that while error needs explaining success is unproblematic. That presumably is ‘explained’ by the brain being ‘in tune’ with the abstract world of logical and mathematical truths. We see here both dogmatism (in the original postulation of the abstract world of truths) and obscurantism (the pseudoaccount of the relation between that abstract world and human practices).
Relativism at 30,000 Feet
31
I think I would rather be a relativist at 30,000 feet than an absolutist at any altitude.34 Acknowledgements I should like to express gratitude to the Economic and Social Research Council for support (grant no. RES-000-23-0088). Versions of this paper have been given as talks to the Student Philosophy Society at Edinburgh, the Department of the History of Science and Technology at the University of Manchester and the Science Studies Program at the University of California, San Diego. I should like to thank the members of the audience on all three occasions for their vigorous critical discussion. I must also thank Celia Bloor, John Henry, Steve Sturdy, Andrew Pickering, and Walter Vincenti for their criticisms and comments. It should not be assumed that they agree with the stance taken in the paper. References Allen, J.E. and Bruce, J. (eds) (1970), The Future of Aeronautics (London: Hutchinson). Anderson, J.D. (1991), Fundamentals of Aerodynamics, 2nd edn. (New York: McGraw-Hill). ––– (1997), A History of Aerodynamics and its Impact on Flying Machines (Cambridge: Cambridge University Press). Barnes, B. (1976), ‘Natural Rationality. A Neglected Concept in the Social Sciences’, Philosophy of the Social Sciences, 6, 115–126. ––– (1982), T.S. Kuhn and Social Science (London: Macmillan). ––– (1983), ‘Social Life as Bootstrapped Induction’, Sociology, 17(4), 524–545. ––– (1984), ‘Problems of Intelligibility and Paradigm Instances’, in Brown (ed.), 113–125.
34 It will be clear that being at 30,000 feet is a relative property and necessarily depends on a frame of reference. It is worthwhile reflecting on what is involved. It might seem obvious that to be at 30,000 feet is to be at 30,000 feet above the surface of the earth. This answer might work in a philosophy seminar but it won’t in the cockpit: things are not that simple. Altitude is measured by barometric pressure. An aircraft altimeter has a sub-scale which allows it to be calibrated against a specified base-line. There are three conventions in use designed to meet different practical circumstances. First, the altimeter may be set using the barometric pressure at sea level (called the QNH); second it may set using the pressure at the airfield (called the QFE); third it may be set using an internationally agreed standard, the so-called International Standard Atmosphere of 1013.2 mb (hPa). The resulting indications on the altimeter are often called ‘altitude’, ‘height’ and ‘pressure altitude’ respectively. There are therefore at least three different things that can be meant by being at 30,000 feet and the differences between them can be a matter of life and death. See, ‘Altimeter Sub-scale Settings’, in Thom, 1988. The principles involved may be simple but they cannot be laughed off by the non-relativist. They are important because every concept has an analogous dependency on conventions.
32
Knowledge as Social Order
Barnes, B. and Bloor, D. (1982), ‘Relativism, Rationalism and the Sociology of Knowledge’, in Hollis and Lukes (eds). Batchelor, G.K. (1967), An Introduction to Fluid Dynamics, Cambridge (Cambridge University Press). Birkhoff, G. (1950), Hydrodynamics. A Study in Logic, Fact and Similitude (Princeton: Princeton University Press). Bloor, D. (1996), ‘Linguistic Idealism Revisited’, in Sluga and Stern (eds). ––– (1997), Wittgenstein. Rules and Institutions (London: Routledge) Bradshaw, G. (1992), ‘The Airplane and the Logic of Invention’, in Giere (ed.). Brown, J. (1984), The Sociological Turn (Dordrecht: Reidel). Bryant, L.W. and Williams, D.H. (1924), ‘An Investigation of the Flow of Air around an Aerofoil of Infinite Span’, Reports and Memoranda of the Advisory Committee for Aeronautics, 989. Busemann, A. (1959), ‘Ludwig Prandtl (1875–l953)’, Biographical Memoirs of Fellows of the Royal Society, 5, 183–205. Campbell Fraser, A. (1911), Selections from Berkeley (Oxford: Clarendon Press). Carnap, R. (1952), The Continuum of Inductive Methods (Chicago: Chicago University Press). Craik, K.W. (1943), The Nature of Explanation (Cambridge: Cambridge University Press). ––– (1966), The Nature of Psychology (Cambridge: Cambridge University Press). Dawkins, R. (1995), River Out of Eden (London: Weidenfeld and Nicolson). Desmond, A. (1989), The Politics of Evolution. Morphology, Medicine, and Reform in Radical London (Chicago: University of Chicago Press). Dickinson, M. (2001), ‘Solving the Mystery of Insect Flight’, Scientific American, June, 35–41. Ellington, C.E. et al. (1996), ‘Leading-edge Vortices in Insect Flight’, Nature, 384, 626–630. Epple, M. (2002), ‘Präzision versus Exaktheit: Konfligierende Ideale der angewandten mathematischen Forschung. Das Beispiel der Tragflügeltheorie’, Berichte zur Wissenschaftsgeschichte, 25, 171–193. Franklin, J. (2002), ‘Stove’s Discovery of the Worst Argument in the World’, Philosophy, 77, 615–624. Frege, G. (1959), The Foundations of Arithmetic. A Logico-mathematical Enquiry into the Concept of Number, trans. J. L. Austin (Oxford: Blackwell). Galison, P. and Roland, A. (eds) (2000), Atmospheric Flight in the Twentieth Century (Dordrecht: Kluwer). Giere, N. (ed.) (1992), Cognitive Models in Science, Minnesota Studies in the Philosophy of Science, 15 (Minneapolis: University of Minnesota Press). Haack, S. (1976), ‘The Justification of Deduction’, Mind, 85, 112–119. Hall, S.G. (1991), Doctrine and Practice in the Early Church (London: SPCK). Hegel, G.F.W. (1977), Phenomenology of Spirit, trans. A.V. Miller (Oxford: Clarendon Press). Hesse, M.B. (1974), The Structure of Scientific Inference (London, Macmillan). Hollis, M. and Lukes, S. (eds) (1982), Rationality and Relativism (Oxford: Blackwell).
Relativism at 30,000 Feet
33
Honderich, T. (ed.) (1995), The Oxford Companion to Philosophy (Oxford: Oxford University Press). Joukowsky, N. (1910), ‘Über die Konturen der Tragflächen der Drachenflieger’, Zeitschrift für Flugtechnik und Motorluftschiffahrt, I(22), 281–284. ––– (1912), ‘Über die Konturen der Tragflächen der Drachenflieger’, Zeitschrift fur Flugtechnik und Motorluftschiffahrt, III(6), 81–86. Küchemann, D. (1970), ‘An Aerodynamicist’s Prospect of the Second Century’, in Allen and Bruce (eds), 24–38. ––– (1978), The Aerodynamic Design of Aircraft (Oxford: Pergamon). Kuhn, T. (1961), ‘The Function of Measurement in Modem Physical Science’, Isis, 52, 161–190. Kusch, M. (2002), Knowledge by Agreement (Oxford: Clarendon Press). Kutta, W.M. (1910), ‘Über eine mit den Grundlagen des Flugproblems in Beziehung stehende zweidimensionale Strömung’, Sitzungsberichte der Königlich Bayerischen Akademie der Wissenschaften, 3–58. Luce, A.A. (1945), Berkeley’s Immaterialism (London: Nelson). Norris, C. (1997), Against Relativism. Philosophy of Science, Deconstruction and Critical Theory (Oxford: Blackwell). Ospovat, D. (1978), ‘Perfect Adaptation and Teleological Explanation. Approaches to the Problem of the History of Life in the Mid-Nineteenth Century’, Studies in History of Biology, 2, 33–56. Owen, P.R. and Maskell, E.C. (1980), ‘Dietrich Küchemann (1911–1976)’, Biographical Memoirs of Fellows of the Royal Society, 26, 305–326. Popper, K.R. (1965), Conjectures and Refutations; The Growth of Scientific Knowledge (London: Routledge and Kegan Paul). ––– (1972), Objective Knowledge. An Evolutionary Approach (Oxford: Clarendon Press). Prandtl, L. (1918), ‘Tragflügeltheorie, 1, Mitteilung’ Nachrichten der Kgl. Gesellschaft der Wissenschaften zu Göttingen, 451–477. Ramsey, A.S. (1935), A Treatise on Hydromechanics, Pt. II (London: Bell). Searle, J.R. (1995), The Construction of Social Reality (London, Allen Lane: The Penguin Press). Sluga, H. and Stern, D.G. (eds) (1996), The Cambridge Companion to Wittgenstein (Cambridge, Cambridge University Press). Thom, T. (1988), The Air Pilot’s Manual, Vol. 4 (Shrewsbury: Airlife Publishing). Tietjens, O. (1931), Hydro- und Aerodynamik nach Vorlesungen von L. Prandtl, II (Berlin: Springer). Vincenti, W.G. (1990), What Engineers Know and How They Know It. Analytical Studies from Aeronautical History (Baltimore and London: Johns Hopkins). Wigner, E.P. (1967), Symmetries and Reflections (Bloomington: Indiana University Press).
This page intentionally left blank
Chapter 2
Relativism: Is it Worth the Candle?1 Trevor Pinch
Relativism, we are told, is logically impossible, a mistake, a matter of self-refutation. Whole books have been written against it (Harris 1992). To many, it is evil and dangerous, a short step away from Nazism. All sorts of slippery slopes lead towards it; nasty regresses follow from it. The sons and daughters of America have been exposed to it, along with rock music, and feminism – this has supposedly led to the closing of the American mind (Bloom 1987). At the end of the cold war, one commentator even targeted it as the greatest threat since communism (Pangle, quoted in Appleby, Hunt and Jacob 1994: 275). Accusations of relativism are often couched within a familiar set of rhetorical tropes. Relativists are frequently described as being on a path or road which is usually dark or will lead to darkness. ‘Every time people go down the relativist road, the path darkens and the light recedes from the tunnel’ (Appleby, Hunt and Jacob 1994: 192). Or, more famously, Hollis and Lukes (1982) in the introduction to their Rationality and Relativism collection talked about the ‘... primrose path to relativism being paved with plausible contentions.’ Phil Kitcher titled his paper when debating these issues with me several years ago, ‘How the Road to Relativism is Paved’.2 Relativism is a road – part of an intellectual journey – which is paved to fool you. It will lead you somewhere you don’t want to go to despite your best intentions to go somewhere else. Along the path, the path on a pilgrim’s progress, may be the ultimate trap – ‘the foul pit of relativism’ – as David Bloor once ironically referred to it. This pit is overshadowed by the glittering ‘High Peaks of Truth’ (Bloor 1976: 142). Relativism is usually presented as something to be tempted by and from which we must be protected – it is as Jarvie put it ‘a simple beguiling doctrine’ which ‘encourages laziness and nihilism’ (Jarvie 1975: 349). While relativists may have perhaps been led up the garden path, relativism itself seems to be a condition beyond your control once you have embarked upon the path or entered the door. In any case your strength on your journey will be sapped as in a case of ‘creeping relativism’, or worse still a case of ‘debilitating relativism’. Relativism is a wasting disease.
1 An earlier version of this chapter was first presented at the Session ‘Relativism, Social Constructivism and the Contemporary Historiography of Science’, HSS Annual Meeting, New Orleans, 12–16 October 1994. 2 Phil Kitcher, ‘How the Road to Relativism is Paved’, presented at the Session ‘Relativism, Social Constructivism and the Contemporary Historiography of Science’, HSS Annual Meeting, New Orleans, 12–16 October 1994.
36
Knowledge as Social Order
What of the people who are relativists, people who dare to advocate such an intellectual heresy? Again the tropes reveal a striking pattern. Relativists are always ‘extreme relativists’; we all know about ‘ranting and rabid relativists’ and what you should watch out for, of course, is a case of ‘rampant relativism’. But it is rare to find the accusation bring levelled in the other direction, of someone being accused of being an ‘extreme realist’ or ‘extreme rationalist’. Likewise I haven’t heard talk of the spread of ‘rampant realism’ or ‘rampant rationalism’ and indeed no-one is accused of being a ‘ranting realist’ or ‘ranting rationalist’. That such phrases are oxymoronic tells us about the aura of proper conduct and deportment which the advocates of philosophical positions such as realism and rationalism are meant to have and display. We trust realists and rationalists to be sensible and moderate and fully in control of their faculties of reason; in short not to be led down paths or to lead us down paths. In contrast the poor relativist is presented as someone easily susceptible to being led (down the path) – not a fully rational person in control of their own faculties of reasoning. A self-proclaimed relativist is someone not to be trusted in scholarly matters; someone who is very far from the ideal type of the analytical philosopher. The cure for relativism is well-known – a strong dose of philosophical salts. The rhetoric around the critique of relativism is not insignificant.3 Language helps constitute the world and it is the critics of relativism who, in their rhetoric, have swallowed this point, hook, line and sinker. The best way to deal with a relativist is not to bash him or her over the head with self-refutation, but rather to point to the illegitimate intrusion of the social or psychological into matters of epistemology. The profane human world has once more polluted the sacred world of epistemology. The path and the paving, and the road and the rant are all things we associate with humans. It is humans who build paths and roads and who rant and rave. A philosophical position should be self-evidently true, requiring the philosopher just to deliver up what may have been hidden; to reveal that which has always been there. That opponents of relativism should systematically employ rhetoric to attack their enemies is, not surprising to students of the sociology of knowledge, it should, however, be something for rationalists and realists to worry about. I recently heard John Searle lecture. Searle is a wonderful exponent of the socalled analytical tradition in philosophy – he is systematic, clear and insightful. What was striking however was the rhetorical means whereby he dismissed people whose position he disagreed with. At one point he mentioned a group of people called ‘social constructivists’ a ‘movement’ which thankfully no-one today (apart from one or two people he hastened to add) believes in – a fashion for the seventies and eighties. How did Searle show the obvious absurdity of such a position? By relating a story of getting your car repaired. When you get your car repaired you want to know what is wrong with it – is the carburettor broken? The last person you want to meet is a deconstructionist car mechanic who claims that the carburettor is a textual phenomenon! The audience burst into laughter – enough had been said. This is the rhetorical phenomenon of what I call the ‘horse laugh’ – first named by 3 For a more detailed examination of the rhetorical constitution of anti-relativist arguments see Edwards et al., 1992; see also Geertz 1984.
Relativism: Is it Worth the Candle?
37
science journalist Martin Gardner in his dismissal of pseudo science. As Gardner noted, ‘A horse laugh is worth a thousand syllogisms’ when it comes to dismissing absurdity (1957). It is the same sort of trick deployed by Alan Sokal in attacking social constructivists in the recent science wars. Scientists have been known to use the ‘horse laugh’ as well in dealing with fringe phenomena.4 Relativism and the Sociology of Scientific Knowledge Relativism is a position which pioneers in the sociology of scientific knowledge (SSK) such as David Bloor, Barry Barnes and Harry Collins have explicitly embraced. Recall Barry Barnes and David Bloor’s 1982 essay, ‘Relativism, Rationalism and the Sociology of Knowledge’, or Harry Collins 1981 ‘Empirical Programme of Relativism’. Relativism was very much in the air in 1970s and 1980s, but is less common these days. Scholars instead talk about ‘situated knowledge’ or knowledge being ‘local’.5 In view of the rhetorical baggage associated with the relativist label this is perhaps a wise course of action. But just what sort of relativism did scholars such as Barnes, Bloor and Collins adhere to? And what sort of relativism is consistent with the fundamental tenets of SSK? In this essay I shall argue that a close examination of the early writings of Barnes, Bloor and Collins shows that they advocated a form of relativism best referred to as ‘methodological relativism’. Methodological relativism is the minimum requirement needed in order to carry out empirical work in SSK. It does not necessarily entail any commitment to stronger forms of relativism such as epistemological or ontological relativism. Two Common Misunderstandings about SSK Before I begin to tease out the different claims about relativism I want to clear the way a little by talking about two common misunderstandings concerning SSK. The first misunderstanding is that SSK is some sort of philosophical position or school rather than an empirical field of investigation. This misunderstanding arose because the early influential writings of Barnes and Bloor were often concerned with programmatic rather than empirical research. Barnes and Bloor themselves (e.g. Barnes and Bloor 1982), however, have made it abundantly clear that their early efforts were designed to make space for a full-blooded empirical sociology of science. And we should remember that in the early 1970s the sociology of scientific knowledge was almost non-existent and widely thought to be impossible. Barnes and Bloor asked what a sociology of scientific knowledge could, and should, look like. The core doctrines of their Strong Programme – causality, impartiality, symmetry and reflexivity – are primarily a set of methodological strictures about how the sociological study of science should be carried out, what sort of explanations it should seek, and so on. That Bloor was trained as a philosopher and successfully 4 On the use of this sort of rhetoric to dismiss cold fusion see Pinch 1995. 5 Feminist epistemology favours these terms. For situated knowledge, see Haraway 1988. For local knowledge see Longino 1997.
38
Knowledge as Social Order
defended his strong program in debates with philosophers, most famously against Larry Laudan (Bloor 1981), should not mislead. The studies of science which have brought the strong programme forward empirically were carried out by historians and sociologists; figures such as Shapin, Pickering and MacKenzie, to name but three of the best known. Also Harry Collins’s programme – although provocatively entitled a programme ‘of relativism’ – was aimed at pushing sociological explanation into the empirical domain of studies of contemporaneous science. The second thing which commentators on SSK have often underestimated, is that SSK has always been a broad church in terms of its philosophical underpinnings and the character of its explanatory task. This was true even of the earliest days of the field when supposedly there was more of a consensus over its programmatic goals. What are nuanced differences within the field becomes for the critic a matter of us all being tarred with the same brush of, say, postmodernity or anti-realism, or more likely today, Latourianism. The field has had over thirty years of healthy debates that mark important distinctions between the work of people like, Latour, Collins, Woolgar, Shapin and Pickering to name just some of the earliest protagonists. For different reviews of the field as it has evolved see, Collins (1983), Shapin (1995) and Pinch (2007). Bearing these two points in mind – the empirical character of the enterprise and the fact that many of the protagonists disagree over key issues – I want now to try and pin down more exactly what can be said about relativism. Symmetry and Methodological Relativism The Edinburgh Strong Program famously advocated the principle of symmetry. This is the claim that the same sorts of sociological explanations should be used for successful and unsuccessful knowledge claims. A failed knowledge claim like phrenology is to be explained in the same way that the successful field of cerebral anatomy is to be explained (Shapin 1979). Barnes and Bloor, in advocating this principle, were concerned to avoid a form of social epidemiology whereby we only use sociological explanations for failed beliefs, thus leaving true beliefs in need of no social explanation. What does this mean in practice? It means that when the historian or sociologist is examining two competing knowledge claims in a symmetrical manner he or she must avoid falling into the trap of saying that one is more true or more valid or more successful in terms of some absolute standards of truth or rationality. What might such a standard in any case be? It might be some epistemological theory, say the correspondence theory of truth adhered to by John Searle (Searle 1995). But this theory is itself a part of a philosophical tradition and different philosophers offer different versions of it and there is contention over its merits. The goal of the ‘naturalistic’ orientation of SSK is not to impose some sort of arbitrary epistemological standard upon the actors under study but to recover their struggles and strategies as they try and bring about a consensus over competing knowledge claims. This ‘bracketing’ of such standards is a form of methodological relativism. This does not mean that one knowledge claim might not actually eventually turn out to be judged true and another
Relativism: Is it Worth the Candle?
39
false. It simply means that for the purposes of this sort of explanatory exercise we have to bracket such epistemological factors. The Strong Programme through the tenet of reflexivity also made it clear that what was sauce for the goose was sauce for the gander. If science was to be explained in terms of social causes, then so too was the sociology of scientific knowledge. SSK ultimately rested upon the same sort of foundations as the sciences. And just as the sciences could be explained causally so too could SSK. The methodological relativism of the Strong Programme was clearly spelt out by Bloor in his original treatise on the sociology of knowledge, Knowledge and Social Imagery. Bloor refers to relativism as the opposite to absolutism. He writes, There is no denying that the strong programme in the sociology of knowledge rests on a form of relativism. It adopts what may be called ‘methodological relativism’. A position summarized in the symmetry and reflexivity requirements … All beliefs are to be explained in the same way regardless of how they are evaluated. (Bloor 1976: 142)
Epistemological factors such as the replication of experiments, the match between theory and experiment, logical proofs and so on are, of course, never far from the surface in examining scientific debates. It is to such factors which scientists themselves appeal during the course of debate. In carrying out these sorts of case study it is usual to show how such epistemological claims when cashed out in practice are themselves not definitive. Negative replications of experiments, for instance, which seem definitive to one side of a controversy, may lack compulsion for the other side. A proof which seems water tight for one group may be contested elsewhere, and so on. It is not that the actors under study do not themselves profess epistemological commitments, it is more that such commitments are rendered in practice in different ways. There is much misunderstanding amongst philosophers on this point, with it frequently being alleged that the sociologists ignore epistemological standards altogether.6 If the actors under study invoke epistemological standards, then the analyst needs also to attend to them. The point, I stress again, is not to engage in some sort of stick game for the sake of it, but to make room for social explanation. Collins would often say that the purpose is to push sociological explanation as far as it will go. Thus the real game for the sociologist or historian is to provide a convincing sociological explanation of some episode of science, say in terms of interests or other social processes. Relativity Revisited – Yet Again Now let us return to matters of relativism. As I mentioned already, relativism was in the air in the late 70s and early 80s. If one revisits the early work in the field one finds explicit reference to the relativistic underpinnings of SSK. For instance, Barry Barnes in the epilogue to his path-breaking, Scientific Knowledge and Sociological Theory, writes ‘... the epistemological message of the work here, can be said to be 6 See, for instance, my debate with the philosopher William J. McKinney, Pinch 1999a, and Pinch 1999b.
40
Knowledge as Social Order
sceptical or relativistic … It is relativistic because it suggests that belief systems cannot be objectively ranked in terms of their proximity to reality or their rationality’ (1974: 154). One of Harry Collins’s earliest published pieces was an examination of relativism and its so-called self refuting character (Collins and Cox 1976) – Collins (and his co-author Graham Cox) conclude that the relativist doctrine is not necessarily self-refuting. Indeed much of the early debate was concerned to expose the fallacies in the argument that if science had social underpinnings this necessarily made the science false and since this also must apply to the claims made by SSK this implied self-refutation. The sort of relativism discussed within SSK in the early days was usually referred to as ‘cognitive relativism’ in distinction to ‘moral relativism’. Sociologists were very familiar with moral relativism – the idea that there was no absolute set of moral values by which to judge human action. Interestingly Collins rejected moral relativism, a position implicitly adopted by many sociologists in dealing with the burgeoning sociology of deviance literature in the 1970s. But cognitive relativism, the idea that there were no absolute standards to be had in judging and evaluating knowledge claims, was worn as a badge of honour at the time. Again it is important to capture the tenor of the times. Scholars were interested in pushing the sociological as far as it would go and since philosophers and sociologists of science such as the Mertonian school had previously dictated there was no room for the social at all in terms of explaining knowledge it was enticing to push on further. Collins was always on the watch for sociologists needlessly boxing themselves in on apriori grounds before seeing whether in practice sociological explanation should be limited. The extent of the ‘more relativist than thou’ move is revealed in an early debate between Collins and Barnes where Collins thought he spotted Barnes backsliding on his relativist claims for Scientific Knowledge and Sociological Theory. The debate between Barnes and Collins was provoked by a passage in the preface. Barnes wrote that: Occasionally existing work leaves the feeling that reality has nothing to do with what is socially constructed ..., but we may safely assume that this impression is an accidental by-product of over-enthusiastic sociological analysis and that sociologists as a whole would acknowledge that the world in some way constrains what is believed to be (Barnes 1974).
People unfamiliar with Barnes’s writing may be surprised by this statement. But Barnes was actually a realist of sorts. He has subsequently developed this position into a form of what he calls physicalism or materialism (sometimes known as metaphysical realism) which presumes the existence of a world independent of people and what they believe about the world (Barnes 1992).7 Barnes distinguishes his minimal or residual realism from what he calls double-barrelled realism where, not only is the existence of an independent external reality accepted, but also some
7 Of course, Barnes is not the sort of realist that would satisfy the Phil Kitchers of this world. Barnes’s position is close to what Kitcher, quoting Devitt (Kitcher 1993: 169) refers to as ‘anti-realism with a fig leaf’.
Relativism: Is it Worth the Candle?
41
specific account of its contents.8 Barnes is against double-barrelled realism, because for Barnes, our knowledge of the externally existing world is inevitably governed by pragmatic and conventional concerns. Barnes then is a single-barrelled realist. He thinks it obvious that there is an independent external reality – for instance that the moon was there long before humans came to know it as the moon, and that it is still there when we don’t look at it. Barnes does not want to accept the specific account of what the moon consists of (this point again seems obvious because scientific accounts of the exact constitution and evolution of the moon have changed dramatically over the last fifty years and seem likely to change again in the future). This does not mean that there is no relationship between our accounts or knowledge of the moon and the moon itself. There is such a relationship but whatever that relationship is, it is not a relationship of correspondence whereby entities in scientific theory simply map onto or mirror that external world. Barnes talked about the denial of this common-sensical form of realism as stemming from sociological ‘over-enthusiasm’. Collins (along with his co-author Graham Cox), perhaps unsurprisingly, took on the challenge of trying to show how even this most basic form of realist commitment might have to be questioned for the purposes of some sociological studies. While agreeing with most of what Barnes had to say, Collins and Cox took Barnes to task for the passage quoted above. They maintained that there might be cases for sociological investigation where even overwhelmingly consensual knowledge about reality should not be taken as necessarily implying a constraint upon what is believed about the world. To illustrate their argument they used the example of a millenial cult studied by psychologist Leon Festinger. In this case, indeed, rather extraordinary claims are actually made about the world, and in particular by, one, Marionne Keech the cult leader who claimed to be receiving messages from extra-terrestrial beings alerting her that the world was about to end in a terrible cataclysm. She gathered her cult (who had by now been infiltrated by Festinger’s team of social researchers) together to await the fateful hour. The hour came and went, but nothing happened – at least nothing directly observable. The world it seems did not end. Within a few hours, however, we find Mrs Keech claiming this as a success for her group: yes, it is they who, with their good actions, have actually saved the world and avoided the catastrophe. Festinger, of course, then went on famously to claim this as a success for Cognitive Dissonance theory whereby a refutation of a belief leads, not to its rejection, but rather to its reinforcement. Collins and Cox point out that if we are following Strong Programme symmetry to deal with the Keech case then we might not want to treat Mrs Keech’s belief, that she had actually saved the world, as a case of irrational belief in the face of the over whelming truth that in fact she did not save the world (but rather saved her own face). To do so, according to Collins and Cox, would be to treat her belief in saving the world as a false belief; something not permitted by symmetry. According to Collins and Cox, if such a case is to be self-consistently treated by SSK then we must not put any demands upon an independent reality in rendering the 8 For most realists such an account would have to be a true account or at least, as my Cornell colleague Richard Boyd likes to talk about it, an account which is ‘truth tracking’.
42
Knowledge as Social Order
particular beliefs of the actors being studied. In other words, to do a fully symmetrical analysis of a case of contested belief, where one proponent believes that her actions have saved the world from disaster, might mean actually suspending our taken for granted beliefs about the world. Thus, for Collins and Cox it is better not to set any limitations on what is possible for the purposes of investigation in the sociology of knowledge. Collins, perhaps unlike Barnes, was at the time interested empirically in just those sorts of cases where highly deviant views about the world are contested.9 The debate I think is instructive because it shows how two self-professed relativist positions can run into opposition. The key difference between Barnes and Collins (as became clear from a later exchange with Law (Law 1977; Collins and Cox 1977), however, turned out not to centre on the strength of their relativism, but rather on the types of explanatory accounts that were permissible. The Strong Program has always looked for causal accounts of beliefs and part of the explanatory apparatus is psychological explanation. Barnes (and Bloor) want to allow psychological causes as explanations for deviant individual beliefs. This is clearest in Bloor’s treatment of Blondlot’s N-Rays in Knowledge and Social Imagery (Bloor 1976). Bloor talks about Blondot’s ‘lapse’ and the claimed discovery of ‘spurious’ N-Rays as a ‘personal and psychological failure of competence by Blondlot and his compatriots’ (Bloor 1976: 25).10 Bloor, like Barnes, points to the danger of sociologists ‘walking into a trap’ of accumulating cases like that of Blondlot. For Barnes and Bloor psychological factors can be evoked to explain the deviance. Collins on the other hand gives no place for psychological explanations and argues that a proper sociological account must deal with deviant individual beliefs, without slipping into the sociology of error. For Collins the failure of Blondlot and Mrs Keech, lay in their failing to convince others of their beliefs rather than in their failure to instantiate commonsensical versions of the world. This failure to institutionalize their beliefs may not result from a psychological failing but rather may be given a sociological explanation in terms of lack of ability to marshal enough resources, convince sceptics, and so on. We thus see that the key difference between Barnes and Collins is not over relativism per se, but rather over the part to be played by psychological explanation. Where does this leave the Strong Program and Collins in terms of relativism? Certainly the symmetry principle of the strong program as Barnes and Bloor acknowledges does entail a form of relativism. Purportedly ‘true’ and ‘false’ knowledge claims must be epistemologically levelled for the purposes of sociological
9 Collins had good reason to be exercised by such cases because at this time he and I were carrying out our sociological study of parapsychology – a study in which we wished to follow the strong programme doctrine of symmetry and treat both proponents and critics of the paranormal in an equivalent fashion. This meant that for the purposes of our study we were not prepared to endorse the standard view that Uri Geller type paranormal metal bending was an example of fraud and delusion. 10 For a rather different more thorough-going sociological treatment of the Blondlot case see Ashmore 1993.
Relativism: Is it Worth the Candle?
43
explanation. But this doctrine, as Barnes shows is consistent with a form of realism.11 What then is the relationship between methodological relativism and realism? For Barnes and Bloor there is an external real world which does constrain human actors, but it seems to be impossible to give a general description of how that world might constrain what can be believed about it except in a few isolated cases of highly deviant individual beliefs where psychological explanations can be invoked to at least partly explain that deviance. To summarize, Barnes is an ontological realist and a methodological relativist. What the strong programme seems to offer is a form of methodological relativism. You can believe that there is one external world and that in some indirect way it does constrain what can be believed about it, but specifics accounts of that world cannot be endorsed. Collins and Methodological Relativism What then of Collins? Collins, unlike Barnes, was not an advocate of realism in regards the natural world, neither was he an ontological relativist.12 For Collins, I suggest, ontological realism was only important when it got in the way and prevented a thorough-going empirical sociological programme of research. Collins is perhaps best described as an ontological agnostic and a methodological relativist. Rereading the early Collins papers one is struck by their methodological tone. There is talk about developing a ‘relativistic heuristic’ (Collins and Cox 1976: 439) or a ‘relativistic frame of mind’ (Collins and Cox 1976: 438) to approach particular empirical episodes. The methodological relativism at the core of Collins’s position becomes clear if one looks at what, for many, is his most notorious statement. It is usually said that Collins’s position is absurd because he believes that ‘the natural world in no way constrains what is believed to be’. (Collins 1981b: 541) Collins, I think, did once harbour hopes for establishing more than a methodological relativism, however, if one examines the full article and passage from which the quote is taken I think it is clear that by this time he was advocating a particular methodological attitude to be taken in empirical work. He took it that this attitude was required for the sociological analysis of deviant beliefs in science such as those professed by Joseph Weber who was on the losing side in the controversy over the existence of high fluxes of gravitational radiation (Collins 2004). The full passage where the quote is taken from reads as follows: The appropriate attitude for conducting this kind of inquiry is to ‘assume that the natural world in no way constrains what is believed to be’ (Collins 1981b: 54).13 11 The realism of the Strong Programme approach can clearly be seen in Barnes et al., 1996. 12 Collins famously was a realist about his own sociological claims but a relativist about the claims made by scientists. 13 Collins is of course, here referring back to the earlier debate with Barnes by putting in quotes part of the original Barnes passage at issue.
Knowledge as Social Order
44
Collins goes on to say: I hope that the detailed empirical work found in this paper and others in the same tradition, will bear out the case for the relativist approach and ensure its adoption as a methodological prescription, even by those to whom it is epistemologically distasteful (ibid.) (my emphasis).
In a later paper by Collins on TRASP (Truth, Rationality and Scientific Progress) the methodological character of the relativism is spelt out even more clearly. Collins writes: ... I want to show that the correct method for social studies of science is the same method as would be accepted by the Radical Programme [the impartiality and symmetry tenets of the Strong Programme] even if a non-relativistic epistemology is accepted (Collins 1981c: 217).
In short by 1981 Collins was to all intents and purposes embracing the same sort of methodological relativism as advocated by the Strong Programme. Barnes and Bloor saw that it was important to stress that they were also realists in regard to the external world – Collins saw those expressions of realism as doing no sociological work so felt that such statements made no difference. Methodological Relativism and the Discovery of Oil To illustrate the force of the methodological relativism advocated, it is salutatory to take an example, not from the esoteric world of science but from the more immediately accessible world of technology. If we think of hard facts about the natural world most of us probably would endorse the fact that oil is formed by the decay of biological material. This biogenic origin of oil on face value seems as hard as the fact that Mrs Keech did not in fact save the world. But it turns out there is a controversy over this fact – a controversy which is not yet satisfactorily resolved. The late Cornell scientist, Tom Gold, proposed an abiogenic theory of the origins of oil in which he argues that hydrocarbons were made deep in the earth at the same time as the planet was formed (Cole 1996; Collins and Pinch 1998). Traces of biological matter found in oil deposits are explained by microorganisms living in the oil. Gold, being an astrophysicist rather than a geologist, was first not taken seriously by many geologists. Gold was not daunted by his initial dismissal. After all he had encountered it before when he had first come up with the seemingly ludicrous explanation that the source of pulsars were rotating collapsed neutron stars. This idea was initially treated as being so crazy that Gold claims he was not allowed to present his work on it at an astrophysical conference. Gold proposed a test for his heretical ideas about the formation of oil and persuaded the oil-starved Swedes to drill a deep well in Swedish granite where no oil should be found. The result of the test was that some modest amounts of oil were found, but critics alleged this oil came from diesel in the drilling mud. Tests of the composition of the oil were contested. Gold, encouraged by these results, drilled again in granite but this time used no oil at all to lubricate the drill head. Again small
Relativism: Is it Worth the Candle?
45
amounts of oil were found – this time the explanation offered by the critics was seepage from other oil fields along fault lines into the granite. There the matter rests for the moment. The SSK study of this controversy by Simon Cole follows the tenets of methodological relativism. The analyst does not need to say in such a case that the biogenic theory is true or false. Such pronouncements are avoided if for no other reason than that if the protagonists themselves cannot resolve the matter it would seem absurd for the analyst to claim to do better. Instead such issues are bracketed. When it comes to technology there seems to be an additional real world ‘trump’ argument that might settle the matter. If a device could be built that obviously worked then surely all opposing arguments would be trumped (Gieryn and Figert 1990)? But establishing that the ‘trump’ case is indeed a good test of the underlying theory depends upon a web of associations. If Gold had found a gusher in Sweden it does seem likely conventional theories of oil would have been hard pressed to come up with an explanation. But ‘trump’ arguments in the world of oil are only trump arguments for particular groups – there are competing groups of experts as to who can testify best to the real world. I recently talked to the CEO of an oil exploration company who has discovered many oil wells and he told me that Gold’s theory was as good as any other in geology. As he put it to me, ‘If geologists really knew where oil was they would all be rich and they aren’t!’ To summarize: methodological relativism means that in this case the SSK analyst is agnostic as to the whether oil is biological in origin. This doesn’t mean that isn’t a consensus over how oil is formed – we can even speak of that consensus as a ‘truth’ to all intents and purposes.14 Conclusion Strong Programme symmetry lies at the core of SSK. All versions of SSK (even Latour and the reflexivists) subscribe at least to symmetry. To operationalize this doctrine sociologically requires us to be methodological relativists. Methodological relativism does not entail epistemological relativism or ontological relativism. Thus criticisms of SSK which condemn it in a broad-brush way for the sins of epistemological or ontological relativism, miss the weight of the work. My paper asks rhetorically whether relativism is worth the candle. If the candle of methodological relativism can light up the social world the answer must be a resounding yes.15 14 One can be a methodological relativist and talk about truth. I am extremely grateful to Donald Campbell for a private communication subsequent to the presentation of this paper at the HSS in New Orleans which clarified my thinking ion this point – see, Campbell 1994 for more discussion of the different forms of relativism. 15 It may not be obvious to my readers, as it was not to at least one member of the audience when this paper was first presented to the HSS in New Orleans, as to why I start this paper dealing with relativism in general, only to end up defending a more limited methodological relativism. Although SSK only depends upon methodological relativism, it is vitally important that accusations of relativism made in a McCarthyite witch-hunt atmosphere (as in the case of some attacks by scientists and others upon SSK) should not become the bench-mark for a witchhunt. The correct answer in such circumstances to the question ‘Are you or have you ever been
46
Knowledge as Social Order
References Appleby, J. et al. (1994), Telling the Truth about History (New York and London: Norton). Ashmore, M. (1989), The Reflexive Thesis: Wrighting Sociology of Scientific Knowledge (Chicago: University of Chicago Press). ––– (1993), ‘The theatre of the blind: Starring a promethean prankster, a phoney phenomenon, a prism, a pocket, and a piece of wood’, Social Studies of Science, 23, 67–106. Barnes, B. (1974), Scientific Knowledge and Sociological Theory (London: Routledge and Kegan Paul). ––– (1992), ‘Realism, Relativism and Finitism’, in Raven et al. (eds), 131–147. Barnes, B. and Bloor, D. (1982), ‘Relativism, Rationalism and the Sociology of Knowledge’, in Hollis and Lukes (eds). Barnes, B. et al. (1996), Scientific Knowledge. A sociological Analysis (London: Athlone). Bloom, A. (1987), The Closing of the American Mind (New York: Simon and Schuster). Bloor, D. (1976), Knowledge and Social Imagery (London: Routledge and Kegan Paul). ––– (1981), ‘The Strength of the Strong Programme’, Philosophy of the Social Sciences, 11, 199–213. Bryant, C.D. and Peck, D.L. (eds) (2007), 21st Century Sociology: A Reference Handbook, Vol. 2 (Thousand Oaks, London: Sage). Campbell, D. (1994), ‘Towards a Sociology of Scientific Validity’, foreword to Kyung-Man K., Explaining Scientific Consensus: The Case of Mendelian Genetics (New York, London: Guildford). Cole, S. (1996), ‘Which Came First, The Fossil or the Fuel?’, Social Studies of Science, 26(4), 733–766. Collins, H.M. (1981a), ‘Stages in the Empirical Programme of Relativism’, Social Studies of Science, 11, 3–10. ––– (1981b), ‘Son of Seven Sexes’, Social Studies of Science, 11, 33–62. ––– (1981c), ‘What is TRASP?: The Radical Programme as a Methodological Imperative’, Philosophy of the Social Sciences, 11, 215–224. ––– (1983), ‘The Sociology of Scientific Knowledge: Studies of Contemporary Science’, Annual Reviews of Sociology, 9, 265–285. ––– (1985), Changing Order: Replication and Induction in Scientific Practice (London: Sage). ––– (2004), Gravity’s Shadow: the Search for Gravitational Waves (Chicago: The University of Chicago Press).
a relativist?’ is to refuse to answer and deny the grounds in which the question is cast. In such circumstances I defend the right of my colleagues to be the most extreme relativists possible. I am grateful to Harry Collins for discussion of this point – see Harry Collins, THES, 30 October 1994.
Relativism: Is it Worth the Candle?
47
Collins, H.M. and Cox, G. (1976), ‘Recovering Relativity: Did Prophecy Fail?’, Social Studies of Science, 6, 423–444. ––– (1977), ‘Relativity Revisited: Mrs Keech – A Suitable Case for Special Treatment?’, Social Studies of Science, 7, 372–380. Collins, H. and Pinch, T. (1998), The Golem at Large: What You Should Know About Technology (Cambridge: Cambridge University Press). Cozzens, S. and Gieryn, T. (eds) (1990), Theories of Science in Society (Bloomington: University of Indiana Press). Edwards, D. et al. (1992), ‘Death and Furniture: The rhetoric, politics and theology of bottom line arguments against relativism’, History of the Human Sciences, 8, 25–49. Gardner, M. (1957), Fads and Fallacies in the Name of Science (New York: Dover Publications). Geertz, C. (1984), ‘Anti Anti-Relativism’, American Anthropologist, 86(2), 263– 278. Gieryn, T, and Figert, A. (1990), ‘Ingredients for a Theory of Science in Society: O-Rings, Ice Water, C-Clamp, Richard Feynman and the Press’, in Cozzens and Gieryn (eds). Haraway, D. (1988), ‘Situated Knowledges: The Science Question in Feminism as a site of Discourse on the Partial Privilege of Perspective’, Feminist Studies, 14(3), 575–599. Harris, J.F. (1992), Against Relativism: A Philosophical Defense of Method (La Salle: Open Court). Hollis, M. and Lukes, S. (eds) (1982), Rationality and Relativism (Cambridge, Ma.: MIT Press). Jarvie, I.C. (1975), ‘Cultural Relativism Again’, Philosophy of the Social Sciences, 5, 343–353. Kitcher, P. (1993), The Advancement of Science (Oxford: Oxford University Press). Krips, H. et al. (eds) (1995), Science, Reason, and Rhetoric (Pittsburgh: University of Pittsburgh Press). Law, J. (1977), ‘Prophecy Failed (for the Actors)!: A Note on “Recovering Relativity”’, Social Studies of Science, 7, 367–372. Longino, H. (1997), ‘Feminist Epistemology as Local Epistemology’, Proceedings of The Aristotelian Society. Pinch, T. (1995), ‘Rhetoric and the Cold Fusion Controversy’, in Krips et al. (eds), 153–176. ––– (1999a), ‘Half a House: A Response to McKinney’, Social Studies of Science, 29, 235–240. ––– (1999b), ‘Final Response to McKinney’, Social Studies of Science, 29, 246–247. ––– (2007), ‘The Sociology of Science and Technology’, in Bryant and Peck (eds), 266–275. Raven, D., Tijssen, L.V.V., and de Wolf, J. (eds) (1992), Cognitive Relativism and Social Science (New Brunswick: Transaction Publishers). Searle, J.R. (1995), The Construction of Social Reality (New York: Free Press). Shapin, S. (1979), ‘The Politics of Observation: Cerebral Anatomy and Social Interests in the Edinburgh Phrenology Disputes’, in Wallis (ed.), 139–178.
48
Knowledge as Social Order
––– (1995), ‘Here and Everywhere. Sociology of Scientific Knowledge’, Annual Review of Sociology, 21, 289–321. Wallis, R. (ed.) (1979), On the Margins of Science: The Social Construction of Rejected Knowledge, Sociological Review Monographs, 27 (Keele: University of Keele Press).
Chapter 3
Who is the Industrial Scientist? Commentary from Academic Sociology and from the Shop-Floor in the United States, ca. 1900 – ca. l970 Steven Shapin
In 1971, Barry Barnes published only his second article – ‘Making Out in Industrial Research’ – appearing in the brand-new journal Science Studies, eventually to be renamed Social Studies of Science. I became Barry’s junior colleague in the next year. He was the Unit’s resident sociologist, and my job was to teach some ‘social’ version of the history of science. The subject of Barry’s paper did not much interest me at the time: its empirical focus was a debate about the socialization of the industrial scientist and its more general themes concerned how one should properly conceive of the socializing process. To my knowledge, Barry never returned to these materials, and it was with some surprise that – many years later – I found myself immersed in them, as I began a project assessing the relations between the authority of knowledge and the imputed character of the knower in 20th century and contemporary America. It was only then that I appreciated the resonances of the topic that Barry addressed so many years ago. During my own career, I have, at various times, been called a historian and a sociologist, even though I was never trained as a sociologist and was not very professionally trained as a historian. As I was leaving the Edinburgh Science Studies Unit in 1989 to take up a position at the University of California, San Diego, Barry asked me in an off-hand sort of way which department I was going to join. I suppose that when I replied it was the sociology department, my face must have betrayed apology, as if to acknowledge my undoubted short-comings. But Barry was immensely re-assuring: ‘Don’t worry’, he told me, ‘I know plenty of sociologists worse than you!’ Even though I am now in a history of science department, I hope Barry still feels that way, for I owe him a lot. ***
Here are two stories about the identity, condition and state of mind of the industrial scientist that circulated in America from the early 20th century to the 1960s or thereabouts. The first story had it that he – almost always ‘he’ during this period – was unhappy, anxious and maladjusted to his lot. Powerfully socialized as a young man
50
Knowledge as Social Order
into a unique set of scientific values associated with the university, he took industrial employment because suitable academic research careers were in short supply or because they were just too low-paid to keep body and soul together for those not keen on an ascetic way of life. He found adaptation to incompatible industrial moral economies difficult, sometimes smouldering with resentment throughout his career. As the twig was bent so grew the tree: socialization into such values was strong and consequential. The industrial scientist deeply disliked, if he did not actively rebel against, the violation of scientific values he found in industry: secrecy, regimentation, hierarchy, constraint, and short-termism. The money – for the money was on the whole good – never made up for it. The scientist was being forced into a gray flannel lab coat, and it just didn’t fit. That was his central problem and it was a problem that industrial research managers would have to deal with as best they could. This is a story about ‘conflict of interest’, though here the texture and vector of conflict are rather different from what they are understood to be in the presentday American research university. In this story, the emotional pull of the unique scientific ethos is so strong that there are grounds for worry that industry, or indeed governmental laboratories which share some of industry’s characteristics, can obtain a sufficient supply of such men, or, when they are recruited, that they can be kept happy and productive. Of course, in the current state of affairs, concerns about ‘conflict of interest’ usually take other forms. Now, university administrators and ethics committees understand that the pull of commercial lucre is so strong that standards of proper academic behaviour are seriously liable to corruption. The historical sources of story number 1 are familiar to sociologists and historians of science. While such sentiments were fairly widely distributed in American culture in the early to mid-20th century,1 the most influential systematic version of this story was elaborated by the late Robert K. Merton in essays of the early 1940s – modified and developed in the 1950s and 1960s by such students and colleagues as Bernard Barber, Norman Storer, Walter Hirsch and Warren Hagstrom – and institutionalized in the canon of American sociology of science. The ‘norms of science’ into which the scientist was socialized became, as Merton said, ‘internalized’, where they formed the scientist’s ‘conscience’ or ‘superego’. Even though Merton was then understandably pre-occupied with such threats to scientific integrity as those posed by the Nazi idea of ‘Jewish physics’ or the Soviet concept of ‘bourgeois genetics’, his initial description of the ‘institutional ethos’ of science also remarked upon the tensions between scientific and commercial values. In science, Merton pointed out, Die Gedanken sind frei, and the ‘rationale of the scientific ethic’ whittles down ‘property rights in science’ to the ‘bare minimum’ needed to secure recognition and esteem to the originator of a scientific idea. Science is public knowledge or it is not science at all, and the very idea of secrecy offends those who have internalized its values:
1 Polemical writing against the obstruction or perversion of pure science by the interests and norms of business appeared in the late 19th and early 20th centuries; see notably Henry Roland’s ‘A Plea for Pure Science’ (1902: 593–613) and Veblen ([1918] 1957).
