Philosophy Department
University of Waterloo
Waterloo, Ontario, N2L 3G1
pthagard@watarts.uwaterloo.ca
© Paul Thagard, 1994
To return to the Science and Disease Articles Table of Contents page, click here.
Abstract
Explanations of the growth of scientific knowledge can be characterized in terms of logical, cognitive, and social schemas. But cognitive and social schemas are complementary rather than competitive, and purely social explanations of scientific change are as inadequate as purely cognitive explanations. For example, cognitive explanations of the chemical revolution must be supplemented by and combined with social explanations, and social explanations of the rise of the mechanical world view must be supplemented by and combined with cognitive explanations. Rational appraisal of cognitive and social strategies for improving knowledge should appreciate the interdependence of mind and society.
Many philosophers, historians, psychologists, and sociologists of science are concerned to explain the development of scientific knowledge. But the kinds of explanations they propose are very diverse. Some philosophers of science prefer logical explanations, in which new scientific knowledge derives logically (inductively or deductively) from previous knowledge. Researchers in cognitive science, including psychologists, computer scientists, and some philosophers, propose cognitive explanations, in which the growth of knowledge derives from the mental structures and procedures of scientists. Finally, sociologists of science offer social explanations, in which factors such as the organization and social interests of scientists are used to explain scientific change.
Are these explanations competitive or complementary? Over the past two decades, since sociologists of knowledge staked claims to what had been the traditional philosophical territory of explaining the growth of scientific knowledge, there has been conflict between proponents of logical and social explanations (see, for example: Barnes, 1985; Bloor 1991; Brown 1984, 1989; Collins 1985). In the meantime, cognitive approaches have emerged with explanatory resources much richer than those available within the logical tradition, but the relation between cognitive and social accounts is rarely specified. Some sociologists are intensely antagonistic toward psychological and computational explanations, even going so far as to propose a ten-year moratorium on cognitive explanations of science (Latour and Woolgar 1986, p. 280). In a similar vein, Downes (1993) attacks what he calls "cognitive individualism" and defends the claim that scientific knowledge is socially produced.
But from a naturalistic perceptive, we can appreciate science as a product of individual minds and as a product of complex social organizations. Not only can we see cognitive and social explanations as providing complementary accounts of different aspects of science, we can look for ways of integrating those explanations, bringing then together in a common approach. Thagard (1993) attempted to integrate the cognitive and the social by comparing scientific communities to systems of intelligent computers. This paper pursues the same goal more generally, by comparing cognitive and social explanation schemas and showing how they can be brought together to form integrated explanation schemas. After describing the structure of explanation schemas and outlining some of the targets of cognitive and social explanations, I present typical schemas for explaining belief change. I then show how integrated schemas can be produced that capture what a unified cognitive/social account of science might look like. To illustrate the unification of approaches, I show how a cognitive account of the chemical revolution can be socially enriched, and how a social account of the early development of science and mathematics can be cognitively enriched. The social categories of Downes (1993) require similar enrichment. Finally, I sketch how a cognitive/social approach offers new perspectives on the question of scientific rationality.
Science is far too complex to be amenable to simple deductive or statistical explanations. But scientific developments can be explained by identifying them as fitting into common patterns of change, a style of explanation that is common in both the natural and social sciences. Consider simple Darwinian explanations of why members of a species have a certain trait, for example why some finches have large beaks. Our knowledge of the laws and historical conditions of evolution is far too sparse to provide a derivation of the trait to be explained, but understanding can nevertheless come by means of a schema such as the following:
Explanation target:
Why does a given species have a particular trait?
Explanatory pattern:
The species has a set of variable traits.
The species experiences environmental pressures.
The pressures favor members of the species that have a particular trait.
So members of the species with that trait will survive and reproduce better than members of the species that lack the trait.
So eventually most members of the species will have the trait.
This structure is obviously an oversimplification of explanations used by evolutionary biologists, but it will serve here to show what I mean by an explanation schema, which consists of specification of a target to be explained and a pattern that provides the explanation. The terms presented in boldface are variables that can be filled in by many different examples. The importance of such schemas for explanation has been discussed (with varying terminology) by various philosophers and cognitive scientists (e.g. Kitcher 1981, 1993; Leake 1992; Schank 1986; Thagard 1988, 1992).
What are the explanation targets in science studies? The most straightforward is belief change, as when we ask why eighteenth century chemists adopted Lavoisier's oxygen theory or why twentieth century geologists accepted plate tectonics. The general explanation target concerns why scientists abandoned their previously held beliefs. But there is much more to the development of science than belief change, for we can ask why conceptual changes took place involving the introduction and reorganization of whole conceptual systems. (See Thagard 1992 for an argument that conceptual change goes beyond belief change.)
