J. van Brakel (1994) Cognitive Scientism of Science. Psycoloquy: 5(07) Scientific Cognition (3)

Volume: 5 (next, prev) Issue: 07 (next, prev) Article: 3 (next prev first) Alternate versions: ASCII Summary
PSYCOLOQUY (ISSN 1055-0143) is sponsored by the American Psychological Association (APA).
Psycoloquy 5(07): Cognitive Scientism of Science

Commentary on Giere on Science-Cognition

J. van Brakel
Department of Philosophy
University of Utrecht
P.O. Box 80.126
3508 TC Utrecht (The Netherlands)



In this review of Cognitive Models of Science (CSS, Giere 1992), I express skepticism about its ability to (dis)solve all foundational issues concerning science and suggest that CSS would do better to redirect its attention to the foundational problems that beset cognitive science itself. Not only CSS but the social and philosophical approaches to science too are, in their extreme forms, caricatures, based on the same scientistic model. What is needed instead is a recognition of the priority of the Manifest Image over the Scientific Image.


Cognitive science, philosophy of science, cognitive models, artificial intelligence, computer science, cognitve neuroscience.
1. Anyone who might want to know what cognitive science has to offer toward a better understanding of science should consult Cognitive Models of Science, edited by Giere (1992, 1993; references to book chapter contributors will be in square brackets). The volume offers a wide variety of approaches that are brought to bear on a cognitive view of science. Unfortunately, however, if one were interested in the more general questions to be raised about the practice and products of science and their justification, this gamut of cognitivist approaches offers very little indeed, apart from a brief, rather acrimonious exchange among Glymour, Churchland, Thagard, and Giere [465-88]. This is a pity, the moreso because, between the lines, several contributors do mention a variety of foundational issues. For example, Carey, in an otherwise interesting contribution about the origin and evolution of everyday concepts, takes for granted that "the existence of rich innate concepts is not in doubt" [90] and Freedman notes that "the same cognitive processes that lead to scientific advances also lead to systematic biases and errors in scientific judgment" [310; cf. Grandy: 204, Gorman: 414]. But neither of these issues (and many others of which more appear below) is addressed in any depth. Perhaps one problem is that asking a number of cognisers what their idiosyncratic, intuitive models of science are is of no use if little attention is given to the confrontation of these idiosyncrasies or the traditions from which they stem. Perhaps another problem is that the scientistic belief in the end of anything nonscientific (including common sense, philosophy, and being human) has as a result that as long as the banner says: "Three cheers for naturalism!", anything goes. (To be sure, there are also many kinds of "noncognitivist" naturalisms as the contributions of, for example, Houts and Haddock [396] and Fuller [427, 430] testify.) For one thing, this leads to trivial statements such as: "scientists who employ heuristic search have a distinct edge over those who work randomly" [Bradshaw: 241] and "theories do not simply develop; they are developed through the cognitive activities of particular scientists" [Giere: xviii].

2. What is Cognitive Science of Science (CSS)? Psychology of science perhaps? Some contributors think so [Houts & Haddock: 367]: "the psychology of science is faced with the task of steering a course between the Scylla of rationalistic epistemologies and the Charybdis of sociological reductionism." Support for this is also found in the recurrent theme that the cognitive turn directs attention to "the `processes' of science, rather than its `products'" [Tweney: 77; cf. Giere: xvi, Nersessian: 3-6, Bradshaw: 239, Darden: 251]. But while the issue of the relative priority of the contexts of discovery and justification is never addressed directly, that of generating alternative hypotheses nevertheless includes an appeal to "criteria for theory assessment" [Darden: 262] taken straight from Scylla. I agree, of course, that we shouldn't oppose "psychologism" by presupposing the authority of deductive logic, because, as Houts and Haddock rightly stress [375], if the laws of logic do not depend on human activity for their authority, then how is it that the laws of logic have any effect on human activity? But from denying the absolute a priori normative authority of deductive logic, nothing follows as to what normative criteria would then govern scientific practice. No doubt investigating the problem-solving capacities of scientists is an interesting psychological research program, but without something about "problem-finding strategies" [252], "solution and failure evaluation strategies," and how all this fits into the surrounding (social) environment, surely no grand claims about the nature of science can be made.

3. One consequence of taking CSS as psychology of science is that it will inherit all the foundational problems of psychology and of cognitive science in particular. The organisers of the conference on which this book is based should be complimented for inviting a wide variety of speakers (but see Bookstein 1993). However, this only brings out more strongly how unclear the presuppositions of CSS are. From a philosophical perspective I found the chapter by Houts and Haddock most rewarding (which is ironic, because they defend a variant of behaviouristic psychology against mainstream cognitivism). They give an excellent summary of the foundational criticisms cognitive psychology must answer and they say that CSS focuses on a decontextualised individual scientist and confuses the mechanics of description with the doings of actors [379-81]: "computer simulations of `scientific discovery' import social conventions of the scientific community through the back door"; representations "acquire their meaning through action or use in the social context of a community of language users"; "because observers may find it useful to describe scientific problem solving as following rules, it does not follow that a scientist solving a problem is necessarily following a rule." Even if such metacriticisms are set aside, CSS has many internal foundational problems as well: Are parallel models better [80]? Are propositional structures psychologically real [xxv]? Should we favour computer models or experiments with humans to model human cognition [411]? Is CSS about human cognition or about an extension of human cognition [252]?