Who is the Industrial Scientist?
51
The communism of the scientific ethos is incompatible with the definition of technology as ‘private property’ in a capitalistic society [...] Patents proclaim exclusive rights of use and, often, nonuse. The suppression of invention denies the rationale of scientific production and diffusion.
And Merton here specifically alluded to late-19th-century litigation between the Federal Government and the Bell Telephone Company which established the inventor’s ‘absolute property’ in his invention and his right to withhold crucial knowledge of it from the public. For mid-20th-century scientists, Merton wrote, this was a ‘conflict-situation’ and, while different scientists were more or less uneasily responding to it in different ways (by taking out patents, by becoming entrepreneurs or, indeed, by advocating socialism), nevertheless the fundamental friction arose from a conflict of values between pure science and commerce. Conflict was always likely to appear whenever science came into contact with institutions whose values differed from those needed for the pursuit of certified objective knowledge and which attempted to enforce ‘the centralization of institutional control’ (Merton 1942: 273, 275, 278; Barber 1952; Hagstrom [1965] 1975; Storer 1966; Storer 1961; Hirsch 1968). From these early statements of the norms of science, there emerged what seems to be a prediction about what empirical research would eventually show, if, indeed, systematic empirical research was deemed necessary to confirming such a matterof-course state of affairs: scientists socialized into this value system would suffer the ‘pain of psychological conflict’ when presented with situations which required or encouraged them to behave in ways that violated the norms they had acquired. To avoid or free themselves of this ‘pain’, it was ‘to their interest’ to conform to the ethos in which they had been socialized. Should the internalized ‘pure science sentiment’ be put under pressure by ‘other institutional agencies’ committed to the application of knowledge and concomitant organizational control, the result will be ‘the persistent repudiation by scientists of the application of utilitarian norms to their work [...] The exaltation of pure science is thus seen to be a defense against the invasion of norms which limit directions of potential advance and threaten the stability and continuance of scientific research as a valued social activity’ (Merton 1942: 260). So it might be deduced that, for both psychological and social-functional reasons, scientists would vigorously resist assimilation into the value-system of commerce – such was the strength of the psychological grip of scientific values and such was the functional dependence of science on the embrace of these values. The scientist in industry, or in the military, would be an unhappy, awkward and possibly even a disloyal figure, in constant conflict with commercial values and organizational structures. As a result of scientists’ unique pattern of socialization, their personalities were such as would not tolerate organizational constraints: scientists were too fiercely independent and mindful of their individual integrity, too sceptical, too hostile to authority structures,
52
Knowledge as Social Order
too loyal to science and too disloyal to local organizational values. Such persons would pose a major problem for the smooth running of commercial organizations.2 The work of Merton, Barber, Storer and, to a lesser extent, of Hagstrom was largely programmatic, but by the 1960s academic sociologists – of whom the most prominent were William Kornhauser at Berkeley and Simon Marcson at Rutgers – were producing extended quasi-empirical studies of the ‘scientist in industry’, centring on the problems of normative ‘conflict’ and assessing the extent to which both industry and the scientists themselves had been obliged to come to some sort of in-practice ‘accommodation’ between value systems that were in such fundamental in-principle conflict (Kornhauser 1962; Marcson 1960). In 1956, the founding of the academic journal Administrative Science Quarterly (ASQ) proceeded from a desire for a ‘general theory of administration’ (Litchfield 1956; Thompson 1956; Parsons 1956), while a number of contributions to early issues – notably the work of Merton’s student Alvin Gouldner on ‘locals and cosmopolitans’ in organizational life – fit within the general framework that diagnosed conflicting loyalties between the scientist and other employees of commercial organizations (Gouldner 1957–1958). Certainly by the early 1950s, the conflicted and unsatisfactory position of the scientist in American industry and in governmental research facilities appeared to commentators of several sorts as an enormous practical problem in the context of the Cold War: how to recruit such people in numbers sufficient to respond to the massive Soviet threat? The low birth rates of the 1930s Depression Era combined with the burgeoning demand for scientists and engineers to drive the Cold War arms race, with the result that the ‘scientist-gap’ and the ‘engineer-gap’ preceded the ‘missile-gap’ in American anxieties during the 1950s and 1960s (for representative practical expressions of such anxieties, see, e.g., Steelman 1947, Vol. 3: 28, 141; Powers 1952: 98; Laughlin 1952; Orth 1957: 58). Few publications from this period dealing with recruitment and with allied problems of retention, creativity, morale and the efficient use of research workers omitted to justify the salience of such problems to the Soviet threat and Federal government agencies, led by the Office of Naval Research, were prime sources of funding for social science work in this area. (Thomas Kuhn’s paper on ‘The Essential Tension’, which was the precursor to key ideas in his seminal 1962 The Structure of Scientific Revolutions, was given in 1959 to a conference on scientific creativity that included contributions from officials in the Pentagon’s Advanced Research Projects Agency, the Air Force Personnel and Training Research Centre, as well as the Dow Chemical Company).3 Already by the late 1940s, the Steelman Report to the President’s Scientific Research Board devoted 2 Here I acknowledge the continuing interest of my former colleague Barry Barnes’s early publications, identifying in the Mertonian story empirical and theoretical awkwardnesses that flowed from a specific theory of socialization, and pointing to an alternative such theory in the work of Howard Becker (Barnes 1971: 157–175; Bloor and Bloor 1982: 83–102; Dennis 1987). However, my purpose here is not the same as Barnes’s: though there is, indeed, criticism of the Mertonian account in this essay, my major impulse here is to show the interest of interpreting it as a historical product. 3 See especially contributions by Maury H. Chorness of the Air Force Personnel Research Laboratory; J.H. McPherson of Dow; and N.E. Golovin of ARPA, who made his concern with the US-USSR creativity-gap quite explicit. The research of one of this volume’s
Who is the Industrial Scientist?
53
much attention to problems of recruitment, creativity and morale, and worried inter alia that scientists socialized into a preference for the freedom, autonomy, openness and collegiality of academic settings might constitute a substantial obstacle to Cold War scientific and technological mobilization against the Soviet threat (Steelman 1947). In this Cold War context, while academic social scientists elaborated a picture of the industrial scientist made-unhappy-through-early-socialization, they had allies in university departments of industrial relations and in business schools, as well as in strands of non-academic cultural commentary. So, for example, the Harvard Business School professor Charles Orth wrote in 1959 that corporate managers completely fail to understand ‘the kind of people who work in the world of science and the influence on their values of the special training which made them scientists’. Socialized in academic environments, scientists all want to do basic research; they are accustomed to, and value, ‘an independence of thought’ and a ‘permissive, lowpressure atmosphere’; they care deeply about the good opinion of their disciplinary colleagues and less about the approval of their corporate superiors. That’s just what scientists are like, whether the cause is academic socialization or the selective recruitment into a scientific career of personality types (Orth didn’t especially care): ‘the resultant personality matrix of the scientist is familiar to anyone who has associated at all closely with these men’. Put these sorts of people in industry, and ‘the fact that they think and behave differently makes life difficult for practically everyone in an industrial organization who must deal with them. Businessmen find it hard to understand almost everything such men say and do’. Indeed, scientists don’t enter industry ‘because they like the idea of working for an industrial organization’; they are drawn to it against their own inclinations, because they need the facilities, resources, and ‘somewhat larger salaries’ that industry can offer and that academia cannot. The scientist ‘typically does not really want to work for an organization’ (Orth 1959).4 Here Orth acknowledged indebtedness to the most widely distributed 1950s picture of the disastrous, but necessary, engagement between scientists and industry, William H. Whyte’s The Organization Man. The centrepiece of Whyte’s book was a set of three chapters on ‘The Organization Scientist’, which identified the pressures brought on scientists in American industry to conform to industrial values, work conditions and structures of authority – pressures which were well on their way to eroding the nation’s capacity for technological and commercial innovation at just the historical conjuncture when those capacities were most needed. For Whyte, as for Merton, the crux of the matter was a conflict of values between science and those institutions called upon to support and enlist science, and in particular a failure on the part of sustaining institutions to comprehend the unique values which alone would encourage scientific geese to lay their utilitarian golden eggs. In the relevant editors (Calvin Taylor) was funded by the Air Force Personnel Laboratory and that of other contributors by the ONR (Taylor and Barron 1963). 4 For insistence that Progressive Era industrial scientists identified very strongly with the companies that employed them and embraced their values, see Patrick McGrath (2002: 19–21).
54
Knowledge as Social Order
sections of The Organization Man, the target was an obtuse, and ultimately selfdestructive, unwillingness on the part of industrial managers to recognize authentic scientific values and to accommodate those values in the organizational life of the industrial research laboratory. His fundamental argument, Whyte said, was that ‘between the managerial outlook and the scientific there is a basic conflict in goals’. The current ‘orgy of self-congratulation over American technical progress’ attributed it overwhelmingly to ‘the increasing collectivization of research’. But this was a mistake that was storing up enormous trouble for the national future. If America effectively organized the individual researcher out of existence, it would eventually put an end to creativity and, by extension, to America’s ability to compete, commercially and militarily (Whyte 1956: 205–206). We are, Whyte wrote, currently lectured by bureaucrats and industrial managers on ‘how the atom bomb was brought into being by the teamwork of huge corporations of scientists and technicians’. Only occasionally does someone have the good sense ‘to mention in passing’ that the creative impulse for the atomic bomb was an individual, unorganized and spontaneous act of genius – indeed, something that ‘an eccentric old man with a head of white hair did back in his study forty years ago’. American industrial management was working remorselessly to ‘mold the scientist to its own image; indeed, it saw the accomplishment of this metamorphosis as the main task in the management of research’. If it succeeded, Whyte judged, it would be committing suicide, for ‘every study’ has demonstrated that the ‘dominant characteristic of the outstanding scientist’ is ‘a fierce independence’ that will not tolerate control, interference or collectivization: ‘In the outstanding scientist [...] we have almost the direct antithesis of the company-oriented man’. American industry’s aversion to the necessary eccentricity of genius was, for Whyte, signaled by a Monsanto recruiting film in which the voice-over announced ‘No geniuses here; just a bunch of average Americans working together’ (1956: 206–207, 211, 214; no such ‘studies’ were actually cited). So that is story number 1, or at least a plausible enough précis of it for present purposes, since the story is still so familiar. There is, however, quite a different story – more precisely, a set of stories – about the identity, condition and state of mind of the American scientist-in-industry during the 20th century, and these stories are not at all well known among academic social scientists and historians. The literature from which these stories emerge is seldom cited and, from what I can tell, scarcely read, by relevant sociologists and historians, with the result that the sensibilities represented there are rarely, if ever, confronted by academic writers. And probably the major reason why this literature is little known and why, therefore, its lessons are little engaged with, is that this commentary emerges not from the academy but from industry itself or, more specifically, from those research managers and administrators whose job it was, from early in the century through the 1960s, to recruit, manage, coordinate and motivate the work of scientists in industry. Sources for story 2 can be found in a wide variety of places. From the early 1950s, industrial consortia and managers established journals of their own in which to trade experiences about problems encountered in the new, and fast-changing, world of commercial research: Industrial Laboratories in 1949, then Research Management in 1957. The I[nstitute of] R[adio] E[ngineers] Transactions on
Who is the Industrial Scientist?
55
Engineering Management intermittently published practical commentary on problems of research management in issues from 1955 to the early 1960s, as did Chemical and Engineering News, The Journal of Industrial and Engineering Chemistry, Mechanical Engineering, The Technology Review, Personnel (the periodical of the American Management Association), and similar periodicals. Conferences and publications on research management convened and sponsored by such Big Business firms as Standard Oil of New Jersey, Standard Oil of California, and the Big Business consortium, the Industrial Research Institute, are further sources for such stories; and, from the 1920s through the 1960s, a small number of the more reflective research executives published books on the subject of organizing industrial research facilities and administering scientists. I call these sources ‘shopfloor’ commentary, even if there must be some pertinent differences in experience and perspective between research managers and the research workers they manage. In marked, if unsurprising, contrast to the academic commentary on research management, shop-floor writing displayed no interest whatever in making points of general sociological interest, in using passages of research management as ‘casestudies’ for any other purpose than coming to some more-or-less robust findings about recurrent problems in and about the industrial laboratory and in proffering some more-or-less plausible practical solutions to those problems. So, for example, a research director at RCA in the early 1950s rejected the pertinence of general theories of administration in terms typical of practical administrators, insisting that the requirements and problems of research management ‘will vary among different units of industry’ and will even depend upon the ‘differing outlooks towards research’ taken by ‘individual research administrators’ (Engstrom 1957: 69).5 As an agricultural research administrator succinctly put it in 1926, ‘It is a condition and not a theory that confronts us’ (Ball 1926: 35). This material is not academic in tone or appearance: there are rarely, if ever, any footnotes or literature references – for writing in this genre that is contemporary with, or subsequent to, the work of Merton and his followers, it is as if such work never existed – and the evident purpose is not apologetic, defensive, or even celebratory, but, rather, in the spirit of trading ‘war stories’ among congenial colleagues. The question here is always practical management, not sociological generalization: how can one make industrial scientists more productive, more creative, happier, more likely to stay with the firm? what forms of organization work best for the industrial research laboratory, or for particular types of laboratories and in different industries and in firms of certain sizes? how does one attract the most able research workers and how can one recognize the signs of ability in potential recruits? That is to say, while the shop-floor literature cannot possibly be confused with academic social science, it is concerned with social arrangements and relations which overlap massively with those of the academic genre. It is warrantably about the same social world and, therefore, one would expect that social problems identified as referentially central in the academic tradition ought to preoccupy writers of the practical literature as well. Unhappy, 5 Engstrom was a radio engineer who had been a director of research at RCA since the 1940s, and in 1951 – when his talk was delivered – was Senior Executive Vice-President of that company.
56
Knowledge as Social Order
value-conflicted industrial scientists were, in part, what theory predicted, but the same unhappy scientists and the same grounds for their unhappiness, should be an enormous practical problem for research managers and their corporate associates. However, with vanishingly few exceptions – exceptions which may dissolve on further investigation – unhappy industrial scientists, made unhappy by the strength of their socialization into unique academic values, just do not exist in the commentary of research managers and allied executives.6 And some of the rare reflections on the transition-process produced by practicing research workers themselves draw attention to disorientation in leaving the more ‘regimented’ world of thesis-research, in which you had just one supervisor to satisfy, for an industrial laboratory in which the lines of control were not nearly so clear. True, orientation to commercial outcomes is not something that the newly-minted Ph.D. was accustomed to, but those who chose industry might either embrace or accept those goals and ‘In general’, as one newly recruited electrical engineer said, ‘industrial research is well-respected in the academic world’ (Gorog 1966: 7, 10–11).7 It is not that the industrial research laboratory is seen as problem-free: far from it; story 2 is about nothing but problems and their possible resolution – problems of recruiting, remunerating, retaining, motivating, organizing and directing the labours of research workers. It was very widely recognized that research workers moving from university to industry may go through processes of adaptation, but such adaptation was usually seen in terms of getting people and their families settled in (schools for the kids, canasta-partners for the wives, etc.); getting research workers familiarized with organizational culture, routines and expectations; and sometimes, indeed, getting them accustomed to a more team-orientated style of work than was typical in academia.8 Nor is it that 6 The closest thing I can find to an expression of role-conflict-through-academicsocialization from an industrial source is Lowell W. Steele, ‘Rewarding the Industrial Scientist: A Problem of Conflicting Values’ (1957). But while Steele was indeed an executive at General Electric, it is relevant that he was not a natural scientist or engineer by training and had no direct shop-floor experience. Steele was an industrial personnel manager at the GE Research Laboratory, in charge of salary practices. He took a Harvard M.B.A. in 1948 (where he may well have encountered the views of the previously mentioned Charles Orth) and, later, an MIT Ph.D. in economics and social science under the supervision of Herbert A. Shepard. Whyte (1956: 211–212) used Steele’s earlier views prominently as an authority on the crushing demands for loyalty and conformity rightly made on scientists by commercial organizations (1953: 471). Steele evidently had a change of heart on this issue between 1953 and 1957. 7 Gorog was a 1964 Berkeley PhD in electrical engineering, then working at RCA. See also Tietjen, ‘The Transition from Graduate School to Industrial Research’ (1966). 8 Horace A. Secrist, Director of Research at The Kendall Company, noted that industrial management ‘often does not fully realize how great an adaptation the scientist does succeed in making when he moves from the university to the industrial laboratory’, drawing attention here not to programmatic norms but to some mundane aspects of adjusting to teamwork (1960: 58–59), and a research manager at Sun Oil noted the importance of showing recruits – through organization charts – where they fit in the corporate scheme of things (Kipp 1960). On the pragmatics of adjustment, see, for example, Warner Eustis (of The Kendall Company), ‘Personnel Policies and Personality Problems’ (Eustis 1948); John A. Van Raalte, ‘Reflections on the PhD Interview’ (1966: 307, 315).
Who is the Industrial Scientist?
57
members of industrial research facilities were necessarily seen as ‘one big happy family’ – though such attitudes were sometimes expressed – or that serious tensions were not recognized to exist between companies’ research functions and such of their other arms as accounting and production.9 In shop-floor commentary, the industrial research laboratory was full of tensions, just as its place in overall corporate culture continued to be problematic throughout my period of interest. But to recognize such tensions and conflicts was much the same sort of thing as it was to recognize endemic tensions and conflicts between, say, firms’ production and marketing divisions or, on the production floor, between supervisors and skilled workers (Steele 1953: 471). Just as these sorts of tensions and conflicts are acknowledged and dwelt upon in internal business commentary, so the research managers whose writings make up story 2 were obsessed by the organizational problems of the industrial research laboratory. It is just that the persistent and consequential problem of socialization so precisely and persistently identified in the academic literature did not exist in shopfloor commentary. Indeed, there are important and pervasive strands of such commentary that portrayed the quotidian realities of industrial research in ways that make academically predicated role-conflict highly problematic. Doing major violence to the heterogeneous nature of this literature, I draw attention here to just a few of these realities and associated values. First, the related questions of autonomy and planning. Autonomy of research work is always a relative matter. If the university was an institution marked by the value it placed on intellectual autonomy – and, indeed, a number of research managers vigorously insisted that it was, and that this marked a fundamental difference between academia and industry10 – nevertheless the substantial reality was that some scientists might choose industry because they would in fact be freer to do the work they wanted. In 1905, Willis Whitney recruited William Coolidge from MIT and, more famously, in 1909, Irving Langmuir from the Stevens Institute of Technology, to conduct fundamental research at General 9 See Tom Burns, ‘Research, Development and Production: Problems of Conflict and Cooperation’ (1964), for a study of tensions between the research facility and production functions in British companies, and for a participant’s idyllic account of relations within the laboratory: ‘In the lab, we’re very happy- sort of happy family relationships. The lab chief must select people on the grounds of getting on with others; they certainly do get on with everybody’ (1964: 123). 10 There are countless examples of such sentiments among industrial research administrators. For a representative example, see C.E. Kenneth Mees (1920: 20–21); also John J. Carty (of AT&T), ‘Science and Business: An Address to the Chamber of Commerce of the United States (1929: 1–2); David F. Noble (1977: 129–131); Willis Jackson (of MetropolitanVickers [UK]), ‘Discussion’ (1957: A.3): insisting that ‘the primary function of university research was in its relation to teaching’, he urged that the goals of academic research should never be defined by any outside sponsor, and ‘a clear distinction should be made between universities and national as well as industrial laboratories’; John A. Leermakers: ‘It is in the universities that true intellectual freedom, so essential to the development of the scientific attitude, is found’ (1951: 2–3). Note, however, that Leermakers was then an associate of Eastman Kodak’s Kenneth Mees, whose views on the freedom required by industrial research are quoted below.
58
Knowledge as Social Order
Electric, if not precisely free as birds, then certainly freed from heavy teaching loads and academic colleagues’ lack of interest in research, with the resultant commercial bonanza from improvements in the tungsten-filament incandescent light-bulb and the incidental benefit of the 1932 Nobel Prize in chemistry for Langmuir (Reich 1985: 69–70, 75, 82–83, 92, 111, 126–127; Kevles 1979: 99–101). The chemist Wallace Carothers, later the discoverer of nylon, reckoned Harvard to be the ‘academic paradise’ for teaching, but Du Pont was able to woo him away in 1928 because, as Hounshell and Smith say, ‘he really did not like to do it. He preferred research’. Carothers cared about the higher salary that industry offered, but worried about the potential loss of ‘the real freedom and independence and stability of a university position’. However, weeks after his move to Du Pont, he had no regrets, writing to a friend: ‘Already I am so accustomed to the shackles that I scarcely notice them. [...] Regarding funds, the sky is the limit. [... E]ven though it was somewhat of a wrench to leave Harvard [...], the new job looks just as good from this side as it did from the other’ (Hounshell and Smith 1988: 230–231). Reflecting on university conditions for research in 1916, Charles Steinmetz, chief consulting engineer for GE, wrote about the ‘false commercialism’, which figured professors’ ‘output’ in their pedagogical production rate, and which ‘wasted the universities’ best assets, its professors’: Thus we find in our colleges men who had shown themselves capable as investigators to do scientific research work of the highest order, overloaded with educational or administrative routine, and deprived of the time for research work. Private industries rarely commit such crimes of wasting men on work inferior to that which they can do: industrial efficiency forbids it (1916: 711–712).
Many commentators, both in academia and industry early in the century, reckoned that research, even of the fundamental sort, might not have a future in the American university because of its primary commitment to teaching and because of its poverty of resources to support research. Autonomy doesn’t mean much then or now – if you can’t get the time or funds to do the research you want to do. Of course, industrial research managers made no bones about the fact that the scientists in their employ were free only within limits, and that those limits were ultimately defined by their company’s commercial concerns. Recognition of that constraint was consistent from the origins of industrial scientific research. In 1919, the physicist Frank Jewett of science-friendly AT&T remarked that ‘The performance of industrial laboratories must be money-making [...] For this reason they cannot assemble a staff of investigators to each of whom is given a perfectly free hand’ (1919: 7).11 And a few years later, his colleague John J. Carty insisted that ‘Unless the work promises practical results it cannot and should not be continued’. Corporate scientists must ask themselves, and be required by their superiors to answer, the fundamental question ‘Does this kind of scientific research pay?’ (1929; quoted in Noble 1977: 115) And in the 1950s, the Director of Research at Minneapolis-Honeywell was one of many in his position underlining the requirement that ‘the results of research must pay’ (Kliever 1952: 6; McGrath 2002: 12–14). Research administrators at OwensIllinois Glass Company noted in the 1960s that there was no point in directing 11 Quoted in Kevles (1979: 100).
Who is the Industrial Scientist?
59
fundamental research if commercial benefit was not reasonably to be expected of it. Research executives were firm in their position: there was no reason for a company to support fundamental research, so to speak, for its own sake, nor for industrial research workers’ enjoying the freedom to pick any kind of problem they pleased: glass companies could not be expected to support employees’ fundamental research in oceanography (Hackett and Steierman 1963: 85–86). Creative people were, of course, always likely to be tempted into ‘fascinating but irrelevant side alleys’, and then it was the supervisor’s task to get them back on organizational track, but not before checking that the by-ways were really devoid of commercial potential (Evans 1969: 24). At the same time, there was no reason for research workers to imagine a conflict of agendas where none necessarily existed: as James Fisk of Bell Labs put it, ‘Because a man considers the needs of the organization that employs him, there is no reason to think that this makes for any lesser contribution in the scientific sense’ (Hackett and Steierman 1963: 85–86; see also, from Bell Labs, Fisk 1959: 163). ‘Our fundamental belief’, Fisk wrote, ‘is that there is no difference between good science and good science relevant to our business’, and by the late 1950s Bell Labs had the Nobel Prizes to support their claim (quoted in Bello 1958: 212–214). Managers commonly acknowledged it to be in the nature of genuine research that the unexpected sometimes turned up, and that these unexpected outcomes might be commercially consequential. Research worthy of the name was always to a degree unpredictable, and expressions of that sort of sentiment were absolutely standard among industrial research managers. In 1950, Kenneth Mees and John Leermakers of the Eastman Kodak Research Laboratory insisted that ‘It is really not possible to foresee the results of true research work’, and in 1958 a GE administrator writing about ‘Free Inquiry in Industrial Research’ defined research as ‘systematic inquiry into the unknown’, drawing from that bland definition the conclusion that the ‘detailed course of a scientific inquiry’ is always subject to unpredictable vicissitudes and, consequently, that ‘a certain amount of freedom on the part of the investigator’ is implied by the very idea of research (Mees and Leermakers 1950: 223; Hebb and Martin 1958: 67).12 Even if such a thing were possible, it was very widely understood that the firm would get few benefits from research whose outcome was wholly predictable. Moreover, it was appreciated that some (not all) research workers were the sorts of people from whom one could not get the best without finding ways to give them their heads. In the ongoing, and highly politicized, debates over the extent to which industrial research, or research of any kind, could and should be planned, some research managers made a distinction between the possibility and even necessity of planning what they called the ‘function’ of research over a number of years – that is, organizing commitments to it and its place in corporate activities – and planning the ‘act’ or ‘conduct’ of research, in which considerable freedom of action was deemed simply necessary (e.g. Kastens 1957; Williams 1947: 5; Hertz 1950: 202; and Brooks ‘Can Science Be Planned?’, in Brooks 1968, ch. 3, especially pp. 59–72). David Noble’s tendentious survey of early American industrial research correctly cites occasional corporate rhetoric pointing to the desirability of tightly 12 Both Hebb and Martin were physicists who moved to the GE Research Laboratory from academic appointments some time after 1955.
60
Knowledge as Social Order
managing the creative process, while eliding any distinction between such rhetoric and shop-floor realities and ignoring a large body of managerial rhetoric that frankly acknowledged limits to any such control.13 Research managers defended their own autonomy by persuading their bosses – to the extent they could – that the research function could not be held to the same system of cost-benefit accountability as other corporate activities. On the one hand, they wanted, and sometimes obtained, financial and temporal flexibility. Kenneth Mees at Eastman Kodak secured a commitment of ten years’ worth of research funding from George Eastman in 1912; a consortium of oil company executives counseled against expecting returns from industrial research in less than five to seven years; and one survey of industrial research managers in the late 1960s found expected average ‘payback’ times from R&D of about four years – with big firms having a more generous time-horizon than smaller ones – though another survey by the management consultancy firm Booz, Allen & Hamilton noted with alarm that less than a quarter of American large companies had any formal method for evaluating research, still less calculating a payback time (see Baum 1947: 17 – I thank Ray Curtin of the Eastman Kodak Company for this reference; see also Mees 1920: 102; Mees 1961: 50; Voorhies 1946: 53–54; Gerstenfeld 1970: 18–23 – for 4-year payback and large versus small company figures; Randle 1959: 131 – for lack of formal evaluating criteria). In 1916, Mees wrote that ‘Those with the most experience of research work are all agreed that it is almost impossible to say whether a given investigation will prove remunerative or not’ (1916: 768). And Charles ‘Boss’ Kettering, when he was recruited from National Cash Register by Alfred Sloan to be director of research at General Motors, made sure that Sloan understood ‘You can’t keep books on research, because you don’t know when you are going to get anything out of it or what it is going to be worth when you get it’ (quoted in Boyd 1957: 118). RAND Corporation economists in the late 1950s drew importantly upon such sentiments in arguing vigorously against imposing rigorous cost-benefit regimes on research and development (see, notably, Nelson 1959: 101–127; cf. Hounshell 2000). On the other hand, allowing research workers a significant amount of company time to do their own research is definitely not the late 20th-century invention of the Silicon Valley high-tech and biotech ‘knowledge economy’. Mees encouraged such autonomous research in the 1910s, and a survey of industrial practices in 1950 found that an allowance of 10 percent to 20 percent – that is, a half-day to a day a week, with company resources to match – was then common, even in American ‘smokestack’ industry (Mees and Leermakers 1950: 235–237; Bichowsky 1942: 106; for the advisability of keeping a ‘slush fund’ – say, 5 percent of the research budget – for individual researchers, with the assent of group leaders, to do with pretty much as they pleased). The president of Dow Chemicals said that he had ‘learned that if a research laboratory is to produce results, the men must be allowed the freedom to be a bit crazy’, and the American Cyanamid research director quoting this remark approved up to 20 percent paid company time for research personnel ‘to work out 13 ‘As the industrial research laboratories grew in size, the role of the scientists within them came more and more to resemble that of workmen on the production line and science became essentially a management problem’ (Noble 1977, ch. 7, 118).
Who is the Industrial Scientist?
61
their own ideas’ (Shepard 1945: 805; from American Cyanamid, quoting Willard Dow). For very highly prized scientific employees that free time could amount to considerably more: the terms that won William Coolidge for GE in 1905 included one-third – in other versions one-half – of his time to continue an existing personal research project (Reich 1985: 75 – for one-third; Birr 1957: 37 – for one-half). When Du Pont tried to recruit the organic chemist Louis Fieser from Bryn Mawr in 1927, Fieser was struck by the research freedom on offer: ‘I never expected to go into industrial work but the thing which makes a decision so difficult in this case is that I don’t have to sell my soul at all; they even said I could bring my quinones along and continue my present work’.14 And a 1952 survey reported that, for some individuals, in some industrial organizations, ‘no plans are made as to how they shall spend their time’ (Anthony 1952: 134–136). While David Hounshell is surely right in identifying this sort of research latitude as a recruiting and retention tactic for some ‘academic elitists’ who saw industrial research ‘as a poorer career option than that offered by a university or a private basic research institute’, he does not claim that all research workers felt that way, and there is abundant evidence of commercial justifications for considerable degrees of industrial research freedom (1996: 26–27).15 Eastman Kodak’s Kenneth Mees became famous in management circles for his celebration of research disorganization. The epigraph to his influential 1920 book on The Organization of Industrial Scientific Research boldly stated that ‘There is danger in an organization chart – danger that it be mistaken for an organization’, the source of which was not a scientist fed up with corporate bureaucracy but one of the founders of American technical management consultancy, Arthur D. Little.16 Disorganization, and the research autonomy consequent on recognition that planning in these areas was naturally constrained, were justified as conditions of commercial functionality. Mees was quite hard-headed enough to insist that ‘the primary business of an industrial research laboratory is to aid the other departments of the industry’, and that its central responsibility was to contribute to the corporate bottom-line 14 Letter from Fieser to James B. Conant, 26 November 1927, quoted in Hounshell (1996: 27); see also Hounshell and Smith (1988: 228, 299 – Fieser ultimately turned down the Du Pont position). But see McGrath (2002: 35, 208 No. 6) for vigorous dissent from Hounshell’s sketch of Conant as a typical academic ‘snob’ about industrial research. Conant did indeed maintain that ‘first-rate’ students should wind up in universities rather than industrial laboratories, but he was himself heavily involved in industrial research and saw nothing awkward in these relationships. 15 Evidently, the then-common practice of allowing industrial research workers free time could be unknown to social scientists commenting on the condition of scientists in industry during the 1950s. A Columbia professor of education, for example, wrote of his impression that such an idea was regarded by industrial research administrators as ‘a fantastic’ and even laughable notion (Bryson 1957: 130). 16 Quoted again in Mees and Leermakers (1950: 244); this so-called ‘second edition’ of 1950 was in fact an almost totally different book than the 1920 original. I have not yet been able to locate the source of this quotation in Little’s publications. Suspicion of the descriptive relevance of organization charts continued to be vigorously expressed by A.D. Little management consultants well after World War II: see, e.g. Kingsbury et al. (1959: 65, 77–78).
62
Knowledge as Social Order
(1920: 67–68), but he rejected the distinction – central to the writings of the British anti-planning Society for Freedom of Science – that it was only ‘pure’ science that was incompatible with planning. Commenting on early writings by Michael Polanyi, Mees said that ‘I take issue with [the] description of applied science as a field in which freedom of science might conceivably be undesirable. I have been engaged in applied science for forty years, and in that period I have come very definitely to the conclusion that the prosecution of applied science in its most efficient form is identical with that of pure science. I don’t think for a moment that it is desirable that applied science should be directed except in times of emergency’.17 Mees’ own muchquoted aphorisms include: ‘When I am asked how to plan, my answer is “Don’t”’, and ‘No director who is any good ever really directs any research. What he does is to protect the research men from the people who want to direct them and who don’t know anything about it’.18 And the only remark by Mees that finds its way into the reference books is his most eloquent condemnation of research control: The best person to decide what research work shall be done is the man who is doing the research. The next best is the head of the department. After that you leave the field of best persons and meet increasingly worse groups. The first of these is the research director, who is probably wrong more than half the time. Then comes a committee, which is wrong most of the time. Finally, there is the committee of company vice presidents, which is wrong all the time (This has been quoted many times, including Kliever 1952: 11; Nelson 1959: 125; Kaplan 2001: 105–106; Jewkes et al., 1958: 138).
This sort of sensitivity to the pathological consequences of attempts to control research was, according to the President of Bell Labs in the 1940s and 1950s, something that ‘All successful industrial research directors know [...] and have learnt by experience’: the ‘one thing a director of research must never do is to direct research, nor can he permit direction of research by an supervising board’ (O. E. Buckley, quoted in Jewkes et al., 1958: 137–138). The same MinneapolisHoneywell research director who insisted on the bottom-line criterion for judging research results concluded his piece in Industrial Laboratories by conceding that, while ‘the amount of freedom or control of research projects is probably the most difficult question in its administration, [...] I tend toward the principle expressed by Thomas Jefferson for government: “The least government is the best government”’ (Kliever 1952: 13). In practical terms, such sensibilities translated into a significant tolerance for the spontaneous and unpredictable emergence of novel research agendas among industrial research workers who enjoyed significant freedom to formulate their 17 See Mees (1947), emphases in original. In these connections, it is interesting to note that, around this time, Mees was collaborating with the English zoologist John R. Baker on a semi-popular history of science: Baker had been a leading light and associate of Polanyi, in the Society for Freedom of Science: see Mees and Baker (1946). 18 Mees’ discussion of a piece by Thomas Midgley in Standard Oil Development Company (1945: 48; cf. Mees and Leermakers 1950: 233). Many research directors felt obliged to take a position – for, against, or qualifying – Mees’ aphorisms: see, e.g. Shepard (1945: 804).
Who is the Industrial Scientist?
63
courses of work. In his own laboratory, Mees had to warn a scientist who developed an interest in high-vacuum pumps and gauges that this was work in no way compatible with the concerns of a photographic company, and one lesson drawn from this story was that scientists in Eastman Kodak’s research laboratory could not do just anything they wanted. But Mees then repeatedly went on to relate that when he saw what a splendid technology was being developed, he secured the resources to spin off a distinct commercial laboratory, one which ultimately became a highly profitable vitamin-producing joint venture with General Mills. Here, as elsewhere, the general lesson that Mees wanted research managers to learn was that autonomy was after all good business. There was no alternative to a high degree of autonomy, and attempts to be unremittingly hard-headed were ultimately self-defeating (Mees 1945: 49; idem 1961: 291–292; Mees and Leermakers 1950: 149; Baum 1947: 47).19 Intriguingly, in the very same year that Whyte’s Organization Man appeared, with its eloquent condemnation of industry’s commitment to crushing scientific genius under the dead weight of teamwork and management control, his own Fortune magazine ran a piece celebrating changed industrial sensibilities towards basic research and its demands for individualism, under a title – ‘geniuses now welcome’ – that gave the lie-oblique to Whyte’s story (Bello 1956).20 A second basis for industrial scientists’ supposed discontent, and consequent roleconflict, was the secrecy that was a necessary feature of commercial research. And, indeed, no research manager commenting on what is now called intellectual property thought that scientists could possibly be allowed freely to publish trade secrets. Their work was understood to belong to the company, and it was for the company to decide what could or could not be published in the open scientific literature (Noble 1977, ch. 6). Scientists joining Du Pont, for example, were obliged to sign non-negotiable agreements that all ‘inventions, improvements, or useful processes’ made while in the company’s employ were the ‘sole and, exclusive property’ of Du Pont, and they agreed not to ‘disclose or divulge confidential information or trade secrets’ (Hounshell and Smith 1988: 300–301). However, in practice many research managers vigorously endorsed the commercial prudence of a quite free publishing policy and argued for the barest minimum of secrecy. The aphorism ‘When you lock the laboratory door, you lock out more than you lock in’ comes not from Robert Merton or Michael Polanyi but – in the same temporal and cultural context – from ‘Boss’ Kettering at General Motors (quoted in Comwell 1948: 301). The free flow of technical information or, at least, the freest flow compatible with broad corporate interests, was widely, if not universally, acknowledged in these circles as a net benefit to all parties.21 19 The scientist concerned was K.C.D. Hickman, and the new company founded to manufacture vitamins A and E using this technology was Distillation Products, Inc. 20 Bello, like Whyte, exempted the industrial research ‘stars’ (Bell, GE, Eastman Kodak) from his criticisms and, while he reckoned that ‘There are still a good many laboratories where signs might still be posted: “No geniuses wanted”’, he applauded big industry’s willingness – perhaps, he thought, prompted by Fortune’s arguments over the years – to develop ‘more flexibility in accommodating unusual personalities’ (Bello 1956: 96). 21 See, for example, the vigorously pro-publication, and wholly pragmatic, views of a research manager at GE (Schmitt 1961).
64
Knowledge as Social Order
Research managers had to advertise their laboratories as workplaces attractive to the most talented research workers, many of whom were found in academia and there was no better way to do this than to encourage their scientists to participate in professional society meetings and to publish in the same journals as their academic disciplinary colleagues. So in 1948 the Director of Research at Sylvania wrote that ‘The reputation of the organization whose research men do publish their findings will be enhanced according to the caliber of the work done’; an oil company research executive agreed that encouraging research workers to publish their results – subject, of course, to ‘adequate patent protection’ – brings ‘substantial indirect value to a company, which gets a reputation for being willing to have its men present papers and for being progressive. It is not just a matter of humouring the man, but can usually be justified from the company viewpoint’; a Vice President of Union Carbide noted that ‘Forward-looking companies have come to realize that they can attract better men if they are willing to permit publication at the earliest moment compatible with a proper regard for economic advantage’; and a research manager at the General Electric Research Laboratory pointed out (without qualification) the likelihood and advantages of reciprocity: ‘Any laboratory that must do good basic scientific research must also encourage open publication of research results because this policy insures access to the work of other laboratories’. For these and many other reasons, a chemical company research director insisted that ‘a liberal company policy on publication’ was simply a ‘must’: ‘Only a blind or short-sighted management curbs or prohibits publication’.22 Even at Du Pont – a company that some academic scientists reckoned to be unduly secretive – official policy was massaged by research directors who recognized that a liberal publication policy was just necessary for attracting first-rate chemists and maintaining their morale. So in the late 1920s two leading Du Pont research managers recruited scientists by assuring them that ‘the work of these [fundamental researchers] shall be published almost without restriction’, and, as Hounshell and Smith document, that open policy with respect to Du Pont’s fundamental research was effectively realized (1988: 223, 301). In addition, publication was understood as a vehicle for interesting disciplinary communities to take up your problems, whether in their relatively pure or relatively applied forms. The more people working away in your area, the better for you (and that is one, sometimes unappreciated, reason why Edward Teller at Livermore in the 1950s urged the freest possible dissemination of information concerning thermonuclear weapons). Patentable findings, of course, had to be legally protected and, while patents had the desired effect of securing property rights to the company, they were understood also as a way of ensuring that the findings were in the public domain, thus satisfying, through a different route, 22 Sylvania (Comwell 1948: 301); Standard Oil, Indiana (Wilson 1949: 277); Union Carbide (McClure 1952: 163, 166); General Electric (Schmitt 1961: 32; uniquely, in my experience, this research manager offered evidence that he had read some works of Robert Merton, citing Merton’s claims about the essential openness of genuine science while completely setting aside Merton’s characterization of what industrial research was like: ibid., 36, 39–40). See also (Mees and Leermakers 1950: 277–280, 313; Hertz 1950: 347–350; Fisk 1959: 164). For a ‘must’ policy on open publication, see (Shepard 1945: 805).
Who is the Industrial Scientist?
65
those scientists who craved publication and consequent recognition from their peers. An RCA research director writing about recruitment problems noted that publication was important to young scientists, but so were patent and incentive awards, as each of these was a sign that the company recognized individual contributions (Van Raalte 1966: 310). Moreover, in fast-moving fields, the trick was not to secure advantage by locking up all possible intellectual property but by keeping one step ahead of the competition. Secrecy in such fields was of little concrete value. And a Sylvania research manager noted that whatever necessity there might be, so to speak, to embargo publication was usually only ‘temporary’: commercial commitment to publication could and should remain as a principle and as a substantial reality (Comwell 1948: 301–302). Such a widely accepted principle could and did coexist with the practical possibility that publication might, at any moment, be prohibited or delayed for commercial reasons – and industrial research workers understood this very well. Some potential recruits to industry bridled at this state of affairs, but many others did not: Reich writes that even with substantial ‘restrictions on communication and publication, Whitney had little trouble finding researchers willing to work at GE’, with even such a prestigious and well-equipped institution as MIT being a rich hunting-ground (Reich 1985: 110).23 Finally, and most interestingly from the point of view of academic theories of socialization, a number of industrial research managers were worried that their newly recruited academic scientists became too quickly and too totally accepting of the values and research agendas of what they took to be corporate, as opposed to academic, culture. At Eastman Kodak, Mees judged it very important that personal credit for research be given to individuals and publication under his name. ‘The publication of the scientific results obtained in a research laboratory is quite essential in order to maintain the interest of the laboratory staff in the progress of pure science. [...] When the men come to a laboratory from the university they are generally very interested in the progress of pure science’, Mees wrote, ‘but they rapidly become absorbed in the special problems presented to them, and without definite direction on the part of those responsible for the direction of the laboratory there is great danger that they will not keep in touch with the work that is being done in their own and allied fields. Their interest can be stimulated by journal meetings and scientific conferences, but the greatest stimulation is afforded by the publication in the usual scientific journals of the scientific results which they themselves obtain’ (Mees 1916: 771; idem 1920: 100 – my emphases; see also Edgerton 1988: 122). The practical problem pointed to here was not the strong and persistent socialization into academic values predicated by sociologists but its opposite – the matter-of-fact willingness of research workers trained in universities to abandon such putatively distinct values.
23 A 1961 survey of publication practices for basic research findings among 174 U.S. companies showed that 14 percent of them published ‘substantially all’ basic research findings; 26 percent published ‘most’; 45 percent published ‘some’; and 16 percent published ‘none’ (National Science Foundation 1961: 11 ff.; quoted in Hirsch 1968: 64).
66
Knowledge as Social Order
And it was that spontaneous abandonment which concerned research managers like Mees – not for moral or ideological, but for wholly practical, reasons.24 One could go on painting this sort of picture of American industrial research from the beginning of the 20th century to the 1960s or so. It would be a picture that, minimally, makes the academically predicated role-conflict problematic and, if it were taken to an extreme, would argue the total illegitimacy of traditionally established contrasts between institutional values, just as it would open up the possibility of a more ‘evolutionary’ than ‘revolutionary’ account of widely imitated practices in contemporary American knowledge-intensive industry. But even at this superficial level of detail the evidence presented here licenses some brief speculations about the relations between strands of rhetoric and institutional realities in this area. First, despite this evidence, I cannot see any reason to dispute all versions of academic theories taking as their task the identification of distinctive values attached to universities and to commercial undertakings. If the question is one that asks what values distinguish academia from industry, then I can see no better response than that which points to academic values clustering around disinterestedness, autonomy, spontaneity and openness, versus commercial values centring on concrete economic outcomes, organization, planning, and the control of intellectual property. In academic institutions, it might plausibly be said, the ‘Mertonian’ values can be publicly celebrated as institutional essence, while in industrial research they are more often asserted tactically as reminders to the uninformed that research is, to a great extent, an uncertain business, not to be subjected to the accountability regimes of other corporate activities. Yet, a theory of ideal-typical differences between institutional cultures is one thing and a description of quotidian realities in complex institutional environments is quite another. Those in the practical business of managing research enterprises have tended to acknowledge the intractable problems of distinguishing between these institutional environments, for theorizing essential differences has been of little concern to them. A paper company research director, given the task of speaking about such things in the mid-1960s, threw up his hands in despair: ‘Neither industrial nor academic research is, as it were, monochromatic; by whatever criteria we choose for judgement, both yield very broad spectra’. He found intra-group differences at least as great as inter-group differences, and thought it ‘quite possible that there are no such unique species as industrial vis-á-vis academic research’ (Nissan 1966: 211–212). Just as nominalism or particularism in these matters seems the natural attitude of the practical actor, so theorizing essential differences has an affinity with the ideological work of distributing value across the institutional landscape. So among academic commentators, most especially in the social sciences, there has been a persistent tendency to elide a distinction between
24 An industrial research director interviewed by Lowell Steele in the early 1950s remarked that ‘he was constantly amazed at the speed at which his scientists became interested in the welfare of the company’. Steele was then aware – though his later change of mind has been noted – that many commentators thought it impossible to get scientists to become ‘company conscious’, but cited such evidence to argue that ‘this criticism is unsound’ (Steele 1953: 471).
Who is the Industrial Scientist?
67
ideal-typifications and accounts of institutional realities. The task of defending and criticizing institutional kinds is thereby made so much easier. So far as possible, apples should always be compared with apples and not with oranges. It may well be that the best response to the questions ‘What makes university science different from industry?’ and ‘What are scientists like?’ is that offered in story 1. For all that, certain obstinate facts remain: (1) by the middle of the 20th century the majority of academically trained American scientists did not work in universities, and neither industrial nor government laboratories had any problems recruiting as many as they wanted, even if there was always a struggle for the best and brightest;25 (2) American universities, certainly at the beginning of the century, and arguably to the present day, were not globally regarded as natural homes for research: most were under-resourced; most had a primary commitment to teaching;26 many experienced cultural, political and religious pressures that seriously compromised any notion that universities, as such, were communities of free, open and suitably resourced inquirers (e. g. Veblen 1957; Kohler 1990). With due respect, the University of Southern Mississippi is not MIT, and, as we have seen, General Electric was well able to attract distinguished researchers away from even MIT because there was more effective freedom of action at the company (Servos 1980, esp. p. 532, for poor research resources at MIT in the early 20th century). Free action in research is a matter of material resources and time as well as of rhetoric and ideals. A comparison between conditions at Harvard and at the scientific laboratory of a small chemical company might sustain story 1, but many highly pertinent comparisons between particular institutions of higher education and many particular industrial research facilities definitely do not. Such comparisons are as available in the 1920s as they remain today. And it is good to remember the restrictions on free research problem25 As early as the late 1920s, James Bryant Conant was noting the change: among chemists awarded the PhD at Harvard between 1907 and 1917, practically all intended an academic career; a decade later, ‘the majority of those who received the advanced degree would enter industrial employment either at once or after a few years’ (1970: 25). Government statistics showed that in the late 1940s 54 percent of all US PhD chemists worked in industrial laboratories, 10 percent in government laboratories and only 33 percent in academia, without specifying the percentage of the latter active in research (cited in Hagstrom 1975: 38). The 1950 Census was the fist to specify personnel in scientific fields other than chemistry, and it showed that of about 150,000 total ‘natural and physical scientists’ – including those without higher degrees – only 33,000 (or a little over a fifth) were working as teachers in institutions of higher education, the only notable exception being mathematicians, 80 percent of whom were in higher education (figures reproduced from National Manpower Council 1953: 47; See also, for example, Hirsch 1968: 4–5, 60; and Hickey 1958: 116–117). 26 Writing in 1914, Kenneth Mees gave reasons why the universities were unlikely ever to become natural homes for scientific research: ‘For the last fifty years it has been assumed that the proper home for scientific research is the university, and that scientific discovery is one of the most important – if not the most important – function which a university can fulfil. In spite of this only a few of the American universities, which are admittedly among the best equipped and most energetic of the world, devote a very large portion of their energies to research work, while quite a number prefer to divert as little energy as possible from the business of teaching, which they regard as the primary function of the university’ (Mees 1914: 618).