Another legitimate explanation target in science studies involves discovery. Why did Antoine Lavoisier discover the oxygen theory in the 1770s? Why did Harry Hess develop the theory of seafloor spreading in the 1950s? While such questions are not open to logical explanations, they are grist for the mills of cognitive and social theorists. Similarly, cognitive and social explanations can be given for why scientists pursue particular scientific research programs. Pursuit is an intermediate stage between the initial discovery or proposal of concepts and beliefs and their eventual becoming accepted. Within that stage, there are many interesting questions to be answered, such as why scientists did particular experiments in particular ways. The rest of this paper will focus on schemas for explaining belief change, but it should not be forgotten that understanding science requires attention to other important explanation targets such as conceptual change, discovery, and pursuit.
Why do scientists acquire new beliefs, sometimes abandoning old ones? My goal in this section is not to answer this question, but rather to characterize the kinds of answers that have been given to it by means of logical, cognitive, and social explanation schemas. For all of these schemas, the explanation target is:
Why did a group of scientists adopt a particular set of beliefs?
But very different kinds of explanatory patterns can be used to answer this question.
For philosophers and others operating within the tradition of Frege and Russell, formal logic provides the central model for understanding knowledge, in a way roughly captured by the following schema.
A. Logical Schema
The scientists had a set of previous beliefs.
The scientists employed a logical method.
When applied to the previous beliefs, the logical method implies a set of acquired beliefs.
So the scientists adopted the acquired beliefs.
This schema can be made more specific by filling in the account of logical method, which might include deduction, confirmation theory, or - the currently most sophisticated candidate - Bayesian probability theory . Recent proponents of logical approaches to scientific change include Gärdenfors (1988), Howson and Urbach (1989), and Levi (1991). The logical positivists who originated this approach to understanding science were not concerned with explaining the growth of scientific knowledge so much as with providing a foundation for knowledge, but logical schemas have more recently been aimed at understanding scientific change.
Cognitive science offers a mentalistic explanatory approach that differs strongly from the antipsychologistic tradition of the logical positivists. It postulates that the human mind contains representational structures and processes that operate on them to produce new structures. These structures include sentence-like beliefs, as well as visual images and various kinds of concepts and schemas. Oversimplifying again, we can roughly capture cognitive explanations of belief change in a group of scientists as follows:
B. Cognitive Schema
The scientists had a set of mental representations that included a set of previous beliefs.
The scientists' cognitive mechanisms included a set of mental procedures.
When applied to the mental representations and previous beliefs, the procedures produce a set of acquired beliefs.
So the scientists adopted the acquired beliefs.
This cognitive schema is more general than the logical one, since the representations and procedures that it invokes need not be those found in formal logic. Nonsentential representations such as diagrams, maps, and other visual images may be included among the scientist's mental representations in addition to sentential beliefs. Mental procedures may be totally unlike methods in deductive and inductive logic and probability theory. For example, in my theory of explanatory coherence, beliefs are accepted on the basis of their coherence with other beliefs, where coherence is modeled computationally by means of connectionist algorithms that perform parallel satisfaction of multiple constraints (Thagard 1992). This theory has been applied to account for belief change in all the major scientific revolutions, and it has also been used by psychologists to account for experimental studies of belief change in nonscientists (e.g. Read and Marcus-Newhall 1993; Schank and Ranney 1992). The cognitive schema thus has a constraint that the antipsychologistic logical schema lacks: the representations and procedures postulated must be plausible parts of human psychology. This constraint rules out computationally intractable logical methods such as deductive closure and psychologically implausible methods such as Bayesian updating.
Unlike logical methods, mental procedures can also explain discovery of new concepts and hypotheses and decisions about pursuit of research programs. Moreover, mental procedures can include ones that we would not want to count as rational, such as motivated inference in which conclusions are affected by thinkers' personal goals (Kunda 1990). Thus the cognitive schema competes with the logical schema for providing understanding of science, since by and large the procedures it postulates are very different from logical methods. In principle, cognitive and logical schemas could be compatible, if human belief change were fundamentally driven by logical mechanisms, but there is abundant evidence that human psychology involves a much broader range of structures and processes than logic describes. (For deduction, see Johnson-Laird and Byrne, 1991; for induction, see Holland, Holyoak, Nisbett, and Thagard, 1986). Different cognitive explanations of scientific development have been offered by Churchland (1989), Darden (1991), Giere (1988), and Langley et al. (1987); for a collection of relevant papers, see Giere (1992).