4. Given these foundational problems of cognitive science, it seems premature to speculate about the priority of psychology as compared with sociology and philosophy, particularly on the assumption that every problem is an empirical one, (Below, I will argue that all three claims to priority are equally unwarranted.) Like all "normal scientists" (in Kuhn's sense), the contributors to CSS rarely if ever have a positive word for philosophy. In comparison, it is surprising how much bending over backwards goes on to please the sociologists of science. For example, Giere stresses that "cognitive models of science need to be supplemented with social models" [xxvi]; cf. Nersessian [6-8], Freedman [310-1], Houts & Haddock [378], Gorman [411].

5. Other contributions suggest that the goal of CSS is more modest than being the Mother of Cognition. Cognitive psychology of science might, for example, provide tools for the empirical (historical, sociological, philosophical) study of science; model simulations might "shed important light on philosophical theories, forcing them to be more precise" [Gorman: 400]; artificial intelligence techniques might well help "to increase philosophical understanding of important episodes in the history of science" [Nowak & Thagard: 301]. The question is how much of this is tool and how much final solution. Assume we were to succeed in reaching "the ultimate goal of [being] able to reconstruct [all] scientific thinking by means of cognitive theories" [Nersessian: 5], what basis would that give to evaluate the achievements of science (including CSS)? Assume we were to completely succeed in reconstructing Hitler's thinking. We would open the door "behind the narrative and formal structures" and reveal "forgotten alternative paths" [Gooding: 52, 64]. But what would this tell us about the morality of Hitler's rationality (cognitions)?

6. But, to repeat, in their proper place, models from cognitive science may be useful in historical reconstruction (as the contribution of Nersessian testifies). Such usefulness may also contribute to the foundational issues in cognitive science referred to above, as when Nersessian [9] finds it necessary to use as tools not only "propositional" representations, but also "mental models" (structural analogs of real-world or imagined situations) and "images" (a mental model from a specific perspective), and when both Nersessian [12, 20] and Darden [261] stress the importance of analogical reasoning, thought experiments, abductive reasoning and other "nonalgorithmic" methods. It's true that "limiting scientific method to the construction of inductive or deductive arguments has needlessly blocked our ability to make sense of many of the actual constructive practices of scientists" [Nersessian: 13], but I have my doubts as to whether current contributions from cognitive science on this score have much theoretical insight to offer in addition to more traditional approaches (such as those of Peirce, Duhem, Campbell, Harre, Hutten see bibliography in Hesse 1966; or compare on one specific point, thought experiments, the recent traditional approaches such as Brown 1991, Horowitz et al. 1991, and Sorensen 1992).

7. This brings me to a last suggestion about the location of CSS in the current history of ideas: should it be seen as on a par with "the new image of scientific practice" [Gooding: 69] that stresses the role of experimentation in science (cf. Gooding et al. 1989, Hacking 1983, Pickering 1984, Radder 1988)? The inclusion of a chapter by Gooding might suggest so. But this would be incorrect, because most work in cognitive science is based on "the impoverished `one-pass' notion of discovery favored by analytical philosophy" [Gooding: 48] and "cannot duplicate the kind of hands-on, craft knowledge possessed by scientists" [Gorman: 409] see also the excellent commentary of Bookstein (1993). In fact, traditional cognitive science (and AI) has little to contribute to an interest in science as a situated praxis, it being essentially non-situated, as the discussions about the frame problem illustrate (Fetzer 1993, Harnad 1993, Hayes & Ford 1993, van Brakel 1992). Of course, the "experimentalists" do share with CSS the internalist approach to science ("today's epistemology of experiment is practiced in complete isolation from, say, technological ethics" [Fuller: 430]).