68
Knowledge as Social Order
choice that were, and remain, endemic in even the best universities. As a member of a sociology department, for example, 1 am free in principle to conduct research in any area I like, but I cannot expect, and I do not receive, credit from my colleagues for work they do not recognize as sociology, nor am I institutionally well placed to obtain support for any research I might propose to do in psychology or economics or particle physics. Moreover, as a research director for the Carrier Corporation noted in the early 1960s, ‘I know of some cases where competent Ph.D. candidates left the universities of their choice [for industry] because they either were not allowed to work on a topic picked by them, or they could not get as advisor the man they wanted in their own field’ (Smith 1962: 261). These, and other, autonomy-restricting states of affairs are not at all uncommon in academia, nor are they inconsistent with saying that universities have an in-principle association with the ideal of research freedom.27 Second, as I have already noted, some academic sociologists, and most explicitly those involved in the early ASQ were in the business of elaborating what they called ‘a general theory of administration’, and in that cause both ideal types and the most robust possible generalizations of organization similarities and differences were central. It is in such a theoretical context that questions like ‘What makes the academy different from industry?’ take on special salience. The shop-floor commentary, by contrast, was not about offering an opposing theory of administration, nor, even, of asking ‘What makes the academy similar to industry?’ Such commentary took problems as they presented themselves to managers concerned to cope with their own quotidian circumstances and to learn from others’ experience if and when such experience seemed warrantably pertinent. The research managers, so to speak, played Simplicio to Galileo’s Salviati: they wanted to know how to deal with this research laboratory; they were not much interested in specifying ‘the nature of science’, ‘the nature of the university’, or ‘the nature of industry’. There is, of course, no escaping the science studies principle that ‘it’s rhetoric all the way down’ – how else would you have access to past organizational realities except through the heterogeneous commentaries of various participants and observers? But one reason why I feel somewhat more confident in the empirical grip of the shop-floor version is that there is no evident commitment to theory-building and no evident apologetic and justificatory flavour to such commentary. If the scientists were being awkward for the reasons predicated by the academic sociologists, I feel reasonably confident that the research managers would have said so. So there are, indeed, problems with the empirical adequacy of the academic story, problems similar to those pointed out in the late 1960s and early 1970s by such British sociologists of science as Barry Barnes, Steven Box and Stephen Cotgrove, and even by such renegade Merton students as Norman Kaplan, whose work in this 27 By the late 1960s, troubled by the changes accompanying Big Science, some academic sociologists nervously broached the idea that, even for the academic scientist, ‘autonomy is a relative matter: his “role set” involves at the very least the judgment of his peers, not to speak of the constraints imposed on him by his need to have a continuous flow of funds for equipment and research assistance – or just plain time to think’ (Hirsch 1968: 67). And, for more full-blooded scepticism, see (Kaplan 1964, esp. p. 114).
Who is the Industrial Scientist?
69
area now appears outstandingly sensitive and prescient (see, notably, Kaplan 1964; also Cotgrove and Box 1970; and Barnes 1971). In this connection, criticism of academic theorizing culpably disengaged from empirical realities can be joined to an historical appreciation of the circumstances in which such a story emerged and secured some local credibility. On the one hand, an academic social science strongly committed to establishing its scientific bona fides was actively seeking for theories of great predictive power, just as it was looking for typologies that allowed practitioners to sort institutions into their species and genuses. To that extent, I consider that much academic social science which dealt with ‘unhappy industrial scientists’ was in large part (not wholly) deducing its objects from theory. At this distance, one has to take on trust the interview-responses recorded by academic sociologists documenting substantial ‘role conflict’ among industrial scientists in the 1960s, but it is not impossible to give alternative plausible interpretations to some of this evidence, and one wonders why a post-World War II survey of scientists’ attitudes saw no problems of bias in posing such questions as this: ‘Aside from money considerations, where do you think a person can get most satisfaction from a career in science – in the Federal Government, in an industrial laboratory, in a university, or somewhere else?’ (Steelman 1947, Vol. 3: 142, 208; my emphases).28 Nor is it any part of my argument that all, or even a certain high percentage, of industrial scientists were happy as clams at high tide – that would be a remarkable state of affairs for members of any organization, much less organizations with such a high degree of organizational uncertainty as industrial research laboratories. Nor is there any reason to ignore occasional – though scarce – expressions of unhappiness among industrial scientists that drew upon cultural repertoires that circulated in the cultural neighbourhood and that might lend some credibility to story 1. However, what is undeniable is the remarkable gap between the world as seen by academic sociologists in mid-century and the same bit of the world as seen by research managers. Moreover, the cultural and political conjunctures from which this academic commentary emerged had characteristics that made the story about role-conflict especially appealing to American social scientists and humanists. If natural scientists and engineers in the Cold War period had institutional options – many could jump from the university to industry or government if they wished – the great majority of social scientists and humanists did not (for such perceptions, see, for example, Hirsch 1968: 53). The university was their natural home in a much stronger sense than it was the natural scientists’ home. And, as we well know, during the Cold War that home, and the values that were most cherished by the humanists, were under serious attack. Many of the social scientists and their allies did indeed defend themselves against McCarthyism and the militarization and commercialization of the academy, but, more to the point, some emerged as the most articulate defenders of science, whether or not they identified what they did as science. So, for example, in 1958, Merton’s colleague Paul Lazarsfeld wrote that college professors as a body tend to be naturally selected from a ‘permissive’ and ‘liberal’ environment: ‘Once he 28 Of all groups responding, 11 percent preferred government, 31 percent industry and 48 percent academia. Note that when industrial scientists alone were asked to put money considerations aside, 58 percent thought industry gave most satisfaction.
Knowledge as Social Order
70
is on campus, economic and social circumstances also militate against a conservative affirmation by the academic man’. But for that minority who came into the academy from a different background, there might be problems adjusting: For a young person coming from a business background, a teaching career involves a break in tradition. The business community has an understandable affinity with the conservative credo, with its belief in the value of tradition and authority, its corresponding distrust of people who critically scrutinize institutions like religion and the family, and its beliefs in the social advantages of private property and the disadvantages of state interference in economic affairs (Lazarsfeld and Thielens 1958: 148–149; with a field report by David Riesman).
Role-conflict was thus naturalized by some sociologists as a feature of the academic identity, whether humanist, social scientist, or natural scientist. The message was in effect ‘We are all in this together’, and Oppenheimer’s struggle in the 1954 security hearings was assimilated to their own. If you wanted to fight external control, and, of course, many humanists and social scientists during the Cold War did not, then damaged scientific allies were nonetheless important allies. The sociologists developed a picture of the nature of the scientific community that invited scientists to consider themselves allies of their non-scientific colleagues. That picture was inter alia a resource for healing emerging fissures in elite American universities. It was a cause as noble as it eventually proved to be futile (see, e.g., Hollinger 2000). References Anthony, R.N. (1952), Management Controls in Industrial Research Organizations (Boston: Graduate School of Business Administration, Harvard University). Ball, C.R. (1926), ‘Personnel, Personalities and Research’, Scientific Monthly 23, 33–45. Barber, B. (1952), Science and the Social Order (New York: Collier Books). Barnes, B. (1971), ‘Making Out in Industrial Research’, Science Studies 1, 157–175. Baum, A.W. (1947), ‘Doctor of the Darkroom’, The Saturday Evening Post, 25 Oct., 15–17, 47, 50, 52. Bello, F. (1956), ‘Industrial Research: Geniuses Now Welcome’, Fortune, Jan., 96– 99,142–150. ––– (1958), ‘The World’s Greatest Industrial Laboratory’, Fortune, Nov., 148–157, 208–220. Bichowsky, F.R. (1942), Industrial Research (Brooklyn, NY: Chemical Publishing Co.). Birr, K. (1957), Pioneering in Industrial Research: The Story of the General Electric Research Laboratory (Washington, DC: Public Affairs Press). Bloor, D. and Bloor, C. (1982), ‘Twenty Industrial Scientists: A Preliminary Exercise’, in Douglas (ed.), 83–102. Boyd, T.A. (1957), Professional Amateur: The Biography of Charles Franklin Kettering (New York: E. P. Dutton).
Who is the Industrial Scientist?
71
Brooks, H. (1968), The Government of Science (Cambridge, MA: MIT Press). Bryson, L. (1957),‘Researchers in Industry’, in Livingstone and Milberg (eds), 129–137. Burns, T. (1964), ‘Research, Development and Production: Problems of Conflict and Cooperation’, in Orth et al. (eds), 112–129 (orig. pub. in IRE Transactions on Engineering Management, March 1961, 15–23). Carty, J.J. (1929), ‘Science and Business: An Address to the Chamber of Commerce of the United States, 8 May 1924’, Reprint and Circular Series of the National Research Council (Washington, DC: National Research Council). Comwell, R.T.K. (1948), ‘Professional Growth of the Research Man’, in Furnas (ed.), 295–307. Conant, J.B. (1970), My Several Lives: Memoirs of a Social Inventor (New York: Harper and Row). Cotgrove, S. and Box, S. (1970), Science, Industry and Society: Studies in the Sociology of Science (London: George Allen and Unwin/New York: Barnes and Noble). Dennis, M.A. (1987), ‘Accounting for Research: New Histories of Corporate Laboratories and the Social History of American Science’, Social Studies of Science 17, 479–518. Douglas, M. (ed.) (1982), Essays in the Sociology of Perception (London: Routledge and Kegan Paul). Edgerton, D.E.H. (1988), ‘Industrial Research in the British Photographic Industry, 1879–1939’, in Liebenau (ed.), 106–134. Engstrom, E.W. (1957), ‘What Industry Requires of the Research Worker’, in Livingston and Milberg (eds), 69–79. Eustis, W. (1948), ‘Personnel Policies and Personality Problems’, in Furnas (ed.), 277–294. Evans, C.G. (1969), Supervising R&D Personnel (New York: American Management Association). Fisk, J.B. (1959), ‘Basic Research in Industrial Laboratories’, in Wolfle (ed.), 159– 167. Furnas, C.C. (1948), Research in Industry: Its Organization and Management (New York: D. Van Nostrand). Gerstenfeld, A. (1970), Effective Management of Research and Development (Reading, MA: Addison-Wesley). Gorog, I. (1966), ‘Successful Adjustment to Industrial Research: How Can University and Industry Assist’, Research Management 9, 5–13. Gouldner, A.W. (1957–1958), ‘Cosmopolitans and Locals: Toward an Analysis of Latent Social Roles’, Administrative Science Quarterly 2, 1–306, 444–480. Hackett, J.W. and Steierman, B.L. (1963), ‘The Organization of a Fundamental Research Effort at Owens-Illinois Glass Company’, Research Management 6, 81–92. Hagstrom, W.O. ([1965] 1975), The Scientific Community (Carbondale: Southern Illinois University Press). Hebb, M.H. and Martin, M. J. (1958), ‘Free Inquiry in Industrial Research’, Research Management 1, 67–83.
72
Knowledge as Social Order
Hertz, D.B. (1950), The Theory and Practice of Industrial Research (New York: McGraw-Hill). Hertz, D.B. and Rubenstein, A.H. (1952), Selection, Training, and Use of Personnel in Industrial Research, Proceedings of the Second Annual Conference on Industrial Research, June 1951 (New York: King’s Crown Press). Heyel, C. (ed.) (1959), Handbook of Industrial Research Management (New York: Reinhold Publishing Corporation). Hickey, A.E. Jr. (1958), ‘Basic Research: Should Industry Do More of It?’, Harvard Business Review 36(4), 115–122. Hill, K. (ed.) (1964), The Management of Scientists (Boston: Beacon Press). Hirsch, W. (1968), Scientists in American Society (New York: Random House). Hollinger, D.A. (2000), ‘Money and Academic Freedom a Half-Century after McCarthyism: Universities and the Force Fields of Capital’, in Hollingsworth (ed.), 161–184. Hollingsworth, P.J. (ed.) (2000), Unfettered Expression: Freedom in American Intellectual Life (Ann Arbor: University of Michigan Press). Hounshell, D.A. (1996), ‘The Evolution of Industrial Research in the United States’, in Rosenbloom and Spencer (eds), 13–85. ––– (2000), ‘The Medium is the Message, or How Context Matters: The RAND Corporation Builds an Economics of Innovation, 1946–1962’, in Hughes and Hughes (eds), 255–310. Hounshell, D.A. and Smith, J.K. Jr. (1988), Science and Corporate Strategy: Du Pont R&D, 1902–1980 (Cambridge: Cambridge University Press). Hughes, T.P. and Hughes, A.C. (eds) (2000), Systems, Experts, and Computers: The Systems Approach in Management and Engineering, World War II and After (Cambridge, MA: The MIT Press). Jackson, W. (1957), ‘Discussion’, The Direction of Research Establishments: Proceedings of a Symposium held at the National Physical Laboratory [Teddington] on 26th, 27th and 28th September 1956 (New York: Philosophical Library). Jewett, F.B. (1919), ‘Industrial Research’, Reprint and Circular Series of the National Research Council, No. 4 (Washington, DC: National Research Council). Jewkes, J., Sawers, D. and Stillerman R. (1958), The Sources of Invention (London: Macmillan). Kaplan, N. (1964), ‘Organization: Will It Choke or Promote the Growth of Science?’ in Hill (ed.), 103–127. ––– (ed.) (1965), Science and Society (Chicago: Rand-McNally). Kaplan, R. (2001), Science Says: A Collection of Quotations on the History, Meaning, and Practice of Science (New York: Stonesong Press). Kastens, M.L. (1957), ‘Research-A Corporate Function’, Industrial Laboratories 8(10), 92–101. Kevles, D.J. (1979), The Physicists: The History of a Scientific Community in Modem America (New York: Vintage Books). Kingsbury, S., Bass, L.W. and Lothrop, W.C. (1959), ‘Organizing for Research’, in Heyel (ed.), 65–99.
Who is the Industrial Scientist?
73
Kipp, E.M. (1960), ‘Introduction of the Newly Graduated Recruit to Industrial Research’, Research Management 3, 39–47. Kliever, W.H. (1952), ‘Design of Research Projects and Programs’, Industrial Laboratories 3(10), 6–13. Kohler, R.E. (1990), ‘The PhD Machine: Building on the Collegiate Basis’, Isis 81, 638–662. Kornhauser, W. (1962), Scientists in Industry: Conflict and Accommodation (Berkeley: University of California Press). Laughlin, D.M. (1952), ‘The Engineer Shortage: Challenge to Management’, Personnel 29, 418–421. Lazarsfeld, P.F. and Thielens, W. Jr. (1958), The Academic Mind: Social Scientists in a Time of Crisis (Glencoe, IL: The Free Press). Leermakers, J.A. (1951), ‘Basic Research in Industry’, Industrial Laboratories 2(3) 2–3. Liebenau, J. (ed.) (1988), The Challenge of New Technology: Innovation in British Business Since 1850 (Aldershot: Gower). Litchfield, E.H. (1956), ‘Notes on a General Theory of Administration’, Administrative Science Quarterly 1, 3–29. Livingston, R.T. and Milberg, S.H. (eds) (1957), Human Relations in Industrial Research, Including Papers from the Sixth and Seventh Annual Conferences on Industrial Research: Columbia University, 1955 and 1956 (New York: Columbia University Press). Marcson, S. (1960), The Scientist in American Industry: Some Organizational Determinants in Manpower Utilization (New York: Harper & Row). McClure, H.B. (1952), ‘External Communication of Research Results’, in Hertz and Rubenstein (eds), 161–176. McGrath, P.J. (2002), Scientists, Business, and the State, 1890–1960 (Chapel Hill: University of North Carolina Press). Mees, C.E.K. (1914), ‘The Future of Scientific Research [Editorial]’, Journal of Industrial and Engineering Chemistry 6, 618–619. ––– (1916), ‘The Organization of Industrial Scientific Research’, Science 43(1118), 763–773. ––– (1920), The Organization of Industrial Scientific Research (New York: McGrawHill). ––– (1945), ‘Discussion of Midgley’, in Standard Oil Development Company (1945). ––– (1947), ‘Discussion [of Michael Polanyi, “The Foundations of Freedom in Science”]’, in Wigner (ed.), 140–141. ––– (1961), From Dry Plates to Ektachrome Film: A Story of Photographic Research (New York: Ziff-Davis). Mees, C.E.K. and Baker, J.R. (1946), The Path of Science (London: Chapman and Hall). Mees, C.E.K. and Leermakers, J.A. (1950), The Organization of Industrial Scientific Research, 2nd edn. (New York: McGraw-Hill). Merton, R.K. (1942), ‘The Normative Structure of Science’, in Storer (ed.), 267–278.
74
Knowledge as Social Order
National Manpower Council (1953), A Policy for Scientific and Professional Manpower: A Statement by the Council with Facts and Issues Prepared by the Research Staff (New York: Columbia University Press). National Science Foundation (1961), Publication of Basic Research Findings in Industry (Washington, DC: National Science Foundation). Nelson, R.R. (1959), ‘The Economics of Invention: A Survey of the Literature’, The Journal of Business 32, 101–127. Nissan, A.H. (1966), ‘Similarities and Differences between Industrial and Academic Research’, Research Management 9, 211–219. Noble, D.F. (1977), America by Design: Science, Technology, and the Rise of Corporate Capitalism (New York: Alfred A. Knopf). Orth, C.D. 3rd (1957), ‘More Productivity from Engineers’, Harvard Business Review 35(2), 54–62. ––– (1959), ‘The Optimum Climate for Industrial Research’, Harvard Business Review 37(2), 55–64. Reprinted in Kaplan (ed.), 194–210. Orth, C.D. 3rd, Bailey, J. C. and Wolek, F. W. (eds) (1964), Administering Research and Development: The Behavior of Scientists and Engineers in Organizations (Homewood, IL: Richard D. Irwin, Inc. and the Dorsey Press). Parsons, T. (1956), ‘Suggestions for a Sociological Approach to the Theory of Organizations (I) and (II)’, Administrative Science Quarterly 1, 63–85, 225–239. Powers, P.N. (1952), ‘Industrial Research Workers and Defense’, in Hertz and Rubenstein (eds), 94–112. Randle, C.W. (1959), ‘Problems of Research and Development Management’, Harvard Business Review 37(1), 128–136. Reich, L.S. (1985), The Making of American Industrial Research: Science and Business at GE and Bell, 1876–1926 (Cambridge: Cambridge University Press). Rosenbloom, R.S. and Spencer, W.J. (eds) (1996), Engines of Innovation: U.S. Industrial Research at the End of an Era (Cambridge, MA: Harvard University Press). Rowland, H. (1902), The Physical Papers of Henry Augustus Rowland, Johns Hopkins University, 1876–1901 (Baltimore: The Johns Hopkins Press). Schmitt, R.W. (1961), ‘Why Publish Scientific Research from Industry?’, Research Management 4(3), 1–41. Secrist, H.A. (1960), ‘Motivating the Industrial Research Scientist’, Research Management 3, 57–64. Servos, J.W. (1980), ‘The Industrial Relations of Science: Chemical Engineering at MIT, 1900–1939’, Isis 71, 531–549. Shepard, N.A. (1945), ‘How Can We Build Better Teamwork within Our Research Organizations?’, Chemical and Engineering News 23(9), 804–807. Smith, J.F.D. (1962), ‘Academic and Industrial Research’, Research Management 5, 257–275. Standard Oil Development Company (1945), The Future of Industrial Research: Papers and Discussion (New York: Standard Oil Development Company). Steele, L.W. (1953), ‘Personnel Practices in Industrial Laboratories’, Personnel 29, 469–476.
Who is the Industrial Scientist?
75
––– (1957), ‘Rewarding the Industrial Scientist: A Problem of Conflicting Values’, in Livingstone and Milberg (eds), 163–175. Steelman, J.R. (1947), Science and Public Policy: A Report to the President by John R. Steelman, Chairman, The President’s Scientific Research Board, 5 vols. (Washington, DC: Government Printing Office). Steinmetz, C.P. (1916), ‘Scientific Research in Relation to the Industries’, Journal of the Franklin Institute 182, 711–718. Storer, N.W. (1961), Science and Scientists in an Agricultural Research Organization: A Sociological Study (Cornell University, Unpublished Ph.D. thesis). ––– (1966), The Social System of Science (New York: Holt, Rinehart and Winston). ––– (ed.) (1973), The Sociology of Science (Chicago: University of Chicago Press). Taylor, C.W. and Barron, F. (eds) (1963), Scientific Creativity: Its Recognition and Development: Selected Papers from the Proceedings of the First, Second, and Third University of Utah Conferences: ‘The Identification of Creative Scientific Talent’ (New York: John Wiley & Sons). Thompson, J.D. (1956), ‘On Building an Administrative Science’, Administrative Science Quarterly 1, 102–111. Tietjen, J.J. (1966), ‘The Transition from Graduate School to Industrial Research’, Research Management 9, 109–113. Van Raalte, J.A. (1966), ‘Reflections on the PhD Interview’, Research Management 9, 307–317. Veblen, T. ([1918] 1957), The Higher Leaning in America: A Memorandum on the Conduct of Universities by Business Men (New York: Sagamore Press). Voorhies, D.H. (1946), The Co-ordination of Motive, Men and Money in Industrial Research, A survey of Organization and Business Practice Conducted by the Department on Organization of the Standard Oil Company of California (San Francisco: Standard Oil Company of California). Whyte, W.H. Jr. (1956), The Organization Man (New York: Simon and Schuster). Wigner, E.P. (ed.) (1947), Physical Science and Human Values (Princeton, NJ: Princeton University Press). Williams, D.L. (1947), Planning of Research and Development (New York: Wallace Clark). Wilson, R.E. (1949), ‘The Attitude of Management toward Research’, Chemical and Engineering News 27(5), 274–277. Wolfle, D. (ed.) (1959), Symposium on Basic Research, Sponsored by the National Academy of Sciences, the American Association for the Advancement of Science, and the Alfred P. Sloan Foundation (Washington, DC: American Association for the Advancement of Science).
This page intentionally left blank
Chapter 4
The Meaning of Hoaxes Harry Collins
One of Barry Barnes’s most well known interventions in the debate about the meaning of knowledge is to make a distinction between social kinds and natural kinds. Natural kinds, like ‘mountain’, refer to something outside society, whereas social kinds, like ‘money’, get all their meaning from the way they are thought about and talked about. I cannot claim to be making any significant contribution to this distinction here but it is my excuse to think about hoaxes. Someone else may be able to say what kind of kinds they are – I’ll just provide some material. In 1996 Alan Sokal, a physicist, published a paper in the cultural studies journal Social Text which he immediately declared to be a hoax, having no serious content (1996). This severely embarrassed the editors of Social Text and, depending on how you view the matter, it also embarrassed a larger or smaller number of social scientists. From November 2001 onwards, the Bogdanov brothers published a series of papers in scientific journals: Classical and Quantum Gravity (Bogdanov and Bogdanov 2001); Annals of Physics (Bogdanov and Bogdanov 2002a); Il Nuovo Cimento (Bogdanov and Bogdanov 2002b); Czechoslovak Journal of Physics (Bogdanov 2001); and Chinese Journal of Physics (Bogdanov 2002). The content of the papers was something to do with ‘string theory’ and the brothers were awarded PhDs on the strength of their work by the University of Burgogne in France. In late 2002, it was widely bruited that their work was a hoax but within a day or two the Bogdanov brothers, who were not the source of the hoax accusation, denied it was any such thing. Embarrassed scientists had to apologize for the hoax accusation. In the meantime, the work was read more carefully and while a few scientists said they thought it contained grains of interest, most agreed that it was bad work – a pastiche at best – that should not have been published.1 Of course, the Bogdanov brothers may or may not have had a hoax in mind; only they know. The obvious consequences for the Bogdanov hoax/non-hoax are roughly the same as for the Sokal hoax. It is just as embarrassing, or even more embarrassing, for a scientific community to discover that it cannot tell whether it has been hoaxed as for it to discover that it has been hoaxed. This, of course, is not as embarrassing as it might be since most published scientific work is not very good anyway, and most is not cited by anyone except its authors. If there is some ‘funny business’ taking place within the big soft underbelly of published work which never makes it far 1 A long exchange among physicists (with a couple of small contributions from the author), can be found at Google, groups, sci.physics.research.
78
Knowledge as Social Order
into scientific consciousness it may not matter very much except to the referees and editors who are immediately involved. Sociologists of science have long known that the acclaimed gold standard of peer review contains a lot of base metal. Nevertheless, the logic of hoaxing remains intriguing and there may be a lesson about the logic (as opposed to the well understood ‘sociologic’), of the peer review system to be drawn out. In the following, when I refer to a ‘hoax’ what I will have in mind is a published hoax of the Sokal type. What are the characteristics of a hoax of this kind? First, to be a hoax a number of readers must initially take the paper to be the ‘real thing’. Ideally, all readers will take it to be the real thing until they are told it is a hoax. (Compare this with a spoof. In the case of a spoof, the writer makes it obvious that it is a pastiche or irony. In the case of a hoax this is not made clear until later – after people have taken it seriously). It follows that a hoax must resemble the real thing. In an ideal hoax the resemblance will be all in the ‘form’ with no resemblance in the content. The hoaxer wants the readers to be ‘taken in’ by the form and thus demonstrate their lack of integrity when it comes to evaluating the content.2 Even if the content has to resemble the real thing to some extent the bigger the deficiency the better; if the content is too good there is no hoax, just a contribution to science. A hoax, then, has to pass initial scrutiny even though the content has to be pretty deficient for the subsequent ‘revelation’ to have a satisfactory sting. We can be fairly sure that the reason this is possible is that a hoax passes its initial hurdle in something of the same way as a confidence trick; the ‘mark’ will do most of the work in ‘repairing’ any more obvious faults because to do otherwise means a large disturbance in life’s routines. In confidence tricks there is usually a positive incentive for repair – for example you get a lot of money, preferably via illicit means, if the offer/person is genuine. But being able to get on with ordinary life without disturbance is also a worthwhile reward. It has been argued that this is why bogus doctors succeed so well (Collins and Pinch 2005). Busy academics also want to get on with their lives hence reviewers and editors are more likely to let something pass than cause a lot of trouble for themselves by exploring the possibility that a contributor is not what it appears to be. What all this helps to establish, if it was not already clear, is that detecting the difference between a hoax and the real thing is not straightforward – that is the very nature of a hoax. It is particularly hard to tell the difference between a hoax and a bad paper because both have the same characteristics: appropriate form with content that is poor but not obviously beyond the pale on first reading. So what is the difference between a hoax and a bad paper? It is the intention of the author. If the author deliberately writes nonsense then it is a hoax; if the nonsense is there even though the author was writing with the best of intentions, then it is not a hoax. As illustrated by the Bogdanovs, the same paper could be a hoax or just (what most people count as) bad work.
2 That incidentally is one of the beauties of Sokal’s hoax; once you forget the content you can see how hilariously he has captured the form.
The Meaning of Hoaxes
79
But wait! Maybe the author is a bit more dead than this. Suppose the author writes nonsense deliberately but never reveals his or her intention. Then the nonsense simply joins all the rest of the bad stuff in the underbelly of science and that is the last we hear of it. If no one knows it is a hoax, then it is not a hoax. Part of the idea of a hoax is to reveal its true nature at some time in the future; if it is not revealed, then what is the point? (I am reminded of that wonderful moment in Doctor Strangelove when an astounded President of the United States asks the Soviet Ambassador why the Russians had not declared the existence of the Doomsday Machine since deterrence is the entire point and no one can be deterred by something they do not know about.) A hoax isn’t a hoax unless you know you’ve been hoaxed. So the hoaxishness depends on the knowledge of the readers, not just on the intention of the author. In the Bogdanov case, for about two days (when I was lucky enough to be deeply involved in conversations about the matter with physicists), the work was taken to be a hoax and was treated as a hoax, including readings of the work accompanied by loud guffaws and ‘How-could-they-possibly-have-got-away-with-that?’s. Those readings ceased when the Bogdanovs denied that their intention was to bring off a hoax; the Bogdanovs preferred to have their work interpreted as either brilliant and original, or failing that, just another contribution to the soft underbelly of bad science. Incidentally, it is almost sure that no-one would have accused the Bogdanovs of hoaxing in the first place if hoax consciousness had not been engendered by the ‘Sokal affair’. The early reports of the Bogdanov ‘hoax’ all referred back to the Sokal affair. Had Sokal not done his work the Bogdanov pieces would almost certainly have disappeared within the body of other non-influential scientific work (unless the brothers themselves had decided to publicize it in some other way). We can also imagine the opposite case. Suppose someone writes nonsense inadvertently, realizes only later that it is nonsense, and announces it as a hoax in order to save face? Maybe that is what Alan Sokal did! (I’m sure it isn’t what Sokal did but it doesn’t spoil the logic of the situation.) Furthermore, some commentators, working in a ‘post-modern’ spirit, declared that Sokal’s paper was not a hoax at all but a genuine contribution which Sokal had declared to be a hoax for mischievous reasons. So where does hoaxishness lie? Where is the locus of hoaxishness? Is it in the author’s intentions, in the author’s declarations about their intentions, or in the community’s beliefs about the author’s intentions? The way this actually works out in practice, at least on the basis of the two cases we have before us, is that since it is well known that the most authoritative view of inner states such as intentions, comes from the experiencer of those states, it is the author’s declarations that count for most. Since Freud we have known that even we can’t be sure about our inner states but, if there is to be an argument about it, outside mental institutions etc, our own view of our inner states still carries more weight than the view of others. In sum: - The ontological locus of a hoax is in the intention of the author: if the author did not intend a hoax there was no hoax.
80
Knowledge as Social Order - The consequential locus of a hoax is in what the audience believes: the hoax only has its intended effect if the audience believes it has been hoaxed. - The epistemological locus of a hoax is the declaration of the author: if the author does not declare that the work was intended as a hoax no-one can prove it.
Try the last point any other way: try arguing that the Bogdanovs really were hoaxing or Sokal really wasn’t – What do you use for evidence? Thus, so long as the authors’ declarations remain unchanged Sokal, as far as we know, hoaxed us and the Bogdanovs did not. So, where are we going with all this? First, a hoax cannot be so damaging to science as many have made it out to be since it depends on the declaration of the author rather than the content of the publication. The damage caused by a hoax ought to be about the same as the damage caused by a bad publication. If the journals are full of bad papers, that’s just about as damaging as the journals being full of hoaxes. There is practically no damage caused by the journals being full of bad papers so any additional damage caused by hoaxes is just to do with the way they are publicized – science remains just as unchangingly bad as it always was. If someone could get the newspapers interested in the fact that they had published a really bad paper it would be just as damaging as someone publishing a hoax. This, effectively, is what happened in the case of the Bogdanovs. Luckily, there are so many bad papers that public interest would be unlikely to remain aroused for long. Second, the whole enterprise of science, notably the idea of scholarly journals, anonymised peer-reviewing, and so forth, depends on the notion that the value of a piece of work is visible within the text alone. Only that way could a peer reviewer judge an anonymous submission to a journal (strangely enough, the hard sciences depend just as much on the idea that `the author is dead’ as the post-modernists – the author is not consulted when a paper is being reviewed). Yet in the case of a hoax, the value depends on much more than the text – it depends on the declarations that the author makes about his or her intentions. Therefore (a) Roland Barthes is wrong about the author being dead in the case of a hoax and (b) the standard model of peer review in science does not work. And ‘b’ is for quasi-logical reasons as well as the well-understood sociological reasons. Having noticed this we are alerted to other similar instances where there is more to an article than what is in it. One such is any paper which rests on statements of statistical confidence. The value of such a result depends on more than what is in the paper – it depends on the history of the work described. If many other statistical tests have been tried on the data before the one that gives a significant result emerges, then the result is not as significant as it looks. Peer reviewers of statistical papers are, then, in an impossible position. In that case too the author is not dead. Less subtly, the reviewing of every empirical paper turns on the belief that the author actually did the experiments or measurements and did not just make them up, as cases of scientific fraud continually remind us. So the reviewing of every empirical paper turns on knowing more than what is in the paper (even as anonymity, which is supposed to prevent the reviewer knowing more than is in the paper, is encouraged).
The Meaning of Hoaxes
81
We return, then, via this quasi-logical excursion to the established sociological point, that in spite of the claims of science to ‘objectivity’ each result we accept rests on a complex web of trust for the author who produced it and therefore the author cannot possibly be dead. To rupture that web of trust deliberately, is in the case of frauds or hoaxes, is to play a dangerous game with the delicate fabric of science. References Bogdanov, G. and Bogdanov, I. (2001), Classical and Quantum Gravity, 18, 4341– 4372. ––– (2002a), ‘Spacetime Metric and the KMS Condition at the Planck Scale’, Annals of Physics, 296, 90–97. ––– (2002b), ‘KMS space-time at the Planck scale’, Il Nuovo Cimento B, 4, 417– 424. Bogdanov, I. (2001), ‘Topological Origin of Inertia’, Czechoslovak Journal of Physics, 51, 1153–235. ––– (2002), ‘The KMS State of Spacetime at the Planck Scale’, Chinese Journal of Physics, 40, 149–158. Collins, H.M. and Pinch, T. (2005), Dr. Golem: how to think about medicine (Chicago: University of Chicago Press). Sokal, A.D. (1996), ‘Transgressing the Boundaries: Toward a Transformative Hermeneutics of Quantum Gravity’, Social Text, 46/47, 217–252.
This page intentionally left blank
Chapter 5
Objectual Practice Karin Knorr Cetina
In this chapter, I want to develop some concepts designed to capture the affective and relational undergirding of practice in areas where practice is creative and constructive. Current conceptions of practice emphasize the habitual and rule-governed features of practice. Though much debate surrounds the exact specification of the relevant rules and habits (see Bourdieu 1977; Giddens 1984; Lynch 1993; Turner 1994; Schatzki 1996; Barnes 2000a, 2001), most authors seem to agree that practices should be seen as recurrent processes governed by specifiable schemata of preferences and prescriptions. Such processes are doubtlessly prominent in many areas of social life; their existence sustains our sense of practices as customary or routinized ways of behaving. However, it is also a characteristic of current times that many occupations and organizations have a significant knowledge base. In these areas, one would expect practitioners to have to keep learning, and the specialists who develop the knowledge base to continually reinvent their own practices of acquiring knowledge. Practice, in this case, would seem to take on a wholly different set of meanings and raise a different set of questions from the ones raised by habitual activities. For example, how can we theorize practice in a way that allows for the engrossment and excitement – the emotional basis – of research work? What characterization of practice might make the notion more dynamic, and include within it the potential for change? Research work seems to be particular in that the definition of things, the consciousness of problems, etc. is deliberately looped through objects and the reaction granted by them. This creates a dissociation between self and work-object and inserts moments of interruption and reflection into the performance of research, during which efforts at reading the reactions of objects and taking their perspective play a decisive role. How can we conceive of practice in a way that accommodates this dissociation? In this chapter, I want to address these questions by taking as my starting point a particular characterization of knowledge-centered practice – one that can be traced back to Heidegger but that also finds support in scientists’ self-understandings of their work. At the core of this characterization lies the assumption that creative and constructive practice – the kind of practice that obtains when we confront non-routine problems – is internally more differentiated than current conceptions of practice as skill or habitual task-performance suggest. The dissociation I have in mind is that between subject and (work) object; though time-differentiation is also important, I think subject-object differentiation captures more directly what happens when work ceases to be habitual procedure. What holds differentiated practice together and gives it continuity is the relationship between subject and object; this chapter
84
Knowledge as Social Order
is a first attempt to find the basis of knowledge-centered constructive and creative practice in a relational rather than a performative idiom. In moving in this direction, I assume that the relational idiom adequately captures the dynamic properties of research. The relational idiom can also carry the reflexive and affective aspects of epistemic practice (see also Barnes 1977). In addition it will bring into focus nonhuman objects, which dominate much of expert work. In the next section, I briefly review knowledge society arguments – those which maintain that professional knowledge activities are expanding in current Western societies and constitute the leading edge of post-industrial work. If these arguments are right – if we are confronted with the growth of knowledge-centered and knowledge-based activities in many areas of social life – epistemic practices may come to dominate other kinds of practice. The specific characteristics of these practices then also become interesting from a practice theory viewpoint. In section 3, I discuss reasons for a relational approach to conceptualizing practice. I will briefly revisit Heidegger’s perspective on these matters, and also present examples of how experts and scientists themselves view constructive practice. To specify how object relationships define the flow of practice one needs to discuss in some detail the notion of ‘object’ relevant to knowledge activities. This will be the topic of the section that follows (4). Knowledge objects differ in important ways from the commodities, instruments and everyday things discussed in the literature; the section spells out these differences, conceiving of epistemic objects as defined by their lack of completeness of being and their non-identity with themselves. The lack of completeness of being of knowledge objects goes hand in hand with the dynamism of research. Only incomplete objects pose further questions, and only in considering objects as incomplete do scientists move forward with their work. In section 5, I turn to the subjects rather than objects in attempting to specify further the relational and affective dynamic of expert practice. This section presents a way of conceiving of the bindingness, reflexivity and mutuality of experts’ objectrelationships as the backbone of this practice. The Knowledge Society Argument (1) What drives one to think about knowledge-creating and -validating or ‘epistemic’ practice? A recent source of concern with knowledge-centered activities are transformation theory arguments. These arguments conceive of current social transitions in terms of a shift from an industrial to a ‘post-industrial’ or ‘posttraditional’ society, in which knowledge is of increased relevance to the economy and other areas of social life. There is a widespread consensus today that contemporary Western societies are increasingly ruled by knowledge and expertise. The proliferation of concepts such as that of a ‘technological society’ (e.g. Berger et al., 1974), an ‘information society’ (e.g. Lyotard 1984; Beniger 1986), a ‘knowledge society’ (Bell 1973; Drucker 1993; Stehr 1994), a ‘risk society’ or ‘experimental society’ (Beck 1992; Krohn and Weyer 1994) embodies this understanding. The recent source of this awareness is Daniel Bell (1973), for whom the immediate impact of knowledge was on the economy, where it resulted in such widespread changes as shifts in the
Objectual Practice
85
division of labor, the development of specialized occupations, the emergence of new enterprises and sustained growth. Bell and later commentators (e.g. Stehr 1994) also offer a great many statistics on the expansion of R&D efforts, R&D personnel and R&D expenses in Europe and the United States. More recent assessments have not changed this argument so much as added further arenas of the impact of knowledge. For example, Habermas’s argument about the ‘technicization’ of the lifeworld through universal principles of cognitive and technical rationality attempts to understand the spread of abstract systems to everyday life (1981). Drucker (e.g. 1993) links knowledge to changes in organizational structure and management practices, and Beck (1992) depicts transformations of the political sphere through corporate bodies of scientists. Finally Giddens, arguing that we live in a world of increased reflexivity mediated by expert systems, extends the argument to the self, pointing out that today’s individuals engage with the wider environment and with themselves through information produced by specialists which they routinely interpret and act on in everyday life (e.g. 1990). The advantage of Giddens’s use of the notion expert ‘system’ is that it brings into view not only the impact of isolated knowledge items or of scientific-technical elites but implies the presence of whole contexts of expert work. These contexts, however, continue to be treated by him and others as alien elements in social systems, elements that are best left to their own devices. Knowledge society arguments consider knowledge as a productive force that – in a post-industrial society – increasingly plays the role that capital and labor played in industrial society. These viewpoints also emphasizes the role of experts, of technology and its associated risks, and of electronic information structures (see also Lash and Urry 1994). But the transition to knowledge societies involves more than the presence of more experts, more technological gadgets, more specialist rather than participant interpretations. It involves the presence of knowledge processes themselves – in the terms chosen here, it involves the presence of epistemic practice. From the present point of view, then, understanding knowledge societies will have to include understanding knowledge practices. In post-industrial societies, knowledge settings are no longer limited to science. To give an example, every major bank employs scores of ‘analysts’ and other ‘specialists’, who research and represent for the bank the world in which the bank moves. Hence research and analysis practices of different kinds penetrate many areas of social life; to some degree, these practices become constitutive of these areas. For example, global financial markets could not exist without the symbolic representations and analyses of trading activities and contextual events that are created by specialists in information provider firms and similar organizations. In other words, the reality of global financial markets is an expert-provided and expert-observed ‘on screen’ reality. The practices of creating this screen reality will need to be analyzed with respect to the specific knowledgeproducing and -validating strategies they implement; as earlier work suggests, knowledge-producing activities in different areas entail different epistemic cultures (see Knorr Cetina 1999). But also at stake is the conceptualization of knowledgecentered practice from a theoretical perspective which is the topic of this chapter.
86
Knowledge as Social Order
The Relational Undergirding of Epistemic Practice (2) I now want to begin to describe this practice, starting with the observation of the dissociative dynamic that comes into play when practice ceases to be a procedural routine. As indicated before, the dissociation relevant here is that between subject and object. What do we mean by this dissociation? How does it come about, and why is it important? The separation between subjects with mental states and independent objects is common to all areas of everyday life. To take an example, a car and its driver are distinct entities in our perception and in much of our experience. Nonetheless, while I am driving, my car becomes what Heidegger calls ‘ready-tohand’ and transparent (1962: 98ff.): it has the tendency to disappear while I am using it. In other words, the car becomes an unproblematic means to an end rather than an independent thing to which I stand in relation. It becomes an instrument that has been absorbed into the practice of driving, just as I, the driver, have been absorbed into the practice of driving – I, too, become transparent. When I engage in this practice, I am oriented to the street, the traffic, the direction I have to take. I am not oriented to the car – unless it malfunctions and temporarily breaks down. Nor am I thinking of myself as separate from the immediate activity. It should be plain that scientific practice, when it is routine or habitual, corresponds to this description. To give an example, consider the following comments of a researcher in a molecular biology laboratory whom I asked about her usage of laboratory protocols (for details see Knorr Cetina 1999, ch. 4): DS: You asked about protocols. We not only work with protocols, we think in terms of them. When I am doing the protocol, pipetting say, I don’t really think about the objects I am dealing with. When it’s a routine, there is, for me, no differentiation between the bacteria that I am using there, and the DNA that I’ll extract and the enzyme that I am placing on to cut the DNA. ‘A thing to do’ is more a protocol than dealing with DNA, it is more in the procedure than in the material.
Or consider another molecular biologist in the same laboratory, who described the practice of cloning in the following terms (see Knorr Cetina 1999, ch. 6): HB: Cloning is perhaps one level below what one calls exciting in the lab. You sit down, you think about a particular construct, and then you clone it. That’s not very different from deciding to dig a hole in the ground and then to dig it – it’s about that exciting.
This sort of practice can perfectly well be described in a performative idiom that conceptualizes it, in Wittgenstein’s terminology, as an ‘ungrounded way of acting’ (1969, No. 110: 17e; see also Dreyfus 1994: 155). Yet it appears equally clear that major portions of knowledge-centered work – those that best epitomize epistemic practice – are not adequately described in just these terms. These portions of scientific practice are not, in the above terms, ‘routine procedures’; they occur when problems arise, or when work is new to a researcher. Consider again an example, the first researcher’s response, at a later point in our conversation, to a question about the protein she was working on:
Objectual Practice
87
KKC: What about your protein? DS: Well, the protein, because it has previously been a problem, the protein is a bit more moody. I think about it, I get more visual, I treat it differently, in one word, I pay more attention to it, it’s more precious. I don’t handle it routinely yet. KKC: How do you visualize it? DS: I see the protein in a certain size in front of me. I visualize why it is precipitating, then I visualize the solution and I visualize the falling out and the refolding process. I also visualize the protein denaturing, streched out and then coming together, and I visualize how it is being shot into the solution and what it is going through when it starts to fold. With the expression, I visualize the bacteria when they grow in a more anthropomorphic way, why are they happy? I try to visualize them shaking around, I visualize aerobic effects, the shaking, how much they tumble around and what could have an effect.
In this second case, the object (e.g. the protein) is no longer ‘invisible’ and undifferentiated, an undistinguishable part of an activity script. Instead, it becomes enhanced and in fact enlarged through the researcher’s strategy of visualizing it and its environment and behavior under various circumstances. It is important to note not only the subject-object differentiation this entails, but also the researcher’s active usage of the means we have to overcome subject-object separation – her deployment of relational resources. Not only does this researcher experience herself as a conscious subject that relates to epistemic objects, she draws upon resources that are entailed in ‘being-in-relation’ in everyday life to help define and continue her research. I take these relational resources to include taking the role or perspective of the other; making an emotional investment (taking an interest) in the other; and exhibiting moral solidarity and altruistic behavior that serves the other person (see also Barnes 2000b). In the present case, DS can be said to take the role and perspective of the protein and of her bacteria; she also imagines the latters’ emotions, engaging in what is perhaps a form of empathy. In the following comment on her protein, DS. indicates her own emotional involvement: KKC: Is it (the protein) like a person? Someone you interact with? DS: No, not necessarily a person. It takes on aspects of some personality, which I feel, depending on if it has been cooperative or not. If it’s cooperative then it becomes a friend for a while, then I am happy and write exclamation marks in my book. But later it becomes material again, it goes back to being in a material state. When it stops doing what I want, then I see a personal enemy and think about the problems.