Sociologists of science tend to focus on different features of science than logical methods and mental procedures. They note that because of their social situations scientists have various interests, ranging from personal ambition to national sentiment. The also note that the development of science depends on part on the social connections that control information flow among scientists and the power relations that make some scientists much more influential than others in determining what science will be done. Amalgamating ideas from various sociologists, we can roughly summarize various social explanations for belief change with the following:
C. Social schema:
The scientists had previous beliefs and interests.
The scientists had social connections and power relations.
Previous beliefs and interests and social connections and power relations lead to acquired beliefs.
So, the scientists adopted the acquired beliefs.
This schema is incompatible with the logical schema which assumes that epistemic matters must be kept isolated from psychological and sociological ones. However, it competes with the cognitive schema only if one assumes that the best explanation of the development of science must be either purely cognitive or purely social. But open-minded cognitivists can easily grant that scientists have the interests, social connections, and power relations postulated by sociologists, and moreover grant that these play some role in the development of science. Similarly, open-minded sociologists can grant that psychological structures and processes can mediate socially affected belief changes. The cognitive schema is incomplete because it fails to note how social relations can affect the spread of beliefs through the group of scientists. The social schema is incomplete because it fails to show how individual scientists came to acquire their beliefs.
A full account of the growth of scientific knowledge will therefore have to integrate the features of cognitive and social schemas, as roughly illustrated by the following:
D. Integrated Cognitive/social Schema
The scientists had a set of mental representations that included a set of previous beliefs and a set of interests.
The scientists' cognitive mechanisms included a set of mental procedures.
The scientists had social connections and power relations.
When applied to the mental representations and previous beliefs in the context of social connections and power relations, the procedures produce a set of acquired beliefs.
So the scientists adopted the acquired beliefs.
As with the previous schemas I presented, a great deal of detail must be added to put this explanation schema to work. To fill in the cognitive side, we must specify the mental representations and procedures that operate on them. To fill in the social side, we must specify the social interests, connections, and power relations. Most importantly, to make the integrated cognitive/social explanation succeed, we must provide a much fuller account of how the cognitive and social features of scientists together determine their belief changes. For example, sociological explanations that appeal to the interests of scientists should be able to draw on Kunda's (1990) account of the cognitive mechanisms by which goals affect the selection of evidence. Her experiments show that in general people do not simply believe what they want to believe, but rather that what they want to believe can influence their recall and use of evidence in more subtle ways that influence but do not fully determine their conclusions.
The question of how to make such integrated explanations work cannot be pursued abstractly, since the balance of cognitive and social factors is different in different historical cases. If the explanation target is why T. H.. Huxley accepted Darwin's theory of evolution by natural selection, cognitive factors such as the explanatory coherence of the theory should predominate, although the social relations of the two friends should not be ignored. On the other hand, if the explanation target is why some nineteenth-century American industrialists embraced Social Darwinism, social factors such as the mesh between their economic interests and the idea of survival of the fittest should predominate, although the cognitive mechanisms of motivated inference must not be ignored. Similarly, the explanation for acceptance of hormonal or sociobiological explanations of behavioral sex differences may have to weight social values more heavily than evidence evaluation (Longino 1990, ch. 6). I now want to look in more detail at two important cases of the development of scientific knowledge: the chemical revolution and the development of the mathematical-mechanistic world view.
In previous work, I offered a cognitive account of the chemical revolution in which Lavoisier's oxygen theory of combustion overthrew the phlogiston theory of Stahl (Thagard 1989, 1990, 1992). This account has two parts, a description of the conceptual changes that took place when Lavoisier developed an alternative to the phlogiston scheme and an explanation, in terms of explanatory coherence, of why the oxygen theory was viewed by him to be superior to the phlogiston theory. Both parts are cognitive, in that conceptual schemes are taken to be organized systems of mental representations, and judgments of explanatory coherence are specified as psychologically plausible computational procedures. My account of the chemical revolution thus instantiates the cognitive schema (B) presented above.
I remarked, however, that my account omitted the social side of the chemical revolution and did not presume to tell the whole story (Thagard 1992, p. 113). What would a social explanation of the chemical revolution look like? My aim in what follows is not to provide a full social account of the acceptance of the oxygen theory, but merely to sketch enough that the compatibility and integrability of social and cognitive explanations becomes evident. (Social treatments of the chemical revolution include McCann (1978), Levin (1984), and Perrin (1987, 1988); other useful sources include Guerlac (1961), Conant (1964), Holmes (1985), and Donovan (1988).) From a social perspective, we can look both at the developments of Lavoisier's own beliefs and also at how these beliefs spread to the larger scientific community.
No scientist is an island: Lavoisier had numerous teachers, friends, and associates who contributed to the development of his ideas. We can mention, for example, Guyton de Morveau who demonstrated to Lavoisier in 1772 that metals gain weight when calcined, Joseph Priestley who showed Lavoisier in 1774 his experiments that mercury when heated forms a red "calx", and his wife, Marie, who translated English articles for him, made entries in his notebooks, and drew figures for his publications. Lavoisier was elected young (25) to the French Academy and participated in its meetings. He also had a smaller circle of chemists with whom he could perform experiments and discuss the defects of the phlogiston theory uninhibitedly at a time when senior chemists such as Macquer would not have approved of the aggressive proposal of an alternative theory. Although his most important publications on the oxygen theory were written by himself alone, he had various other joint publications including the influential Methods of Chemical Nomenclature (1787), written with Berthollet and Fourcroy.
Lavoisier's broader social situation also contributed to his work. His substantial income as a tax farmer meant that he had ample resources and time to conduct his experiments (although this position ultimately led to his execution during the French Revolution). According to an early biographer: "His great wealth, his excellent education, his mathematical precision, his general views, and his persevering industry, all contributed to ensure his success" (Thomson 1813, p. 81). Understanding of the differences between the spread of oxygen theory in France and England requires appreciation of the institutional differences between the two countries, which McEvoy summarizes:
The difference between Lavoisier's corporate view of knowledge and Priestley's individualistic epistemology highlights the difference between the institutional organization of French and British science in the late eighteenth century. In the highly organized and centralized community of France, the pressures of formal education, centralized learned societies, employment opportunities, and a competitive system of reward and recognition meant that aspiring French chemist had little choice but to follow the intellectual lead of the academicians in Paris. In contrast, the organization of English science was much weaker, comprising fewer educational institutions, decentralized societies, little employment opportunity, and a looser congregation of amateurs with closer ties to entrepreneurial industry than their French contemporaries. Thus, whereas the highly integrated community of state-subsidized French theoreticians provided fertile ground for the flowering of paradigmatic conformity during the Chemical Revolution, the dissemination of Lavoisier's theory in England met with a more varied resistance. (McEvoy 1988, pp. 210-211).
Thus a full explanation of the development of the oxygen theory should not be limited to conceptual development and belief revision as in my cognitive account. Nevertheless, there is no incompatibility between that account and the relevant social information. No matter how much is said about how Lavoisier gained information from his associates or about how his social situation inclined him to act in certain ways, there remains the problem of describing how his conceptual system developed and changed as he formed and adopted the oxygen theory of combustion, rejecting the phlogiston theory that he had held as a young chemist. As displayed in the integrated Cognitive/social explanation schema (D), cognitive and social explanations of conceptual change can coexist.
Mind and society thus both contributed to the development of the oxygen theory, but they do not tell the whole story either. The experiments of de Morveau, Lavoisier, Priestley and others were a very important part of the development of eighteenth century chemistry: neither mental nor social construction can fully explain why experiments on combustion and calcination gave the results they did. The growth of scientific knowledge is a function of mind, society, and the world. The difficult task for science studies, including naturalistic philosophy of science, is to create a synthetic account of how mind, society, and the world interactively contribute to scientific development.
The social side of the chemical revolution becomes even more prominent if one addresses the question of how other scientists besides Lavoisier came to adopt the oxygen theory. Contrary to the common view that adoption of a revolutionary theory only comes when the proponents of the previous theory die off, the oxygen theory was almost universally adopted in France and (more slowly) in England by scientists who had to abandon their previous phlogiston beliefs. A cognitive explanation of this switch goes roughly like this. Through personal contact with Lavoisier or his disciples, or through reading his argumentative publications, scientists began mentally to acquire the new scientific conceptual scheme. The new mental representations enabled them to understand Lavoisier's claims and to appreciate that the oxygen theory has greater explanatory coherence than the phlogiston theory. This appreciation is part of a cognitive process that led them to accept the oxygen theory, abandoning the phlogiston theory and its conceptual scheme.
From a social perspective, we want to know much more about how information spread from scientist to scientist. Diffusion of the oxygen theory was slow, even in France (Perrin 1988). Members of Lavoisier's immediate circle such as Laplace were fairly quick to adopt his views, but the majority of French chemists came around only in the late 1780s and early 1790s. According to Perrin, nearly all converts initially resisted Lavoisier's theory, but underwent a conversion lasting several years. The duration of conversion has both a social and a cognitive explanation: the cognitive explanation is that developing a new conceptual system and appreciating its superiority to the old one is a very difficult mental operation; the social explanation is that information flow in social networks is far from instantaneous. Lavoisier and his fellow antiphlogistinians worked to improve the flow, by giving lectures and demonstrations, by publishing articles and books, and by starting a new journal, Annales de chimie. It is also possible that different scientists had different interests that made them resistant to the new theories, although I know of no documentation of this. It is certainly true that different scientists had different initial beliefs and cognitive resources. My cognitive account of Lavoisier cannot be automatically transferred over to all the other scientists, since they had different starting points and associated beliefs. (For discussions of cognitive diversity, see Giere (1988), Solomon (1992), and Kitcher (1993).) In principle, we would need a different cognitive account for each scientist, although these accounts would have a great deal in common, since the scientists shared many concepts and beliefs, not to mention similar underlying cognitive processes.