From the subject-object differentiation and the relational definition of the situation DS reaps, one imagines, insights, clarity about next moves, epistemic dividends. In other words, DS uses relational mechanisms as resources in articulating and ‘constructing’ an ill-defined, problematic, non-routine and perhaps innovative epistemic practice. When Heidegger analyzed our instrumental being-in-the-world as a form of unself-conscious but nontheless goal-directed employment of equipment in its referential context he also pointed out what happens when equipment becomes problematic
88
Knowledge as Social Order
(1962: 98ff.): then we go from ‘absorbed coping’ to ‘envisaging’, ‘deliberate coping’ and to the scientific stance of ‘theoretical reflection’ on the properties of entities. This characterization recaptures the ones I have given, with perhaps one difference. Heidegger came to characterize knowledge in terms of a theoretical attitude that entails a ‘withholding’ of practical reason. He gave an important characterization of how the project of science appears derivative of the primordial stance of taking things for granted in everyday life. But at the same time, his characterization provided less than an adequate account of knowledge processes – in which the presence of equipment is massive, instrumentality prevails, and theorizing rather appeals to us as being itself a form of practice. With the notion of a theoretical attitude, Heidegger brought back into the picture the subject-object differentiation which he had wanted to drive out of the philosophical discourse with his definition of ‘Dasein’ as a form of concerned coping (‘Self and world belong together in the single entity, Dasein. Self and world are not two entities, like subject and object ...’ cf. Dreyfus 1994: 67ff.). But perhaps as a consequence of his larger project, Heidegger never quite gave situations of subject-object distantiation the same consideration and attention that he gave to concerned coping. Theoretical knowledge, for Heidegger, remained a form of ‘thematizing’ that objectifies objects from a position of detachment. Heidegger did not develop the idea that this ‘detachment’ simultaneously makes possible relationships in which one can dwell and which can be extended and unfolded through relational mechanisms and resources. I take the position that Heidegger’s detachment should rather be recast in terms of the notion of differentiation (between subject and object); that differentiation entails the possibility of a nexus between differentiated entities which provides for our integration in the world (for a form of being-in-the-world); and that this form of being-in-relation also defines a form of ‘practice’ – in particular, it defines epistemic practice. What is an Object? Objects Characterized by a Lack in Completeness of Being (3) The ‘alien’ tissue element of epistemic practice that now needs further discussion is that of an object.1 How can we characterize knowledge objects and why do they require special attention? One reason for discussing epistemic objects is that our everyday notion of an object (take Heidegger’s ‘hammer’) would seem to contradict the features of objects that scientists and other experts encounter. To spell out these features I want to start from a suggestion by the historian of biology Rheinberger, who means by ‘epistemic things’ any scientific objects of investigation that are at the center of a research process and in the process of being materially defined. Objects 1 If there is one aspect of knowledge cultures on which received viewpoints on science and expertise and the newer studies of science and technology agree then this is that knowledge cultures centrally turn around object worlds to which experts and scientists are oriented (for the new sociology of science, this has been emphasized particularly by Callon (e.g. 1986) and Latour (e.g. 1993). For interesting attempts to work with these ideas by historians and sociologists of science see for example Pickering (1995), Wise (1993) and Dodier (1995). For an important study of individuals’ attachment to computers see Turkle (1995). Thevenot’s (e.g. 1994) concepts provide perhaps the most general sociological perspective on the issue.
Objectual Practice
89
of knowledge are characteristically open, question-generating and complex. They are processes and projections rather than definitive things. Observation and inquiry reveals them by increasing rather than reducing their complexity. Rheinberger also emphasizes that what an object is at present to some degree depends on how its future develops (see below). Rheinberger is interested in the historical structure of research programs which oscillate around objects of knowledge that escape fixation (e.g. 1992). Building upon Rheinberger’s ideas, I want to characterize objects of knowledge (‘epistemic objects’) in terms of a lack in completeness of being that takes away much of the wholeness, solidity and the thing-like character they have in our everyday conception. The everyday viewpoint, it would seem, looks at objects from the outside as one would look at tools or goods that are ready-to-hand or to-betraded further. These objects have the character of closed boxes. In contrast, objects of knowledge appear to have the capacity to unfold indefinitely. They are more like open drawers filled with folders extending indefinitely into the depth of a dark closet. Since epistemic objects are always in the process of being materially defined, they continually acquire new properties and change the ones they have. But this also means that objects of knowledge can never be fully attained, that they are, if you wish, never quite themselves. What we encounter in the research process are representations or stand-ins for a more basic lack of object. From a theoretical point of view, the defining characteristic of an epistemic object is this changing, unfolding character – or its lack of ‘object-ivity’ and completeness of being, and its non-identity with itself. The lack in completeness of being is crucial: objects of knowledge in many fields have material instantiations, but they must simultaneously be conceived of as unfolding structures of absences: as things that continually ‘explode’ and ‘mutate’ into something else, and that are as much defined by what they are not (but will, at some point have become) than by what they are. The idea that ‘every component of an organism is as much of an organism as every other part’, uttered by a scientist to whom a particular plant had exploded in that way, can perhaps capture the idea of an unfolding ontology. The unfolding ontology of objects foregrounds the temporal structure and, to put it into the original Freudian terms, the ‘Nachträglichkeit’2 in definitive existence of knowledge things (their post-hocness), which is difficult to combine with our everyday notion of an object. I will argue in the next section that it is the unfolding ontology of these objects which accommodates so well the structure of wanting, and binds experts to knowledge things in creative and constructive practice. There are other characteristics. Epistemic objects frequently exist simultaneously in a variety of forms. They have multiple instantiations, which range from figurative, mathematical and other representations to material realizations. Take the case of a detector in a high energy physics experiment. ‘It’ continually circulates through a collaborating community of physicists in the form of partial simulations and calculations, technical design drawings, artistic renderings, photographs, test 2 Freud illustrated the principle for some mental disorders: some childhood experiences turn out to have been profoundly disorienting and disorder-promoting only after a person develops a mental disorder, which may happen decades after the experience occurred.
90
Knowledge as Social Order
materials, prototypes, transparencies, written and verbal reports, and more. These instantiations are always partial in the sense of not fully comprising ‘the detector’. ‘Partial objects’ stand in an internal relation to a whole. The instantiations I have listed should not be conceived of as a halo of renderings and preparatory materials anticipating and representing another object, ‘the real thing’. It is ‘the real thing’ itself that has the changing ontology which the partial objects unfold. But do physicists not mean, by a detector, the physical machine after it has been built and when it is complete and running? Is the object not always an intended, an imagined whole? My point here is simply that as an intended object, a detector is an endlessly unfolding project consistent with the above circumscription of an epistemic object as marked by a lack in completeness of being. We should also consider that the boundaries of a technical instrument such as a ‘running detector’ are still highly problematic: only parts of the instrument tend to be operational at any one time, the physical machine will not run without remote controls, without computers and other equipment connected to it, and the instrument exists for most practical purposes mainly in the form of detector-(component)-measurements, representations and simulations (it is literally put behind lead walls and inaccessable while it is running). Finally, even when such an instrument is officially declared ‘finished’ and ‘complete’, the respective experts are acutely aware of its faults, of how it ‘could’ have been improved, of what it ‘should’ have become and did not. The ‘finished’, working detector, then, is itself always incomplete, is itself simply another partial object. The notion of an imagined object captures the ontological difference between current instantiations and a possibly more complete, ideal, or in another sense extended object. The imagined object might itself be instantiated in design drawings that project a future or hidden state. In this sense a concrete, imagined object is also a partial object, albeit one that stands in relation to an available, occurrent object-state as an object that marks the difference to this state. As historical studies show, scientists sometimes map out ideal objects in publications even when current techniques are not able to produce them (e.g. Borck 1997: 6).3 But imagined objects can also split and divert current practice by projecting a new possible object, one that calls into question current concerns or simply departs from them in lateral ways. This is how current practice often gets constructively extended into new strands. To return now to the partial object, I do not conceive of it as a gliding replacement for any presumed ‘real’ object in the sense of a referent. Partial objects, like epistemic objects in general, do not derive their immediate practical significance from the real. The point I want to draw attention to is the signifying force of (partial) epistemic objects by virtue of the internal articulation of these objects. Consider a transparency containing a curve which indicates the increasing ‘downtime’ of, say, a computer over its lifetime. The curve does not just ‘represent’ the unspecific experience that the instrument needs repairs over time. It specifies the exact way in which repair incidences accumulate. It may show that there is a small but steady increase of such incidences in the first years, followed by a steep and bumpy downtime-increase 3 Borck’s example refers to Einthoven’s publication of an ideal graphic registration of the heart in 1895 as an emblem of the curves of later electrocardiographs.
Objectual Practice
91
during midlife, and a slow increase in a generally high incidence of repair shutoffs during older age. From the curve, one can try to decide at what points to replace the instrument. This will make apparent the need for further information, for example about the level of downtime that is acceptable to a project – the curve is telling, but not (ever) telling enough. What one can decide is what points of the curve to explore further to obtain the missing information. For example, one can calculate the cost of data losses through downtime before and after a steep decline in repair incidences. The signifying force of partial objects (of epistemic objects in general) resides in the pointers they provide to possible further explorations. In this sense these objects are meaning-producing and practice-generating; they provide for the concatenation and constructive extension of practice. One can also say the significance of these entities resides in the lack they display and in the suggestions they contain for further unfolding (for a more complicated theoretical physics example, see Merz and Knorr Cetina 1997: 918). Thus in creative and constructive practice (partial) epistemic objects have to be seen as transient, internally complex, signifying entities that allow for and structure the continuation of the sequence through the signs they give off of their lacks and needs. Their internal articulation is important for the continuation of epistemic practice; not just their differe(a)nce to other objects, as in a Saussurean linguistic universe. I do not see partial epistemic objects as elementary units into which a complex whole is decomposed, but rather as complex links which extend a practical sequence at least partly through being unfoldable into equally complex sublinks. An example from everyday life might be a computer equipped with the relevant software. The computer can be ‘unfolded’ into signifying screens and subscreens, which stimulates in users an epistemic and affective relationship with the instrument (see also Turkle 1995). I have been emphasizing the unfolding, dispersed and signifying (meaningproducing) character of epistemic objects, and particularly their non-identity with themselves, to bring out the divergence of this idea of an object from everyday notions of material things. I must now add a word about the role naming plays in relation to these objects. The point I want to remind us of is a simple one: a stable name is not an expression and indicator of stable thinghood. Rather, naming, in the present conception, is a way to punctuate the flux, to bracket and ignore differences, to declare them as pointing to an identity-for-a-particular-purpose. I tend to think that one can see a stable name for a sequence of unfolding objects as a way of translating between different time zones, among others personal and institutional time zones. For example, when a sequence of objects and partial object-states is called a ‘Liquid Argon Calorimeter’, it is brought into accordance with projectfinancing requirements, work organization principles, institutional career tracks, and so on. A typical example of constantly changing or unfolding objects (also familiar from everyday life) are computer programs. In expert programming, authors write, run and update the code to suit their own, changing interests. At the same time they serve a community of users for whom they may issue the code in ‘versions’, ‘updates’, program ‘family’ members and so on (see Merz 1997). The packaging of progressive modifications in recognizable ‘versions’ and ‘updates’ requires a special effort, which the author makes taking into account user needs. The notion
92
Knowledge as Social Order
of unfolding refers to the evolution of a sequence of which certain segments (and possibly other segments) are gathered together by applying identical names to them. The process of naming and that of unfolding (and dispersion) are independent of each other, and might even stand in contradiction to one another. Epistemic Practice as Sustained by Object Relations (4) A number of suggestions about how epistemic practice might be conceived have been implicit in the discussion so far. I now want to address these more directly and systematically. I limit myself to two features of epistemic practice: its underlying relational dynamic and the lateral branching out of this practice. The first feature pertains to a kind of practice that is dynamic, constructive (creative) and perhaps conflictual. As indicated before, contemporary accounts favor a conception of practice in terms of habits and routines. As a consequence, these authors seek to explain practice (understood as practices) as an appeal to the embodied acquisition of preferences, perceptual schemes and dispositions to react, and by an appeal to shared tacit rules. The former is more the Bourdieu and Dreyfus line of thinking; in the latter case, the nature of the rules, and their exact relation to practical activity, lies at the core of controversies (see Bourdieu 1977; Dreyfus and Dreyfus 1986; Lynch 1993; Turner 1994; Rouse 1996; Schatzki 1996). In both cases, practice requires participants to have learned something which they subsequently deploy or enact in concrete situations. In contrast, I see epistemic practice as based upon a form of relationship (see also Knorr Cetina 1997; Greenberg and Mitchell 1983) that by the nature of its dynamic transforms itself and the entities formed by the relationship. What sort of relationship? Consider once more epistemic objects as described before. I want to maintain that the open, unfolding character of knowledge objects uniquely matches the ‘structure of wanting’ with which some authors have characterized the self. I derive this idea from Lacan (e.g. 1975), but it can also be linked to Baldwin (1897: 373ff.) and Hegel.4 Lacan derives wants not as Freud did from an instinctual impulse whose ultimate goal is a reduction in bodily tension, but rather from the mirror stage of a young child’s development. Wanting or desire is born in envy of the perfection of the image in the mirror (or of the mirroring response of the parents); the lack is permanent, since there will always be a distance between the subjective experience of a lack in our existence and the image in the mirror, or the apparent wholeness of others (e.g. Lacan and Wilden 1968; Alford 1991: 36ff.). One can also attempt a rendering of the lack in a representational idiom that is closer to the present concern. Accordingly, wants are always directed at an empirical object mediated by representations – through signifiers, which identify the object and render it significant. But these representations never quite catch up with the empirical object, they always in some aspects fail (misrepresent) the thing they articulate. They thereby reiterate the lack rather than eliminate it.5 To relate this now to epistemic objects, the point I made before is that the representations experts 4 Baldwin’s and Hegel’s notions of desire are summarized by Wiley (1994: 33). 5 In putting it this way I draw on Baas’s rendering of Lacan’s notion of a thing – albeit without claiming that my reading here is correct (1996: 22f.).
Objectual Practice
93
come up with in their search processes are not only partial and inadequate, they also tend to imply what is still missing in the picture. In other words they suggest which way to look further, through the insufficiencies they display. In that sense one could say that objects of knowledge structure desire, and provide for the continuation and unfolding of object-oriented practice. Let me say a little more about what it is that the notion of a structure of wanting offers; one has some explaining to do when turning to a sociologically arcane language such as the one I choose. The Lacanian ideas I use serve to specify objectual relations, which I see as the touchstone of a practice centered on epistemic objects, as relationships based upon a form of mutuality: of objects providing for the continuation of a chain of wantings, through the signs they give off of what they still lack; and of subjects (experts) providing for the possibility of the continuation of objects which only exist as a sequence of absences, or as an unfolding structure. What need not concern us further is Lacan’s account of the lack in subjectivity as rooted in the child’s narcissistic relationship to him/herself rather than to a lost person, or his explanatory trope of the mirror stage. One need not find the Lacanian account of the mirror stage persuasive in order to find the idea of a structure of wanting plausible. The latter is a convenient way to capture the way wants have of continually searching out new objects and of moving on to them – a convenient way, if you wish, to capture the volatility and unstoppability of desire. With regard to knowledge the idea of a structure or chain of wantings brings into view whole series of moves and their underlying dynamic rather than isolated reasons, as the traditional vocabulary of motives, intentions and actions does. It also suggests a libidinal dimension or basis of knowledge activities – which is ignored or denied when we conceive of science and expertise as cognitive endeavors. I believe that the existence of such a dimension is borne out by the intensity and pleasurability of objectual relations as experienced by experts. It is also ‘in tune with’ ontological reorientations towards ‘experience’ etc. in the wider society as diagnosed by some (Welsch 1996). The notion of a knowledge society is not at odds with, for example, that of an experiential society, or with a turn toward a more visual and visually simulated world – what it is at odds with is an arid and overly cognitively tilted notion of knowledge. The conduct of expertise has long harbored and nourished an experiential mentalité, if ‘experience’ is defined, as I think it should be, as an arousal of the processing capacities and sensitivities of the person. The conjunction of the relational and libidinal dimension gives practice a flavor and quality distinctively different from that of routines and habits. It remains for me to add a note about the lateral and angular branching off of strands of practice. The notion of unfolding when applied to practice can easily be understood as a forward-pointing sequence of steps driven by the interlocking dynamic of a structure of lacks and wantings. However, this would ignore the frequent splitting of activities into different strands, and the possible displacements of one strand by another. Such lateral shifts imply the transference of wants and relational substance from one chain of objectual involvements to another. As the study of science shows, processes of inquiry rarely come to a natural ending of the sort where everything worth knowing about an object is considered to be known. The idea of a structure of wanting implies a continually renewed interest in knowing
94
Knowledge as Social Order
that appears never to be fulfilled by final knowledge. But it also implies that interest may turn elsewhere, that it jumps the rails of one line of practice and continues on a different track in a somewhat different direction. The angularity of epistemic practice, its continual lateral divergence from itself, needs further discussion which I cannot offer here. Suffice to say that angular splitoffs add a disruptive element to the conception of practice I advocate, an element of conflictual breaks not generally recognized in current conceptions of practice. Summary and Conclusion (5) The notion of a knowledge society suggests that knowledge-centered practice focused on epistemic objects becomes a prominent part of all areas of social life. I have characterized the objects involved (which may be natural things, instruments, scientifically generated objects, etc.) in terms of their unfolding ontology, the phenomenon that they may exist simultaneously in a variety of forms, and their meaning-generating connective force. These ideas also suggest a notion of practice that is more dynamic, creative and constructive than the current definition of practice as rule-based routines or embodied skills suggests. The challenge we face, with the present argument, is to dissociate the notion of practice somewhat from its fixation on human dispositions and habits, and from the connotation of iterative procedural routines. I propose to conceive of the backbone of practice in terms of a relational dynamics that extends itself into the future in creative and also in disruptive ways. This relational dynamics does not simply mean the existence of positive emotional ties between individuals and non-human objects. We can theorize the sort of object-relations addressed in this chapter better through the notion of lack, and of an interlocking structure or chain of wantings, than through positive ties and fulfillments. The notion of a structure of wanting entails the possibility of a deep emotional investment in objects; an involvement that is at the same time congruent with the many flavors and orientations of this investment. Epistemic environments cannot be understood, I want to maintain in concluding, without understanding expert-object relationships. Knowledge-centered work shifts back and forth between the performance of ‘packaged’ routine procedures and differentiated practice as described in this chapter. It is with respect to differentiated practice that a relational idiom becomes plausible and may help in conceptualizing chains of activity. It may also become relevant to object-oriented practice outside knowledge contexts. In a knowledge society, objects in many areas of social life begin to display the kind of internal complexity and dynamic extendability that they have in science and expertise. Computers, financial instruments (Zelizer 1994), sophisticated sports equipment are typical examples – these appear on the market in continually changing versions, they are both ready-to-hand and subject to further development and investigation. As objects in everyday life become high technology devices some of the relational aspects of their existence in expert contexts also carry over into daily life. Some of the problems these devices raise in everyday contexts may well have to do with the relational demands they make and for which some lay users may not be prepared. Conversely, the appeal these objects have for some
Objectual Practice
95
users may also consist in the relational opportunities they offer (for computers, see Turkle 1984; 1995). When epistemic objects become epistemic everyday things, the relational approach I have advocated may also become relevant to understanding daily work activities and instrumental action. References Alford, C.F. (1991), The Self in Social Theory (New Haven: Yale University Press). Baas, B. (1996), ‘Die phänomenologische Ausarbeitung des Objekts a: Lacan mit Kant und Merleau-Ponty’, Riss 1, 19–60. Baldwin, J.M. ([1897] 1973), Social and Ethical Interpretations of Mental Development (New York: Arno Press). Barnes, B. (1977), Interests and the Growth of Knowledge (London: Routledge). ––– (2000a), Understanding Agency: Social Theory and Responsible Action (London and Thousand Oaks, CA: Sage). ––– (2000b), ‘The Macro-micro Problem and the Problem of Structure and Agency’, in G. Ritzer and B. Smart (eds). ––– (2001), ‘Practice as Collective Action’, in Schatzski et al (eds). Beck, U. (1992), Risk Society. Towards a new Modernity (London: Sage). Bell, D. (1973), The Coming of Post-Industrial Society. A Venture in Social Forecasting (New York: Basic Books). Beniger, J.R. (1986), The Control Revolution (Cambridge: Harvard University Press). Berger, P., et al. (1974), The Homeless Mind. Modernization and Consciousness (New York: Vintage Books). Borck, C. (1997), ‘Herzstrom’, in V. Hess (ed.). Bourdieu, P. (1977), Outline of a Theory of Practice (Cambridge: Cambridge University Press). Callon, M. (1986), ‘Some Elements of a Sociology of Translation: Domestication of the Scallops and the Fishermen of St. Breuc Bay’, in J. Law (ed.). Dodier, N. (1995), Les Hommes et les Machines (Paris: Editions Métailié). Dreyfus, H.L. (1994), Being-in-the-World. A Commentary on Heidegger’s ‘Being and Time’ Division I (Cambridge, Mass.: The MIT Press). Dreyfus, H.L. and Dreyfus, S.E. (1986), Mind Over Machine: The Power of Human Intuition (Oxford: Blackwell). Drucker, P.G. (1993), Post-Capitalist Society (Oxford: Butterworth Heinemann). Giddens, A. (1984), The Constitution of Society: Outline of the Theory of Structuration (Cambridge: Polity Press). ––– (1990), The Consequences of Modernity (Stanford: Stanford University Press). Greenberg, J.R. and Mitchell, S.A. (1983), Object Relations in Psychoanalytic Theory (Cambridge, Mass.: Harvard University Press). Habermas, J. (1981), Theorie des kommunikativen Handelns (Frankfurt/M.: Suhrkamp).
96
Knowledge as Social Order
Heidegger, M. (1962), Being and Time (New York: Harper). Hess, V. (ed.) (1997), Die Normierung von Gesundheit. Messende Verfahren der Medizin als kulturelle Praxis (Husum: Matthiesen Verlag). Horwich, P. (ed.) (1993), World Changes. Thomas Kuhn and the Nature of Science (Cambridge, MA: The MIT Press). Knorr Cetina, K. (1997), ‘Sociality with Objects. Social Relations in Postsocial Knowledge Societies’, Theory, Culture & Society 14(4), 1–30. ––– (1999), Epistemic Cultures. How The Sciences Make Knowledge (Cambridge, MA: Harvard University Press). Krohn, W. and Weyer, J. (1994), ‘Society as a Laboratory: The Social Risks of Experimental Research’, Science and Public Policy 21(3), 173–183. Lacan, J. (1975), The Language of the Self (New York: Delta). Lacan, J. and Wilden, A. (1968), Speech and Language in Psychoanalysis (Baltimore: Johns Hopkins University Press). Lash, S. and Urry, J. (1994), Economies of Signs and Space (London: Sage). Latour, B. (1993), We Have Never Been Modern (Cambridge: Harvard University Press). Law, J. (ed.) (1986), Power, Action and Belief: a New Sociology of Knowledge? (London: Routledge and Kegan Paul). Lynch, M. (1993), Scientific Practice and Ordinary Action (New York: Cambridge University Press). Lyotard, J.-F. (1984), The Postmodern Condition (Minneapolis: University of Minnesota Press). Merz, M. (1997), ‘Multiplex and Unfolding: Computer Simulation in Particle Physics’, paper presented at the Annual Meeting of the Society for Social Studies of Science, Tucson, Arizona, October. Merz, M. and Knorr Cetina, K. (1997), ‘Deconstruction in a “Thinking” Science’, Social Studies of Science 27(1), 73–111. Pickering, A. (1995), The Mangle of Practice. Time, Agency and Science (Chicago: University of Chicago Press). Rheinberger, H.-J. (1992), ‘Experiment, Difference, and Writing: I. Tracing protein synthesis’, Studies in the History and Philosophy of Science 23(2), 305–441. Ritzer, G. and Smart, B. (eds) (2000), Handbook of Social Theory (Thousand Oaks, CA: Sage). Rouse, J. (1996), Engaging Science: How to Understand its Practices Philosophically (Ithaca: Cornell University Press). Schatzki, T. (1996), Social Practices: A Wittgensteinian Approach to Human Activity and the Soul (New York: Cambridge University Press). Schatzki E. et al. (eds) (2001), The Practice Turn in Contemporary Theory (London: Routledge). Stehr, N. (1994), Arbeit, Eigentum und Wissen. Zur Theorie von Wissensgesellschaften (Frankfurt/M.: Suhrkamp). Thevenot, L. (1994), ‘Le régime de familarité. Des choses en personne’, GenPsis 17, 72–101. Turkle, S. (1984), The Second Self. Computers and the Human Spirit (New York: Simon and Schuster).
Objectual Practice
97
––– (1995), Life on the Screen (New York: Simon and Schuster). Turner, S. (1994), The Social Theory of Practice: Tradition, Tacit Knowledge and Presuppositions (Chicago: University of Chicago Press). Welsch, W. (1996), ‘Aestheticization Processes: Phenomena, Distinction and Prospects’, Theory, Culture and Society 13(1), 1–25. Wiley, N. (1994), The Semiotic Self (Chicago: Polity). Wise, N. (1993), ‘Mediations: Enlightenment Balancing Acts, or the Technologies of Rationalization’, in P. Horwich (ed.). Wittgenstein, L. (1969), On Certainty (New York: Harper & Row). Zelizer, V. (1994), The Social Meaning of Money (New York: Basic Books).
This page intentionally left blank
Chapter 6
Producing Accounts: Finitism, Technology and Rule-Following Donald MacKenzie
[Finitism’s] core assertion is that proper usage is developed step by step, in processes involving successions of on-the-spot judgements. Every instance of use, or of proper use, of a concept must in the last analysis be accounted for separately, by reference to specific, local, contingent determinants. Finitism denies that inherent properties or meanings attach to concepts and determine their future correct applications (Barnes 1982: 30). [N]othing in the rule itself fixes its application in a given case (Barnes 1995: 202).
An important theme in the work of Barry Barnes has been ‘finitism’. The theme emerged from Barnes’s work in the sociology of science, and in particular from his engagement with the philosophy of science of Mary Hesse (1974), and first became clear when Barnes made explicit the finitist aspects of the work of T.S. Kuhn (Barnes 1982). Subsequently, Barnes built finitism into the core of his approach to social theory and to the analysis of social order, for example elaborating Wittgenstein’s famous account of rule-following. In this chapter, I will argue that Barnes’s finitism throws light on a domain quite different from that in which it was developed: accounting, and specifically financial reporting. Many of the relevant phenomena are familiar to practising accountants and to researchers on accounting, but this chapter suggests that finitism has the virtue of systematically making them salient. It argues that the distinction between finitism and the theory of meaning to which it is opposed – extensional semantics – helpfully clarifies fundamental issues to do with accounting. After this short introduction, a second section of this chapter briefly recaps the main tenets of finitism, and gives a couple of simple illustrations of their application to rule following. The third section argues, very briefly, that finitism offers an appropriate viewpoint from which to analyze financial reporting. The more extensive fourth and fifth sections illustrate the analysis by drawing on an empirical case study of accounting and financial reporting in a UK mid-market company. The sixth section is the chapter’s conclusion. Finitism and Rule-Following From the viewpoint of this paper, finitism is best seen as a theory of the application of terms to instances or particulars. Consider a term ‘A’, which could be an everyday
100
Knowledge as Social Order
word such as ‘walk’ or ‘red’; a mathematical term such as ‘converge’ or ‘polyhedron’ (Lakatos 1976); or an accounting term such as ‘depreciate’, ‘asset’ or ‘finance lease’. An apparently common-sense view of the application of terms to instances is to conceive of terms as having fixed meanings. Once we have decided, individually or collectively, what a term means, then the infinite universe of entities, processes, activities, states of affairs and other particulars is divided up into instances of A and of not-A: red entities and entities that are not red; walking and activities that are not walking, such as running; polyhedra and entities that are not polyhedra; purchases of capital assets and activities that are not purchases of capital assets; ‘finance leases’ and states of affairs that are not ‘finance leases’ such as operating leases,1 and so on. Although many people hold this view of meaning quite informally, it is known technically as ‘extensional semantics’. A term’s ‘extension’ – the ‘set of things of which it is true’ (Barnes 1982: 31) – is fixed in advance of usage of the term. It may sometimes be difficult to determine whether a newly-encountered particular is an instance of A or not, but if extensional semantics is correct the difficulty is merely empirical. In contrast, finitism denies that the universe of all the instances or particulars that may ever be encountered should be thought of as divided into instances of A and of not-A. All we ever have – as individuals or as an entire culture – is a finite set of past applications of ‘A’ to particulars. When a new particular is encountered, the difficulty is more than the empirical one of determining its properties: we need to decide whether it is sufficiently like the previous particulars we have classed as A to warrant that classification. No two empirical entities or activities are ever entirely identical; there are always differences between them that could be pointed to as well as similarities; ‘every situation is in detail different from every other’ (Hesse 1974: 12). Extensional semantics and finitism are neatly summarized by Barnes in the diagram I have reproduced as Figure 6.1. Note that finitism goes beyond the assertion that meanings are social conventions; that assertion is entirely compatible with the extensional-semantics view that once the meaning of ‘A’ is chosen the instances to which it correctly applies are then fixed. Rather, in a finitist perspective every application of a term to an instance is implicitly a decision. Not only is the extent of similarity to previous particulars classed as A always in principle contestable, but those previous instances are always revisable: we may decide that one or more previous applications of ‘A’ were mistaken.
1 A lease can be in effect a way of borrowing money to buy an asset, and regulators have been concerned that such leases – ‘finance leases’ – appear on a corporation’s balance sheet and so enter into calculations of the extent of a corporation’s borrowing and the level of return on its assets. See IAS [International Accounting Standard] 17, ‘Leases’ (IASB 2005: 887–914).
Producing Accounts
101
Note: Redrawn from figure in Barry Barnes, T.S. Kuhn and Social Science, 1982, The Macmillan Press Limited, reproduced with permission of Palgrave Macmillan.
Figure 6.1
Extensional semantics and finitism
Of course, applications of a term frequently do not feel like decisions. Often, our sense of analogy with previous instances of ‘A’ is overwhelming, and we feel that this object just is red; that figure is plainly not a polyhedron; this expenditure is obviously an expense, not purchase of an asset; that transaction is clearly a finance lease not an operating lease. Finitism in no sense denies the prevalence of such instincts, without which the smooth functioning of accounting or any other practice would be impossible. It does, however, assert that they can be analyzed without appeal to the ‘intrinsic meaning’ of the term in question. Training and habit, for example, are powerful sources of classificatory instincts. An area in which finitism is of particular importance – both in social theory (Barnes 1988 and 1995) and in the case of accounting – is following a rule. According to finitism, the application of a rule to a specific situation is always in principle problematic, just as the application of a term to a particular is problematic. Consider, for example, the sixth commandment, ‘Thou shalt not kill’. It seems a straightforward enough prohibition, but consider its application to enemy soldiers, enemy civilians (as ‘collateral damage’), terminally ill people in great pain who have expressed a wish to die, human foetuses, animals (in experiments), animals
102
Knowledge as Social Order
(for food), and so on. What the commandment implies for each of these has been the object of fiercely varying instincts. Murder may seem too easy a case with which to illustrate rule-finitism, but consider a harder case: chess. It is around a millennium old, so humanity has had plenty time to refine the rules of chess, and it lacks not just murder’s moral and legal load but also its real-world complexities. Chess is a ‘micro world’ (see, e.g., Collins 1990): a limited, artificial domain, deliberately stripped of ambiguity, one that has, for example, been relatively easy to automate. Here, surely, we should find rules to which extensional semantics applies. Consider, though, the puzzle reproduced in Figure 6.2. White, to play, must deliver checkmate in a single move. Even the best of chess players could stare at this puzzle for many hours and not see how to solve it. The solution is that white moves its most advanced pawn to the eighth rank and replaces it with the rook that currently blocks the diagonal from the white bishop to the black king. The latter is thus exposed to check, and has no flight square: the rook controls the eight rank. It is checkmate.
Note: Puzzle composed by Richard Haddrell.
Figure 6.2
White to play and mate in a single move
Producing Accounts
103
Any chess player will very likely respond that the ‘solution’ is not a legal move: it is contrary to the rules of chess. Consider, however, the pawn-promotion rule as it stood in the Laws of Chess in May 2005: ‘When a pawn reaches the rank furthest from its starting position it must be exchanged as part of the same move for a queen, rook, bishop or knight of the same colour’ (FIDE 2005a, rule 3.7.e). The solution to the puzzle is, surely, a reasonable interpretation of what it is to ‘exchange’ the pawn for a rook. Indeed, in the summer of 2005 the rule was changed with the apparent aim of blocking this interpretation. The word ‘new’ was added, so it now reads ‘exchanged ... for a new queen, rook ...’ (FIDE 2005b, rule 3.7.e, emphasis added). However, the solution to the puzzle might still under some circumstances be argued to be allowable. Imagine the game is being played with an old wooden set, but that some pieces have been lost and the white rook currently on the board is a modern plastic replacement. Is that not ‘a new ... rook’? One can, of course, imagine a rule of interpretation being added to the laws of chess to specify what ‘new’ means in the context of the pawn-promotion rule. If finitism is correct, though, we would simply be entering a regress, for the rule of interpretation would itself contain terms that stand in need of interpretation. At first sight, the finitist analysis of rule-following may appear to be a recipe for anarchy. Not so. Chess does not degenerate into chaos because moves akin to the puzzle’s ‘solution’ can, given sufficient imagination, be made out to be compatible with its rules. The puzzle is artificial: the mental habits of the experienced chessplayer are such that a move analogous to the ‘solution’ would never be considered. If a player in a chess tournament made a move like the ‘solution’, he or she would be regarded by other players as ignorant, cheating or possibly even mad. Should a position similar to Figure 6.2 arise in a game against a computerized chess program, it would be impossible to input the ‘solution’, at least unless one had access to the program code and had the skills to modify it. As we shall see, factors such as these – experience and habit; the constraints that arise from other people; and the constraints that arise from things – are also important in making accounting the orderly process that it largely is. Finitism and Accounting The relevance of finitism to the analysis of accounting has been argued at length elsewhere (Hatherly, Leung and MacKenzie forthcoming), so let me summarize briefly. A fundamental process in bookkeeping and accounting is applying a term to a particular such as a transaction, an event or an item (that is, determining what kind of particular it is, which often takes the concrete form of assigning a code from an organization’s ‘chart of accounts’: see Figure 6.3), and ‘measuring’ it by assigning to it a numerical, normally a monetary, value. Such ‘primary’ acts of classification and measurement are the foundation of accounting: there is a sense in which everything else is built on top of them.
104
Knowledge as Social Order
11305 Plant & Machinery - Depreciation 21101 Stock - Raw Materials 21407 Stock - Finished Goods - Slow Moving Stock 23101 Trade Debtors 50401 Various - Sales 51930 Other - Stock Variance - Revaluation
Figure 6.3
Examples of codes in the case-study firm’s chart of accounts
Measurement is ‘the process of determining the monetary amounts at which the elements of the financial statements are to be recognised and carried in the balance sheet and income statement’. Plainly, it ‘involves the selection of the particular basis of measurement’ (IASB 2005, 50). Classification is at least as consequential. For example, if an item is bought, the transaction needs classed either as an expense (in which case the expenditure is a direct deduction from a corporation’s income for the corresponding period) or as purchase of a capital asset, in which case it becomes an entry into the corporation’s balance sheet, and is set against income only gradually, in the form of depreciation.2 The application of accounting terms to particulars cannot be made arbitrarily. In all developed countries, the acts of classification and measurement that generate a corporation’s balance sheet, income statement and cash-flow statement are tightly rulebound and regulated. For example, listed companies in the U.K. and other countries in the European Union have to follow the 2,000 pages of International Financial Reporting Standards (IFRSsTM) and International Accounting Standards (IASs) set by the International Accounting Standards Board (IASB 2005), while their counterparts in the U.S. have to follow similar but even lengthier standards laid down by the Financial Accounting Standards Board.3 The prominent role of standards in accounting and financial reporting makes the finitist analysis of rule-following relevant. Accounting standards are typically a mixture of ‘rules’ and higher-level ‘principles’ (along with supplementary material such as definitions, explanations, worked examples, interpretations, and so on). The European standards, for example, use bold type to identify principles. The distinction between a ‘rule’ and a ‘principle’ is an interesting and important one. However, for reasons of space the discussion that follows will not distinguish systematically between the two: the basic rule-finitist argument – that a rule in itself does not determine how it is to be applied to any given situation – is equally applicable to a principle. 2 ‘An asset is not recognised in the balance sheet when expenditure has been incurred for which it is considered improbable that economic benefits will flow to the entity beyond the current accounting period. Instead such a transaction results in the recognition of an expense in the income statement’ (IASB 2005: 49). 3 See www.fasb.org.
Producing Accounts
105
An Accounting System To investigate empirically how finitism might apply to accounting, I conducted a small empirical case study of accounting in a UK mid-market listed company. I interviewed half of the members of its small finance department, and also its warehouse and logistics manager and its vice-president for information services. Each interview was around 90 minutes long, and all interviews were tape-recorded and transcribed in full. The chief financial officer (CFO), who plays the most central role in the firm’s accounting, was interviewed twice. Because they are clearly an important audience for financial reports, I am also beginning a small set of interviews with auditors not connected to the case-study firm, and this is drawn on in a very limited way (I hope these interviews will indirectly permit me to address the issue of typicality discussed below). The case study was thus interview-based rather than observational. However, each interview at the case-study firm was conducted in the office or at the desk of the member of staff being interviewed, and I asked the interviewees responsible for data entry into the firm’s computer system to show me on their screens what they did and how, and also to show me examples of the many paper documents that also play an important role. It is, of course, a limited case study, and may not be typical. Most obviously, the very act of permitting me detailed access demonstrates the firm’s confidence in the probity of its accounting. Its auditors shared this confidence: they had made no adjustments at all to its draft accounts in the previous three years. It is also noteworthy that the degree of automation of the firm’s business activities was high. All its main commercial activities are coordinated, planned and tracked using a unified relational4 database – an Enterprise Resource Planning (ERP) system – and much of its accounting is done using the ERP system’s accounting subsystem. The application of the ERP system was clearly successful: although interviewees spoke frankly, and sometimes noted features of the system that were awkward or that they wished it had, I encountered none of the bitter complaints of unusability recorded, for example, in Button and Harper (1993). The resultant efficiency of its accounting helps explain how a firm with a mid-market valuation can operate with a finance department of limited size. Let me begin with the work of the accounts administrator I interviewed. She and her colleagues do the firm’s bookkeeping. As already noted, this is done using the accounting module of the ERP system, although the traces of accounting’s preelectronic past are still there in the terminology used for the relevant parts of the relational database: ‘purchase ledger’, ‘general ledger’, and so on. As I sat with the
4 ‘Relational’ databases, introduced from the 1970s onwards, eliminated many of the technical restrictions in earlier hierarchical database design. In a relational database, ‘[a]ny table can be accessed directly without having to access all parent objects ... any tables can be linked together, regardless of their hierarchical position’ (Powell 2006: 9).
106
Knowledge as Social Order
administrator, she processed an electricity bill the firm had just received, applying to it the chart-of-accounts category ‘Heat & Light – Electricity’. She did not need to ponder whether this was the correct classification, and she knows the appropriate code (86101) by heart. Measurement was equally straightforward: she simply entered the total stated on the bill into the appropriate box on the ERP screen. However, even this most ‘obvious’ of primary acts of classification and measurement occasionally gives rise to a problem. Because value added tax (VAT) can be reclaimed, a record is maintained in the database of the amounts of VAT that have been paid. The ERP system automatically applies the prevailing rate of VAT to calculate the component of the electricity bill that is VAT. Sometimes, though, that number is not exactly the same as the VAT amount stated on the bill, presumably because of differences in rounding algorithms. If that happens, the accounts administrator manually adjusts the number in the ERP system so that it is the same as on the bill. Other acts of assigning terms to particulars are less obvious than applying the code for ‘Heat and Light – Electricity’ to an electricity bill. The administrator showed me an expenses claim she had just received from one of the firm’s employees. Colleagues from overseas branches of the firm had been visiting, and he had classed the payment for the dinner he had had with them as ‘business entertainment’, corresponding to the chart-of-accounts code 62101 (see Figure 6.3). The expenses claim was entirely legitimate, but the administrator told him she was going to change the classification. For his original classification to be valid, external customers would have had to be present, but she knew that those present had all been employees. A particular classificatory decision that has to be made by the accounts administrator, her colleagues and the financial controller (who sits at a nearby desk, is responsible for all day-to-day transactions, and is consulted on instances where the appropriate classification is not apparent) is whether a transaction is an expense or purchase of an asset (as noted, the subsequent accounting treatment of the two will be quite different). In the case of large purchases, which need higher-level approval, the decision will already have been made: an appropriate meeting will formally have approved a proposed transaction as purchase of a fixed asset. At other times, the decision is taken more informally. Often, this happens when a staff member who is ordering an item walks over to the finance department with a requisition form (which is the immediate context in which much of the coding of purchases in done). Amongst the considerations that the accounts administrator and the controller told me they took into account are the nature of the item, how long it is going to last, how much it costs, the consequences of the classification (for example, the fact that an item classed as a fixed asset becomes an entry in the fixed-assets register, and in consequence its location thereafter needs to be tracked), the amounts remaining in the relevant budgets, and specific guidelines for the classification of certain categories of items such as information technology. The controller does a final check of the primary acts of classification and measurement that her colleagues perform. When the accounts administrator and the controller’s other colleagues classify a transaction, it is placed in a ‘hold table’ within the ERP system. While it would be far too time-consuming for the controller to inspect every entry in such a table in detail, she scans through the entries before permanently updating the relevant computerized ledgers. ‘[B]ecause of experience’,
Producing Accounts
107
she has a rough idea what to expect: for example, she is familiar with the names of regular suppliers and the typical amounts of purchases from them, and she can anticipate predictable fluctuations such as the dip in travel expenditure around Christmas. If the entries or totals in the hold table look odd to her – ‘if they are very much out’ – she will ask herself, for instance, whether ‘something [has] gone in there that shouldn’t really be in there?’ and check in detail before ‘posting’ the record to update the ledgers. Day after day, week after week, the accounts administrator, the controller and their colleagues perform thousands of acts of classification and measurement: applying terms – codes from the chart of accounts, cost centre codes, and so on – to particulars; entering into the ERP system the appropriate monetary amounts corresponding to a given particular; and checking the results of such acts. Some indication of the volume of such acts was conveyed to me by a set of 14 large leverarch files on a shelf close to the desks of the accounts administrator and controller. The files contained the paper documents – purchase orders, bills, invoices, expenses claims, and the like – corresponding to the previous month’s acts of classification and measurement (although the firm’s processes are highly automated, it cannot entirely abandon the materiality of paper). As noted, primary acts of classification and measurement of the kind performed by the accounts administrator and her colleagues are the foundations of accounting, yet they have almost never been the subject of examination in the literature of accounting or adjacent social sciences (the limited exceptions include Suchman 1983 and, especially, Anderson, Hughes and Sharrock 1989: 123–139). ‘Routine’ they may be, but aspects of these acts are nonetheless skilled. In particular, while the firm’s ERP system can process the results of acts of accounting classification, it cannot itself perform such acts. While, as noted in the conclusion, one can imagine their automation – perhaps via a neural network – accounting classification is still a human preserve in all the contexts of which I am aware. Some of the knowledge needed to perform primary accounting classification is contextual: knowing whom the attendees at the dinner would have been, or understanding the nature of the goods or services bought from a specific supplier. Other aspects of it come from training and experience in bookkeeping and accounting. Other employees of the firm also have the contextual knowledge, but because they lack training in accounting they do not normally do the most crucial classificatory work: that, for example, is why a member of staff drawing up a purchase order needs to carry it or send it to the finance department for coding. The ERP system is, however, much more than simply a giant pocket calculator aggregating the results of primary acts of classification and measurement. For example, the system enforces double-entry bookkeeping, the basic discipline that underpins accounting (and that is, according to the classic argument of Weber, Sombart and Schumpeter, the basis of the rational business enterprise).5 In double entry, every transaction is treated as a debit from one account and credit to another (so a sale for cash, for instance, is a credit entry in a trader’s sales account and – confusingly for the untrained – a debit entry in his or her cash account). Although each transaction 5
For a re-evaluation of the classic argument, see Carruthers and Espeland (1991).
108
Knowledge as Social Order
is entered into the case-study firm’s ERP system only once, the system automatically makes the balancing entry, maintaining the fundamental property of double-entry accounts: that the sum of debit balances is equal to the sum of credit balances. As well as enforcing double-entry bookkeeping, the firm’s ERP system also supports modern accrual accounting. In this, ‘the effects of transactions and other events are recognised when they occur (and not as cash or its equivalent is received or paid)’ (IASB 2005: 37). In consequence, a company’s profit in a given period, calculated on an accrual basis, is not the difference between the money it has received and paid out (which is what it would be in ‘cash accounting’), but the difference between the revenue earned in the period and the costs incurred in earning that revenue, even if (for example) some or all of the cash receipts will not arrive until a later period and cash expenditures have been made in an earlier period. The ERP system is set up so that specific entries into it (often made by non-financial staff, in particular storekeepers) trigger a transaction becoming a cost or a revenue. For example, when an item or set of items the firm has ordered arrives in one of its stores, its shipment labels are checked against the corresponding documentation,6 and, if they tally, a storekeeper will fill in a ‘purchase order receipt’ note via the ERP system. In the words of the CFO, that is ‘when it becomes a liability’ (the nature of that liability will already have been determined at the point at which the purchase order is coded, and is in this case an aspect that would later be recognized as a cost when the item in question is sold or used). Similarly, when an entry is made into the ERP system indicating that a customer’s carrier has uplifted a batch of the firm’s products, the system ‘will immediately recognize the invoice for the goods that have gone in the van out the door without actually us getting paid’. Again, the necessary coding will already have been done. Just as the apparently routine nature of much classification hides the skill involved, so the apparently routine nature of the process triggering the recognition of an expense or a revenue disguises the knowledge and experience needed to handle its day-to-day contingencies: the arrival in the warehouse of goods for which there is no purchase order, or for which a member of staff has prematurely ‘closed’ the purchase order; paperwork that should be numbered ‘13367B’ but because of poor print quality is read as ‘133678’; and so on. The case-study firm’s warehouse staff members are able to handle such contingencies: they know what to do when they occur, and the implementation of the case-study firm’s ERP system allows them to do it (the inadvertent elimination of the capacity to handle such contingencies – contingencies that are familiar to those directly involved in the business process in question, but whose existence is sometimes not grasped by system developers or implementers with an idealized view of the process – is a major source of problems in automation of business processes in other contexts).7 It is, however, correct to say that the ERP system ‘supports’ accrual accounting, not that it ‘enforces’ it. A significant amount of intervention, mainly by the financial controller, is necessary to ensure that accrued expenses (the cost of goods and services 6 Contents of shipments are also sometimes checked. 7 This is, for example, a main plank of the argument for a sociotechnical, even an ethnographic, approach to system development. See, for example, Suchman (1983).
Producing Accounts
109
that have been received or consumed, but for which no invoice or bill has been received) and prepayments (the cost of goods and services that have been paid for but not yet received) are recognised in the appropriate financial period. Electricity, for instance, is a classic example of an accrued expense: the bill is sent only after it has been consumed, and it may not arrive until after the end of the accounting time period in which the electricity has been consumed. The bonuses earned by sales staff are another example of an accrued expense: they are obligations that accumulate throughout the time period over which they are calculated, even if they are paid only after it ends. Many maintenance contracts are examples of prepayments: they are paid for at the start of the time period they cover. Another aspect of accounting that the firm’s ERP system both supports and enforces is segregation of duties. Access to the system is password-controlled, and there is a matrix of ‘read’ and ‘write’ permissions for each staff member. The more obvious aspect of segregation of duties is to create barriers to fraud by ensuring, for example, that the payment corresponding to a purchase order raised by one person can be authorized only by a different person. More subtly, however, access controls are set up in such as way that the ERP system ‘solidifies’ the myriad applications of terms to particulars by members of the finance department. The access-control matrix has the form of two triangles, one inverted. ‘Read’ permissions increase as one ascends the firm’s organizational chart: those at the bottom can read little; those at the top have wide read access. ‘Write’ permissions, in contrast, diminish: for example, neither the CFO nor the chief executive have the ‘write’ permissions needed to amend a primary act of classification, measurement or recognition that has become a record in the ERP system. In addition, if a record is amended, the ERP system creates an audit trail of what has been changed and by whom. Of course, these structuring effects of the ERP system work only if those who maintain the access-control matrix do so appropriately, and if passwords are not shared (the latter is hard to police, though auditors have simple but useful ways of checking for password-sharing). Furthermore, the access controls build into ERP systems such as that in use at the case-study firm are unlikely to be robust enough to survive the full gamut of sophisticated ‘attacks’ considered in the case of systems on which national security depends (see MacKenzie 2001, chapter 5). ‘I have no doubt there could be loopholes: there are loopholes in every system’, says the vicepresident for information services: ‘It’s about ensuring that our security procedures help us to discover and rectify potential loopholes before they are exploited by others’. The sophistication that would be needed to find and to exploit them – like the sophistication needed to get a chess program to accept a non-standard move – put the relative ‘business risk’ at an acceptable level. Following a Rule The firm’s ERP system receives and processes inputs of the kind discussed above, and (as just discussed) it supports and enforces an appropriate segregation of duties, thus ‘solidifying’ primary acts of classification, measurement and the recognition
110
Knowledge as Social Order
of revenue and expenses. It also has a ‘report generator’ facility that can produce a wide variety of reports both for purposes internal to the firm and for external financial reporting. For example, in the period immediately prior to the case study, International Financial Reporting Standards (IFRSs) had become applicable for the first time, and the firm’s business planning manager had devoted considerable effort to ‘building a balance sheet in the format for the new IFRS’: that is, constructing a mapping from the chart of accounts to the categories of an IFRS balance sheet (and sometimes subdividing existing chart-of-accounts codes), so that the reportgenerator facility could be used to produce such a balance sheet. The business planning manager had done his work well, but it would be naïve to think that adequate financial reporting could be achieved simply by using the ERP system to aggregate primary acts of classification, measurement and recognition, and then pushing the appropriate report-generator button. If one does that, the figures will be nonsensical, according to the CFO. Important adjustments to the aggregate results of primary acts of classification, measurement and recognition are necessary in order to produce adequate accounts, and they go beyond the adjustments, discussed above, that the financial controller makes for accrued expenses and prepayments. The necessary adjustments are typically not made by using the ERP system directly: they are performed separately, normally using Excel spreadsheets, and then uploaded into the system. They are thus not subject to the system’s discipline, but are subject to discipline of a different kind: the rules of financial reporting. As already noted, for a listed company in the European Union, the relevant rules are those of the ‘International Financial Reporting Standards’ (IFRSs) and ‘International Accounting Standards’ (IASs). Let me concentrate on two standards – one longstanding, one newer – that the case-study firm has to follow.8 The first is the longstanding (though recently revised) IAS 2, ‘Inventories’. It occupies twenty pages, but its central principle – set out in bold – is its clause 9: ‘Inventories shall be measured at the lower of cost and net realisable value’ (IASB 2005: 662). One issue in following IAS 2 is knowing what the firm’s inventory consists in. Inventory is recorded in the ERP system, but the contents of the relevant registers cannot simply be assumed to correspond to the physical entities in its warehouses. This reconciliation is the province of the warehouse manager and his staff. The items in the warehouses are normally in packages labelled with a barcode corresponding to a record in the ERP system, so the manager and his colleagues can perform the ‘stock check’ by walking around a warehouse with a barcode scanner. Inevitably, though, there are records on the ERP system for which no corresponding physical object can be found. The degree of agreement, however, is typically over 99 percent. Investigation can resolve many of the residual discrepancies, and those that remain are an example of a pervasive issue: ‘materiality’. No accounting system, or set of accounts, is likely ever to be judged perfect. ‘We always find errors’, one 8 These are the two standards the following of which was discussed most by interviewees. This may indicate that interviewees saw them as raising particular issues of judgement, but similar issues could also be identified when interviewees discussed the following of other standards, such as IAS 19 (‘Employee Benefits’).
Producing Accounts
111
auditor (not connected to the case-study firm) told me. An accountant or auditor must judge if a discrepancy is ‘material’: that is if (in the words of IAS 1, ‘Presentation of Financial Statements’) it could ‘influence the economic decisions of users taken on the basis of the financial statements’. There is no set quantitative threshold, so materiality must be ‘judged in the surrounding circumstances’ (IASB 2005: 610, emphasis in original deleted). However, a stock discrepancy of well below 1 percent is most unlikely to be judged ‘material’, and the warehouse manager and his staff are indeed able to demonstrate to the firm’s auditors that the ‘reconciliation is ... good’ and they ‘have control of [their] stock’. To follow IAS 2 obviously requires more than this, though: the ‘cost’ and ‘net realisable value’ of inventory both need to be measured. The firm produces large numbers of items, and it would not be sensible for it to try to calculate an actual cost for each individual item. Hence the ‘cost’ that it measures in following IAS 2 (and also uses for other purposes) is a ‘standard cost’: in the words of the CFO, ‘effectively your best estimate of what you think the costs are based on what you’re paying people just now and all the rest of it’ (‘all the rest of it’ includes not just items such as the cost of raw materials and of services, but also physical contingencies such the effectiveness of production processes and the results of product testing). As invoices and the like subsequently arrive and are processed by staff members and by the firm’s ERP system, variances from this estimate are calculated automatically and scrutinized: ‘we ... review ... our standard costs each quarter ... to make sure they’re still in line with reality’ (the level of consistency between estimates and outcomes is indicated by the firm’s goal that variances from its estimates of ‘gross margin’ should be less than a percentage point). So ‘cost’ is tied tightly to the aggregate of the corresponding primary acts of classification, measurement and recognition, but it is not reducible to that aggregate. Measuring the ‘net realisable value’ of inventory requires different judgements: will the products that the firm has in store be sold, and if so at what price? Following IAS 2’s rules for valuing inventory thus means, in the words of the CFO, ‘looking at things like stock obsolescence provisions. And slow-moving stock’. The resultant decisions are consequential. For example, the firm’s auditors are (in the words of the financial controller) ‘very fussy about say the stock provision. They always want to see all the backup for that’. Every month, the CFO and another member of his staff meet with the relevant business planners and managers to discuss which items of stock should be classed ‘obsolescent’ or ‘slow-moving’. For each product, the amount in inventory is compared with ‘the number of units that the sales people see that they can sell over the next twelve months’. If the latter is less than the former, says the CFO, ‘we will make a call on that to say, right, is this just a temporary blip ... or what is it? ... [W]e say to the sales team, “can you justify to us why we shouldn’t provide for everything that’s not going to be sold in the next twelve months?” ... [W]hy are these parts not saleable?’ Is the cause likely to be temporary, or will it be permanent? (To ‘provide’ for obsolescence or slow movement is to reduce the measured value of inventory by an appropriate amount; the amount is the ‘provision’). Such decisions are often not entirely clear-cut. In making provisions, the CFO thus has to ask himself: ‘do I take a light touch or a heavy touch?’ In the first two
112
Knowledge as Social Order
inventory-provisioning meetings in a quarter, he tends to take a ‘light touch’ in regard to stock that seems not to be selling: ‘I’ll give them [sales staff] the benefit of the doubt in that first period’, says the CFO. That changes at the third meeting, as the time by which financial results need reported draws near, because ‘at the end of the day I have to support the numbers’. If at that point he can still see no clear evidence that sales of the relevant item have picked up or will pick up, he will insist on making a provision for slow movement or obsolescence. Valuing inventory according to IAS 2 is thus a task that requires skilled judgement, but it is a familiar one: the CFO has much experience of doing it. However, not all the rules the case-study firm has to follow are familiar in this sense. Like other listed companies, it had just had to move from following the rules of U.K. Generally Accepted Accounting Principles (GAAP) to following International Financial Reporting Standards (IFRSs). The case-study firm’s ‘corporate compliance manager’ had taken primary responsibility for this demanding transition. The new standard on which I will focus (IFRS 2, ‘Share-based Payment’, or its UK equivalent FRS 20) requires a firm ‘to reflect in its profit or loss and financial position the effects of share-based payment transactions, including expenses associated with transactions in which share options are granted to employees’ (IASB 2005: 143, emphasis in original). Like many others, the case-study firm issued considerable amounts of call options on its shares to its employees, and to follow IFRS 2 it thus had to measure the value of these options (a ‘call option’ is a contract that gives its holder the right, but does not impose the obligation, to buy an underlying asset at a set price on or up to a given future date). IFRS 2 specifies in some detail how to measure the value of an option. It contains ‘Application Guidance’, which for example lays down that if ‘market prices’ of ‘share options granted to employees’ are not available (and they usually are not, because the detailed terms of these options will generally differ from those that are traded publicly) ‘the fair value of the options granted shall be estimated by applying an option pricing model’ (IASB 2005: 159). The standard does not, however, say which option-pricing model is to be used; nor, for example, does it specify in detail how to estimate expected share-price volatility (a key parameter in option-pricing models). The compliance manager also did not find within the standard an entirely clear answer to a particular question that exercised her: the interaction between the treatment of options that had been issued before and after the standard came into force. Because (unlike IAS2) IFRS 2 was new, there was no established practice within the firm for the compliance manager to turn to in following it. As well as herself studying the standard and the associated guidance in detail, she therefore also took careful account of the extensive discussions that were going on within the accountancy profession about how to follow IFRSs. One advisory firm had set up an area of its website ‘that was totally dedicated to IFRS, so that was very useful ... they also had a news feed, so any little snippet from articles ... that referred to IFRS or how companies were dealing with their IFRS projects ... popped up’. She found a detailed worked example, also issued by a professional advisory firm, particularly helpful. It was of a company that had a wide variety of types of share-based payments. It showed how the company accounted for them under existing GAAP, ‘and it took you
Producing Accounts
113
through the process of it implementing IFRS for those different types of share-based payments. And that was much more real than perhaps what was in the standard’. In general, says the compliance officer, ‘worked examples enhanced my understanding of what they, the words [in IFRSs], were trying to tell you to do’. The difficult task of being sure that one was following IFRS 2 correctly was eased by the case-study firm’s auditors (‘they ... have technical departments … that are reviewing other companies’ sets of accounts and I suppose advising clients that were in the process of doing IFRS implementation as well’) and by hiring a specialised advisory firm that was valuing options for many other companies. The latter firm confirmed that the case-study firm was indeed following IFRS 2: ‘we’re drawing on [the advisory firm’s] experience to ensure that we had ... come to the right conclusions’, for example that the decision to use the Black-Scholes-Merton optionpricing model (see MacKenzie 2006) was the appropriate one in the circumstances. The advisory firm also helped to calculate the appropriate volatility figure to input to the Black-Scholes-Merton model. The case-study firm’s stock-market listing was recent, making the historical volatility of its own share price unreliable. Under such circumstances, the guidance to IFRS 2 permitted ‘the historical volatility of similar entities following a comparable period in their lives’ (IASB 2005: 164) to be considered. The advisory firm compiled data on companies similar to the case-study firm that had been listed for longer. There was a large group that had similar volatilities and some outliers, so the latter were eliminated and the mid-point volatility chosen to be an appropriate estimate of share-price volatility. Conclusion No empirical case-study – especially not such a modest one as this – can prove finitism correct and extensional semantics wrong. When the contingencies of the application of terms to particulars are demonstrated, as they have been here, the proponent of extensional semantics can always argue that they arise from poorly formulated or improperly grasped meanings, rather than from inherent features of concept-application. A comprehensive rule book of accounting, applicable without ambiguity, is possible, such a proponent might argue – we simply have not yet discovered it. So all that can be claimed is that the findings of this case-study are consistent with finitism, not that they show it to be correct. However, the nature of rule-following in the examples discussed above is just as finitism would predict. Applying IAS 2, for example, involved the CFO judging matters such as the strength of the cases that others were putting to him about likely sales of inventory, and what IFRS 2 implied was not fully evident from the standard itself: it had to be determined in part from how others were applying it. The usefulness of the worked example (‘much more real than perhaps what was in the standard’) is textbook finitism, akin to the central role of the ‘paradigm’ in the sense of the ‘accepted problem-solution in science’ (Barnes 1982: xiv). The finitism of primary accounting classification – such as which chart-ofaccounts code to apply to a transaction – is perhaps less evident than the finitism
114
Knowledge as Social Order
of rule-following, for the correct answer often seems ‘obvious’. However, this ‘obviousness’ is a trained, habitual response, not a logically-necessary one, as is brought to light when an intelligent non-accountant (such as the staff member who thinks that a dinner over which business is discussed is thereby ‘business entertainment’) classes things differently. Even in the simplest case discussed here – an electricity bill – one could imagine someone who did not know the sort of things that are classed as ‘maintenance’ choosing code 82101 (‘Maintenance – property’). The scope of finitism is thus wider than the spectrum of acts of classification and measurement that accountants acknowledge as involving ‘judgement’. Much of the classificatory work of the accounts administrator, for example, does not involve judgement in that specific sense, but – as emphasized – it is still skilled. For all the case-study firm’s commitment to efficiency, it has not sought to automate this classificatory work, and this may well be wise: the automation of the finitist application of terms to particulars is problematic, for the reasons explored in Collins (1990). One can, as suggested above, imagine an automated system that would search purchase orders for ‘cues’ to use to code them, but it seems unlikely that such a system would improve upon the performance of an experienced human being with a good understanding of the activities of his or her organization and of the other contingencies of his or her context. It is worth re-emphasizing that a finitist viewpoint does not imply that the application of terms to particulars is arbitrary, random or chaotic. As noted, the casestudy firm’s accounting process were orderly and impressively efficient. Rather, what finitism suggests is that the sources of orderliness are not to be found in the terms themselves but in what is done with them. As Bloor puts it: According to meaning finitism, we create meaning as we move from case to case. We could take our concepts and rules anywhere, in any direction ... We are not prevented by ‘logic’ or by ‘meanings’ from doing this ... The real sources of constraint [are] our instincts, our biological nature, our sense experience, our interactions with other people, our immediate purposes, our training, our anticipation of and response to sanctions, and so on through the gamut of causes starting with the psychological and ending with the sociological (Bloor 1997: 19–20, emphasis in original).
For example, the ‘audience’ for financial reports clearly matters. One key audience is the firm’s auditors; another is the financial markets. More than ten stock analysts cover the firm intensively, for example making detailed predictions about upcoming financial results. Says the CFO: If you’re a public company your numbers are in the public gaze and people are anticipating and ... analysts are trying to second guess what the company is going to do all the time, and it’s meeting that expectation, whether it’s going to be good, bad or indifferent ...
While ‘the auditing profession ... never ... hit you hard for being too prudent’, undue prudence is discounted by the market, which adjusts its expectations accordingly with potentially adverse consequences for the company if those expectations are not met:
Producing Accounts
115
If the market keeps on saying ... ‘they’re always sort of underestimating what they’re going to do’, they will build that into their figures ... to say, ‘we know they always come good’. And of course … the quarter that they think you’re going to come good, then … you don’t come in at that level, [that] expectation, confidence is lost and all the rest of it. So people like no surprises. So that’s what you’re trying to even ... out over the piece.
Hence, for the CFO, the importance of consistency in accounting decision-making. Neither over-prudence nor over-optimism are consistent with his duty (and the obligation imposed upon him by the Companies Act) to give a ‘true and fair view’ of the company’s economic situation. He seeks, for example, to make provisions with a hand that is neither ‘too heavy’ (over-prudent) nor ‘too light’ (over-optimistic): ‘what we’re certainly trying to do is come up where it’s middle of the road’. In the context of accounting and financial reporting, Bloor’s list could indeed be taken as a programme for empirical research that is much-needed: research of the kind sought by Cooper and Robson when they call for ‘examining not only the development of accounting rules, but also how they are interpreted, implemented and audited’ (2006: 428). In terms of this chapter’s specific findings, however, I would add ‘the technological’ to Bloor’s ‘gamut of causes’. Thus the case-study firm’s ERP system ‘partitioned’ the work of accounting into (a) primary classification, measurement and recognition, typically performed by less senior staff, the results of which the system was designed to solidify by constraining and making visible any subsequent alterations; and (b) secondary adjustments, provisions, and the like, made by more senior staff. The ERP system the case-study firm uses is a well-regarded one – ‘a tried and proven and trusted financial application’, in the words of the vice-president for information services – and its financial module is integrated internally with its other modules, which means that from an auditing viewpoint, ‘you don’t have to prove the interfaces between the systems’. The firm’s auditors can thus for practical purposes rely upon the system’s technical trustworthiness, which is: Absolutely essential. When you, from [an] audit point of view to know that when … you push a button that information gets put into the general ledger and the sales ledger or purchase ledger, in the right way, the right time, gets flown through to the P&L [profit and loss account or income statement] and the balance sheet every time is absolutely essential.
A trustworthy technical system ‘means that the whole audit cycle is much, much quicker’. The speed of the cycle in the case-study firm was impressive: its auditors had approved its most recent annual accounts in less than a month after the end of the reporting period. All modern firms of any size will almost certainly be found to be doing their accounting using technical systems, albeit often stand-alone systems rather than integrated ERP systems. If those stand-alone systems are judged by auditors to be trustworthy (and there is some indication in my interviews that not all will be judged to be so), then the technical ‘partitioning’ of accounting found in the casestudy firm may not be unusual. That suggests a hypothesis re auditing: that unless there is reason to suspect fraud, auditors will be content to deal with primary acts of
116
Knowledge as Social Order
classification, measurement and recognition on a sampling basis, perhaps delegating the work to more junior staff, while concentrating their attention on adjustments made by technical means (such as Excel) that are external to the main accounting system. One former auditor (not connected to the case-study firm) did indeed tell me that as an auditor one concentrates ‘on what they do with Excel’. Of course, that one data point doesn’t confirm the hypothesis, but if his response were found to be typical it would indeed suggest that technical systems play a structuring role. Whatever the fate of that specific hypothesis, I hope this chapter has shown that finitism can throw interesting light on accounting. Whether finitism or extensional semantics is correct is not just a social-science question: it has a bearing upon policy as well as. How to regulate accounting is a major topic of debate, especially (but not exclusively) in the U.S. If extensional semantics were correct, then regulation could be improved by formulating more exact definitions and writing ever-tighter rules. If finitism is correct, that path is endless, and the regulation of accounting needs to involve factors that are hard to define, but important nevertheless: a recognition of the importance, and skilled nature, of apparently ‘routine’ bookkeeping; the integrity of accountants, and their sense of being members of a profession as well as employees of a firm; the trustworthiness of technical systems; the independence of auditors; and so on. Accounting is crucial to the economic governance of modern societies, but it cannot be made orderly by meanings and rules alone. Acknowledgements This chapter reports work that is part of a UK Economic and Social Research Council Professional Fellowship on ‘Social Studies of Finance’ (RES-051-27-0062). I am deeply grateful to the case-study firm, especially to its CFO, for allowing me to conduct the research reported here. References Anderson, R.J., Hughes, J.A. and Sharrock, W.W. (1989), Working for Profit: The Social Organisation of Calculation in an Entrepreneurial Firm (Aldershot, Hants: Avebury). Barnes, B. (1982), T.S. Kuhn and Social Science (London and Basingstoke: Macmillan). ––– (1988), The Nature of Power (Cambridge: Polity). ––– (1995), The Elements of Social Theory (London: UCL Press). Bloor, D. (1997), Wittgenstein, Rules and Institutions (London: Routledge). Button, G. (1993), Technology in Working Order: Studies of Work, Interaction and Technology (London: Routledge). Button, G. and Harper, R.H.R. (1993), ‘Taking the Organisation into Accounts’, in Button (ed.), 98–107. Carruthers, B.G. and Espeland, W.N. (1991), ‘Accounting for Rationality: DoubleEntry Bookkeeping and the Rhetoric of Economic Rationality’, American Journal of Sociology, 97, 31–69.
Producing Accounts
117
Collins, H.M. (1990), Artificial Experts: Social Knowledge and Intelligent Machines (Cambridge, MA: MIT Press). Cooper, D.J. and Robson, K. (2006), ‘Accounting, Professions and Regulation: Locating the Sites of Professionalization’, Accounting, Organizations and Society, 31, 415–444. FIDE (Fédération Internationale des Échecs) (2005a), ‘FIDE Laws of Chess’, www. fide.com, accessed 16 May 2005. FIDE (2005b), ‘FIDE Laws of Chess’, www.fide.com, accessed 11 October 2005. Hatherly, D., Leung, D. and MacKenzie, D. (forthcoming), ‘The Finitist Accountant: Classifications, Rules and the Construction of Profits’, in Pinch, T. and Swedberg, R. (eds). Hesse, M. (1974), The Structure of Scientific Inference (London: Macmillan). IASB (International Accounting Standards Board) (2005), International Financial Reporting Standards (IFRSsTM) 2005 (London: IASB). Lakatos, I. (1976), Proofs and Refutations: The Logic of Mathematical Discovery (Cambridge: Cambridge University Press). MacKenzie, D. (2001), Mechanizing Proof: Computing Risk, and Trust (Cambridge, MA: MIT Press). ––– (2006), An Engine, not a Camera: How Financial Models shape Markets (Cambridge, MA: MIT Press). Pinch, T. and Swedberg, R. (forthcoming), Living in a Material World: On the Mutual Constitution of Technology, Economy, and Society (Cambridge, MA: MIT Press, under consideration). Powell, G. (2006), Beginning Database Design (Indianapolis, IN: Wiley). Suchman, L.A. (1983), ‘Office Procedure as Practical Action: Models of Work and System Design’, ACM Transactions on Office Information Systems, 1, 320–328.
This page intentionally left blank
Chapter 7
Power and Legitimacy Mark Haugaard
When Barnes wrote On the Nature of Power (1988), it was explicitly an exercise in sociological theory as opposed to normative political philosophy. This contrasts with the majority of power theorists for whom analysing the ‘is’ of power is inextricably tied to the normative ‘ought’. While I consider Barnes’s sociological strategy legitimate, in this article I wish to analyse some of the normative implications of Barnes’s sociological insights. In particular I wish to explore some of the normative criteria whereby we can distinguish legitimate from illegitimate power. In the power literature there are two broad extremes of analysis. Conflictual theorists tend to assume that power is some kind of noxious phenomenon which is always exercised against people’s interests. Within this perspective power is a form of domination. The opposing view is a consensual one, in which power is perceived as enabling. Power is what gives us a joint capacity for action. The former, which is the dominant view in the literature, includes Dahl (1957), Bachrach and Baratz (1962), Lukes (1974 and 2005) and Mann (1986 and 1993). The consensual, minority, position, includes Arendt (1958 and 1970) and Parsons (1963). Relative to the ‘is’ ‘ought’ dichotomy, Arendt would largely (though not exclusively) be an ‘ought’ theorist, while Parsons was decidedly an ‘is’ theorist – although it has to be acknowledged that his ‘is’ did implicitly legitimate the workings of US democracy against the critiques of Mills (1959) and others (for instance, Hunter 1953). In The Nature of Power, Barnes is working within the consensual tradition and building upon Parsons’s insights but without the handicap of the latter’s structural functionalism, which is replaced by a blend of social theory derived from Durkheim, Kuhn and Rational Choice theory. In the eyes of Barnes, Parsons’s great insight is the fact that power is primarily a capacity for action. Even in the case that it is exercised over others this ‘power over’ is a subset of the more significant category of ‘power to’. Also, Parsons correctly observed that power is not a zero-sum phenomenon which exists ‘out there’ in the world as a pre-given entity. Rather, power is created by social order. There are essentially two halves to Barnes’s argument. The first concerns the creation of power based upon some kind of consent, which derives from the ideas of Kuhn and Durkheim, while the second half concerns illegitimate power and is largely framed in terms of rational choice. For Barnes providing conceptual space for illegitimate power is significant because of Parsons’s failure to do so (Giddens 1968). In this article I intend to concentrate on the first part of his argument. In particular, I wish to use that consensual part of his argument as a springboard for distinguishing between legitimate and illegitimate power. If we take on board the central insight
120
Knowledge as Social Order
that power is joint capacity for action, by what criteria can we distinguish this kind of power from power as domination? In answering this question I shall be moving from ‘is’ to ‘ought’ and back again through a process which parallels Rawls’s account of reflective equilibrium. For Rawls (1993: 28) reflective equilibrium entails moving from our intuitions to principles of justice and back again. If the latter deliver counterintuitive outcomes then they are revised in this light. Moving the other way, the rules thus derived are used to systematize our intuitions into a coherent perspective which enables us to think about justice abstractly. In the exercise undertaken here, I wish to move back and forth from empirical sociological observations concerning the nature of power to abstract normative principles. What makes this conceptual move possible, is the fact that the ‘is’ which we are dealing with are social actors whose behaviour is normatively governed, thus it has an ‘ought’ already embedded in it. We are distilling from our sociological observations of actors implicit normative principles which are universally at work in the production of power. Using a similar process I will also try to tease out normative principles from the work of other power theorists, including Lukes, Gramsci and Foucault. Lest there be a misunderstanding, I am not claiming that this is an exercise which Barnes would have approved of. In fact, I suspect that he would be highly suspicious of my introduction of social norms into the model and, as he would see it, the strength of his model lay in the empirical, not in the normative field, where I now propose to transpose some of his insights. Foucault once observed that he was interested in making use of Nietzsche and, similarly, I would argue that the best compliment you can pay to an author is to use their work to accomplish tasks which they never anticipated. If, in this endeavour, the work I use groans as it is bent into new shapes, like oak being steamed and bent to form the frames of a ship, the outcome of the exercise should be evaluated relative to the finished product – we judge a ship by how well she sails, not by the stress of her oak frames. If we observe with Barnes that power is a capacity for action and thus desirable, while we simultaneously also concur with Lukes and others that power frequently manifests itself as domination, we have a conceptual problem: by what criteria do we distinguish the two? What distinguishes legitimate from illegitimate power? There are two distinct answers here: a sociological empirical one, and a normative evaluative one. The former analyses the viewpoint of social actors as an empirical fact, while the latter concerns itself with how society should be constituted. While it is the latter which interests us for the purposes of this article, I will begin with the former as it contains the key to the latter. As an empirical fact, legitimate power is power which is based upon the consent of the actors involved. In the work of Parsons (1963) consent derived from shared system goals, while in Barnes it comes from a common interpretative framework (Barnes 1988). Sociological, de facto, consent is not, however, always the basis for normative legitimacy. As a political scientist in a Weberian mode, I may observe that in a certain society patriarchy was (or is) legitimate but that is not the same as actually endorsing patriarchy as normatively legitimate. This distinction is crucial to understanding Gramsci and Lukes. The objective of hegemony is to create a position of domination in which the subaltern actors consent to their own
Power and Legitimacy
121
domination. For Gramsci the power of the bourgeoisie rested upon their capacity to make their acts of domination appear legitimate in the eyes of the dominated. Lukes based his conceptualisation of three-dimensional power upon Gramscian notions of hegemony. Those who suffer from ‘false consciousness’ are actors who, essentially, consent to their own domination because they do not know what their real interests are. In the second edition, this is integrated with the work of Bourdieu, whereby false consciousness is a process of domination where consent is rooted in an imposed habitus derived from the bourgeoisie (Lukes 2005). Thus, domination is made appear ‘natural’. While the subaltern, or dominated actors, give de facto consent to power, it is Gramsci’s and Lukes’s view they should not be doing so. The latter is a normative evaluation whereby normative and sociological legitimacy are out of sync with each other. In Foucault we also find a similar disjuncture of consent. Actors within an episteme or discourse formation share consent to an interpretative world-view. They even engage in exercises of power over each other that presuppose a shared epistemic consensus. As observed by Clegg (1989), these are the rules of the game which decide what constitutes victory or defeat. Truth in this case performs an integrative function, whereby consent is created. As I have observed previously (Haugaard 1997 and 2003), truth is something a ‘rational actor’ cannot disagree with, consequently the only alternative to the ‘true’ is ‘madness’ and ‘deviance’, which constitute subject positions that inherently disempower. This kind of strategic use of truth creates de facto legitimacy but not normative legitimacy. Of course, in Foucault this is/ought distinction is never made explicitly, but his critique depends upon it. He is not simply describing power and truth, which, as a fact, actors consent to, but power and truth that they should not be consenting to. The implicit heroes of the story are those who resist by not consenting to the order of things. Why are they to be applauded? Is it simply because they resist or because they resist for normatively desirable reasons? Maybe I should not presume to answer on Foucault’s behalf; my answer would be that they should be applauded only if the resistance is for reasons of normative desirability. What Gramsci, Lukes, Foucault and Barnes have in common, curiously enough, is that the basis of consent is derived from social knowledge; it is only in Barnes that this consent is not simply a ruse for domination. How do we distinguish de facto consensus from normatively desirable consent? The easy and unsatisfactory answer is of course in terms of some kind of speaker’s privilege. Traditional Leninist discourse took this form and, arguably, appeals to ‘false consciousness’ are implicitly structured this way. In order to look for a more satisfactory path, I wish to begin from Barnes’s perspective. In The Nature of Power, Barnes argues that power is a capacity for action. There is natural power and social power. The former is derived from physical objects while the latter comes from social order. Barnes’s perception of social order is strongly influenced by Durkheim and Kuhn (the former receives greater emphasis in The Elements of Social Theory, 1995, while the latter is key in The Nature of Power, 1988). Social order is a self-fulfilling prophecy whereby the orderedness of the social world is a reflection of our expectation of social order. Barnes criticizes Parsons and Hobbes for viewing the pre-social agent as a virtual savage beast in need of
122
Knowledge as Social Order
restraint either through a massive internalisation of norms or coercion. In contrast, for Barnes human beings are viewed as having a predisposition to socialization. Partly influenced by Chomsky and Piaget, Barnes sees us cognitive beings who, through socialization, acquire common interpretative frameworks akin to paradigms in Kuhn’s work. As in Durkheim of The Elementary Forms of Religious Life, our socialization is a process of categorization of the world (1995: 95). Barnes explains the construction of social knowledge by analogy between selfreferring and nonself-referring natural objects (1988: 46–54). In order to decide whether or not a sphere is a sphere, we examine it carefully. In contrast if we wish to know if a particular outcrop of rock is the summit of a mountain we look not only at it but all the other surrounding protruding rocks in order to reach a decision. The surrounding rocks constitute a ring of reference which makes the summit a summit – it is not self-referring. Compare this to a cup, what makes a cup a cup, is not intrinsic to it, but is constituted by the ring of reference which surrounds it. In this case the ring of reference is the fact that social actors believe it to be a cup. This shared perception of things self-validates. As a whole this is a shared paradigm which enables actors ‘to do’ things (power as capacity) which they could not otherwise. They co-ordinate their actions by sharing a particular view of social order and the more successful they are in using this knowledge the stronger, hence more effective, the ring of reference becomes. In this context it is worth observing that paradigms are, of course, theoretically close to the idea of epistemes and discourse formation in Foucault, so I would argue the same logic applies. Shared epistemes give actors a capacity for action which they would not otherwise have. I think this is what lies behind Foucault’s claim that power should not be considered only negatively, as saying ‘no’, but also positively (Foucault 1979: 194). A shared system of knowledge gives us a capacity for action, which enables us to do things which we could not otherwise. Foucault also asserts that there is no escape from power (Foucault 1980: 52), which is entirely correct: to the extent to which we are social actors, we interpret and categorize the world and there is no escape from this. Social actors impose, to use Foucault’s words, an ‘order of things’ upon the world, or to echo the title of Foucault’s Professorship at College de France, these use ‘systems of thought’ by which they make sense of the world. Obviously, there is nothing inherently insidious in the fact that we cannot escape from this kind of power/knowledge. Although, it has to be recognized that in some instances it can normatively wrong – my current task is to distinguish the two. In Barnes’s perspective, the power of an individual is constituted through rings of reference. What makes leaders powerful is the fact that they are each surrounded by a ring of reference which constitutes their power. They are not intrinsically powerful – even if in speech we tend to say so and so is powerful. To use an example (my example, not Barnes’s), what distinguishes the actual Napoleon from the ‘napoleons’ who are found in psychiatric institutions is not internal to them but the fact the former (unlike the latter) had a substantial ring of reference which validated his power. In order to distinguish cups from the power of John as a leader, we should distinguish between allocative and authoritative resources. Allocative resources refer to our ability to control the physical object world while authoritative resources have to do with our command over people (Giddens 1981: 51). Authoritative resources
Power and Legitimacy
123
are nonzero-sum relative to their use. If actors in authority misuse their authoritative power, the ring of reference which surrounds them contracts and becomes weaker, while other uses can strengthen the ring of reference – a virtuous spiral occurs. Thus the more people believe in someone’s leadership the more powerful they become. Conversely, because power is externally rooted in rings of reference, once actors decide to withdraw their power, the powerful become powerless. Of course, that is not to say that illegitimate leaders cannot stay in power but this entails artificially reinforcing rings of reference. Barnes uses rational choice theory to explain how illegitimate leaders can maintain their rings of reference even when those who constitute those rings of reference wish to withdraw their consent. The powerful use techniques such as divide and rule to make sure that the less powerful cannot withdraw their ring of reference simultaneously thus rendering the dominators powerless. However, I wish to methodologically bracket this analysis for a moment. As was observed by Arendt (1970) and Parsons (1963), authoritative resources do not exist in isolation from other types of resources. A subdivision of allocative resources is coercive resources. If coercive resources are used as a substitute for authoritative resources a deflation of authoritative power resources may take place. The virtuous cycle (whereby authoritative power resources are increased) is connected with the growth of legitimacy. The more legitimate a regime is, the more authoritative power that regime has. To argue from a normative angle, there is a positive correlation between legitimacy and perceptions of justice. It can plausibly be argued that the success of democratic institutions is not simply linked to some kind of a developmental process whereby people demand greater levels of justice but that just institutions are functional to society because they lead to higher levels of authoritative power. As this power is not lost when spent but actually increases in direct correlation to levels of legitimacy, this virtuous circle of power makes such a system more effective, hence more competitive, than systems which achieve their goals using coercive resources. This rather fortunate correlation between authoritative resources and effectiveness should not lead us to a simplistic Hegelian developmental view because legitimacy and justice are not identical. A person reared in a Panopticon might view a system in which those in authority demanded they behave as an automaton as legitimate but, viewed it from the outside, we would say that such subjection of self to another can hardly be considered just. We might wish to say our Panopticon person was suffering from ‘false consciousness’ but, as already argued, that is not the route I wish to take. Arendt claims that power and violence are opposites (1970: 56). This is not simply an empirical claim (as suggested so far) but also a normative one. According to Arendt, violence is by nature instrumental, in the sense that it is a means to an end but never an end in itself. Consequently it always needs justification. Legitimate power, in contrast, ‘needs no justification’ (Arendt 1970: 51–52). Violence and war are always justified relative to their opposite: peace. All wars are fought for peace, while peace is never pursued for an end other than as an end in itself (Arendt 1970: 51). According to Arendt, ‘power is in the same category; it is, as they say, “an end in itself.”’ (Arendt 1970: 51). By power, of course, she means legitimate power. What makes violence different from legitimate power? Unlike the latter, violence does not presuppose the consensual agency of other. If you stab a person, shoot them
124
Knowledge as Social Order
or bomb them from 30,000 feet, the other is only a physical object. You are acting upon their body. If a person rapes another, what distinguishes this from love is the total disregard for other as a social agent. Other is a physical object from which (not whom) gratification can be derived. It is not interaction but action upon the other. It is action which ignores the reciprocal relationship between the thing signified at the centre of a ring of reference. What makes violence normatively wrong is not simply its physicality but its denial of the agency of other. Let us use a contrasting thought experiment to illustrate the point. Imagine that a person bites a lump of flesh out of another human; imagine that a domestic dog bites a piece out of its owner’s leg and; imagine that a wolf bites the leg of an explorer. While the latter is unfortunate for the poor explorer, it is not a normative issue, while the dog incident is somewhat so and the person is definitely so. What distinguishes the person, the dog and the wolf, in descending order, is their level of socialization. A human is a fully socialized agent, the domesticated dog somewhat so, and the wolf not at all. What makes violence a normative issue is not the sheer physicality, but the apparent disregard for the agency of other when the latter might be expected – a wolf is not expected to have regard for the social agency of other. In Barnes’s terms, the wolf is not part of the ring of reference of the explorer, while the domestic dog is part of its owner’s ring of reference making the latter what they are. Following the work of Bourdieu, the use of the term ‘symbolic violence’ has become increasingly popular. Here, of course, all physicality is absent and what I am attempting to theorize can be seen in its pure form. In symbolic violence meaning is imposed upon other with total disregard to their social agents. Let us for instance say, that a certain group are classified as ‘primitives’ or they are spoken for (possibly even from altruistic motives, in the way 19th century evangelists tended to when they were ‘saving natives’). What is normatively reprehensible about this is the imposition of meaning upon other without any regard to their self-perception. It is an action upon them, about them, which disregards them as reciprocal social beings. It is an act where a group insist upon constituting a ring of reference which those at the centre of the ring of reference reject. In that sense it is the reverse of the Napoleon versus the ‘napoleon’ example: the ‘napoleons’ insist that they are Napoleon but they are not because there is no ring of reference to validate this, while the for the ‘native’ the ‘civilized’ constitute a ring of reference which gives them identity but which is foisted upon them. What does it mean to disregard the social agency of another? Reversing this, what does it mean to be a social agent? Following the cognitivist interpretation of social agency suggested by Barnes, being a social agent is essentially to be a worldcreating agent. The world out there exists made of physical things composed of atoms (or so the scientists tell us according to their world-creating capacity) which have no meaning until we impose it upon them. The tree outside my window becomes what it is, more than a collection of atoms, a tree (in fact, a Scots pine) through my imposition of meaning upon it. This meaning is not something which I personally invented but comes from an interpretative horizon which I share with others or, in Barnes’ terms, a common ring of reference. What makes me a social agent, integrated with others, is the fact that the meaning which I impose upon reality is similar to that of others.
Power and Legitimacy
125
As argued by Wittgenstein in the private language argument (Wittgenstein 1967), meaning creation is not a solipsistic act but an inherently social one. Regarding the other as social agent entails having regard for the capacity of other to impose meaning upon the world. Symbolic violence, in contrast, is an imposition of meaning upon other which disregards their ‘world-creating’ capacity. In a way it makes the other a solipsistic agent. The latter is particularly normatively reprehensible when it is the self of the other who is the object that is given a signifier or meaning (‘native’ etc). As argued by Barnes, a shared interpretative framework gives us a capacity for action that enables us to do things which we could not otherwise. What makes power legitimate is regard for the interpretative being of other. Authority, which is the personification of legitimate power in a social role, is essentially meaninggiven. What makes a politician a politician, a policeman a policeman, and a father a father to a child, is a certain meaning-giveness of those roles. As Arendt correctly observes (Arendt 1970: 45), if a father starts inflicting violence upon the child he has lost authority. In fact, I would say, he has ceased to act as a father. They are like the psychiatric ‘napoleons’, for whom others are unwilling to constitute a ring of reference. To return to the example of rape, what makes the rape of a child by a father particularly reprehensible is it that it violates an additional set of categories of meaning – those of father and child – which have a semi-sacred status. If an elected politician starts to accept bribes or act like a king, this is a violation of what it means to be a politician within the democratic system. In these cases their legitimate power decreases and they lose authority by violating a system of meaning. Of course, they may seek to make up this deficit in legitimate power by resorting to violence. However, in that act they disregard the world-creating capacity of other and thus violate the social norms which empower them. What I have been doing here is to move from the empirical fact that legitimate power and violence exist in inverse proportion towards a normative conclusion. This is a process akin to what Rawls describes as ‘reflective equilibrium’ whereby the theorist moves from intuitions to rules and back again. By a similar process, we move back and forth between a sociological is to a normative ought. The reason that this process works is that the sociological is derives from the normative intuitions of social actors. We can also distil the essence of normative legitimacy implicit in the work of Foucault from the way he pays particular attention to acts of resistance. The type of resistance which he had in mind was not simply a resistance to specific outcomes. He was not interested in the resistance of Marx to the work of the bourgeois economists, because he saw this as a form of shallow conflict (Foucault 1980: 262). What made it shallow, I would argue (Haugaard 1997: 41–97), is their mutual adherence to a shared system of meaning. They all, for instance, adhered to the labour theory of value and also regarded social agents as utility maximizers. The conflicts which really interested Foucault were deep conflicts over meaning, or conflicts over the right to create the world. What happens in these conflicts is that one method of ‘world creation’ silences another. In Discipline and Punish, what particularly interested him was the capacity of agents to impose their meaning of crime and the criminal upon the world. Was the latter someone who attacked the body politic, as perceived in the ‘sovereign model’? Were they an agent who had committed an act which could be
126
Knowledge as Social Order
found on a tabula of crimes, as in the Classical model? Or were they a specific type of deviant individual, as in the modern panoptical system? What interested him was how the advocates of each system of meaning tried to impose their version of reality upon the world and how the subjects of these systems of domination tried to resist. In other words, what was crucial was the conflict and ultimate victory of one system of world-creation over another. The appeal to truth was an attempt to reify one system of meaning over another and, thus, to ensure victory. What was reprehensible about the use of truth was that it silenced other possible acts of world creation. We find a similar process at work in Gramsci’s characterization of hegemony, in which the bourgeoisie impose their meaning of the world upon the proletariat. For instance, Gramsci saw the imposition of a single Italian language, derived from language of the Tuscan elite, upon the whole of Italy as precisely such an instance of the imposition of world-creation upon the working classes (Ives 2004: 33–62). To borrow conceptual vocabulary from Bourdieu, hegemony entails an act of symbolic violence whereby the interpretative horizon of one social group, the bourgeoisie, is elevated above the habitus of other groups, the subaltern classes, and, as a consequence, imposed upon them. In other words, the world-creating capacity of the habitus of the bourgeoisie is given validity as the only mode of legitimate expression and, as a consequence, those to whom this does not express their ‘life world’ are forced to use an alien mode of world creation, or habitus, in order to be heard. Again, I would argue that what makes these acts of legitimacy creation (de facto, sociological legitimacy) inherently normatively illegitimate is an act of nonreciprocal interaction with the system of meaning of other. As I have already mentioned, legitimate power includes ‘power over’. When a parent tells a child to go to bed or when the elected leader of a country gives an order, this may constitute legitimate power even it meets with resistance. The issue is where does the resistance come from? Is it simply over the outcomes or does it derive from a contest over meaning itself? I would argue that, if the former is the case, legitimacy is preserved, while the latter places a question mark over it. In social interaction one has to distinguish between conflicts which reproduce the existing order of meaning and those in which meaning is contested. The democratic process is premised upon relative agreement upon meaning, while outcomes are subject to struggle and conflict. Imagine two parties contesting an election. Both party A and party B field candidates who stand for election; individuals vote; the votes are counted, and the party which obtains the most votes wins. There is clearly conflict here but it is conflict over outcomes. A and B agree on what constitutes a ‘party’, as well as ‘voting’, ‘counting’ and ‘winning’ and thus reconstitute shared rings of reference – they are like the ‘cup’ example in Barnes. If party B, which lost the election, were to seize power in any case they would both be guilty of a kind of self-contradiction whereby the meanings which they endorsed prior to the outcome being known (going into the election) were violated after the election and they would also be deliberately silencing the discourse which A subscribed to. In this case, I would argue that defaulting can be normatively legitimately prevented through coercion because it is the use of violence to sustain an already agreed upon ‘world creating’ cognitive framework. To echo the language of Rousseau, if the general will is a system of meaning which is shared by the social actors, being constrained
Power and Legitimacy
127
into outcomes (including undesired ones) by those meanings is an instance of being ‘forced to be free’, thus legitimacy does not decrease. Only when meanings are violated in order to achieve ‘victory’ legitimate power decreases. Arguably, the manner in George W. Bush ‘won’ the first presidential election, is such an instance when the categories of meaning became contested and the legitimacy of the system decreased. Part of the reason why the creation of consensus upon meaning is normatively absolutely fundamental for legitimate power is that this moment of interactive consent is also the moment of social inclusion. When the way in which we make sense of the world is consented to by other, other has made us a ‘social being’, rather than a solipsistic agent. If we refer to the Kantian dictum of treating others as an end in themselves, rather than a means to an end, recognizing the meanings which other reproduces in an act of structuration is the most fundamental act of recognition of their being in the world as an incontestable presence. Meaning is not created singly. Part of the process of socialization is learning how to dovetail your individual acts of structuration with those of society as a whole. The entire ontological security of a social agent depends upon the fact that others confirm-structure most acts of structuration. Or in Barnes’s language, it is learning what are the appropriate rings of reference which others share. It is learning a shared cognitive system. In The Nature of Power Barnes asks himself the question whether it is possible to be mistaken concerning authoritative leadership (1988: 51). If I come to a group and mistakenly think Fred, rather than John, is the leader is this not a contradiction if we hold that leadership is constituted through rings of reference? After all, I am part of a ring of reference for Fred. Barnes answer is that if John is the leader this entails that he has a substantially larger ring of reference than I stand for singly. When I learn of my ‘mistake’ what I learn is that others constitute a ring of reference for John, not Fred. In a way, my mistaken belief is akin to a moment of ‘private language’, which is an act of social deviance. Of course, if there are many other ‘deviants’ like myself, then Fred will become the leader, thus I am no longer mistaken. Social conformity and competence entails learning to confirm the meanings which others attach to our actions. One of the reasons that Garfinkel’s breaching experiments elicited such strong reactions was precisely because they contested common taken for granted meanings, and undermine the ontological security of their subjects. As Giddens argues (1984: 61–63), based upon the observation of Bettleheim, what was central to the breaking down of the Jewish inmates of the camps was undermining their habitus. The prison guards would constantly violate the norms of interactive behaviour, with the result that most of the prisoners ceased to be able to act as social agents. When they reached the stage of being unable to make eye contact, they usually died shortly afterwards. There was, however, a minority who managed to resocialize according to the ‘norms’ of the camps, and thus equipped themselves ontologically for survival in this entirely different social world. If we think of social agency as an internalisation of interpretative framework, the forced rejection of the interpretative framework of another is equivalent to erasure of their existence. George Orwell once wrote that the ultimate form of torture is to
128
Knowledge as Social Order
teach someone that two and two make five. Bentham argues that the Panopticon can be used for experiments upon children to teach them that two and two make five and that the moon is made of green cheese. When we read this we recoil from the Panopticon, as a kind of devil’s artifice, which embodies Orwell’s understanding of torture. Why is this torture? Not only has the other become a pure physical object but also their social being itself has become something plastic. In Madness and Civilization Foucault argued that modern power was based upon total non-reciprocity (1971: 249). Reason confronted unreason not in a dialogue but as a monologue of reason about unreason. Again I would argue that the implicit normative force of this critique of modernity is a sense that non-reciprocity is the ultimate form of power as domination, while power that is derived from the merging of social agents’ interpretative horizons constitutes legitimate power. According to Habermas’s version of deliberative democracy, reason is absolutely central to the democratic process. In essence, what made the democratic revolution possible was the idea that unconstrained reason could be used to derive legitimacy (Habermas 1984). While I would in some respects agree with Habermas in this assertion, what it misses is the important moment that exists prior to the moment of reason, which the act of interpretation itself. As has been argued by Mouffe (2000: 49), the act of meaning-giving itself presupposes an exclusion of alternative meanings. However, unlike her, I would not accept from this conclusion that rational discourse is, as a consequence, inherently a form of domination. The latter move is one that is illegitimately made because she assumes, like so many theorists though not Barnes, that power is always a form of domination, rather than a joint capacity for action. Earlier in the article I said that I would make Barnes’s work groan and to a certain extent the very act of using his ideas to reach normative conclusions is one such act of steaming and bending of timber. Out of respect for him and his work, I would like to return to two aspects of his work in this regard: the use of rational choice and the way in which developing these normative conclusions changes our perspective of his analysis. After analysing Barnes’s characterization of the consensual basis of power, I contrasted this with violence. While I do not think he would disagree with this (1988: 118–121), it should be noted that he primarily uses various rational choice strategies to explain how powerful actors can retain their ring of reference when those who constitute it might wish to withdraw their consent. What these strategies share with what I have said above concerning normatively reprehensible uses of power/knowledge, is a manipulation and disrespect for the agency of other. Through strategies, such as divide and rule or the natural fear that the less powerful have of being first to oppose a tyrant, the rings of reference are maintained. However, to the extent to which the holders of authority claim legitimacy which is actually not theirs but manipulated in this Machiavellian manner, I would argue that this qualifies as symbolic violence because it is an insistence upon a version of reality which is not reciprocated. This lack of reciprocity is demonstrated by the need to use these subversive rational choice strategies. As already stated, Barnes deliberately methodologically brackets normative considerations. However, in so doing the legitimacy which social actors confer
Power and Legitimacy
129
upon power/knowledge relations is taken as an uncontested given. In many respects what has prompted this article is the insight that there is remarkable convergence between consensual and conflictual theories, which is not immediately obvious given their implicit normative assessment of the sociological facts. Foucault argues that epistemes or systems of thought are indistinguishable from power relations. Lukes and Gramsci observe that hegemonic ideas reinforce relations of domination. However, all three of them do so as a form of social critique. Barnes observes that paradigms, social knowledge and mutually reinforcing rings of reference are the essence of power but, in essence, argues that this constitutes a form of empowerment. He does, of course, make conceptual space for illegitimate power through coercion and divide and rule, but he does not entertain the idea that illegitimacy could be specifically embedded in the cognitive frameworks which also empower. This normative observation has a sociological impact in the sense that it alerts us to the fact that the source of consensus may not be a shared desire for empowerment but the strategic use of knowledge by some over others. Consequently, the shared cognitive frameworks that Barnes describes gain a slightly different character. Of course, this should not lead us to some kind of paranoid position in which the fact that cognitive frameworks are central to the constitution of power relations entails the conclusion that all knowledge is a ‘will to power’ as domination. Balance is what is required and what I have done in this article is to begin to tease out the criteria which can be used to distinguish legitimate from illegitimate power/knowledge relationships and thus to achieve balance. References Arendt, H. (1958), The Human Condition (Chicago: University of Chicago Press). ––– (1970), On Violence (London: Allen Lane The Penguin Press). Bachrach, P. and Baratz, M.S. (1962), ‘The Two Faces of Power’, American Political Science Review, 56, 947–952. Barnes, B. (1988), The Nature of Power (Cambridge: Polity Press). ––– (1995), The Elements of Social Theory (London: UCL Press). Bourdieu, P. (1990), The Logic of Practice (Cambridge: Polity Press). Clegg, S.R. (1989), Frameworks of Power (London: Sage). Dahl, R.A. (1957), ‘The Concept of Power’, Behavioural Science, 2, 201–215. Durkheim, E. (1995 [1921]), The Elementary Forms of Religious Life (New York: The Free Press). Foucault, M. (1970), The Order of Things (London: Routledge). ––– (1971), Madness and Civilization: A History of Insanity in the Age of Reason (London: Tavistock). ––– (1979), Discipline and Punish: The Birth of the Prison (Harmondsworth: Penguin). ––– (1980), Power/Knowledge: Selected Interviews and Other Writings 1972–1977, Gordon, C. (ed.) (Brighton: Harvester Press). ––– (1981), The History of Sexuality Volume 1: An Introduction (Harmondsworth: Penguin).