Thus there is much more to a social account of the chemical revolution than was present in my cognitive explanation of Lavoisier, but the expanded social account must coalesce with cognitive descriptions of Lavoisier and all the other scientists whose beliefs and conceptual systems changed.
Despite the antagonism that some sociologists display toward psychology, many sociological explanations of scientific developments can usefully be supplemented by cognitive explanations. As an illustration, consider the sociological account of some essential features of early modern mechanistic thought given by Richard Hadden. His abstract provides a summary (Hadden 1988, p. 255):
A sociological explanation is offered for certain features of the mathematical-mechanistic world view. Relations of commodity production and exchange are seen as providing an analogy of 'abstraction' for such a world view. The mediation between social relations and content of science is provided by commercial reckoners who contributed a new meaning to ancient mathematical concepts and thus paved the way for the notion that all sensually intuitable events are explicable in terms of the motion of qualitatively similar bodies.
The explanation target here is the emergence in the fifteenth and sixteenth century of the view that nature can be understood mechanically and mathematically.
Hadden argues that social relations involving commercial arithmetic provided an analogy for how nature could be understood. "The crux of my argument is that a view of the conditions of the period gets projected onto all of nature and eventually human society as well." (Hadden 1988, p. 257.) Just as in the early modern European economy the sensible properties of commodities such as bread and shoes could be abstracted into exchange values, so the sensible properties of all physical objects could be ignored in favor of their mechanical and mathematical properties. Hadden provides evidence that such developments as the replacement of ancient concepts of number were influenced by commercial concerns. For example, Simon Stevin, who was among the first to introduce the notion of decimal fractions, was very much concerned with practical mathematical problems.
Without evaluating the plausibility of Hadden's Marxian account, we can readily see that it presupposes cognitive processes. His explanation of the emergence of new mathematical ideas assumes that "social relations provided analogies and metaphors which were refined technically by thinkers whose concerns involved, at first, the reckoning up of calculable social relations." (Hadden 1988, p. 271). Thinkers such as Stevin, Hadden conjectures, used commercial social relations as analogs to develop ideas about mathematics and science. Although Hadden's documentation of Stevin's use of analogy is sparse, later uses of social analogies in science have been well established. Darwin, for example, came up with the idea for natural selection by reading Malthus on political economy (Darwin 1958). It has also been conjectured that Lavoisier's innovative concern with conservation of matter may have been influenced by his tax farmer's familiarity with the balance sheet.
Hadden says nothing about how analogical thinking actually works, but this is where cognitive science has much to offer, since the topic has been thoroughly investigated over the past decade using psychological experiments and computational models. The process most relevant to Hadden's account is analogical mapping, in which some of the content of a source analog is transferred to a target analog. In Hadden's case study, the target analog involves the mathematics and physics of objects, and the antecedently understood source analog involves commercial and social objects. According to the theory of mapping of Holyoak and Thagard (1989), people's cognitive processes in mapping from one domain to the other requires simultaneous satisfaction of semantic, structural, and pragmatic constraints. This is not the place to go into detail on cognitive theories of analogical thinking. The key point is that such theories exist, and in fact are presupposed by sociological explanations such as Hadden's that see analogy as the mediating factor between social relations and the development of science. Cognitive theories of analogy are not alternatives to Hadden's account: the social and economic relations he discusses are an important, ineliminable part of the story. Rather, cognitive explanations supplement the social ones by describing the mental processes of the thinkers who made the transition to new ideas.
Latour and Woolgar (1986) pursue their extreme anticognitive stance by speaking only of how scientists use "inscriptions" to produce other inscriptions, as if all that mattered to the process of scientific development were the social relations of scientists and the paper they shuffle around. They clearly miss an important part of what is going on when the cognitive representations and processes of scientists enable them to read what has been written, develop and test new hypotheses, and produce new writings. Like Hadden, Latour and Woolgar can only gain from cognitive models that provide a crucial supplement to their social accounts of what laboratory scientists are doing. As Bloor (1991, p. 168) pointed out in the second edition of one of the books that spawned the sociology of scientific knowledge, sociologists would be "foolish" to deny the need for a background theory about individual cognitive processes.