130
Knowledge as Social Order
Giddens, A. (1968), ‘“Power” in the Recent Writings of Talcott Parsons’, Sociology, 2, 257–272. ––– (1981), A Contemporary Critique of Historical Materialism (London: Macmillan). ––– (1984), The Constitution of Society (Cambridge: Polity Press). Goffman, E. (1971), The Presentation of Self in Everyday Life (Harmondsworth: Penguin). Gramsci, A. (1971), Selections from Prison Notebooks (London: Lawrence and Wishart). Habermas, J. (1979), Communication and the Evolution of Society (Boston: Beacon). ––– (1984), The Theory of Communicative Action: Vol. I, Reason and the Rationalization of Society (Cambridge: Polity). ––– (1987), The Theory of Communicative Action: Vol. II, The Critique of Functionalist Reason (Cambridge: Polity). Haugaard, M. (1997), The Constitution of Power (Manchester: Manchester University Press). ––– (2002), Power: A Reader (Manchester: Manchester University Press). ––– (2003), ‘Reflections on Seven Forms of Power’, European Journal of Social Theory, 6, 87–113. Hunter, F. (1953), Community Power Structure (Chapel Hill: University of Carolina Press). Ives, P. (2004), Language and Hegemony in Gramsci (London: Pluto). Laclau, E. and Chantal, M. (1985), Hegemony and Socialist Strategy (London: Verso). Lukes, S. (1974), Power: A Radical View (Basingstoke: Palgrave Macmillan). ––– (2005), Power: A Radical View, 2nd edn (Basingstoke: Macmillan). Mann, M. (1986), The Sources of Social Power, Vol. 1 (Cambridge: Cambridge University Press). ––– (1993), The Sources of Social Power, Vol. 2 (Cambridge: Cambridge University Press). Mills, C. Wright (1959), The Power Elite (New York: Oxford University Press). Morriss, P. (2006), ‘Steven Lukes on the Concept of Power’, Political Studies Review, 4, 124–135. Mouffe, C. (2000), The Democratic Paradox (London: Verso). Parsons, T. (1963), ‘On the Concept of Political Power’, Proceedings of the American Philosophical Society, 107, 232–262. Rawls, J. (1993), Political Liberalism (New York: Columbia University Press). Weber, M. (1948), From Max Weber: Essays in Sociology, Gerth, H.H. and Mills, C.W. (eds) (London: Routledge). Wittgenstein, L. (1967), Philosophical Investigations (Oxford: Blackwell).
Chapter 8
Barnes on the Freedom of the Will Martin Kusch
1. It is a great pleasure for me to contribute to a Festschrift for Barry Barnes. Of writers in the sociology of scientific knowledge he is second to none, and equal only to David Bloor and Harry Collins, in his profound and lasting influence on my thinking. Barnes’s important paper on social institutions (1983) as well as his books on power (1988) and social theory (1995) have provided central threads for much of my work during the past 15 years. And I continue to learn from him to this day. In this chapter, I want to discuss some of the central themes of Barry Barnes’s recent book, Understanding Agency: Social Theory and Responsible Action (2000). Understanding Agency is perhaps to date Barnes’s most ambitious and important work: it outlines a new field of sociological inquiry, the sociology of free will; it suggests a fascinating functionalist account of the language games in which we invoke free will, responsibility and agency; it reflects on the effect of the new biotechnologies on our self-image; and it provides a devastating critique of the social theories of Roy Bhaskar, Antony Giddens, and Jürgen Habermas. Put in a nutshell, one might characterise Barnes’s project by saying that he does for the freedom of the will what Durkheim tried to do for religion. Unsurprisingly, Barnes himself describes Understanding Agency ‘as an attempt to revive a Durkheimian view of human beings’ (2000: x). One of the many impressive features of Barnes’s oeuvre is the case with which he assimilates and uses scholarship of other disciplines. Understanding Agency is no exception. Barnes draws extensively on the work of sociologists, biologists, psychologists and philosophers. This gives his overall argument a high degree of empirical depth and conceptual sophistication. And it makes it easy for readers of other fields to find entry points into the debates around social theory. My own entry point is the relationship between philosophical and sociological studies of the freedom of the will. My interest in this relationship was triggered by Barnes’s choice not to consider the extensive philosophical controversy over freedom of the will. In order to justify this choice, Barnes writes that philosophers’ and sociologists’ concerns in this area are ‘importantly different, but not obviously so’ (2000: 15). I am not convinced. And thus I shall try to explain some ways in which a greater engagement with the philosophical literature can help to bring Barnes’s position, and some of its tensions and difficulties, into sharper relief. My discussion has five parts. I shall begin by asking what a sociological study of the freedom of the will could be; and how it might be thought to differ from a philosophical investigation. I shall propose that the difference is smaller than Barnes
132
Knowledge as Social Order
implies. Subsequently, I shall introduce the standard philosophical taxonomy of possible positions regarding the freedom of the will. I shall develop this taxonomy in a way that highlights the need for further refinement of Barnes’s own spectrum of opponents. The centrepiece of my study will be a critical discussion of Barnes’s arguments for ‘a sociological version of compatibilism’ (2000: xi). I shall suggest ways in which some of these arguments might be resisted. I shall conclude by listing what philosophers should nevertheless learn from Barnes’s fascinating book. As this preview already indicates, my engagement with Barnes’s book will be primarily critical. I dare adopt this critical attitude in the belief that the only way to honour a serious thinker is to engage in detail with his or her arguments – and this regardless of whether such exercise ends in agreement or disagreement. 2. What is a sociological study of freedom of the will, and how does it differ from a philosophical investigation? Barnes draws the boundary in a longish footnote early on: The approach here differs from that of the philosophers who have extensively discussed the issues of free will and responsibility. In fact it differs in a way that could invite confusion if detailed use were made of their work, because sociological and philosophical aims are importantly different, but not obviously so. This is why there is very little reference in this book to the enormous philosophical literature on free will and related topics. A quick way of conveying the different orientations involved is to note the enormous problems philosophers find with the notion of moral luck (Williams 1981). That someone may be morally fortunate seems completely straightforward and unproblematic to me, and, more important, to be so just so long as a naturalistic perspective is taken on the relevant issues (2000: 15).
This is not very helpful. The issue of moral luck is not a central issue in philosophical debates about the freedom of the will. For instance, the recent 638-page long Oxford Handhbook of Free Will (Kane 2002), containing 26 essays by all the leading authors in the field, makes no mention of the topic. And it is not obvious to me that one must have a problem with moral luck simply in virtue of being a philosopher. But perhaps the distinction between philosophical and sociological studies of free will can be drawn in some other way. Here are some obvious candidates. Perhaps sociology differs from philosophy in that the former is descriptive and the latter normative-prescriptive. In the present case this might mean that sociologists describe the function of talk of free will and responsibility, whereas philosophers prescribe how we should think about freedom and responsibility. There are passages in Understanding Agency where Barnes leans towards this view. For instance, he chastises Bhaskar and others for letting their normative-political agendas get the better of their duty to give correct descriptions of the function of references to free will (2000: 31, 89). These passages contrast with others, however, that express a different attitude. Barnes himself does not abstain from making normative statements. He not only evaluates competing social theories, he also passes judgement on the ways social actors think and act in everyday life. He voices sympathy for the ‘robust compatibilism’ of ordinary discourse, declaring compatibilism to be ‘perfectly defensible’ and having ‘merits’ (2000: 4, 114).
Barnes on the Freedom of the Will
133
Moreover, Barnes judges incompatibilism to be usually ‘sustained in a formally unsatisfactory form’ (2000: 114). There are other, and prima facie more promising, ways in which a sociological study of the freedom of the will might be distinguished from philosophical concerns. A first proposal is to suggest that a sociological investigation relates this issue to familiar themes in the social sciences and their philosophy. The most relevant of such themes are the dispute between ‘individualism’ and ‘collectivism’ (or ‘communitarianism’, or ‘holism’), and the distinction between ‘explanation’ and ‘understanding’. According to the second proposal, sociology investigates the role or function of talk of freedom of the will, responsibility and agency in both everyday life and in sociological theories. Let us call a study of the former an investigation into ‘voluntaristic discourse’. A sociology of free will might give detailed microsociological accounts of various features of such discourse, and it might offer functionalist explanations for why such discourse exists at all. Sociology might also investigate which features of such discourse are ubiquitous or universal, and which are local and contextual. And it might try to identify the causes of changes in voluntaristic discourse. The study of references to freedom of the will, responsibility and agency in sociological theories might be construed as a critical exercise. Perhaps it turns out that most currently popular theories invoke notions of freedom and responsibility that are not defensible. Finally, the third way is that of the sociology of scientific knowledge. The sociologist might want to understand the role of different conceptions of the will in various scientific disputes. No doubt, the history of science provides the sociologist with plenty of interesting material here. To mention just one example close to home, the famous ‘Imageless Thought Controversy’ in early-twentieth-century Germanspeaking lands was in good part a conflict between psychologists who adhered to incompatibilist views and psychologists who advocated compatibilism (Kusch 1999). In Understanding Agency, Barnes focuses on the first and the second way. In passing he also shows an interest in the third – at least as far as biology is concerned. Relating the problem of freedom of the will to the dispute between individualism and collectivism is perhaps the most important idea of Understanding Agency; surprisingly enough, no-one has seen the significance of this link before. This link will receive closer attention later in this chapter. Here it is more to the point to emphasise that Barnes is also developing the second way in considerable detail. He gives a functionalist account of voluntaristic discourse; he studies the various uses of concepts like ‘will’, ‘responsibility’ and ‘agency’, and he ponders questions of ubiquity and variation. He also reflects in detail on the uses of these concepts in wellknown social theories. He criticises these theories for their alleged individualism and incoherent position regarding the freedom of the will. And yet, one might still wonder whether the existence of these three projects really justify the view that, regarding the freedom of the will, ‘sociological and philosophical aims are importantly different’. Relating the free-will issue to the debate between individualism and collectivism should be a philosophical concern as much as a sociological interest. To suggest otherwise is to propose that (Anglo-American)
134
Knowledge as Social Order
philosophers are right to stick to their predominantly individualistic orientation. And it is to remove the controversy over individualism from the philosophy of the social sciences. Similar remarks are in order as far the second project is concerned. It might be thought that philosophers do not concern themselves with messy empirical issues like the voluntaristic discourse of everyday life. It is true that many analytic philosophers studying the freedom of the will would not think of their work as including such an empirical study. Again, however, appearances are, at least to a degree, deceptive. It is the standard lore of all philosophy in this area to start the investigation from a consideration of our so-called ‘intuitions’ about freedom in various real and imagined scenarios. (Assume you had been drugged with heroine and killed someone. Would your action have been free? And what if you had drugged yourself?) There is some debate over the status of such intuitions, but at least there is agreement that they are spontaneously produced and widely shared judgements. The point that is important for us here is that a study of these intuitions is in good part an empirical investigation, and that in studying these intuitions philosophers are investigating aspects of our voluntaristic discourse. After all, it seems natural to assume that voluntaristic discourse and our intuitions about freedom and responsibility are inseparable phenomena. Needless to say, my aim here is not to collapse sociology into philosophy or philosophy into sociology. Of course there are important methodological distinctions between philosophers’ and sociologists’ ways of approaching a given topic: philosophers do not send out questionnaires, sociologists tend to avoid formal logic. And the two fields share only some of their respective bodies of theoretical knowledge: few philosophers read Giddens, few sociologists Kripke. And yet, for all that, there is no reason to believe that philosophical and sociological investigations into the freedom of the will have principally different aims and objectives. At least it is far from obvious whether a sociological study can – without loss – ignore the philosophical discussion. 3. Abstract claims are cheap. I shall now try to make my case in a more concrete fashion. I begin with terminology. Barnes follows other writers in social theory by using a plethora of terms and distinctions in order to capture the main options regarding the freedom of the will. He distinguishes ‘compatibilism’ from ‘incompatibilist forms of voluntarism’, and ‘dualism’ and ‘exceptionalism’ from ‘monism’ and ‘naturalism’. Dualism is equated with incompatibilism, and monism with compatibilism. The incompatibilism is a dualist since she exempts free actions from the causal-deterministic order; this move makes her an ‘exceptionalist’ as well. The monist knows of only one order, namely the natural causal order: hence the monist is also an advocate of ‘naturalism’. I believe that this classification is unduly complex and that it leaves some important options in the dark. The received classification of positions in the philosophy of freedom of the will can help to bring this out (Figure 8.1).
Barnes on the Freedom of the Will f
Compatibilism (i.e. Soft Determinism)
vs.
Incompatibilism
Hard Determinism (Often: Illusionism)
Figure 8.1
135
vs.
Libertarianism
The main philosophical views on freedom of the will
Compatibilism allows that an action can be free despite it being causally determined. Only certain kinds of causes, coercion for example, remove an action from the realm of freedom. Incompatibilism insists that freedom and determinism do not mix. An action cannot be both causally determined and free. Incompatibilism comes in two forms. ‘Libertarianism’ holds that freedom of the will is possible since certain phenomena are exempt from causal determinism. Such phenomena are, first and foremost, acts of choosing how to act. ‘Hard determinism’ responds to the incompatibility of freedom and determinism in a less compromising fashion. According to hard determinism there is no freedom of the will at all; there is only causal determination. The attribute ‘hard’ is used in order to distinguish this form of determinism from compatibilism. Compatibilism is a ‘soft’ form of determinism since it allows causal determination to go (at least sometimes) hand-in-hand with freedom of the will. Hard determinism leaves no space for this possibility. Hard determinists do of course recognise that we often speak as if libertarianism or compatibilism were true. Some hard determinists therefore call for a reform of language. More typically, hard determinists advocate ‘illusionism’: they regard our libertarian or compatibilist ways of speaking and thinking as useful illusions. They believe that people will be more likely to act cooperatively if they believe themselves to possess internal states of choice and rationality. Applying this terminology to Understanding Agency, it is clear that Barnes’s main target – what he calls ‘incompatibilist forms of voluntarism’ – is libertarianism. The position defended in the book, so we are told, is ‘a sociological version or compatibilism’. Hard determinism or illusionism are never identified as options. This is problematic for two reasons. On the one hand, it seems that important strands of thought in the history of social theory are forms of hard determinism. I am thinking of course of important schools of thought within Marxism. It is surprising that Barnes shows no interest in these views. On the other hand, it is not clear to me whether Barnes is right to describe his own position as a form of compatibilism. As we shall see, much of what he says is in fact compatible [!] with hard determinism. Indeed, Barnes even adopts a strong form of illusionism. Be this, for the moment, as it may, note that hard determinism is monistic and naturalistic. Monism and naturalism are for Barnes what puts any theory of humans into good stead.
136
Knowledge as Social Order
4. Barnes links the debate over the freedom of the will to the controversy between individualism and collectivism. For him the main opponents are individualistic libertarianism and collectivistic compatibilism. The individualistic libertarian submits that the ability to make free choices is grounded in particular capacities of the individual. Barnes’s collectivistic compatibilist presumes that ‘free agent’ is a social status; that attributions of this status have causal explanations; and that such attributions are made in order to causally influence others. Closer inspection reveals that we actually have to deal with two different oppositions here. One opposition is that between individualism and collectivism. The other is that between ‘internalism’ and ‘projectivism’ – this too is a distinction that has been invoked in the philosophical debate over freedom of the will (Double 2002). Let me explain. The first opposition can be identified by means of the question: Can we make sense of different positions (regarding the freedom of the will) by focusing on socially isolated individuals, or do we need to introduce groups into the picture? The second difference involves the following query: Are ‘free action’ and ‘free agent’ statuses that we (as individuals or groups) project upon behaviours and individuals, or are these terms used in order to indicate the presence of internal states in the agent? It is not immediately obvious that an individualistic version of, say, libertarianism must be internalistic, or that a collectivistic version of say, compatibilism, must be projectivistic. At least in the philosophy of language, some well-known individualists accept projectivism insisting that statuses can be projected by the individual upon herself (Blackburn 1984). Moreover, some collectivists believe that communities can have shared internal states; such states may be the interlocked dispositions of community members (Horwich 1998). I hasten to add that my unravelling of Barnes’s two bundles – projectivistic collectivistic compatibilism and internalistic individualistic libertarianism – is more than pedantry. A good part of Barnes’s case for his compatibilism is the obvious unattractiveness of the only seriously considered alternative. And thus it seems appropriate to wonder whether other, possibly more attractive, positions have been neglected. Here I can only list four positions that might repay closer consideration: • •
•
•
Collectivist internalistic libertarianism: the freedom of the will is grounded in collective capacities; Individualistic projectivist compatibilism: ‘free agent’ is a status that individuals can project upon themselves; their acts of so doing can be explained causally; Individualistic hard determinism: there is no free will; the causal determination of human behaviour can be explained without invoking groups as causal agents; Collectivist hard determinism: there is no free will; the causal determination of human behaviour can be explained only by allowing for groups as causal agents.
5. Barnes’s argument for his own position – collectivist projectivist compatibilism – is complex. One important strand of the argument is Barnes’s attack on individualism.
Barnes on the Freedom of the Will
137
Here his primary target is Rational Choice Theory. Against the latter, Barnes insists that individual calculation always involves knowledge and that knowledge is a ‘trans-individual resource’: ‘Rationally calculated actions make use of knowledge, but knowledge is a trans-individual resource, collectively generated; evaluated and standardised and made available to individuals who for the most part take it on trust as the knowledge of “their society”’ (2000: 51). Barnes also rehearses familiar Wittgensteinian ideas against individualist solutions to the problem of rule-following (2000: 54). And he refers to his own previous work which contains many important considerations in favour of collectivism (e.g. 1983, 1995). To my mind at least, Barnes’s position concerning this issue is entirely convincing. Barnes’s arguments for compatibilism and projectivism demand more attention in the present context. One can discern the following five lines of reasoning. First, a ‘robust’ or ‘easy-going compatibilism’ characterises ‘most everyday [voluntaristic] discourse’ (2000: 4, 111). We usually allow that an action can be caused and chosen at the same time. For instance, we often regard actions under the influence of alcohol as free (in some sense). Furthermore, our legal practices assume that responsibility and causation typically go together: ‘to be responsible for something an agent must have caused it’ (2000: 9). Since responsibility and freedom of choice are closely related – in general to be responsible is to be free – our legal practices too display a commitment to compatibilism. The fact that compatibilism dominates our voluntaristic discourse constitutes a prima facie argument in its favour. Second, although our voluntaristic discourse is dominated by compatibilism, we sometimes are willing to advocate forms of libertarianism. Barnes speaks of the ‘dualism of everyday voluntarism’ (2000: 25). Our voluntaristic discourse is occasionally characterised by a contrast between voluntary and caused actions, or between choice and causation. This tendency is confirmed by dictionary entries according to which voluntary actions are free, unconstrained, unpredictable, spontaneous and not caused (2000: 4). Moreover, agents ‘acting under causal constraint are often said not to be … responsible’ (2000: 5). Nevertheless, the appearance of libertarianism in our everyday voluntaristic discourse does not constitute a prima facie argument in its favour. The asymmetrey in the treatment of everyday compatibilism and libertarianism can be defended by noting that ‘whatever incompatibilism is to be found in our practical life is typically sustained in a formally unsatisfactory form, as is the case, for example, in many legal contexts’ (2000: 114). Barnes gives one example of such ‘formally unsatisfactory form’. In many legal settings it is assumed that choices are in principle uncaused – that they are free in a libertarian sense. At the same time lawyers and judges have to reckon with the possibility that various causes – from upbringing to drugs – might impinge upon the choice. The standard ‘solution’ is to suggest that causal circumstances can, to an extent, constrain or restrict the free will. Moreover, in any given case, the degree of deserved blame depends on the extent to which the free will was constrained (ibid.). What makes this legal practice ‘unsatisfactory’ is that ‘sadly… measurement of the relevant “constraints” is still a little difficult’ (ibid.). In other words, it is close to impossible to make sense of the required ‘degrees of freedom’.
138
Knowledge as Social Order
Third, as far as freedom of the will is concerned, projectivism is superior to internalism. We will never be in a position to determine statuses on the basis of internal states since the states in question are invisible and contestable. This can be seen in more than one way. Legal contexts show that the interest in assigning a specific status – say ‘guilty’ – drives the attribution of states, and not vice versa: The move that is supposed to be made is from state to status, but it is clear that strong back-pressure, to say the least, exists from status to state. Prior concerns with status can and at times do shape and structure attributions of state… The most immediately obvious features of courtroom decisions actually raise the question of whether concepts like ‘responsibility’, ‘choice’, ‘free will’, ‘agency’ and so forth might not be secondary features of the institution of responsible action, mere rationalising accompaniments of procedures moved by pragmatic expediency (2000: 14).
Moreover, work in ‘attribution theory’ demonstrates that our willingness to attribute specific states varies with social circumstances. Confronted with the very same type of helpful behaviour, we are prone to interpret it differently depending on the social status of the one who is helping us. If a person of perceived high social status assists us in some task, we are likely to attribute to her a desire to be kind. If a person of perceived low social status assists us, we tend to conclude that she was coerced by the circumstances. In other words, we are biased to see high-status agents are internally motivated, and biased to regard low-status agents as externally determined (2000: 35). Furthermore, work by C. Wright Mills (1940) and others suggests that different cultures impute different mental states as motives of action. This invites accepting ‘the conceptual relativism implicit in sociological and ethnomethodological work, and to acknowledge that different vocabularies of states may be deployed in different ways, all of which stand as viable means of making sense of action’ (Barnes 2000: 43). Barnes insists on this point against the very attribution theorists that he uses in support of his anti-internalism. Attribution theorists speak of our ‘biases’ in attributing different mental states to people in different social circumstances. Talk of bias implies that there is all unbiased and singularly correct way of attributing states. Attribution theorists believe that psychology will make progress in identifying the real mental states; and that here psychology will follow the model of the physical sciences. Allegedly the physical sciences have progressed towards ever more accurate conceptions of invisible theoretical entities. Barnes rejects this comparison and the interpretation of the natural sciences it involves: ‘The epistemological and ontological status of [science’s] current theoretical entities remains an unsettled matter in the philosophy of science, and there is nothing to suggest that these entities, unlike the long sequence of their since modified or discarded predecessors in the history of science, are in any sense final’ (2000: 42). Fourth, from the viewpoint of collectivistic projectivistic compatibilism it is possible to provide what philosophers would call an ‘error theory’ of individualistic libertarianism. Sociological compatibilism explains why we are biased in favour of individualism, and why we are prone to overlook the ways in which we project statuses. According to Barnes’s analysis our everyday voluntaristic discourse is the medium in and through which we influence one another. We do so by manipulating
Barnes on the Freedom of the Will
139
the so-called ‘deference-emotion system’ (Scheff 1988). Each one of us monitors the extent to which even the most minute reactions of others either enable us to feel pride or force us to feel shame. We enable others to feel pride when we maintain and preserve their most basic status: the status of being a responsible and free agent. Such status cannot be acquired once and for all. Like all statuses is has to be recreated from moment to moment by others’ actions. What leads others to grant us this status is that we are susceptible to their requests and wishes. This gives the seemingly paradoxical result that to be free is to be susceptible to the influence of others. Barnes summarises the point as follows: ... our sense of the free will of an agent derives from her susceptibility to others, the kind of susceptibility implied in accounts of the deference-emotion system; our characterisation of an action as chosen identifies it as the kind of action that is open to modification through use of the system, that is, through symbolic communications and the evaluations they convey (2000: 69).
All this points in the direction of compatibilism: attributions of freedom to an agent are compatible with her being under the causal influence of others. Why then are we tempted to opt for internalism, individualism and libertarianism? Barnes suspects that the underestimation of mutual susceptibility is a universal feature of voluntaristic discourse in all cultures (2000: 60). The cause of this underestimation is the concern to prevent the deference-emotion system from being artfully manipulated. No-one wants to be manipulated by others. And hence we inadvertently slide towards the view that a free agent are lacking in susceptibility. We find it hard to acknowledge our fundamental susceptibility because the most salient types of susceptibility are deceit and betrayal. Rightly defending ourselves against the latter, we mistakenly blind ourselves to the prevalence of the former (2000: 144). Individualism is also encouraged by another feature of voluntaristic discourse: in it human beings ‘assign rights and responsibilities to each other as separate, independent unit’, or as ‘discrete points’ (2000: xi). It is easy to overlook that these discrete points are constituted in and through interaction. Indeed, the projectivistic workings of our voluntaristic discourse are hard to grasp. It is not easy to understand that something is money only because of our collective belief that it is money. And it is difficult to comprehend that ‘status… is a property located at a point in a system of relations but constituted by the system as a whole’ (2000: 65–66). We are prone to overlook the system by situating its powers into the point itself. And thus we attribute to invisible mental states what is in fact constituted by collective action. Finally, once individualism and internalism are in place, libertarianism is hard to avoid; individualism and internalism lead us to think of causal influences as manipulative and coercive, and thus as detrimental to freedom. Up to a point, Barnes’s error theory concerns all of us: all human beings share the tendency ‘to underestimate the extent and importance of their own sociability’ (2000: 152). But some of us are worse off than others. Ordinary participants or members in voluntaristic discourse are trapped in the mistaken interpretations presented above, and this regardless of whether or not they tend towards compatibilism, hard determinism or libertarianism. The same is true of much of contemporary social
140
Knowledge as Social Order
theory (Bhaskar, Giddens, Habermas) since it fails to recognise the problematic origins of its individualistic and internalistic libertarianism. Barnes feels slightly uneasy about the way his analysis of voluntaristic discourse overrules the actors’ own views. On the one hand, he defends himself with the thought that ‘everyday activity may be described better from the outside than by those who enact it – better, that is for some purposes’ (2000: 72). On the other hand, Barnes is worried that his analysis might potentially undermine the very institution that is its concern: To become more aware at the level of reflection of what we all conspire to ignore at the level of ordinary interaction cannot be without risk. Following Erving Goffman, we could say that our very sociability is just what blinds us to that sociability, and that fear of undermining the whole discourse of status and responsibility, and weakening the relations of mutual respect that link us all together, makes us reluctant to pull at the blindfold (2000: 153).
Fifth, the libertarianism of modern social theory is problematic in more than one way. Rather than to reflect critically on the role of individualism, internalism and libertarianism in voluntaristic discourse, Habermas, Giddens and Bhaskar take over these positions as if they were self-evident and without alternatives. Moreover, all of these social theorists make the dubious assumption that a progressive social theory must be libertarian, that is, that political liberation is incompatible with thoroughgoing determinism. In other words, determinism is seen as ‘inherently fatalistic and reactionary’. This judgement is arguably wrong: the same ‘kind of political work, [that] at one time in one context, was done using notions of free agency or free will has, elsewhere and at other times, been done by a determinism that denied their existence’ (2000: 31). A related tendency can be discerned particularly clearly in Habermas: ‘approved actions [are] credited to agency and individual autonomy, and deplorable actions to constraints and externalities (just as is commonplace in everyday life)’ (Barnes 2000: 89). The most fundamental problem with the libertarianism of social theory, however, is that it is never really carried through. The idea of a full-blown libertarian social theory can avoid incoherence only by forgoing all predictive aspirations. If actions were free in a strong libertarian sense, then actions could not be predicted. This is a price that social theorists are unwilling to pay. Thus, when it comes to developing their position in more detail, social theorists compromise their programmatic libertarian position. Libertarian social theorists want to have their cake and eat it too: they want to see themselves as champions of freedom of the will, and yet they want to predict choices with accuracy and certainty (2000: 31). 6. Barnes’s five arguments for compatibilism are important, intriguing and complex. I am not sure that I have taken their full measure. Nevertheless, and on danger of getting the wrong end of the stick, I shall make some critical comments. I present these in an explorative spirit, hoping that they may lead to further work developing and strengthening Barnes’s position. First of all, I cannot help noting that several of Barnes’s arguments are at least structurally similar to familiar arguments in the philosophical debate over the
Barnes on the Freedom of the Will
141
freedom of the will. Take for instance the idea that in everyday discourse one can find both compatibilist and libertarian views; that this establishes a prima facie case in favour of both; and that the prima facie case for libertarianism ultimately fails to pass muster because of formal difficulties in developing it further. This style of argument is the standard lore of philosophical inquiry that starts from intuitions. The philosopher finds intuitions pulling her in, say, two different directions, A and B. She investigates the consequences of these intuitions and finds that intuitions pulling her in the direction of B lead to difficulties of some sort or other. Intuitions pulling her in the direction of A do not cause such problems. This is likely to convince her that the A-intuitions are to be kept but that B-intuitions should be overruled and explained away. A particularly compelling way of doing so is to present an ‘error theory’ of Bintuitions: this might be a theory of cognitive illusions, or ideology. Maybe creatures like us are cognitively biased to form misleading intuitions of type B, or members of a certain social class or group are socially determined to think in this way. It is also clear that Barnes’s arguments for projectivism and against internalism would be familiar to philosophical readers. The philosophical debates over the nature of folk psychology, meaning and mental content is full of similar ideas. In these debates it is often asked whether it is adequate to assimilate invisible mental states to the theoretical entities of the natural sciences, and whether our talk of mental states is either factual or projective. In the philosophical debate over the freedom of the will one finds these concerns as well (e.g. Double 2002). Finally, it should be obvious that the relationship the freedom and predictability of an action is an old philosophical chestnut. Again, then, I do not see why Barnes feels that to make contact with the philosophical literature on freedom of the will ‘could invite confusion’ (2000: 15). 7. Turning to a more substantial issue, there is something simultaneously ingenious and troubling about Barnes’s treatment of everyday libertarianism. The peculiar nature of this treatment can be best brought out by contrasting it with the philosophical method of ‘reflective equilibrium’ (Rawls 1999). Take the work of a moral or political philosopher. If Rawls is to be believed, then her work goes something like this. She starts with her moral intuitions. She tries to build a systematic normative theory of moral actions on the basis of as many of her more important moral intuitions as possible. Once she has arrived at the theory, she checks it against her intuitions. This testing might identify tensions between some of her intuitions and her theory. In response, she might either give up these intuitions, or refine the theory so that it can do justice to the intuitions. The resulting theory in turn might contradict some other intuitions, and thus the process has to be repeated. It comes to an end only once the moral philosopher has achieved ‘reflective equilibrium’ between her intuitions and her theory. Barnes’s case is not like the case of the moral philosopher. We get closer to his situation if we replace the moral philosopher by a sociologist of morality. The sociologist of morality does not try to formulate a normative moral theory. He is trying to come up with a descriptively correct theory of, say, how people speak about moral issues and why they do so. Perhaps the explanation lies with the function of moral discourse. Now, can the theory-building of the sociologist of morality follow the model of reflective equilibrium? I think the answer must be negative. Although
142
Knowledge as Social Order
the sociologist of morality builds a theory about morality, he does not check his theory against his moral intuitions. And thus his theory-building cannot lead him to give up any of these intuitions. His moral intuitions and his sociological theory of morality simply are not on the same level or of the same order. If there are intuitions for him to check, then these will be intuitions about the social function of moral discourse. But not only are these different from moral intuitions (as to good and bad, right and wrong), they are also not as important to him as moral intuitions are to the moral philosopher. The latter has little alternative to starting from his moral intuitions. But the sociologist of morality is able to do better than consult his intuitions about the social function of moral discourse: he can go out and empirically observe that discourse. If this is a correct account of the difference between moral philosopher and sociologist of morality, what are we to say about Barnes’s insistence that the libertarianism of everyday discourse should be seen as something of an illusion? Barnes provides us with a theory that accounts for the function of our voluntaristic discourse. Voluntaristic discourse is a discourse in and through which we project the statuses of ‘free and responsible agents’ and ‘free actions’. And it is a discourse in and through which we influence each other causally. The discourse has both libertarian and compatibilist tendencies, even though the latter are more prevalent than the former. This is not a theory about whether or not libertarian or compatibilist intuitions are true: it is a theory about what kind of discourse sustains and creates these viewpoints. And thus this theory cannot entitle us to favour one viewpoint over the other. There is no contradiction between our libertarian intuitions – intuitions to the effect that free actions are uncaused – and the fact that in voluntaristic discourse we influence each other causally. At the most, Barnes’s theory can tell us that a good many of the actions that we take to be free (in the libertarian sense) are actually not free (in the libertarian sense). But this does not show that libertarianism itself is a mistake. Here is a different way to arrive at the same point. Barnes offers a theory that shows us the causal role of speech acts in which we ascribe the status of ‘free and responsible agent’ to others. And he takes it as obvious that the existence of such theory provides an argument for compatibilism; he takes it as evident that a causal explanation of voluntaristic talk, of references to free will and responsibility, vindicates the view that free will and causal determinism are compatible. But the conclusion does not follow from the premises. To show that speech acts referring to free will are compatible with causal determinism is not yet to show that free will is compatible with causal determinism. Compare: to show that speech acts referring to God are compatible with causal determinism is not yet to show that God is subject to causal determinism. The libertarian call happily accept Barnes’s premise and yet deny his conclusion. All this is not to deny that Barnes has a case against individualism and that his error theory regarding everyday individualism is indeed successful. What makes the difference between the two cases – individualism here, libertarianism above – is that everyday individualistic intuitions are threatened if it turns out that individualistic renderings of knowledge, rule-following and calculation lead to contradictions or other conceptual problems. Such analysis, combined with an account that shows
Barnes on the Freedom of the Will
143
how our very participation in institutions blinds us to their effects, can serve to fatally undermine these individualistic intuitions. In other words, individualistic intuitions and Barnes’s sociological theory of our voluntaristic discourse are on the same level. Barnes is not just giving us an account of how references to the autonomous individual are collective phenomena. He also shows that the socially isolated, autonomous individual is unable to achieve knowledge, follow rules and carry out rational calculations. At this point it might be thought that I am not giving Barnes his due. After all, in the case of everyday libertarianism he too goes beyond the mere causal explanation of references. He points out everyday libertarianism often takes a ‘formally unsatisfactory form’ in that it implies the existence of ‘degrees of freedom’. Choices are more or less free depending on the extent of constraint put upon them. And the libertarian has difficulties explaining how such degrees are measured. The underlying picture is that of a man whose freedom of movement can be greater or smaller, depending on how heavy or tight are his shackles (2000: 114). Unfortunately, Barnes’s critique of libertarianism is not decisive. The problem is real enough, but it is not peculiar to libertarianism. All voluntaristic discourse faces the problem of having to declare agents not just responsible or lacking in responsibility simpliciter, but of having to declare agents more or less responsible. And since the concepts of responsibility and freedom are inseparable, degrees of responsibility call for degrees of freedom. The libertarian tries to capture these degrees in the way described above, the compatibilist must find some other way. If Barnes wants to tip the balance in the latter’s favour, he must show us a way in which the compatibilist can avoid these ‘formal’ problems. Turning next to Barnes’s defence of projectivism, here too I feel that a determined opponent would not have to give up quite yet. No doubt, the ‘back-pressure’ from status to state is real enough. The evidence of attribution theory to this effect is impressive and overwhelming. It is also easy enough to agree with C. Wright Mills’s observation that different cultures impute to some degree different states as motives of action. But it is still a long way from here to the projectivistic conclusion according to which our voluntaristic talk is performative, and performative only. Cognitive psychologists and most philosophers of mind would agree that different cultures do not coincide fully in the list of mental states that they are ready to attribute to individuals. Nevertheless, they would maintain that the variation is heavily constrained, and that at the very least the ‘core triangle’ of belief, desire and action is universal. And they have assembled an impressive array of arguments and experimental results to back up their position (cf. Kusch 1999, Part II, for a review). To be sure, my own views are closer to Barnes than to this consensus academicus but I feel that Barnes has not done enough to make our side compelling. Barnes writes that ‘for all that it appears to refer to the internal states of individuals, voluntaristic discourse is actually the vehicle of human sociability’ (2000: 2). The determined opponent might still insist that the ‘for all that’ and ‘actually’ are misplaced: voluntaristic discourse refers to the internal states of individuals and is the vehicle of human sociability. Although I do not wish for a moment to defend the social theories under attack in Barnes’s book – indeed his criticism of them is an argumentative gem – I want to at least briefly comment on the alleged incompatibility of libertarianism and
144
Knowledge as Social Order
prediction. Barnes seems to share the view of his opponents according to which actions or choices that are free in a libertarian sense could not be explained or predicted. The idea seems to be that prediction and explanation presuppose causal determination. But this is not the case. There are third options between determinism and randomness, and libertarians have been quick to seize upon them. Let me only mention one such libertarian proposal here, a proposal that has the advantage of clarity and detail (Ekstrom 2000). Laura Ekstrom argues that a decision is free (in the libertarian sense) if, and only if, it was caused by a particular range of deliberative events: Why did the free agent decide in that way? Because of reasons x, y, z, and so on. Why did those reasons lead him to decide as he did? The determinist would answer: Because of a deterministic causal law linking such reasons to such a decision. But the proposed account answers: Because the agent exercised his evaluative faculty in a particular way. Why? For reasons that inclined but did not necessitate a particular outcome to his deliberation process. These causal statements report necessary, but not strictly sufficient, conditions for a decisively formed preference. In order to be explicable, the decisively formed preference need not be necessitated. But in order to be free, the decisively formed preference must not be necessitated. It may be tempting to suppose that a preference cannot be explained if it might have been otherwise, given all of the considerations of the agent and holding fixed the natural laws. But this supposition is only … a ‘demon of determinism’. Not all explanation is deterministic (2000: 111).
What holds for explanation applies of course to prediction as well. Prediction need not be deterministic. The occurrence of different considerations can raise the probability of a decision in favour of one preference rather than another, without necessitating that this decision will result (ibid., 114). This, it seems to me, is a position that would enable social theorists to have the cake and cat it too. 8. In section 3 above, I pointed out that Barnes’s own taxonomy of positions does not leave space for a distinction between compatibilism and hard determinism. I voiced the suspicion that Barnes’s position might therefore turn out to belong more, or as much, to the latter than to the former category. Having gone over Barnes’s arguments in some detail, I am now in a position to substantiate my suspicion. Barnes holds that actions are always caused, and always caused deterministically. He also insists that central elements of our voluntaristic discourse are based on what we might call ‘social illusions’: we social agents are blind to the ways in which we are susceptible to the influences of others; we fail to see what the formula ‘could have done otherwise’ really amounts to (it is tantamount to ‘could have been influenced by the deference-emotion system’): we are mistaken in believing that voluntaristic discourse refers to internal states; we are wrong in thinking (at least sometimes) that chosen actions cannot have causes; and we are confused in thinking that actions can ever be explained in terms that are central to our voluntaristic discourse. Barnes commits himself to the last-mentioned view in speaking about the attempts of geneticists and biotechnologists to bring human behaviour into the scope of their theories: ‘… it can be perfectly natural and forgivable for them to consider that their
Barnes on the Freedom of the Will
145
accounts face competition from “voluntaristic explanations”’ (Barnes 2000: 118). The implication is that geneticists and biotechnologists should learn to know better. It is clear that this position is a rather strong form of illusionism. And it is hard to see what remains of freedom of the will – even a compatibilist version – once we adopt this illusionism. What remains is that in our voluntaristic discourse we are often willing to talk and think along compatibilist lines. Is this not enough? Does this not give us a prima facie case compatibilism? I fear it does not. Barnes’s very general illusionism undermines all respect for the actors’ categories and untutored philosophical preferences. One cannot dismiss so much of the agent’s perspective as illusion and yet see it as giving us a prima facie case for compatibilism. 9. In this chapter, I have commented, from a philosophical perspective, on Barnes’s work on freedom of the will. I have done so in an effort to identify ways in which Barnes’s position could be challenged, and where it needs further elaboration and development. In particular I have suggested that it might have been useful if Understanding Agency had employed the standard philosophical taxonomy of possible positions, and if it had been sensitive to philosophical concerns with projectivism, illusionism or error theories, and third options between determinism and randomness. I hasten to add that I regard none of these difficulties as insurmountable, and that I look forward to seeing my critical comments disarmed by further work in this area (perhaps even by myself). The approach taken in this chapter might easily give rise to one mistaken impression of my views. Since I have one-sidedly emphasised ways in which Barnes’s argument could have been strengthened by making contact with the philosophical literature, the reader might be left with the suspicion that I place philosophy above sociology, and that I do not see a need for philosophers to learn from Barnes. Nothing could be further from my mind. And thus I conclude by listing, at least telegraphically, what I regard as the three most important lessons (for philosophers) of Understanding Agency. The first lesson is that it is enormously fruitful to relate the debate over freedom of the will to familiar themes in philosophy of the social sciences in general, and the controversy between individualism and collectivism in particular. To my mind, making this connection potentially transforms many traditional ways of approaching the topic. As Barnes’s book demonstrates effectively, compatibilism has much to gain from an association with collectivism, and libertarianism is weakened if it can be shown to have a natural affinity towards individualism. It is extraordinary that these links have not been salient prior to Barnes’s work. The second lesson is Barnes’s work on the social constitution of statuses. Some philosophers working on freedom of the will have also proposed projectivist options, but none of this work has achieved the sophistication and clarity of Barnes’s suggestions. Projectivism has met with great interest in the philosophy of meaning as well as in meta-ethics. Taking its full measure in the philosophy of free will can do no better than start from a careful consideration of Barnes’s suggestions. The third lesson is a new way of looking at ‘the data’. I have emphasised similarities between Barnes’s empirical investigation of voluntaristic discourse, and philosophers’ reflections on their intuitions. It is no retraction of these earlier
146
Knowledge as Social Order
comments if I conclude by stressing the differences. Philosophers should – wherever possible – follow Barnes’s model and focus not just on their intentions, but investigate the discourse in which these intuitions are rooted and generated. In proceeding in this way philosophers can gain a critical historical perspective upon the very existence of these intuitions. There is no need to leave this source of insight to sociologists. References Barnes, B. (1983), ‘Social Life as Bootstrapped Induction’, Sociology, 17, 524–545. ––– (1988), The Nature of Power (Cambridge: Polity Press). ––– (1995), The Elements of Social Theory (London: UCL Press). ––– (2000), Understanding Agency: Social Theory and Responsible Action (London: Sage). Blackburn, S. (1984), ‘The Individual Strikes Back’, Synthese, 58, 281–301. Double, R. (2002), ‘Metaethics, Metaphilosophy, and Free Will Subjectivism’, in Kane (ed.), 506–528. Ekstrom, L.W. (2000), Free Will: A Philosophical Study (Boulder, Col: Westview). Horwich, P. (1998), Meaning (Oxford: Clarendon Press). Kane, R. (ed.) (2002), Oxford Handbook of Free Will (Oxford: Oxford University Press). Kusch, M. (1999), Psychological Knowledge: A Social History of Philosophy (London: Routledge). Mills, C.W. (l940), ‘Situated Actions and Vocabularies of Motive’, American Sociological Review, 5, 904–913. Rawls, J . (1999), A Theory of Justice (rev. ed., Cambridge, MA: Belknap Press). Scheff, T.J. (1988), ‘Same and Conformity: The Deference-Emotion System’, American Sociological Review, 53, 395–406. Williams, B. (1981), Moral Luck (Cambridge: Cambridge University Press).
Chapter 9
Agency, Responsibility and Structure: Understanding the Migration of Asylum Seekers to Ireland Steven Loyal
From the late 1970s, debates in the burgeoning field of social theory centred on a putative tension between structure and agency. Despite considerable intellectual investment, the theoretical vacillation between these positions resulted in little by way of progress – at least if progress in this context would indicate an emerging consensus around a range of new concepts and methodological procedures. The abstract dualism that so excited the collective imagination of the theoretically inclined social scientists remained unresolved, leaving protagonists to move into more fruitful or fashionable areas. One of the few noteworthy contributions from within ‘social theory’ has come from Barry Barnes whose account of agency depicts human beings as fundamentally social creatures who use the concept of agency to confer responsibility upon one another. This paper looks at questions of how agency and structure have been incorporated in various theories of migration by using the arrival of asylum seekers into Ireland as a case study. Its central purpose is to show to what extent Barnes’s discussion of agency can help us to account for and understand such processes. Migration Reflecting the structure-agency debate more generally, theories of migration have variously emphasized individual and structural reasons for migration. It is useful to distinguish three major approaches to the study of migration (Wood 1982; Massey et al.,1993; Castles and Miller 1998). i)
Microeconomic /Rational Choice Theory Approaches
ii)
Historical-Structural Approaches
iii)
Social Network/Household Strategy Approaches
Perhaps the most influential approach to migration was instituted by E.G. Ravenstein in The Laws of Migration ([1885] 1976). Listing ‘[b]ad or oppressive laws, heavy taxation, an unattractive climate, uncongenial social surroundings, and even
148
Knowledge as Social Order
compulsion (slave trade, transportation)’, Ravenstein argues there are many reasons for migration, but, ‘none of these currents can compare in volume with that which arises from the desire inherent in most men [sic] to “better” themselves in material respects’ ([1885] 1976: 286). Migration involves a relocation of human beings initially within and later between countries, and is motivated by economic needs. More recently, Ravenstein’s model has been reworked through a rational-choice model which takes as its point of departure the independent individual, rationally weighing the costs and benefits of leaving one area for another, in order to maximize his or her satisfaction. This utility is typically measured against a set of fixed economic preferences usually measured by wage differentials. In some rationalchoice approaches, a push-pull framework has also been introduced within which it is argued that the individual chooses to migrate. Push factors include demographic growth, low living standards, lack of economic opportunity and political repression, while pull factors include demand for labour, availability of land, good economic opportunities and political freedoms. In diametric contrast historical-structural approaches have placed greater emphasis on macro causes for migration. Microeconomic approaches, they argue, lend an unwarranted veneer of freedom to the analysis of the agent’s reasons to migrate, when in reality the migrant has no real alternative but to migrate. Drawing on Marxist political economy and world systems theory, a link is made between immigration and the structural economic requirements of modern industrial economies, and an emphasis placed on social forces which constrain individual action. Like the rational-choice model it criticizes, this framework also looks to ‘push and pull’ factors to explain migration however, these are no longer seen as the context or background within which the individual chooses to migrate, but the very cause of migration itself. The causes of migration are primarily related to structural factors – the macro needs of the capitalist system for labour power: the ‘movement of workers [is] propelled by the dynamics of transnational capitalist economy, which simultaneously determines both the “push” and the “pull”’ (Zolberg 1989: 407). Historical-structuralist approaches have in turn been criticized for paying scant attention to the socioeconomic factors that motivate individual actors. Consequently, there remains ‘a conceptual discontinuity between the units of analysis (systems of production and associated classes) and that which is being explained (the movement of people)’ (Wood 1982: 306–307). A third approach, which in part arose in order to remedy the deficiencies of the microeconomic and structural standpoints focuses on social networks and household strategies (Wood 1982; Massey et al. 1987; 1993; 1994). As Massey argues, migration is a cumulative process whose operation is governed by six principles: that migration originates historically in transformations of social and economic structures in sending and receiving societies; that once begun, migrant networks form to support migration on a mass basis; that as international migration becomes widely accessible, families make it part of their survival strategies and use it during phases of life-cycle when dependence is greatest; that individual motivations, household strategies, and community structures are altered by migration in ways that make further migration more likely; that even among temporary migrants, there is an inevitable process of settlement abroad; and that among settlers, there is a process of return migration (Massey et al. 1987: 6–7).
Agency, Responsibility and Structure
149
In addition to these three major explanations for the causes of migration there have also been attempts to provide broad, overarching categorizations of various types of migration. In essence, these aim to distinguish chosen from forced migration by focusing on the will (power) of the migrant. Thus, for example, Petersen talks of five broad classes which he designates as primitive, forced, impelled, free and mass migration: In primitive migrations the activating agent is ecological pressure, in forced migration it is the state or some functionally equivalent social institution. It is useful to divide this class into impelled migration, when migrants retain some power to decide whether or not to leave, and forced migration, when they do not have this power. Often the boundary between the two, the point at which the choice becomes nominal, may be difficult to set. Analytically, however, the distinction is clear-cut, and historically it is often so (1958: 258).
Refugees and Asylum Seekers Reflecting the changing patterns of global migration which emerged particularly after the mid-1970s and again following the collapse of the Soviet Union, there has been a stronger focus on refugees and asylum seekers as a specific modality of migration. This raises new questions concerning how migration should be understood. According to Zolberg, refugees in the developing world arise as a by-product of two major historical processes: the formation of new states and confrontations over the social order in both old and new states. These, he argues, ‘are analytically distinguishable, but often combined in reality to generate complex and extremely violent conflicts’ (1989: 416). Most discussions of refugee movement have of necessity taken the juridical definition of refugees outlined by the United Nations in 1951 as their point of departure. This defines a refugee as: persons who are outside their country because of a well founded fear of persecution for reasons of race, religion, nationality, membership of a particular social group or political opinion, is outside the country of his nationality and is unable, or owing to such fear, unwilling to avail himself of the protection of that country.
Kunz expands on this definition: With a different past and with motivations at variance with those affecting voluntary migrants, the refugee moves from his homeland to the country of his settlement against his will. He [sic] is a distinct social type … Mass flight, individual escape, or any of the varied types of forced departure all may lead to refugee existence and give rise to refugee movements. It is the reluctance to uproot oneself, and the absence of positive original motivations to settle elsewhere, which characterizes all refugee decisions and distinguishes the refugee from the voluntary migrants (1973: 130).
Although jurisprudence and most legal and technical discourse have taken as a first principle the existence of responsible individual agents able to act autonomously by
150
Knowledge as Social Order
virtue of the powers of reason and free, refugee discourse, interestingly, does not. Theoretical approaches which examine the motivation of refugees designate their movement in terms of the very negation of this principle: as ‘forced’, or ‘impelled’, rather than as freely chosen. In fact the label ‘forced migration’ has increasingly become acceptable as a by-word for refugee studies. In refugee discourse ‘push’ and ‘pull’ factors become recalibrated so that ‘push’ factors correspond to causal processes which force the migrant to leave his or her home country, whilst ‘pull’ factors designate criteria which cajole the migrant towards the country of destination. Asylum in Ireland Considerations of why actors migrate are of course not solely confined to academics with their various theories of migration. Other groups and sectors within society, including the State and government, the police and courts, the media, as well as everyday members of society, also employ accounts of why individuals migrate. By examining the different interpretations and reactions to the arrival of a number of asylum seekers to Ireland between 1992 and 2002, we can see how these various explanatory practices were employed. In 1992, Ireland only received 39 applications for asylum. By 1999, this figure had risen to 7,724 and to 11,634 by 2002. In total, there were approximately 50,000 applications for asylum in Ireland between 1992 and 2002. In comparative terms, however, this figure is very low, constituting about 1.5 per cent of all asylum applications made in Europe. The formal procedure for deciding whether or not an asylum seeker gains refugee status in Ireland is determined at the Refugee Applications Centre in Dublin, which is part of the Department of Justice, Equality and Law Reform. Asylum cases are heard by an interviewing officer during a long interview held at the centre. The interview, which usually lasts between four to six hours, is driven by the principle of establishing contradictions and minor inconsistencies in the asylum seeker’s account. Prima Facie, the mechanisms and procedure for determining whether an individual should be granted refugee status seem straightforward. The interviewer investigates the social, political and economic situation of the country of origin of the asylum seeker. If there appear to be no grounds for identifying any political conflict, or believing the asylum seeker’s story of why he or she left their country of origin, then the applicant is refused refugee status and implicitly classed as an economic migrant seeking to escape poverty. In reality, however, the task is more complex. Firstly, many of the countries from which asylum seekers arrive are characterized by both widespread poverty economic underdevelopment, but also by acute political instability and democratic deficits. Secondly, the UN category of persecution includes within its definition both those who are being persecuted for belonging to particular social groups based on race, religion etc. and those holding certain political opinions. The applicability of the latter criterion is for obvious reasons difficult for the investigating officer to ascertain. The categorization of the asylum seeker as a refugee or an economic migrant is predicated on the possibility of distinguishing political persecution from other causes of migration. ‘Expert’
Agency, Responsibility and Structure
151
government bureaucrats are expected to decide the ‘correct’ classification of an asylum seeker by attributing various motives to him or her. To complicate matters in contrast to most other EU states, where officers examining asylum seekers have a legal background or university degree most immigration officers in Ireland are retired Gardaí or former civil servants who generally have little knowledge, understanding or experience in the field of asylum claims or knowledge of the applicant’s country of origin. Finally, there is of course a crucial question of asylum law or the norms it stipulates, and how it is interpreted and applied to specific cases and concrete instances to asylum seekers. Given that norms cannot apply themselves and do not have determinate applications as Wittgenstein (1958) and Garfinkel (1967) have both shown, and Barnes (1995) has written incisively on, the role of the collective interaction in specifying what the norm ‘really implies’ is of some interest though will not be expanded upon here. Nevertheless, these decisions are made and these have important implications for the asylum seeker. Despite the existence of these formal mechanisms for adjudicating asylum claims, the growing number of asylum seekers who came to Ireland during the 1990s met with increasing racism and hostility. In a number of newspapers, asylum seekers were variously portrayed as ‘illegal’,1 ‘criminal’, ‘drug-pushers’, ‘bogus’, ‘fraudsters’, ‘rapists’ and presented as a collective drain on social and economic resources, or as responsible for the housing crisis, continuing unemployment and the lack of social facilities in Ireland (Pollack 1999; Loyal and Staunton 2001). The central claim made by the media was that these migrants were falsely posing as asylum seekers forced to leave their country of origin when in fact they were ‘really’ ‘economic migrants’ who had entered Ireland in order to exploit its comparatively generous welfare system. This binary categorisation of ‘genuine’ refugees on the one hand and ‘economic migrants’ or ‘bogus’ refugees on the other also structured political discourses. Statements from the Irish Government that the Irish welfare regime was attracting ‘economic migrants’ became commonplace. John O’Donohue, the former Minister of Justice, Equality and Law Reform, in a speech to the Irish Business and Employers Confederation stated: In the early years of this decade and prior to that, our relatively high unemployment rates and low social welfare payments ensured that illegal immigrants invoking the asylum convention targeted the more prosperous countries – even small ones like Denmark and Finland. Let us be clear about it. Our current economic boom is making us a target (1999).
Similarly Noel O’Flynn, a TD for Cork, announced in more populist rhetoric: We’re against the spongers, the freeloaders, the people screwing the system. Too many are coming to Ireland and too many to Cork in my view … I’m saying we will have to close the doors. The majority of them are here for economic reasons and they are thumbing
1 Though it should be noted that, strictly speaking the term ‘illegal asylum seeker’, as used by the media, is a non sequitur since all individuals under international law are legally entitled to apply for asylum.
Knowledge as Social Order
152
their noses up at Irish hospitality and demanding everything under the guise of the Geneva Convention while the taxpayer is paying for it all (Irish Times, 29 January 2002).
According to these government and media discourses, asylum seekers who arrived in Ireland did not do so because of political persecution – which ‘pushed’ or caused them to leave their home country – but for economic gain, principally to exploit Ireland’s comparatively generous welfare regime. They were in effect ‘pulled’. In this voluntaristic discourse, as autonomous individuals, these migrants were responsible for their actions, in so far as they could have acted otherwise. Similarly, many Irish citizens condemned asylum seekers for choosing to migrate to Ireland for economic gain and such views were expressed in various opinion polls. By contrast with these negative discourses, various Non Governmental Organizations (NGOs) including the Irish Refugee Council, Amnesty International and the Irish Council for Civil Liberties accounted for the motivation of these asylum seekers in an overwhelmingly causal discourse. Drawing on the Geneva Convention, the asylum seeker’s actions were causally explained in terms of political persecution. Asylum seekers, it was argued, were neither ‘economic migrants’ nor ‘bogus’ but ‘genuine asylum seekers’ who had come to Ireland because they were forced to do so. Ireland, as a signatory to the UNHCR, was obligated to confer refugee status open them. Understanding Agency As Barnes notes in Understanding Agency (2000), the notion of choice has been central to most sociological frameworks. Rational choice theory, for example, takes the autonomous, independent, calculating individual with fixed preference schedules as its starting point. Equally, other schools of social theory, influenced by the work of Kant, have postulated an agent capable of utilizing his or her internal capacities to act autonomously and with the power of reason. For Kant, an individual agent equipped with reason, can always ‘act otherwise’. Kant connects this belief with the advocacy of a marked separation between the human and the physical world. Similarly, Parsons, Bhaskar, Habermas and Giddens all hold that for an individual to possess agency is for her to possess internal powers and capacities, which, when exercised, make her an active entity. Thus Giddens, for example, defines the agent as someone who can ‘always do otherwise’ (1977). These thinkers all subscribe to the view that there is a fundamental difference between the human and the natural world. Agency becomes framed within a dualistic vision in which the power of individual reason is distinguished from any naturalistic, biological or structural mechanism, which would otherwise determine or overwhelm it. Humans are regarded as irreducibly active individuals whose actions cannot be fixed or determined by structural factors. Moreover, although all of these thinkers had at one point acknowledged and insisted on the importance of social interaction and social practices in reality an individualistic rationalism tacitly permeates in their work. This residual and persistent individualism, however, means that their accounts of agency – and on a broader basis, social order, the development of knowledge, norms and rules and moral action – remain fundamentally flawed. Individualistic theories, for example, cannot account
Agency, Responsibility and Structure
153
for the creation, transmission and use of knowledge since this presupposes both communal trust and authority and social recognition. Nor as Wittgenstein (1958) and Garfinkel (1967) have both demonstrated, can it provide a feasible explanation of rule-following. But what explains the reluctance of many theorists to abandon the individual as the bedrock for their sociological analysis? For Barnes, there are two related explanatory factors. The first relates to the moral and political considerations of the thinkers involved. The second, that social theorists have unwittingly imported the concept of ‘agency’ from everyday discourse where its essentialism and individualism plays a crucial performative function in social interaction. That is, they have conflated the ‘analytical’ with the ‘folk’ concept of agency. In everyday usage the concept of agency and choice display two consistent features in their usage. Firstly, they are only applicable to human beings and not to animals or physical processes. Secondly, although human actions can be described within a causal idiom, choice is generally believed to exist where causation is absent; chosen actions are contrasted with those that are compelled or caused. Here, the agent responsible for something must be its cause and yet must not be acting as the effect of some prior cause; that is he or she is the uncaused cause of his or her intended actions. Moreover, in everyday life the discourses of agency and causation are both tied to the notion of responsibility. Agents who act voluntarily are said to be responsible for their actions. Conversely, those who act under (causal) constraint are seen to be exempt from responsibility. Reference to responsibility entails two different concepts – ‘status’ and ‘state’. Status refers to the externally visible system of social institutions and social relationships wherein individual persons are reciprocally accountable to one another. By contrast, ‘state’ refers to how everyday actors regard one another as rational, capable of reasoning and making independent judgments, that is, as having a ‘normal’ mental state. People deploy voluntaristic discourse in their interactions in order to acknowledge each other as having the status of individuals carrying rights and obligations. However, in so doing they reify this socially constructed status into a natural phenomenon. For Barnes, the concept of agency therefore functions inter alia to assign responsibility. This explains the individualistic and essentialist bias inherent in everyday discourse. Moreover, that members count something as having a status is what constitutes it having a status, so that status classifications have self-referring and self-validating properties. Equally, when individuals treat one another as autonomous, they ‘become’ autonomous. Barnes argues that an agent’s action can be made intelligible in terms of two fundamental institutions: the institution of responsible action whereby, as an uncaused cause, the individual is responsible for the consequences of his or her action, or in terms of the institution of causal connection, whereby he or she is not responsible. No empirical fact of the matter, however, allows us to discern which institution is the correct one to be employed. Attributions about what is going on inside an individual’s head or to invisible states will always remain problematic. Drawing on Durkheim, Barnes posits a fundamentally social account of humans. Humans, he argues, are interdependent social creatures who affect each
154
Knowledge as Social Order
other profoundly through their interaction. Here, he employs Goffman’s analysis of face, in which participants in social interactions have an overriding concern with maintaining face or their standing in the eyes of others and modulate their actions with reference to signals of acceptance, recognition and deference. Scheff regards this concern with status and recognition as part of a bio-social deference-emotion system (1988). In addition, a basic competence that social individuals expect of each other in these ongoing interactions is, as Garfinkel argues, accountability – of individuals being able to give an intelligible account of their actions to others (1967). This double determination, that humans are both mutually susceptible to the evaluations of others, and simultaneously mutually accountable, forms the basis for Barnes’s account. This, however, leaves us with a paradox. On the one hand it has been argued that responsible agents referred to in everyday discourse understand one another as autonomous individuals possessing free will. Yet, Barnes also claims that responsible individuals are mutually accountable and susceptible, and that this can be explained in a causal idiom. In order to resolve this paradox, Barnes argues that we should see the imputation of rationality in everyday discourse as a collective misrecognition of members’ accountability. Individual rationality should not be interpreted as ‘reason’ or ‘ratio’, but as making an action ‘intelligible’ or ‘reasonable’: rationality is accountability reified into the concept of an internal individual power. In turn, attributions which refer to free or chosen actions should be seen as denoting the susceptibility of humans, which has again been reified. Here, the characterization of an action as chosen identifies it as a kind of action that is open to modification, through reliance upon the deference-emotion system and symbolic communication. Following this re-description, the empirical antecedents of actions which are relevant to their recognition as chosen by members become whatever indicates or implies that an action was performed by a moral, susceptible agent and was modifiable through symbolic communication. New instances of actions may be classified on the basis of empirical analogy with existing or past instances of actions. The institutions of responsible action and causal connection are not, however, simply alternative sense-making systems for the ex post facto rationalization of action, but function pragmatically to increase members’ knowledge of when to orient towards others in order to modify their actions. That is, they generate knowledge of when agents may be treated as occupants of social statuses sensitive to the expectations directed towards them, capable of acknowledging responsibilities and exercising rights. From a deterministic point of view, the conventional contrast between chosen and caused actions can then be rendered as a contrast of two classes of caused actions. The contrast does not, however, concern the presence or absence of causation but the degree of resistance of a caused process to variation. A deterministic view will acknowledge that an action which is ongoing because of the operation of a set of constantly operating causes might nonetheless be modified if an additional cause is brought to bear upon it. Having discussed how we should understand agency – primarily as a concept used by social actors in their everyday life to allot statuses and confer responsibility upon one another – Barnes also provides a compressed account of social structure by extending his discussion of social statuses. What, he asks, has prompted social
Agency, Responsibility and Structure
155
theorists to talk of social structure or social system? There is, he argues, a strong intuitive sense that social structure refers to something external, pressing upon individuals and forcing them to act as they otherwise would not. However, he argues, the referents in question are not external to the collective at all. On the contrary, social structure is constituted as a distribution of self-referring knowledge carried by the relevant collective of responsible agents. Although hypostatized into an external object, social structure is constituted and wholly internal to a collectivity of social individuals. Institutional order exists as members’ knowledge of it and it is constituted by that knowledge. This almost Hegelian view of social structure is, however, difficult to identify and analyze for two reasons. Firstly, this knowledge is not separate from the members that carry it. Like knowledge of the rules which constitute an essential aspect of it, social structure is self-referring knowledge, sustained by a collective with nothing external to point to. Secondly, knowledge exists as a delocalized phenomenon, and this makes it difficult to grasp. To see how social structure is manifested, it is therefore necessary to look to the distribution of knowledge of the collective as a whole. Thus, each member makes such a small contribution to the order that s/he can overlook this contribution and treat the order as pure externality. Rethinking Asylum Seekers How helpful, then, is Barnes’ work on agency in helping us to explain both how migration should be conceptualized and the controversy surrounding the arrival of asylum seekers to Ireland? It has, I think, numerous virtues. I will begin by outlining its strengths relative to existing theories of migration and the problem of agency and structure. Barnes is right to point to the excessive individualism and rationalism which characterizes much sociology, despite the claim by many thinkers to have incorporated social actors and social practices within their analyses. He is equally correct to argue that this excessive dependence on individualism means that these approaches cannot adequately account for the production and the reproduction of the social order. Such criticisms are no less relevant for assessing the aporias characterizing most theories of migration. Despite their different presuppositions, frameworks and levels of analysis, contemporary theories of migration share similar theoretical and empirical limitations. In his theory of agency Barnes also correctly reasserts the importance of social statuses as vital foci for the location of responsibilities and for assigning praise and blame in sociology. And it is with a modified Parsonian notion of social status that he also defines social structure. The latter is conceived as a distribution of self-referring knowledge carried by the relevant collective of responsible agents, such that the institutional order exists as members’ knowledge of it. This conception of social structure as consisting of members’ knowledge offers an interesting, though not fully developed, solution to the problem of relating agency to structure. It correctly argues that the constraint acting upon human beings is not some inexorable, opaque force,
156
Knowledge as Social Order
but is composed of other human beings and their interdependence, an argument he shares with Elias ([1939] 2000). Moreover, Barnes is right to be reluctant to use the conceptual vocabulary of agency and structure. This conceptual couplet has misled rather than facilitated any theoretical development. It could in fact be argued that the failure to undertake a sociogenesis of the question of agency and structure in sociological analysis has perhaps been the greatest hindrance to resolving the debate (Loyal 2003). The very questions the debate poses and its social conditions of possibility have scarcely been interrogated. The task of sociology for Barnes is instead to understand human beings as responsible agents engaged within their environment through the mediation of their inheritances of knowledge within an uncompromisingly monistic and naturalistic framework. Importantly, by orienting his approach towards human beings naturalistically and seeking to understand how humans behave, Barnes shows how causal naturalistic and biological accounts of action are perfectly compatible with members’ own accounts of each other as free agents, and he does so, moreover, on an empirical basis underlain by certain epistemological presuppositions concerning the social nature of knowledge. Here the social order is simply regarded as an extension of the natural order. Barnes shows how genetic explanations are fully compatible with social explanations. To an extent his remarks on agency, like Wittgenstein’s, are ‘really remarks on the natural history of human beings’ (1958: 415). Historical-structuralist theories have downplayed the role of social actors in migration. Conversely, microeconomic approaches have overemphasized the importance of agency and decision making in their approach. Neither has developed an adequate conceptual framework for studying migration because they are framed according to political and metaphysical concerns, rather than scientific or empirical criteria. There is an urgent need to reconceptualize and understand migration by emphasizing the fundamentally social nature of humans. That is, there is a need for accounts of migration which place a greater emphasis on social processes and which look to family, peer and larger group interactions, for example, in order to explain why people migrate. There have, of course, been theories of migration which have looked to the household context or to social networks as important frames within which to interpret the decision to migrate (Bach and Schrami 1982; Wallerstein 1979; Massey et al. 1987; Stark and Bloom 1985; Portes and Borocz 1989), but these have continued to include a rigidly rational and economic content within their categories. Although there has been some recognition of the importance of status competition and distinction as a factor in explaining macro processes of migration (Piore 1979), an in-depth analysis of micro interactions has been, on the whole, absent. The one exception, and perhaps the most systematic attempt to emphasize the social nature of humans and their pursuit of strategies of honor as an integral part of the migration process, is found in the work of Abdelmalek Sayad (2004). In addition, questions concerning asylum seekers, like those concerning migration more generally, have invariably been politically saturated and have taken on a Manichean quality in which an all powerful state fights against a powerless individual. It is therefore even more important in analyses of such phenomena that we understand the migrants’ actions naturalistically, monistically and empirically.
Agency, Responsibility and Structure
157
Part of this involves avoiding the confusion of ‘folk’ and ‘analytical concepts’ of agency, something that has been prevalent in studies of ‘race’ and ‘identity’ (Banton 1979; Brubaker and Cooper 2000). What Bourdieu calls ‘practical categories’ play a crucial performative role in everyday life and sociologists should be analyzing the genesis of these categories rather than adopting them uncritically as tools for analysis. That is, sociologists should examine how actions are explained by social individuals in terms of agency rather than simply explaining them in terms of agency. A further important merit of Barnes’ analysis of agency and choice is that it usefully de-psychologizes the analysis of action and its motivation and takes us away from a reductionist and metaphysical subjectivism. Following Wittgenstein, Garfinkel and Mills, who each independently adumbrated the fundamentally social nature of knowledge and social order, rather than attempting to interpret what the deeper motives of actors, or in this case asylum seekers, ‘really’ are or explaining their motives in terms of ‘private’ mental states, sociologists need to study these processes empirically by approaching conduct socially and from the outside. As Mills argues in his paper on situated actions and vocabularies of motives: ‘[r]ather than fixed elements “in” an individual, motives are the terms with which interpretation of conduct by social actors proceeds. This imputation and avowal of motives by actors are social phenomena to be explained. The differing reasons men give for their actions are not themselves without reasons’ (1940: 904). Motive imputation itself needs to be studied. To paraphrase Bourdieu, we can say that attributions classify and they classify the classifier (1984: 6). Moreover, motive attribution depends upon the existence of socially accepted vocabularies of motives which vary historically, but also across cultures and across situations within given cultures. Hence, if we look at the US historically, religious motives have gradually become replaced by ‘individualistic, sexual, hedonistic, and pecuniary vocabularies of motive’ (Mills 1940: 910). Similarly, Freudian and Marxist theories of imputation which posit unconscious drives and economic motives to explain action, themselves need to be explained. The structured distribution of alternative explanatory concepts in terms of their specific carriers and their historical genesis both need to be unearthed. If, as Barnes argues, an agent’s actions can be made intelligible in terms of two fundamental institutions – the institution of responsible action and causal connection – then so it is that asylum seekers have had their actions interpreted according to these two competing institutions both by the government bureaucrats and everyday members. In terms of the institution of responsible action the migrant is seen as having chosen to leave his or her country of origin, that is, having exercised his or her ‘agency’. With reference to the institution of causal connection, the migrant can be seen to have been forced to leave his or her country of origin or compelled by structural forces. Both of these discourses are integrally tied to the notion of responsibility and cannot be understood independently of it, since the attribution of refugee status is designed to provide for a claim to some entitlement, that is, it has important situational consequences. The decision to migrate is therefore of normative interest and it is invariably interpreted to gauge whether praise or blame should be attributed to it. It is an inherently moral and contentious matter and subject to evaluative inquiry. If a
158
Knowledge as Social Order
migrant is seen to have freely chosen to come to Ireland for his/her own economic gain, then s/he merits the opprobrium or negative treatment which is associated with the status of an illegal ‘economic migrant’. By characterizing an action within the institution of responsible action, various members attempt to respectify the ‘asylum seeker’ as an ‘economic migrant’, which is in effect to accord him or her a deviant status. Moreover, we can argue that these statuses are not simply givens, somehow seen to exist independently of specific communities of knowledgeable actors. They are, as Barnes rightly notes, elements of an institutional order constituted as a self-referring distribution of knowledge. Whether an individual migrant acquires the status of an ‘asylum seeker’, ‘economic migrant’, etc., is constituted by the surrounding ring of belief sustained by interacting individuals that treat him or her as an asylum seeker, economic migrant, refugee etc. Although the issue of the classification of people and how this affects the people classified is of immense importance, it will only be briefly touched on here. It is because of these morally charged accusations of deviancy and the implications they have for the treatment of migrants that asylum seekers sometimes employ ‘defence-accounts’, devices which are ‘designed specifically to avoid or reduce blame allocations’ (Scott and Lyman 1970: 138–139). These are techniques aimed at neutralizing attributions of blame. Thus in response to accusations of ‘really’ being economic migrants, many asylum seekers respond by accounting for their actions in terms of a lack of choice and by implication, lack of responsibility: Actually, I didn’t decide, I didn’t have any choice but to travel and go to a safe country and here I am. I didn’t choose Ireland because in the situation that was, at the time, anyone couldn’t have any choice to go anywhere … nobody helped me, like any organization to say ‘look we are taking you to a safe country, so which country are you choosing?’ I just left. You don’t know which … just go and go and be wherever you end.
The grounds for whether the institution of casual connection or voluntary action is used to account for migration and the corresponding inferences made by many of the groups seem opaque. Since the act of migration accedes to either account indifferently, it could be argued that the actual account given is purely a matter of what members of the recipient country are disposed to give it. This assumes that accounts of action as caused or chosen invariably have less to do with the action itself and more to do with those who account for it in one way or another. As Mills notes, the sociology of knowledge can usefully provide an account of how various theories of motive arise and which group carries which theory, or makes which imputation. Hence, imputations and attributions made by various groups within Irish society, both sympathetic and unsympathetic to immigration, have often been deployed on a purely expedient basis tied to moral-political interests and concerns. Various politicians and various newspapers have portrayed asylum seekers in terms of the institution of responsible action and have interpreted their actions within a framework emphasizing choice and agency, in order to impugn blame. By contrast, various NGOs working on refugee issues, as well as the asylum seekers themselves, have emphasized the institution of causal connection in order to exempt asylum seekers of
Agency, Responsibility and Structure
159
blame and responsibility. Everyday members appear to disagree on which institution is most apposite for characterizing the act of migration, largely in accordance with their position in social space. A number of individuals in the lower fractions of the working class and those living in socially deprived areas, have tended to employ the institution of responsible action, regarding asylum seekers as ‘spongers’ arriving to exploit Ireland’s welfare system. As one asylum seeker noted: In inner city areas these people are thinking that refugees and asylum seekers are their competitors, or in competition with them. Other people, they don’t talk or act in such a way. They ask millions of questions and they don’t tell you anything about themselves... nothing. [In] Simple curiosity they can be dismissive. They can say ‘you are a refugee, why did you come here? When are you going back, when will you leave Ireland?’
How are the negative reactions, including xenophobia and racism, by many native Irish citizens towards migrants to be understood? It could be argued that within the context of an increasingly individualized society, and especially following the onset of the Celtic Tiger during the 1990s, the dominant vocabulary of motives in Irish society has been framed within a discourse of economic and pecuniary criteria in which people are increasingly judged to have acted instrumentally. And migrants are simply inserted within this framework. Or it could be argued that the negative reactions from many indigenous Irish citizens towards incoming migrants can also be seen as processes aimed at social closure and as means for status groups to maintain a monopoly of access to social and economic resources (Weber 1971). For Barnes, however, as we noted above, attribution is not simply an ex post facto activity governed by moral-political interests, but has an empirical criterion that everyday agents use to distinguish chosen from caused actions. Underneath these political and social interests, there is a basis for this attempt to render actions as chosen or caused which takes its inspiration from the universal propensity in all cultures to assign rights. How accurate is this view with reference to explaining the treatment of migrants? Although there may be some basis for it, I think there is a need to qualify what we can call Barnes’ positive account. This follows from the two strains within the sociology of knowledge that he draws upon: between an analysis which draws on Durkheim and emphasizes the ‘cult of the individual’ and the ideology of individualism, on the one hand, and a standpoint derived from Mannheim which emphasizes group political interests, on the other. Thus, Barnes’ argument seems to imply that only social theorists have employed a moral-political or normative framework when accounting for action and that, on the whole, everyday members use a less evaluative empirical/pragmatic criterion. However, the above example illustrates that the practical grammars of everyday actors are equally evaluative moving from status to state. This, of course, is an illegitimate move within the context of assigning responsibility since: ‘we are supposed first to identify the state of free agency and only then assign the status of responsibility’ (Barnes 2000: 1). It is, nevertheless, the dominant characteristic. Hence, although Barnes is right to claim that the susceptibility of individuals is the basis for assigning agency and responsibility on some broad ontological basis, substantively and on occasion these processes, as the case of asylum seekers in Ireland shows, tend to be
160
Knowledge as Social Order
reconfigured or reprocessed through group interests and various worldviews. Thus, although the everyday interpretation of agency may have an empirical criterion as its basis and be tied to the assignment of responsibility, in the case of asylum seekers, this usage seems to be qualified by members drawing on dominant political and media discourses. To an extent this is not surprising, since the state, government and media provide powerful interpretative frameworks which permit people to think about social issues in definite ways (Hartmann and Husband 1972; van Dijk 1987). The state and media are important in shaping social categorization and classification since not all classificatory schemes are equally likely to gain social recognition or become institutionalized, codified and legally sanctioned. Given its dominant position in social space, and the field of power, the state with its ‘monopoly of the legitimate use of physical and symbolic violence over a given territory’ (Bourdieu 1998: 40) and the media, have greater power to impose their system of classifications and discourse. However, it could be argued in response to this qualification that processes of moral-political interpretation or moving from status to state may be particularly salient especially when discussing issues concerning the attribution of deviant statuses as is the case with asylum seekers (Berard 1998). In addition, it could be argued that interest group accounts such as those advocated by Mills are more concerned with the retrospective character of accounts than with their prospective character. This qualification to Barnes discussion of imputation and attribution does not, however, compromise the extraordinary importance and compelling nature of his overall account. It merely points to the need for its occasional supplementation. Conclusion Barnes’s work has undergone various shifts during its development, though there are also some significant continuities. His recent discussion of agency and determinism in fact builds upon his earlier discussion of the same issue outlined in his first book, Scientific Knowledge and Sociological Theory (1974). Other factors which have remained consistent throughout his writings have been the range and type of influences on his work, and relatedly, his overall thought-style. In many ways his work, like that of Wittgenstein, Durkheim, Garfinkel and Kuhn, whose writings he draws upon, can be read in terms of what Mannheim calls a ‘conservative thought style’ (1986). Such a thought-style emerged in reaction to the excessive individualism and rationalism of the Enlightenment (Mannheim 1986; Barnes 1994). It emphasized the importance of the social nature of humans which, rather than being conceived as a weakness of the species – as it is for liberals who valorize the sovereignty and autonomy of the individual – it constitutes the very basis of the social order and social life itself. This of course implicitly parallels and augments Durkheim’s famous dictum concerning the ‘precontractual basis of contract’. Yet Barnes manages to bypass the failures inherent in more radical versions of this conservative thought-style; especially their strident emphasis on particularity and a tendency to share the same evaluative oppositions which characterize the natural law thought style they oppose. In that sense his work transcends the false oppositions which have become inherent and
Agency, Responsibility and Structure
161
permanent features of most sociological thinking since the Enlightenment and the reaction to it. Instead of dualism, or promulgating the subjective or objective, Barnes emphasizes monism and intersubjectivity; instead of postulating abstract individuals, he emphasizes concrete sociality; instead of contrasting the natural and the cultural worlds, he emphasizes both their continuity and discontinuity. In an age pervaded by a thoroughgoing individualism and its concomitant ethics, a consistent and logical reference to the social nature of human beings in sociology has been especially difficult to achieve. This is evident in the residual individualism characterizing most social theory. By contrast, the individual does not simply move in and out of various social statuses: reference to an individual is reference to a status which has been assigned. It is perhaps in light of these intellectual obstacles that we should see Barnes’s profound contribution to social theory; that is, an extension and development of the legacy of two of the greatest sociological thinkers who consistently and with steadfast rigour and clarity, emphasized the importance of the modality of the social – Durkheim and Goffman. And it is in relation to their work, with its emphasis on the interaction order, that he takes us beyond the false opposition between agency and structure. References Bach, R. and Scharmi, L. (1982), ‘Migration, Crisis and Theoretical Conflict’, International Migration Review, 16(2), 320–341. Banton, M. (1979), ‘Analytical and Folk Concepts of Race and Ethnicity’, Ethnic and Racial Studies, 2(2), 127–138. Barnes, B. (1974), Scientific Knowledge and Sociological Theory (London: Routledge and Kegan Paul). ––– (1988), The Nature of Power (Cambridge: Polity). ––– (1994), ‘Cultural Change: The Thought-Styles of Mannheim and Kuhn’, Common Culture, 3, 65–78. ––– (1995), Elements of Social Theory (London: UCL Press). ––– (2000), Understanding Agency: Social Theory and Responsible Action (London: Sage). Berard, T. (1998), ‘Attributions and Avowals of Motive in the Study of Deviance: Resource or Topic?’, Journal for the Theory of Social Behaviour, 28(2), 193–213. Bourdieu, P. (1984), Distinction: A Social Critique of the Judgment of Taste (London: Routledge and Kegan). ––– (1998), Practical Reason (Cambridge: Polity Press). Brubaker, R. and Cooper, F. (2000), ‘Beyond Identity’, Theory and Society, 29(1), 1–47. Castles, S. and Kosack, G. (1973), Immigrant Workers and Class Structure in Western Europe (London: Oxford University Press). Castles, S. and Miller M.J. (1998), The Age of Migration (Basingstoke: MacMillan). Douglas, J. (1970), Deviance and Respectability: The Social Construction of Moral Meanings (New York: Basic Books).
162
Knowledge as Social Order
Elias, N. ([1933]1983), The Court Society (Oxford: Blackwell). ––– ([1939] 2000), The Civilizing Process (Oxford: Blackwell). Garfinkel, H. (1967), Studies in Ethnomethodology (Englewood Cliffs: PrenticeHall). Giddens, A. (1977), Studies in Social and Political Theory (London: Hutchinson). Hartmann, P. and Husband, C. (1972), ‘The Mass Media and Racial Conflict’ in Denis McQuail (ed.). Kiberd, D. (1999), Media in Ireland: the search for ethical journalism (Dublin: Open Air). Kunz, E. (1973), ‘The Refugee in Flight: Kinetic Models and Forms of Displacement’, International Migration Review, 7, 125–146. Lee, E. (1966), ‘A theory of Migration’, Demography, 3(1), 47–57. Loyal, S. (2003), The Sociology of Anthony Giddens (London: Pluto Press). Loyal, S. and Staunton, C. (2001), ‘The political economy of asylum seekers in Ireland’, Irish Journal of Sociology, 10(2), 33–56. Mannheim, K. (1986), Conservatism (London: Routledge and Kegan Paul). Massey, D. et al. (1987), Return from Aztlan: The Social process of International Migration from Western Mexico (Berkeley: University of California Press). ––– (1993), ‘Theories of International Migration: A Review and Appraisal’, Population and Development Review, 19, 431–465. ––– (1994), ‘An Evaluation of International Migration Theory: The North American Case’, Population and Development Review, 20, 699–751. Mcquail, D. (1972), Sociology of Mass Communications. (Harmondsworth: Penguin). Mills, C.W. (1940), ‘Situated Actions and Vocabularies of Motives’, American Sociological Review, 5, 904–913. Petersen, W. (1958), ‘A General typology of Migration’, American Sociological Review, 23(3), 256–66. Piore, M. (1979), Birds of Passage (Cambridge: Cambridge University Press). Pollack, A. (1999), ‘An Invitation to Racism? Irish daily coverage of the refugee issue’, in Kiberd (ed.). Portes, A. (1997), ‘Immigration Theory for a New Century: Some Problems and Opportunities’, International Migration Review, 31(4), 799–825. Portes, A. and Borocz, J. (1989), ‘Contemporary Immigration: Theoretical Perspectives on Its Determinants and Modes of Incorporation’, International Immigration Review, 23, 606–630. Ravenstein, E. G. ([1883] [1885] 1976), The Laws of Migration (New York: Arno Press). Sayad, A. (2004), The Suffering of the Immigrant (Cambridge: Polity). Scheff, T. (1988), ‘Shame and conformity: the deference-emotion system’, American Sociological Review, 53, 395–406. Scott, M. and Lyman, B. (1970), ‘Account, Deviance and Social Order’, in Douglas (ed.). Stark, O. and Bloom, D. (1985), ‘The New Economics of Labor Migration’, The American Economic Review, 75, 173–178.
Agency, Responsibility and Structure
163
Van Dijk, T. (1987), Communicating Racism: Ethnic Prejudice in Thought and Talk (Newbury Park Calif: Sage). Wallerstein, I. (1979), The Capitalist World-Economy (Cambridge: Cambridge University Press). Weber, M. (1971), The Interpretation of Social Reality (London: Joseph). Wittgenstein, L. (1958), The Philosophical Investigations (Oxford: Blackwell). Wood, C. (1982), ‘Equilibrium and Historical-Structural Perspectives on Migration’, International Migration Review, 16(2), 298–319. Zolberg, A. (1983), ‘The Formation of New States as a Refugee-Generating Process’, Annals of the American Academy of Political and Social Science, 467, 24–38. ––– (1989), ‘The Next Waves: Migration Theory for a Changing World’, International Migration Review, 23, 403–430.
This page intentionally left blank
Chapter 10
Against Maladaptationism: or What’s Wrong with Evolutionary Psychology?1 John Dupré
This chapter is not directly concerned with topics for which Barry Barnes is mainly known, but it is indebted to him nonetheless. The chapter updates criticisms of evolutionary psychology that I have been intermittently engaged in for the last 20 years or so with reference to much more recent work I have been doing on genomics. The latter has been carried out in collaboration with Barry from the start, when we were co-applicants for the ESRC Centre for Genomics in Society (Egenis), University of Exeter grant that set up the centre of which I am now Director, and in which I have been involved in a multidisciplinary study of genomics for the last five years. This collaboration, after considerable delay for which I am solely responsible, will shortly result in a book coauthored by Barry and myself on the sociology and philosophy of genomics (Barnes and Dupré forthcoming). So what is in fact the central topic of the chapter, recent developments in genomics, is the one around which the work of Barry and myself has been most closely interconnected. This chapter also reflects my long-term interest in critical science studies. Here the connection is more diffuse. However the sociology of scientific knowledge provides a fundamental perspective from which to ask awkward questions about what one is really doing in proposing critical analyses of allegedly scientific projects. Like many philosophers I have sometimes been tempted to seek a God’s-Eye view from which to pontificate about what is good and bad about particular scientific projects. However, on more mature reflection I’m quite content to frame my arguments in more naturalistic terms: evolutionary psychology, specifically, fails to meet the standards generally agreed among enthusiasts for Western science, and fails to cohere with scientific views that are much more deeply grounded in our epistemic standards than
1 A version of this chapter was presented at Princeton University as a Decamp lecture, in February 2005. I am grateful for comments on that occasion by Robert Wright, though I fear he will not think I have properly appreciated them. The support of the Economic and Social Research Council (ESRC) is gratefully acknowledged. This work was part of the programme of the ESRC Research Centre for Genomics in Society (Egenis).
166
Knowledge as Social Order
it is. Moreover its pronouncements may very well be bad for us.2 That’s enough to be getting on with, I think. My title may surprise some. One of the common accusations against evolutionary psychology is that it is panadaptationist, or Panglossian, it supposes that everything about us is an adaptation, perfectly shaped for its conditions of life by the all-powerful hand of natural selection. But there is no mystery here, and both criticisms may be correct. The explanation is just that evolutionary psychology, or the variant of it with which I shall be concerned, supposes that we are adapted to the environment of the Stone Age. Thus we are adapted, perhaps Panglossianly adapted, by natural selection, but not to the environment in which we have the misfortune to find ourselves but to one long past. Hence the maladaptation. To take one familiar example, in the Stone Age fat and sugar were rare and excellent sources of energy, so we became adapted to consume them voraciously whenever the chance presented itself. But in a world full of Krispy Kreme donuts and deep-fried Mars bars this trait is highly maladaptive and leads to heart disease, diabetes, and all the other woes of the age of obesity. The brand of Evolutionary Psychology I’m considering today is the programme developed by Leda Cosmides, John Tooby, David Buss and a few others, and popularised with great effect by Stephen Pinker and, with rather more sophistication, by Robert Wright.3 It is not the only version, but it is the most prominent. It can be quickly summarised. It holds that our minds consist of a large set of modules, shaped by natural selection to solve particular problems set in our evolutionary history. These are problems from the period known as the Pleistocene, roughly the one to two million years preceding the emergence of human civilisation, which I shall refer to loosely as the Stone Age. The view that we are adapted to the Stone Age rather than to modern life is the maladaptationism of my title. It is quite surprising that we should be, in this way, systematically maladapted. We are, after all, probably the most successful large organisms in the history of life, and this success has accelerated as the conditions of our existence have diverged ever further from those of the Stone Age. It is surprising, at least, that this should have happened if we are systematically adapted to a quite different environment from the one in which we appear to have thrived so spectacularly. Fortunately, there is no good reason to accept this maladaptationist thesis. It is based on bad biology: an obsolete view of genetics and a dubious and probably unsupportable view of evolution. There is much else wrong with evolutionary psychology, and its errors have been thoroughly documented by myself and others.4 I shan’t discuss here, for instance, the Panglossianism mentioned above, the assumption that any feature of an organism, including the cognitive structure of the human mind is likely to be an 2 This possibility has been confirmed by work carried on in collaboration between Egenis and the Exeter Department of Psychology, showing that sexist assumptions, in particular, are liable to be strengthened by exposure to claims that sexual differences are biologically grounded (Morton et al. 2006). 3 The classic source is Barkow et al. 1992. See also Buss 1999. More popular accounts are Pinker 1997, and Wright 1994. 4 See Dupré 2001.