Downes (1993, p. 452) accuses me and others of cognitive individualism, "the thesis that a sufficient explanation for all cognitive activity will be provided by an account of autonomous individual cognitive agents." Obviously, I do not hold this position, and in fact have given a battery of arguments why psychological reductionism in science studies is bound to fail (Thagard 1993). But the kind of anticognitive view that Downes seems to prefer in alliance with Latour, Woolgar and Collins is also bound to fail. Downes distinguishes three levels of social aspects of science, each of which can be shown to have an essential cognitive component.
The first level is the "public embodiment of scientific theories", which includes the textbooks, research papers, instruments, and other shared property of the scientific community. These clearly exist outside the mental representations of individual scientists, and naturalistic science studies cannot ignore their significance. But part of this significance is cognitive: the use of textbooks, papers and instruments by scientists presupposes their mental capacities to read, write, plan, design, and in other ways produce and use such tools. The public embodiment of scientific knowledge would be pointless if scientists did not have the cognitive processes to understand and produce the embodied objects. Use of external representations such as books and diagrams means that the thought of each scientist does not have to rely entirely on his or her own internal mental representations; but internal representations are needed to comprehend the external ones.
Downes' second level is social interaction, such as is found in complex laboratory work where no one researcher is entirely responsible for the ultimate result. This level is indeed of great importance, as is clear from research in fields like psychology, where most research is collaborative, and experimental physics, where almost all work is collaborative. But the importance of collaboration and social interaction speaks only against the most implausible forms of psychological reductionism and provides no support for purely social accounts. Understanding of how scientists work with each requires in part understanding of how they communicate with each other, which in turn requires cognitive theories of how they represent information and use language and other means such as diagrams to convey information to each other. Level 2 is undeniably social, but it is also undeniably cognitive.
Downes' third social level depends on the claim that the activities of scientists only make sense when taken in the context of a broader scientific community. The difference between someone performing an experiment and someone else doing the same physical motions in a play depends on the fact that the former is part of a community of experimenters. We can grant this social distinction, but cannot help but notice that there are obvious cognitive distinctions too. The mental representations of the trained scientist are drastically different from those of the actor who is merely mouthing lines, since the scientists has absorbed an enormous amount of both declarative and procedural knowledge in the course of training. The ability of the experimenter to plan experiments and interpret the results cannot be explained purely in terms of social context, but must also make reference to mental structures and procedures.
My arguments that Downes' three social levels each have a crucial cognitive aspect are in no way an attempt to explain them psychologically. We can appreciate social aspects of science at each of these levels while simultaneously appreciating relevant cognitive aspects. Figure 1 diagrams four simple models of the relations between psychological and sociological explanations of science. In (a), psychological reduction, only psychological explanations of science are admitted, and any social aspects are also to be explained psychologically. Sociological reduction, (b), is the equally extreme view that science and the psychological have purely social explanations. Model (c), social production, is a slightly less extreme view that simply ignores the psychological in giving social explanations of science. By far the most plausible model is (d), which has the development of science being explained both socially and psychologically, with the relation of the social and the psychological being interpenetration rather than reduction. Sometimes we need the social to help explain the psychological, as when the development of Lavoisier's beliefs is seen as in part the result of his circle of friends. And sometimes we need the psychological to explain the social, as when cognitive processes of analogy thinking are used to help understand how social structures can suggest scientific theories.
*** Insert Figure 1 about here.
The best strategy for naturalistic studies of science is neither psychological reductionism nor sociological reductionism, but an integrated approach that takes both the cognitive and the social seriously. To conclude, I want to argue that such an approach can be normative as well as descriptive.
When the sociology of scientific knowledge arose in the 1970s with its implication of supplanting logical explanation schemas with social ones, philosophers were aghast. Since Frege, philosophers in the analytic tradition viewed incursions of psychology into epistemology as assaults on rationality. Incursions of sociology seemed even worse, especially given the rampant relativism of some sociologists such as Woolgar (1988). However, as epistemology and philosophy of science have come to take psychology more seriously, it has become obvious that psychologism requires new theories of rationality, but need not embrace irrationalism or relativism. For example, Giere (1988), Goldman (1986), Harman (1986), and Thagard (1988, 1992) all use psychology to challenge traditional logic-based conceptions of rationality while opening up new territory for rational appraisal.
Similarly, taking the social context of science seriously does not entail relativism. Goldman (1992, p. 194), Kitcher (1993), and Solomon (in press) have outlined how social practices, like cognitive processes, can be subject to rational appraisal, for example concerning the extent to which they promote reliable beliefs. Logical explanation schemas carry rationality with them for free, since any beliefs that are inferred logically are presumably warranted. With cognitive and social explanations the matter is more complicated. We have to ask first what is the best cognitive and social account of a scientific development, and only then raise the question whether the cognitive and social processes invoked are ones that, from a broader view, promote the ends of science. In pursuit of the first question, philosophers of science can ally themselves with psychologists, sociologists, and historians of science who, lacking an appetite for the second question, may choose to leave concern for rationality in philosophy, its traditional home. But rational appraisal of social practices and organizations has barely begun.