Against Maladaptationism
167
optimal response to some conditions at some point in evolutionary history, or the even less defensible obverse assumption, that if something would have been a good idea, it almost certainly evolved. I shan’t consider the controversial issue of the modularity of the mind, often described by Cosmides and Tooby as analogous to a Swiss Army knife. And I won’t say anything about the endemic evidential weaknesses of the project, the ways in which evolutionary speculations or conveniently hand-picked animal analogies so often make up for thin and controversial empirical grounding of the claims about what has actually evolved in the human case. Given the widespread criticisms of Evolutionary Psychology and, perhaps more importantly, the fact that there are other perhaps more credible evolutionary approaches to understanding human behaviour, it may well be wondered why I should spend any time on it at all. So let me offer some broader context. I think quite generally that the ways in which evolution can illuminate biological questions are often misunderstood. Evolution provides one essential kind of explanation of biological phenomena, but its ability to predict or discover phenomena is limited. Attempts to do so generally involve extremely simplistic evolutionary models, and their apparent outputs can be almost entirely traced to these simplifications. The one important exception to this sceptical suggestion is the extent to which evolution legitimates comparative biology. Detection of homology, the common evolutionary origin of a feature, can provide defeasible but valuable clues about function. Despite the hype about the human genome project, it has been well understood from the start that the most interesting information that might come from genome sequencing technology was comparative, the ability to detect similar and evolutionarily related genomic elements in different biological contexts. An extraordinary difficulty for any form of evolutionary psychology is that there are no relevant species for evolutionary comparison. To the extent that cognitive mechanism evolved, as evolutionary psychologists propose, several million years after the division of the human lineage from that of the chimpanzees, and given that everyone agrees that all contemporary humans belong to one species, this lack is indisputable. Evolutionary Psychology proposes to fill this gap by claiming to know when we evolved our distinctive psychology and what were the environmental conditions that our ancestors faced there, and by offering a priori arguments about what would be the best psychological mechanisms to deal with those conditions. Such arguments then provide epistemological depth to thin and controversial evidence about what humans are actually like. I claim that all of these steps are invalid. We don’t know when our distinctive psychology evolved and much of it is likely to be well adapted to contemporary conditions. We don’t really know a great deal about the conditions of the Pleistocene and even if we did this would provide the most doubtful grounds for inferring anything about our adaptive responses to them. Psychology should be empirical not a priori. As I have indicated, my main focus will be on the first part of the argument. Much of this, however, is an excuse to discuss some remarkable developments in recent biology which should quite generally invite reconsideration of some common broad assumptions about the mechanism of evolution. Complexities emerging from recent molecular biology point to a much wider range of possible evolutionary mechanisms than have been widely recognised. Though in one way this makes evolutionary theory an increasingly rich and exciting field, it also makes attempts
168
Knowledge as Social Order
to infer biological fact from evolutionary theory increasingly risky. And while the plurality and complexity of evolutionary mechanisms greatly increases the resources for evolutionary explanation, it correspondingly decreases the possibilities for evolutionary prediction. In fact, and perhaps for this I should apologise, I won’t have a lot to say at all directly about evolutionary psychology. I do hope, however, to show what an increasingly implausible project it is becoming in the light of recent biology. As evolutionary thinking begins to catch up with the revolution in molecular biology, the decades old evolutionary theory on which Evolutionary Psychology has been built can now be seen to be of largely antiquarian interest. Why are we thought to be adapted to the conditions of the Stone Age? Let me quote Leda Cosmides and John Tooby: Our species spent over 99 per cent of its evolutionary history as hunter-gatherers in Pleistocene [Stone Age] environments. Human psychological mechanisms should be adapted to those environments, not necessarily to the twentieth-century industrialised world. The rapid technological advances of the last several thousand years have created many situations, both important and unimportant, that would have been uncommon (or nonexistent) in Pleistocene conditions. Evolutionary theorists ought not to be surprised when evolutionarily unprecedented environmental inputs yield maladaptive behaviour (Cosmides and Tooby 1987: 280–281).
Here we see not only the explicit maladaptationism but also the implicit panadaptationism: ‘human psychological mechanisms should be adapted to those environments’. Why should we not expect that psychological mechanisms have adapted to more contemporary conditions? It is not enough to say merely that our ancestors have spent more time in the Stone Age. Our ancestors have also spent a great deal more time as single-celled organisms. But this does not show that we are adapted to life in the primordial slime. The answer widely assumed by Evolutionary Psychologists is that there has not been enough time since the Stone Age for us to have adapted significantly to more recent conditions (and, of course, that there was enough time for our early human ancestors to adapt to the conditions they encountered, whatever those were). So how much time is enough? How fast is evolution? It is still commonly held, and underlies this part of the Evolutionary Psychologists’ argument, that evolution consists in change in gene frequency. The whole story goes something like this. Psychological adaptation amounts to the existence of neurological structures in the brain. These structures are built by genes. The necessary genes are acquired by random mutation of existing genetic material and selection of advantageous mutations. Since a random mutation is almost certain to be disastrous unless its consequence is fairly similar to that of the unmutated state, each mutation is assumed to provide only a small change. A series of these small changes, each of which will take a substantial number of generations to reach fixation in the population, can eventually produce complex adapted structures. Richard Dawkins (1996) gives a celebrated illustration of this way of thinking in his discussion of the evolution of the eye in The Blind Watchmaker. Provided we can think of 1,000 or 10,000 steps between no eye and fully functional eye, geological
Against Maladaptationism
169
time is long enough for each of these steps to have appeared by chance mutation and spread to fixation through the population. Dawkins in fact seems to think that this development is almost inevitable, but we need only assume that it is possible. I’m not at all sure whether, if this picture is right, a million or two years is long enough for the evolution of the human mind. Our ancestors two million years ago had brains about one third of the size of our present brains, so it is reasonable to assume, as Evolutionary Psychologists generally do, that important contemporary human neurological structures evolved in those two million years. For evolutionary psychology this amounts to the generation of genetically determined neurological structures, mutable only by thousands of generations of genetic trial and error. One crucial idea behind this argument, then, is that adaptive traits are carried over by genes from the periods in which they evolved. And the random mutation and selective retention of genes is a process that requires thousands of generations. So let me begin by saying something about genes. For the reason just noted, genes figure prominently in Evolutionary Psychological writing. Although they reasonably enough protest when accused of holding that genes determine behaviour, they do generally hold that genes determine psychological mechanisms.5 To quote Robert Wright: ‘They boil down to genes, of course (where else could rules for mental development ultimately reside?)’ (1994: 9). So what are genes? It is not sufficiently widely known how difficult this question has become to answer.6 One possible answer goes back to the history of genetics and the Mendelian research programmes, particularly on fruit flies, of Morgan, Müeller, and others. This programme investigated hypothetical factors that were the heritable causes of differences between organisms. It became clear that these causes had something to do with chromosomes, and experiments on linkage, correlations between inherited traits, enabled the mapping of these factors as quasi-spatially related. When Crick and Watson famously published the chemical structure of DNA it was natural to suppose that these hypothetical factors could finally be identified with concrete material objects, parts of chromosomes or, that is, sequences of DNA molecules. This conclusion has turned out to be highly problematic, however. Certainly the phenotypic differences studied by classical geneticists could generally be identified with differences in the DNA sequence somewhere in the genome. Alternative bits of sequence with identifiable phenotypic effects are referred to as alleles so that, for example, we can talk about alleles for blue eyes or brown eyes which, more or less, follow the familiar Mendelian laws. However we should not assume that these alleles are readily identifiable objects. We can see this by looking at what may be the most important upshot of allelic selection, elimination of genomic errors. Medical genetics, because it is concerned precisely with harmful genomic errors, retains a strong connection with the tradition of classical genetics. But, to take one 5 This point seems more equivocal in some more recent work by Cosmides and Tooby, including some remarks in their contributions to Barkow et al. (1992). But without it it is quite unclear how they can maintain the Stone Age adaptation story that is at the core of their programme. 6 For discussion of the problems see Moss 2003; various papers in Beurton et al. 2000; and Dupré 2005.
170
Knowledge as Social Order
of the best known genetic diseases, cystic fibrosis, there is no well-defined object referred to by the expression ‘gene for cystic fibrosis’. Cystic fibrosis is caused by a dysfunction in a protein that controls ion transfer across cell membranes. Over 100 mutations have been identified in the genomic region that codes for this protein, with different mutations determining varying severity of symptoms in cystic fibrosis patients. The gene for cystic fibrosis, then, is a set of errors. Though in this case it would be correct to say that any of these mutations causes cystic fibrosis, it would be highly misleading to describe the unmutated sequence as a gene for not having cystic fibrosis. As a matter of fact a very similar story can be told about the gene for blue eyes. Again, this is not a piece of DNA that somehow produces the blueness of eyes, but any of a range of errors in the DNA sequence that subverts the production of brown pigment in the eyes. And again, though there are therefore genes that cause blue eyes, it is at best misleading to think of the functional alleles at blue eye loci as causes of brown eyes. The complexity of causal paths from bits of DNA to features of organisms makes the project of correlating things of these two kinds largely futile. Many different bits of DNA sequence and much else besides are involved in the normal production of a phenotypic trait. We can confidently assert that a bullet in the head was the cause of death, but it is problematic to suggest, except under very unusual circumstances, that the absence of a hole in the head is the cause of someone staying alive. One might say that genes for brown eyes are parasitic on genes for blue eyes: if there were no identifiable effects of mutations in the relevant bit of DNA there would be no classical genes for either blue or brown eyes. And this follows merely from the quite uncontroversial point that the Mendelian concept applies only to differences. There is an irony here with some of the more acrimonious debate around Evolutionary Psychology. Critics of EP have suggested that Evolutionary Psychologists are involved in providing genetic explanations for human differences, between males and females or between homosexuals and heterosexuals, for example, and thereby reifying what are in fact superficial and malleable distinctions. Evolutionary Psychologists retort indignantly that their central concern is with the genetic basis for human universals. But in the only really clear sense of the word ‘gene’ there are no genes for universals. And that by definition, since genes are defined only by the differences they cause. Am I suggesting that there is no genetic basis for normal development? Only, admittedly, in the pedantic sense that there are no well-defined entities, answering to the concept of genes, involved in normal development. But we can then say that there is a genomic basis for development: parts of the genome are crucially important. So why not just call those parts of the genome genes, and stop quibbling? To answer this we need to look a bit more closely at the quite different concept of the gene employed in genomics. When analysts of data from the human genome project report that there are about 23,000 genes therein, this estimate has nothing to do with relations to phenotypic traits. Very roughly speaking, what they mean is a sequence of coding DNA between a signal to start transcribing (that is, generating RNA sequence that may later be translated into amino acid sequence that may become part of a functional protein or enzyme) and a signal to stop transcribing. It is instructive that for several years
Against Maladaptationism
171
there was serious controversy, with wildly different estimates circulating, about this number, though now something like consensus seems to be developing. A closer look at genomic activity makes it easy to see why this was not a simple matter of counting. The number of proteins produced in human cells is a more controversial issue than the number of genes, but estimates start at around 100,000 and range up to around a million. Obviously this indicates that a gene, in the sense used by molecular biologists, can be involved in the production of many proteins. In fact these molecular genes are known usually to consist of alternating segments, known as exons and introns. In the simplest case, after the gene has been transcribed into RNA the introns are edited out and only the exons are translated into a protein. But in many or most cases different sequences of exons are composed by different editing processes, genes are ‘altenatively spliced’, and the same gene may give rise to many different proteins. In some cases products from parts of other genes, even the introns from other genes, are included in the splicing process. And further modifications to proteins occur after the edited RNA product has been translated into a protein. Cases are known in which several hundred different protein products are derived from the same gene. Why does all this complexity matter? In the first place it contributes to dislodging a picture of the genome that still informs a good deal of thinking about evolution, not least human evolution, of the genome as some kind of blueprint or programme for the production of an organism. It begins to suggest, instead, something quite different, a repository of informational resources upon which the cell can draw in making a huge range of functional products. I think this is still misleading, because like the blueprint or programme metaphors it aims to replace, it sounds too static. The genome is located in a complex structure and the various forms that this structure adopts in the life of a cell are important to its functioning and its interactions with other components of the cell. A biologist colleague7 likes to define the genome as ‘a space in which genetic events happen’. But for now the important point is to dislodge the metaphors that somehow suggest that the whole organism is somehow encoded in the genome, the idea that Lenny Moss (2003) has appropriately related to the preformationist tradition in the history of biology. To get beyond this picture we need to look at another bit of biological dogma, the demise of which is perhaps less universally acknowledged, the so-called Central Dogma of molecular biology. This dogma holds that information flows only from DNA to RNA and finally to amino acid sequence, never in the other direction. This may sound like nothing more than a characterisation of the basic steps in the production of functional proteins, but in fact it is widely used to lend support to the preformationist picture of the genome: since information only flows outwards from the DNA, it must all be contained in the DNA. At any rate, interpreted in anything more than the narrowest sense just indicated, it is completely false. In a way it seems obviously false. For what matters to the functioning of a cell is that the right functional products get produced at the right time and in the right place and this is certainly affected by changes in the chemical environment of the cell. Still, if 7
Steve Hughes.
172
Knowledge as Social Order
one held to the view of the genome as a programme, one could think of the cellular environment as something fully controlled by the genome and thereby effecting the appropriate expression of the necessary bits of DNA sequence at that point in the programme. To see that this picture cannot be sustained we need to look at more direct ways in which the central dogma is mistaken. The most important, or at least the best understood, of these is methylation. This process, which has led some to refer to methylation as the fifth base in the DNA code, is a modification of the DNA structure that suppresses the expression of modified bits of sequence. It is a process that occurs throughout the life of a cell, and is certainly one of the crucial determinants of gene expression. It is clear that methylation is not just part of a system by which the genome regulates itself. For example, it has been established that maternal care or its absence affects methylation patterns in the brains of juvenile rats and, in turn, this affects the disposition of the affected rats to provide such care when adult (Weaver et al. 2004). While it was once thought that methylation was removed during the production of gametes, it is becomingly increasingly clear that this is by no means always the case. This leads us to what might well be called the Central Dogma of evolutionary biology, a dogma closely related to the previous Central Dogma, and one that has also become wholly untenable, the assumption that the only thing that is inherited is DNA. And it is essential to note that it is not crucial whether this inheritance is transmitted through the germ plasm: the sequence of events sketched for rats is an entirely feasible mechanism of intergerational inheritance. One of the points of problematising the gene concept is to raise the question what kinds of genomic difference are in fact the important targets of selection. It seems increasingly likely that the importance of selection between alleles has been greatly exaggerated in recent evolutionary theory and it may indeed turn out, as most geneticists believed in the heyday of classical genetics, that this is largely a process of error elimination and not one capable of creating major evolutionary novelty. Getting rid of the idea of genes for traits, too easily interpreted as objects with the specific function of causing those traits, should at least raise doubts about this idea. Perhaps more importantly there is increasing awareness of a much greater range of possibilities for genomic changes that may provide far more promising bases for major evolutionary change. We have seen that a DNA sequence comprising a set of exons and introns may provide the basis for production of a large number of distinct products. Which, if any, of these it produces at a particular point of development and in a particular tissue will depend on a wide variety of factors: chemical modification of the genome, as for example in methylation; structural changes such as greater or lesser condensation of the chromatin; and the chemical species present in that cell at that time. There are parts of the genome capable of initiating cascades of developmental changes, and interestingly, these genetic triggers are generally extremely ancient, found in very distantly related organisms. As I shall explain further in a moment, some of these factors, and not merely those consisting of DNA sequence, can also be passed on to offspring. Genomes themselves evolve in a great diversity of ways. Recombination, the result of random sampling from the genomes of two parents in producing an offspring, is standardly recognised as an important source of variation. But there are
Against Maladaptationism
173
many other processes. Whole chromosomes and even genomes can be duplicated. These duplications are thought to be important in providing redundant genetic material in which large changes of organisation or sequence can occur without loss to the organism of essential functions. Smaller parts of genomes can be duplicated within or across chromosomes by inserting copies of themselves into the genome. Retrotransposons, a very important class of such genomic elements, which constitute a substantial proportion of many genomes, appear to be important in the functional reorganisation of genomes. And, much more commonly than was once supposed, DNA from other organisms can be inserted into genomes. It is interesting to reflect, in the light of some of these facts, on the surprise that is sometimes still expressed at the claim that human genomes are, say, 99.4 per cent identical to those of chimpanzees. One may well wonder what exactly this means, but without worrying about that, we can certainly wonder why we should care. No doubt if the genome were a blueprint this would be quite surprising. If the blueprints for two ships, say, are 99.4 per cent identical (without again worrying exactly what that means) we might expect two pretty similar vessels. But if we were told that they were made of an almost identical set of raw materials, or that they had identical engines, we would have no such preconceptions. The fact is that even if the genomes were 99.9 or 100 per cent identical, nothing much would follow as to the degree of similarity of the organisms of which they were the genomes. It is an interesting and important discovery that parts of genomes are very strongly conserved through very long periods of evolutionary time, and substantial proportions of our genomes are almost identical to parts of the genomes of worms and even bacteria. This does not tempt us to wonder whether we are really rather similar to worms or germs, though it does direct us to look for bits of chemical machinery that we may share with very different creatures. We may here recall some prescient remarks of Francois Jacob in 1977: ‘What distinguishes a butterfly from a lion, a hen from a fly, or a worm from a whale is much less a difference in chemical constituents than in the organization and the distribution of these constituents’ (1977: 1165). In sum, genomic function is a very different matter from genetic sequence and it is genomic function that provides the differences on which natural selection can work. It is likely that creation of entirely new bits of sequence has been greatly overstated as a central process in evolution and that redeployment of existing genomic resources may be much more important in producing large evolutionary changes. At any rate, research in genomics is opening up a wide range of possibilities for thinking about evolutionary change, and we should certainly not be committed to seeing the evolutionary process solely in the terms developed over 50 years ago. Let me now come at the topic from a rather different direction, the philosophical analysis of evolution itself. Much of this has been focused on the so-called units of selection problem, the question what exactly does natural selection select. Thirty years ago it was fairly uncontroversial that the primary objects of selection were individual organisms and perhaps also groups of organisms. Then came Richard Dawkins’s notoriously successful popularisation of the ideas of G.C. Williams, and a great many people were convinced that ultimately the only possible unit of selection was the gene, understood by Dawkins in a broadly Mendelian way as a difference in the DNA sequence that made a difference to the phenotype. The crucial premise for
Knowledge as Social Order
174
this move was the claim that only genes were inherited. Whereas organisms invariably perished in an evolutionarily trivial length of time DNA, in Dawkins’s colourfully hyperbolic term, was immortal. The structure of DNA was passed on intact from parent to child and hence that structure was potentially immortal. Indeed, molecular biologists have been able to find large chunks of DNA more or less identical between ourselves and plants or even bacteria, and thus presumably preserved across the aeons of evolutionary time from our distant common ancestors. But, impressive though this point may seem, the view that only DNA is inherited is quite unsustainable. And this is one of the reasons why most philosophers of biology, and some prominent evolutionists, have never been much convinced by the gene selection theory. It is easy to show that Dawkins’s gene selection theory provides an inadequate model of evolutionary processes even if it is conceded that inheritance is solely mediated by DNA, and most philosophers concerned with these issues have accepted a pluralistic answer to the units of selection: selection acts on objects at a range of different scales, including genes, organisms, and very possibly groups of organisms. But we should not concede this view of inheritance. Broadening our understanding of inheritance suggests a much more radical rethinking of the units of selection problem that has been developed under the rubric of Developmental Systems Theory, or DST.8 DST, I shall suggest, provides a context in which we can understand the significance for evolution of the recent advances in genomics. The Central Dogma of evolutionary theory stated that the only transgenerational vehicle of inheritance is the genome. The negative phase of DST provides a fundamental critique of this dogma. In very brief summary, it asks the question whether there is anything unique about DNA that justifies its privileged status in evolutionary models, and offers a negative answer. It has been claimed that DNA is unique in its ability to replicate itself. But DNA requires a range of other structures and substances for replication and, with similar access to other resources, including DNA there is a wide range of structures that successfully replicate themselves in the course of development. The genome has been conceived as a privileged source of information. But it is easy to show that from an informational perspective the status of the genome is symmetrical with other contextual resources through which information is conveyed. Just as the cellular environment provides a channel for conveying information about the genome, the genome provides a channel for conveying information about its environment. Though the issue in not uncontroversial, DST has placed a strong burden of argument on those who wish to show how the genome has a unique status in biological organisation. The positive claim of DST brings us back to the unit of selection. For DST this is the full life cycle of the organism. DST looks at the whole set of resources that are necessary for the reproduction of the life cycles of organisms, and the means by which parent organisms facilitate the availability of these resources for their offspring. This picture, of course, retains the basic Darwinian idea that evolutionary change is driven by the differential success that organisms have in launching, during their own life cycles, life cycles of organisms similar to themselves. By rejecting the picture of evolution as essentially no more than a sequence of 8
The clearest summary of DST I know of is Griffiths and Gray 1994.
Against Maladaptationism
175
gradually changing gene pools, this move makes room for the reintegration of development into evolutionary models. It is plain that the restriction of inheritance to the genome cannot be right in the case of human reproduction. For a modern human in a developed modern society to successfully launch and sustain the life cycle of another modern human in that same society many other resources must be provided: maternal care in infancy, schools, hospitals, and much else. And these resources affect the course of development. Despite debates about the importance of innate underlying cognitive structures, it is impossible to deny that a human developing with access to the full range of such developmental resources will acquire a range of capacities – from reading and writing, to appropriate table manners and locally appropriate dress sense – not available to one denied these resources. Such cultural developmental resources are not unique to humans. Birds must provide nests, termites must construct mounds, and beavers must build dams if they are to be successful in reproducing their respective kinds. No doubt there are greater or lesser innate dispositions displayed in these acts of provision – lesser for birds than for termites, for instance. But often the experience of exploiting the resource will also provide some of the information necessary for reproducing the resource when they become parents. Many species of birds, for instance, learn by imitation the songs necessary for attracting reproductive partners. These developmental resources fully external to the bodies of the reproducing and reproduced life cycles are of obvious importance in human evolution as current human reproduction involves a vast infrastructure of resources that are maintained and improved upon by successive generations. What is less obvious is that the genome is not a unique bearer even of internal heritable information. A consequence of the critical work of DST has been to dismantle the conceptual firewall that some have tried to construct around the genetic to preserve its privileged place in evolutionary models. In reality, the minimum physical material passed on to an organism in reproduction is a single cell. The female egg contains a vast set of chemical materials. Though the production of many of these chemicals depends on genomic resources, the genome contains resources that could in principle produce an unimaginably large set of different chemical environments. The transmission of one particular set is potentially a transgenerational transmission of information of a complexity not incomparable with the transmission of the genome itself. And as I have mentioned, functionally important modifications of the genome itself, such as methylation, are also transmitted to an extent that remains unclear and, indicatively of its pivotal ideological role, highly controversial. An obvious consequence of transmission outside the body is that this sort of inheritance is Lamarckian. By this I refer (with apologies to Lamarck)9 to the inheritance of acquired characteristics. Schools, for instance, allow the acquisition of characteristics that can be transmitted to future generations. A consequence, or ideological function, of the dogma of only genetic inheritance is to emphasise the 9 Lamarck, of course, had a complex position in which this view was embedded and which I don’t want to endorse; and many others, such as Darwin, were fully committed to the inheritance of acquired characteristics.
176
Knowledge as Social Order
intellectual iniquity of Lamarckism. The phenomenon of methylation is one of a range of recent biological insights that threaten to open more fully the Lamarckian Pandora’s Box. This is more controversial terrain than I have so far ventured into, but some recent results are suggestive. I have already mentioned the research on maternal care in rats affecting gene expression in the brain of pups. Rats deprived of maternal care in infancy grow up more fearful and show stronger hormonal response to stress than normally nurtured rats. Closer to home, there are data showing that low birth weight children born during the Dutch famine of 1944–1945 not only had increased susceptibility to various later life illnesses, but passed this susceptibility on to their children, and epigenetic effects such as abnormal methylation patterns provide a plausible explanation. One other well documented case is the effect of social status on the production of dopamine receptors in the brain. Higher ranking Macaque monkeys turn out to be less susceptible to cocaine addiction than monkeys that they socially outranked. This difference was traced to the fact that exposure to lower ranking monkeys, but not to higher ranking ones, effected changes in the expression of genes in the monkey’s brains, specifically to the production of dopamine D2 receptors (Morgan et al. 2002). I don’t know whether these changes are heritable, but certainly mechanisms exist whereby they could turn out to be. Less controversial are effects of the environment that are not directed, but affect the rate of evolutionary change. There is work going back to Barbara McClintock that shows that the activity of retrotransposons, genetic elements that replicate themselves throughout the genome, is increased when plants experience stress. This will tend to cause genomic reorganisation that can provide material for rapid evolutionary change. There is, at any rate, considerable evidence that these elements, which constitute a very substantial proportion of most genomes including ours, have important effects on gene expression, and can have decisive effects in early embryogenesis, which is of course the point at which the largest effects on development can be expected. Perhaps most intriguing of all are the small, non-protein-coding RNAs that are proving to be omnipresent in cells and to have vital, diverse, but very partially understood, functions. They appear able to bind to DNA, inhibiting its expression, they can control the activity of protein coding RNAs, and some can even bind to proteins, altering their behaviour. It would be impossible to begin to describe the intriguing findings that are beginning to emerge from this incipient research field, but it seems clear that it represents an entirely new level of cellular control. Small RNAs also have the ability to move between cells, and may prove to have important communicative functions between tissues. And, of course, a set of these RNA fragments is part of what is transmitted with the maternal ovum in reproduction. A very radical and heretical view of evolution, most forcefully presented by Mary Jane West Eberhard (2003), suggests that adaptation is, in the first instance, a process of organismic response to the environment, facilitated by the developmental plasticity of organisms. Genomic adaptation follows. So, far from selection among genes being the primary force behind adaptation, it is largely a consequence of phenotypic adaptation. Perhaps then we are well adapted to modern life, but our genomes are still catching up.
Against Maladaptationism
177
I have done no more than gesture at some of the extraordinary insights that are currently emerging in molecular biology. Why should we care? Recall the basic argument underlying the Stone Age origin of the human mind. Essentially the mind is a product of the genome. Behaviour, to be sure, responds differentially to environmental circumstances, but the basic structure of the mind is laid down in the genes. This is a thoroughly bottom up picture. Genes, as the dogma has it, produce RNA, which produces proteins. Proteins provide the predetermined neurological structure that then interacts in a determinate way with environmental contingencies. We are not, as the Evolutionary Psychologists insist, exactly programmed to be rapists, but given the right set of stimuli in which our Stone Age minds calculate rape as the best reproductive strategy, rapists we become. And finally, the most fundamentally bottom-up part of the picture is the model of evolution that claims that evolutionary history is, in essence, no more than a sequence of genomes, each slightly modified to improve on its predecessor. I have tried to indicate that this picture is entirely obsolete and unrelated both to contemporary molecular biology and the most plausible understanding of the evolutionary process. Certainly these bottom-up processes are important, but equally important are simultaneous top-down processes. The environment does not just shape the human mind in the uncontroversial sense of filling in gaps in a preexisting structure - speaking English rather than French, or knowing which social rules to monitor for cheats. As shown by the high status monkeys who just say No to drugs, social factors can influence the expression of genes in the brain, and basic brain chemistry. Gross morphology can affect the shape of cells which can affect the chemical functions within cells, a process that has been found to be very significant in early cell differentiation (Folkman and Moscona 1978). DNA produces RNA, but while some of that RNA contributes, in very complex ways, to the coding for protein sequences, other bits feed back on the function of DNA or on the splicing and translation of coding RNAs. Proteins also feed back on the expression of DNA or contribute to the physical structure of DNA, also an essential determinant of gene expression. Hence in reproduction it is not just a set of genes that is passed on to descendants, but an exquisitely complex and dynamic chemical system of which the genome is just one vital interacting part. And to the extent that organisms shape the environment in which their offspring are found either purposefully, as is carried to by far the highest level in our own species, or simply as a by-product of their characteristic behaviours, this will also affect the developmental sequence of chemical environments in the differentiating cell lines. How effective at tracking exogenous changes in the environment such a system will prove to be over evolutionary time is not to be settled by abstract calculations on the trajectories of naked DNA. Certainly there can be maladaptive time lags in evolutionary processes, but these are to be discovered empirically rather than proved a priori. Let me summarize these conclusions by returning to the question of the universality and diversity of human nature. Evolutionary psychologists respond to the accusation that by seeing human nature in the genes they are reifying differences between people, by insisting that their primary concern is with the common genetic inheritance that we have all inherited from our Flintstone ancestors. Still, where
178
Knowledge as Social Order
there are evident differences between people, as for example between homosexuals and heterosexuals, these must be located in the genes, and silly stories are made up about Stone Age homosexual shamans providing for their nephews and nieces. It is true that we are an unusually genetically homogeneous species, so perhaps these differences are not so great. An important exception, of course, is the difference between men and women. Since the genetic difference between a human male and a human female on some common measures exceeds that between a human male and a male chimpanzee it is not surprising that Evolutionary Psychologists have portrayed men and women almost as if they belonged to different species – perhaps even came from different planets. The picture I have sketched is neutral on the uniformity of human nature, in large part because I am sceptical about the usefulness of the very concept of human nature. Of course there is a vast amount of human biology common to the human species, some of which we are just beginning to understand. The problem with the concept of human nature is that it tends to suggest fixity, indeed something like a traditional human essence. It is true that we have mostly learned to reject biological essences as incompatible with evolution but still, in the time scales that matter to us, this may seem a pedantic concern. And indeed, though they certainly admit that we have evolved and are probably evolving still, Evolutionary Psychologists perfectly illustrate this effective essentialism with the claim that human nature is, as far as matters to us, stuck in the Stone Age. I object that there is no reason to suppose that we are stuck in the Stone Age, and indeed that we are very likely quite well adapted to the twenty-first century. And it is possible that we may soon be adapted to something quite different. Is this an assertion of the blank slate view of the human mind so violently denounced by Stephen Pinker? Not at all. Human development is a much more complex process than the crude genetic determinism supposed by Pinker, but its very complexity may make it difficult to change in predictable ways. My point is just that organisms in general, and ourselves in particular, are much more subtle and interesting than the antiquated biological picture I have criticised suggests. In evolutionary time there are many ways in which they may respond to changing environments: partly by changing their genomes, though probably the important changes to the genome amount to the redeployment of existing genomic resources, and perhaps more importantly still by changing environmental factors that elicit new employments of existing resources both genomic and more widely biological. Because for our species many changes in this last category can be purposefully effected, it is possible that significant evolution could have happened very rapidly. And the great differences in all except genomes between ourselves and our nearest relatives clearly point to the conclusion that it has. This evolutionary flexibility is, from a proper perspective, inextricably connected to developmental flexibility. It is not, as the accusation of blank slateism suggests, a trivial matter of producing whatever developmental outcomes we might like to imagine, but nor is the production of new developmental outcomes something ineluctably barred by genetic fate. Human differences, the diversity, however much there may be, in actual human developmental outcomes is our best clue as to the diversity of outcomes that might be achieved with a will and a better understanding of human development.
Against Maladaptationism
179
No doubt there are many kinds of time lags. Much of our genomic machinery is inherited from simple organisms billions of years ago, though it seems to be rather adaptable to new uses. At the opposite extreme rapid social change produces developmental obsolescence. People of my generation are surely less well-adapted to the age of information technology than will be today’s teenagers. And it may be that we have deeply engrained tendencies of behavioural development that stem from exigencies of some part of our evolutionary history. But if so this needs to be empirically demonstrated in detail, not proved by a priori argument. And even if such atavistic defects are demonstrated, there is no reason to suppose that they are somehow immutable. One message of this chapter is a sceptical one. The more we understand of contemporary biology, the more we see how much we don’t know. We still understand very little of the development of the simplest organisms, let alone the most complex. How, as our knowledge of molecular biology and ontogeny develop, these will bear on more refined understandings of the process and tempo of evolution is perhaps even more difficult to discern. The biology underlying Evolutionary Psychology, at least can be confidently rejected as based on assumptions that have unravelled in the last couple of decades. Conclusions drawn from so rickety a base about a matter as important to us as Human Nature should be rejected not only for their epistemological worthlessness, but because groundless guesses in this area can be extremely dangerous. Let me end where I began, with the Krispy Kreme donuts. There can of course be no doubt that biological facts about humans, a part even of human nature, are engaged in the attraction of many of us to fat and sugar. These are, after all, good sources of energy that we are physiologically equipped to exploit. But what is interesting about the case is the diversity of human responses to the omnipresence of these resources. Obesity is not an inevitable response to the overabundance of cheap calories. In fact, and not inexplicably, obesity seems to arise most strongly where overabundance intersects with poverty, that is among poor people in rich countries. Unfortunately such observations do not differentiate between the hypothesis of a fixed psychology responding to varying circumstances and a variable psychology developing in response to varying environments. So how do we choose between these alternatives? Are we genetically programmed fat-guzzlers sucked inexorably towards the donuts, or are we blank slates, haphazardly imprinted with the culture of Mars bars or a healthy bourgeois love of broccoli? Of course we are neither. The way to break down the dichotomy between these equally hopeless alternatives is to begin to appreciate the intricate hierarchy of upward and downward interactions between objects and structures at all levels of the biological hierarchy. In doing so we dispense with the stultifying dogmas I have mentioned in this chapter, and we see the importance of a perspective on evolution that encompasses the diversity of processes susceptible to selective change. And, finally, we can begin to understand the vast changes in human behaviour that have occurred over the last few thousand years without seeing ourselves either as formless lumps of psychoplasm or atavistic relics from the mists of prehistory.
180
Knowledge as Social Order
References Barkow, J.H. et al. (1992), The Adapted Mind: Evolutionary Psychology and the Generation of Culture (Oxford: Oxford University Press). Barnes, B. and Dupré, J. (forthcoming), Genomes and What to Make of Them (Chicago: Chicago University Press). Beurton, P. et al. (eds) (2000), The Concept of the Gene in Development and Evolution (Cambridge: Cambridge University Press). Buss, D.M. (1999), Evolutionary Psychology: The New Science of the Mind (Boston: Allyn and Bacon). Cosmides, L. and Tooby, J. (1987), ‘From Evolution to Behavior: Evolutionary Psychology as the Missing Link’, in Dupré (ed.), 277–306. Dawkins, R. (1996), The Blind Watchmaker (New York: W.W. Norton). Dupré, J. (ed.) (1987), The Latest on the Best: Essays on Evolution and Optimality (Cambridge, MA: MIT Press). ––– (2001), Human Nature and the Limits of Science (Oxford: Clarendon). ––– (2005), ‘Are There Genes?’, in O’Hear (ed.), 193–210. Folkman, J. and Moscona, A. (1978), ‘Role of Cell Shape in Growth Control’, Nature 273(5661), 345–349. Griffiths, P.E., and Gray, R.D. (1994), ‘Developmental Systems and Evolutionary Explanation’, The Journal of Philosophy, 91(6), 277–304. Jacob, F. (1977), ‘Evolution and Tinkering’, Science, 196, 1161–1166. ––– (2001), ‘Complexity and Tinkering’, Annals of the New York Academy of Sciences 929(1), 71–73. Morgan, D. et al. (2002), ‘Social Dominance in Monkeys: Dopamine D2 Receptors and Cocaine Self-administration’, Nature Neuroscience, 5(2),169–174. Morton, T.A. et al. (2006), ‘We Value What Values Us: The appeal of Identityaffirming Science’, Political Psychology, 2006: 2, 823–838. Moss, L. (2003), What Genes Can’t Do (Cambridge, Mass.: MIT Press). O’Hear, A. (ed.) (2005), Philosophy, Biology and Life, Royal Institute of Philosophy Supplements (Cambridge: Cambridge University Press). Pinker, S. (1997), How the Mind Works (New York: W.W. Norton). ––– (2002), The Blank Slate: The Modern Denial of Human Nature (London: Allen Lane). Weaver, I.C. et al. (2004), ‘Epigenetic Programming by Maternal Behavior’, Nature Neuroscience, 7(8), 847–854. West-Eberhard, M.J. (2003), Developmental Plasticity and Evolution (New York: Oxford University Press). Wright, R. (1994), The Moral Animal: Evolutionary Psychology and Everyday Life (New York: Vintage).
Index absolute knowledge 15, 19, 26–7, 30 absolute truth 16, 19–20, 25–6 absolutes 16–20, 28 absolutism 15–19, 25–6, 29–30, 39, see also anti-relativism accountability 10, 60, 66, 154 accounting 99–116 accounts administrators 105–7, 114 adaptation 14, 28, 50, 56, 166, 169n5, 176, see also maladaptationism; panadaptationism aerodynamics 22, 23n23, 25, 27 aerofoil theory 25n26 aeronautics 13–31 agency 147–61 alleles 169–70, 172 anti-realism 19nn, 38, 40n7 anti-relativism 17, 28–30, 36n3, see also absolutism asylum seekers 147–61 auditors 105, 109–11, 113–16 autonomy 140, 143 and migration 149–50, 152–4, 160 and research management 53, 57–8, 60–1, 63, 66, 68 Barnes, B. 1–10 and agency 147, 151–61 and extensional semantics 100–1 and finitism 99–101 and freedom of will 131–46 and genomics 165 and kinds 77 and power 119–29 and realism 40–1, 43–4 and relativism 21n18, 37–8, 40–4 and socialization 49, 52n2, 68, 122 basic research 53, 61, 63, 65n23 belief 1–3, 6–7, 10, 13, 15, 29–30, 38–43, 79–80, 127, 139, 143, 158
Bell 51, 59, 62, 63n20 Bloor, D. 3–4, 13–31, 35, 37–9, 42–4, 114–15, 131 Bogdanov, G. and I. 77–80 Bourdieu, P. 2, 92, 121, 124, 126, 157, 160 bourgeoisie 121, 126 business, see accounting; industrial scientists causal connection 153–4, 157–8 causal determination 135–6, 144 causal determinism 135, 142 causation 137, 153–4 chess 102–3, 109 circulation 22–7 classification 21, 124, 135 accounts 100–1, 103–4, 106–10, 113–16 people 151, 153–4, 157–8, 160 cognitive relativism 15, 17–18, 40 Cold War 35, 52–3, 69–70 collective action 2, 8–10, 139 collectivism 8, 10, 133, 136–8, 145 Collins, H. 9, 37–44, 77–81, 114, 131 compatibilism 132–40, 142–5 computers 88n1, 90–1, 94–5, 103–6 Conant, J.B. 61n14, 67n25 conscience 14, 50 consent 119–21, 123, 127–8 constructive practice 83–4, 89, 91 contingencies 30, 108, 111, 113–14, 177 correspondence theory 19–21, 24, 38, 41 Dawkins, R. 13–14, 18, 25–6, 29, 168–9, 173–4 decision-making 122 and accounting 100–1, 106, 111–13 and industrial scientists 61 and migrants 144, 149, 151, 156–7 deference-emotion system 139, 144, 154 design 26–7, 30n33, 31n34, 90
182
Knowledge as Social Order
determinism 134–6, 139–40, 142, 144–5, 154, 160, 178 deviance 40, 42–3, 121, 126–7, 158, 160 discourse 2, 8, 10, 19n15, 30, 126 everyday 132, 141–2, 153–4 moral 141–2 ordinary, see everyday discourse rational 128 refugee 150–60 voluntaristic 133–4, 137–40, 142–5, 152–3 dissociation, between subject and object 83, 86 DNA 86, 169–74, 176–7 dogma 171–2, 174–5, 177, 179 domination, power as 119–21, 123, 126, 128–9 Douglas, M. 2 Du Pont 58, 61, 63–4 Dupré, J. 165–80 Durkheim, E. 2, 4, 29, 119, 121–2, 131, 153, 159–61 Eastman Kodak 57n10, 60–1, 63, 65 economic migrants 150–2, 158 Edinburgh Strong Programme, see strong programme enterprise resource planning, see ERP system epistemic objects 84, 87–95 epistemic practice 84–8, 91–4 ERP system 105–11, 115 everyday discourse 132, 141–2, 153–4 evolution 166–9, 171, 173–9 evolutionary models 167, 174–5 evolutionary psychology 165–79 evolutionary time 173–4, 177–8 experience 2, 7 and accounting 103, 106–8, 112–14 and immigration 151 and industrial science 55, 56n6, 60, 62, 68 and maladaptationism 175 and objectual practice 86–7, 89n2, 90, 92–3 and relativism 15 extensional semantics 99–102, 113, 116 external world 6, 41, 43–4
false consciousness 121, 123 financial controller 106, 108, 110–11 finitism 7, 21n20, 99–116 forced migration 149–50 Foucault, M. 2, 9, 120–2, 125, 128–9 free agents 136, 139, 144, 156 freedom of will 131–46 Garfinkel, H. 2, 8, 10, 151, 153–4, 157, 160 GE 56n6, 58–9, 61, 63nn, 65 gene expression 172, 176–7 General Motors 60, 63 genes 168–74, 176–8 genomics 165–80 Gramsci, A. 120–1, 126, 129 Habermas, J. 85, 128, 131, 140, 152 habitus 2, 121, 126–7 hard determinism 135–6, 139, 144 Haugaard, M. 119–29 hegemony 120–1, 126, 129 Heidegger, M. 83–4, 86–8 hoaxes 77–81 Hollis, M. 35 horse laugh 36–7 Hounshell, D.A. 58, 61, 64 Hughes, S. 171n7 human nature 177–9 humanists 69–70 ideal fluids 22–4, 27n30, 28 idealism 4, 17–21, 24, 27–8 illegitimate power 119–20, 123, 126, 129 illusionism 135, 145 imagined objects 90 imposition of meaning 124–5 of world creation 126 imputation 15n6, 154, 157–8, 160 incompatibilism 132–5, 137 individualism 10, 63, 133–4, 136–40, 142–3, 145, 152–3, 155, 157, 159–61 industrial scientists 49–70 inheritance 156, 169, 172–5, 177 inner states 79 internalisation 122, 127 internalism 136, 138–41 intuitions 120, 125, 134, 141–3, 145–6 Ireland 147–61
Index Joukowsky, N. 22–6, 27n30 justice 120, 123 Kaplan, N. 68–9 Knorr Cetina, K. 83–95 knowing subjects 17–18 knowledge, see Introductory Note and detailed topics knowledge activities 84, 93 knowledge-centered practice 83, 94 knowledge societies 84–5, 93–4 Küchemann, D. 26–7 Kuhn, T.S. 52, 99, 119, 121–2, 160 Kusch, M. 9n7, 21n20, 131–46 Kutta, W. 22–6, 27n30 Lacan, J. 92–3 learning 14–15, 83, 127 Leermakers, J.A. 57n10, 59 legitimacy 119–29 legitimate power 119–20, 123, 125–9 libertarianism 135–45 life cycles 148, 174–5 logic 1, 5, 78, 114, 134 Loyal, S. 147–61 Lukes, S. 6, 35, 119–21, 129 MacKenzie, D. 7, 38, 99–116 maladaptationism 165–80 Mannheim, K. 159–60 Marxist thought 2–4, 9, 125, 135, 148, 157 Massey, D. 148 materialism 2, 4, 18–19, 25, 40 matter 18nn Mazzotti, M. 1–10 meaning 20–1, 77–81, 83, 91, 99–101, 113–14, 124–8, 141, 145 meaning finitism, see finitism measurement 21n19, 23, 26, 80, 103–11, 137 Mees, C.E.K. 57n10, 59–63, 65–6, 67n26 Mertonian School 4, 9, 50–3, 55, 64n22, 66 methodological relativism 8, 37–9, 43–5 methylation 172, 175–6 migration 147–61 Mills, C.W. 119, 138, 143, 157–8, 160 monism 5–6, 10, 134–5, 156, 161 moral discourse 141–2 moral intuitions 141–2
183
moral luck 132 moral philosophers 141–2 moral relativism 40 morality 141–2 motives 138, 143, 151, 157–9 mutations 168–70 Napoleon/napoleons 122, 124–5 natural selection 166, 173 naturalism 4–5, 14–15, 20, 25, 28–30, 38, 132–5, 152, 165 object relations 84, 92–4 objectivity 3, 5, 29, 81 objects, see epistemic objects objectual practice 83–95 oil, discovery of 44–5 ontological relativism 37, 45 organizations and industrial scientists 49–70 Orth, C. 53, 56n6 panadaptationism 166, 168 partial objects 90–1 physicists 4, 17, 58, 59n12, 77, 79, 89–90 Pinch, T. 35–45 Popper, K.R. 5, 20n16, 28n31 power 119–29 Prandtl, L. 23–6, 28 primary acts 104, 106–7, 109–11, 115–16 projectivism 136–9, 141, 143, 145 psychology, evolutionary 165–79 random mutation 168–9 rational choice theory 119, 123, 128, 137, 147–8, 152 rationality 6, 10, 29–30, 38, 40, 85, 135, 154, see also rational choice theory Ravenstein, E.G. 147–8 RCA 55, 56n7, 65 real world 19n13, 43, 45 realism 4, 19, 24, 36, 40–4 recognition and accounting 108–11, 115–16 reflective equilibrium 120, 125, 141 refugee discourse 150–60 relativism 5–6, 8, 13–31, 35–45, 138 research management, see industrial scientists responsibility 147–61
184
Knowledge as Social Order
responsible action 10, 138, 154, 157–9 responsible agents 142, 154–6 rhetoric 17, 35–7, 59–60, 66–8, 151 rings of reference 122–9 RNA 170–1, 176–7 rule following 99–116 scientific community 70, 77 scientific knowledge, sociology of 3–6, 8, 10, 28, 37–40, 131, 133, 165 scientific objectivity, see objectivity Searle, J. 20, 36, 38 self-referring knowledge 155 Shapin, S. 3–4, 38, 49–70 situated knowledge 37 sociability 9–10, 139–40, 143 social actors 120, 122, 125–9, 132, 154–7 social agents 124–5, 127–8, 144 social knowledge 121–2, 129 social life 7–9, 83–5, 94, 160 social order, see Introductory Note and detailed topics social theory, see Introductory Note and detailed topics socialization 49–53, 56–7, 65, 122, 124, 127 sociology, see Introductory Note and detailed topics Sokal, A. 37, 77–80 speech acts 10, 142
SSK, see scientific knowledge, sociology of; strong programme Steele, L.W. 56n6, 66n24 stock 104, 110–12 Stone Age adaptation 166, 168–9, 177–8 strong programme 3, 6, 13, 15, 19nn, 28, 37–45 structure and agency 147–61 subject-object differentiation 83, 87–8 subjectivity 29, 92–3 susceptibility 10, 36, 139, 144, 154, 159, 176 symbolic violence 124–6, 128, 160 symmetry 6, 37–9, 41–2, 45, 174 torture 127–8 unfolding 89, 91–4 universality 5–6, 26–7, 177 violence 123–6, 128 viscosity 22, 24, 25n26 voluntaristic discourse 133–4, 137–40, 142–5, 152–3 vortices 23–7 Whyte, W.H. 53–4, 56n6, 63 will, freedom of 131–46 Wittgenstein, L. 2, 7, 21, 29, 86, 99, 125, 137, 151, 153, 156–7, 160 world creation 124–6