Solomon (in press) has made the audacious proposal that the scientific community should be taken as the important unit of cognitive processing, rather than the individual scientist. She contends that a scientific community may reach a consensus that can be judged to be normatively correct from an empirical perspective, even though not even one individual scientist in the community made an unbiased judgment. While the view that she calls "social empiricism" is a useful antidote to past neglect of social aspects of rationality, it swings too far in that direction. My Integrated Cognitive/Social Schema in section 3 allows various cognitive and motivational biases to influence the judgments of scientists. But if these biases are as dominant as Solomon suggests, it becomes mysterious how the community collectively reaches a consensus based on empirical success rather than communal delusion. On the other hand, if scientists share cognitive processes such as those postulated by my theory of explanatory coherence (Thagard 1992), then their convergence on the empirically successful theory despite their disparate individual biases becomes intelligible. Individual evaluations of the merits of competing theories are not all there is to rationality, but they are an indispensable part of it.
A key conclusion to draw from the interdependence of cognitive and social explanations of scientific change is that appraisal of cognitive and social strategies must also be linked. Cognitive appraisal should take into account the fact that much scientific knowledge is collaborative, so that we should evaluate particular cognitive strategies in part on the basis of how well they promote collaboration. Conversely, social appraisal should take into account the cognitive capacities and limitations of the individuals whose interaction produces knowledge. Determining how to facilitate the growth of scientific knowledge, like the more descriptive task of explaining this development, depends on appreciating the complex interdependencies of mind and society.
* Thanks to Paul Rusnock, Cameron Shelley, Miriam Solomon, and Jim van Evra for comments on an earlier draft, and to Kathleen Gorman for research assistance. This work is supported by a grant from the Social Sciences and Humanities Research Council of Canada.
Requests for reprints should be sent to Paul Thagard, Philosophy Department, University of Waterloo, Waterloo, Ontario, N2L 3G1. Electronic mail: pthagard@watarts.uwaterloo.ca.
The cognitive literature on analogy is vast. Holyoak and Thagard (in press) provide a comprehensive introduction. On mapping, relevant works besides our own include Gentner (1983), and Falkenhainer, Forbus, and Gentner (1989). Thagard, Holyoak, Gochfeld, and Nelson (1990) discuss retrieval of analogs. Research on analogy has been very interdisciplinary, as psychological experiments and theoretical ideas have given rise to computational models that have suggested new psychological experiments.
Barnes, B. (1985), About science. Oxford: Blackwell.
Bloor, D. (1991), Knowledge and social imagery (second ed.). Chicago: University of Chicago Press.
Brown, J. R. (ed.), (1984), Scientific rationality: The sociological turn. Dordrecht: Reidel.
Brown, J. R. (1989), The rational and the social. London: Routledge.
Churchland, P. (1989), A neurocomputational perspective. Cambridge, MA: MIT Press.
Collins, H. (1985), Changing order: Replication and induction in scientific practice. London: Sage Publications.
Conant, J. (1964), Harvard case histories in experimental science. Cambridge, MA: Harvard University Press.
Darden, L. (1991), Theory change in science: Strategies from Mendelian genetics. Oxford: Oxford University Press.
Darwin, C. (1958), The autobiography of Charles Darwin and selected letters. New York: Dover.
Donovan, A. (1988), "The chemical revolution: essays in reinterpretation", Osiris 4(second series).
Downes, S. M. (1993), "Socializing naturalized philosophy of science", Philosophy of Science 60: 452-468.
Falkenhainer, B., Forbus, K. D., & Gentner, D. (1989), "The structure-mapping engine: Algorithms and examples", Artificial Intelligence 41: 1-63.
Gärdenfors, P. (1988), Knowledge in flux. Cambridge, MA: MIT Press/Bradford Books.
Gentner, D. (1983), "Structure-mapping: A theoretical framework for analogy", Cognitive Science 7: 155-170.
Giere, R. (1988), Explaining science: A cognitive approach. Chicago: University of Chicago Press.
Giere, R. (ed.), (1992), Cognitive models of science. Minneapolis: University of Minnesota Press.
Goldman, A. (1986), Epistemology and cognition. Cambridge, MA: Harvard University Press.
Goldman, A. I. (1992), Liaisons: Philosophy meets the cognitive and social sciences. Cambridge, MA: MIT Press.
Guerlac, H. (1961), Lavoisier - The crucial year. Ithaca, NY: Cornell University Press.
Hadden, R. W. (1988), "Social relations and the content of early modern science", British Journal of Sociology 39: 255-280.
Harman, G. (1986), Change in view: Principles of reasoning. Cambridge, MA: MIT Press/Bradford Books.
Holland, J. H., Holyoak, K. J., Nisbett, R. E., & Thagard, P. R. (1986), Induction: Processes of inference, learning, and discovery. Cambridge, MA: MIT Press/Bradford Books.
Holmes, F. (1985), Lavoisier and the chemistry of life. Madison, WI: University of Wisconsin Press.
Holyoak, K. J., & Thagard, P. (1989), "Analogical mapping by constraint satisfaction", Cognitive Science 13: 295-355.
Holyoak, K. J., & Thagard, P. (in press), Mental leaps: Analogy in creative thought. Cambridge, MA: MIT Press/Bradford Books.
Howson, C., & Urbach, P. (1989), Scientific reasoning: The Bayesian tradition. Lasalle, IL: Open Court.
Johnson-Laird, P. N., & Byrne, R. M. (1991), Deduction. Hillsdale, NJ: Lawrence Erlbaum Associates.
Kitcher, P. (1981), "Explanatory unification", Philosophy of Science 48: 507-531.
Kitcher, P. (1993), The advancement of science. Oxford: Oxford University Press.
Kunda, Z. (1990), "The case for motivated inference", Psychological Bulletin 108: 480-498.
Langley, P., Simon, H., Bradshaw, G., & Zytkow, J. (1987), Scientific discovery. Cambridge, MA: MIT Press/Bradford Books.
Latour, B., & Woolgar, S. (1986), Laboratory life: The construction of scientific facts. Princeton, N.J.: Princeton University Press.
Leake, D. B. (1992), Evaluating explanations: A content theory. Hillsdale, NJ: Erlbaum.
Levi, I. (1991), The fixation of belief and its undoing. Cambridge: Cambridge University Press.
Levin, A. (1984), "Venel, Lavoiser, Fourcroy, Cabanis, and the idea of scientific revolution", History of Science 22: 303-320.
Longino, H. (1990), Science as social knowledge: Values and objectivity in scientific inquiry. Princeton: Princeton University Press.
McCann, H. (1978), Chemistry transformed: The paradigmatic shift from phlogiston to oxygen. Norwood, NJ: Ablex.
McEvoy, J. G. (1088), "Continuity and discontinuity in the Chemical Revolution", Osiris 4: 195-213.
Perrin, C. E. (1987), "Revolution or reform: The chemical revolution and eighteenth century concepts of scientific change", History of Science 25: 395-423.
Perrin, C. E. (1988), "The chemical revolution: Shifts in guiding assumptions". In A. Donovan, L. Laudan, & R. Laudan (eds.), Scrutinizing science: Empirical studies of scientific change. Dordrecht: Kluwer, pp. 105-124.
Read, S., & Marcus-Newhall, A. (1993), "The role of explanatory coherence in the construction of social explanations", Journal of Personality and Social Psychology 65: 429-447.
Schank, P., & Ranney, M. (1992), "Assessing explanatory coherence: A new method for integrating verbal data with models of on-line belief revision". In Proceedings of the Fourteenth Annual Conference of the Cognitive Science Society. Hillsdale, NJ: Erlbaum, pp. 599-604.
Schank, R. C. (1986), Explanation patterns: Understanding mechanically and creatively. Hillsdale, NJ: Erlbaum.
Solomon, M. (1992), "Scientific rationality and human reasoning", Philosophy of Science 59: 439-455.
Solomon, M. (1994). "Social empiricism," Nous 28: 325-342
Thagard, P. (1988), Computational philosophy of science. Cambridge, MA: MIT Press/Bradford Books.
Thagard, P. (1989), "Explanatory coherence", Behavioral and Brain Sciences 12 : 435-467.
Thagard, P. (1990), "The conceptual structure of the chemical revolution", Philosophy of Science 57: 183-209.
Thagard, P. (1992), Conceptual revolutions. Princeton: Princeton University Press.
Thagard, P. (1993), "Societies of minds: Science as distributed computing", Studies in History and Philosophy of Science 24: 49-67.
Thagard, P., Holyoak, K., Nelson, G., & Gochfeld, D. (1990), "Analog retrieval by constraint satisfaction", Artificial Intelligence 46: 259-310.
Thomson, T. (1813), "Biographical account of M. Lavoisier", Annals of Philosophy 2: 81-92.
Woolgar, S. (1988), Science: The very idea. Chichester, Sussex: Ellis Horwood.