8. In addition to inheriting the foundational problems of cognitive science, concerning problems stemming from traditional philosophy of science and epistemology CSS has nothing much to say, being neutral or ambiguous with respect to all interesting questions. As an example, consider Thagard's connectionist net ECHO, in which scientific hypotheses are represented by nodes and the weights correspond to the degree of explanatory coherence. Tweney suggests that ECHO and similar networks "amount to a complicated way to compute a box score!" and because Thagard based his networks on descriptions written by the winners, it's not surprising that they win [87]. Freedman uses ECHO to model "social factors" [311], showing that "it is possible to generate distinct mental representations of the same hypotheses and evidence" [329], because ECHO contains "various adjustable parameters that can materially affect the outcome of the calculation" [Giere: xxiv]. Hence "it could be argued that ECHO, like BACON, merely reproduces the results of social negotiations... these programs... may actually illustrate and support some sociological analyses" [Gorman: 410-1]. In particular, "ECHO does not provide externally valid evidence for the superiority of explanatory over alternate accounts of how scientific controversies are resolved" [410]. Some of these criticisms are acknowledged: "The most serious limitation is that the input provided to the program was devised by one of us (Nowak)... and the question of historical bias naturally arises" [Nowak & Thagard: 300]. Perhaps ECHO is an interesting toy, but it solves no substantive problems. Analogous points could be made about the issue of incommensurability, which CSS writers both applaud and deny, and analyse in a variety of ways [8, 11, 20, 68, 90-5, 123, 179, 298, 312; see also the excellent index] or about Savage's proposal for "a foundationalist theory of sensory cognition that could be confirmed by cognitive science" [228]. Not unexpectedly, the mechanics of cognition (or the human brain for that matter) equally subserve universal rationality and relativism (as well as any third alternative that can be thought of).

9. Some of the work in CSS is interesting, but would be better redirected towards an explicit contribution to the foundational problems that beset cognitive science, instead of making unjustified claims about dissolving foundational questions in favour of piecemeal cognitive engineering. It is not that I propose a return to a priori philosophy of science or a conversion to post-modern sociology. All three approaches are, in their extreme caricatures, enlightened by the same scientistic model. The first thing that is needed to get a better perspective on science (and technology) is to acknowledge the priority of ordinary human forms of life over the scientific picture of the world. One way to put this is to say that everything worth talking about depends on the Manifest Image, the latter expression covering a family of images overlying a plurality of similar forms of (human) life (for "manifest" versus "scientific" image see Sellars 1963; for "forms of life" see van Brakel 1994). It's the Manifest Image that underlies the Scientific Image, and not the other way round. There are many different reasons why the Manifest Image should be given priority (van Brakel 1993). This is not the place to dwell on them, but I will give a few hints.

10. Attempts by philosophers of science in the earlier part of this century to provide a picture of science unified by one method have failed. Attempts to specify a notion of reduction so that all sciences could be fitted into one world picture have failed. Attempts by philosophers to cover up these failures by talking about supervenience have failed as well. (CSS assumes these problems can be solved by stipulation.) The unity, pluralism, and pragmatism of the Manifest Image cannot fail, because it sustains everything. Cross-culturally, only something like the Manifest Image is shared.

11. The progress of science is built on a projectable sequence of projectable predicates (including "meta" predicates), but at the most fundamental level progress depends on intuitions and categories entrenched in the prevailing Manifest Image. Think of logic: at the meta-meta-meta-level we use ordinary English to discuss its relevance, vitality, and validity. Discussions about the status of set theory are rampant with appeals to informal intuitions. Take quantum mechanics. As Bohr stressed, there is no way of testing the predictions of quantum mechanics without appealing to macroscopic objects and colloquial language to describe experiments.

12. All scientific categories derive their existence in part from normative methodological criteria which govern communication between scientists ("cognitive scientists still need to explain why scientists modify their cognitive constraints" [Freedman: 329]). These criteria are grounded, in the end, in the Manifest Image and include, in the end, the goals of science as part of a substantial Moral Image. Giere writes: "the only form of rationality that exists is the instrumental use of empirically sanctioned strategies to achieve recognized goals" [484]. Does this mean that "recognized goals" is outside the realm of rationality? Would he agree to substitute "morality" instead of "rationality" in the quotation given?

13. Of course, in the twentieth century, the Manifest and Scientific Images cannot be neatly separated. I do not deny that the Scientific Image (among other theoretical images) strongly influences the Manifest Image. But I would repeat that notions like "cause," "explanation," "reasonable," "truth," "right," "relevant," and so on, are firmly and finally grounded in the pretheoretical Manifest Image -- which is not to say that these notions are themselves incapable of change and variation. One might say: "Science could explain, for example, how explanation, communication, and normativity are possible among humans." Of course, but how is the request for such an explanation finally justified? Moreover, there will be alternative "sciences" to offer explanations of whatever is considered relevant to explain. To make a judgement with regard to these alternatives we are committed, explicitly or tacitly, to making judgements on issues such as "deciding which features of science we value most," "rightness," "appropriate to the circumstances," and so on. Judgements on such issues are always grounded in the Manifest Image and cannot be "bootstrapped out of it." The only way out would be to appeal to (the ideal of) exactly one best method of inquiry and exactly one best end of enquiry which gives THE answers. Such answers are only possible in the Brave New World and its congeners.


Bookstein, F.L. (1993) Geometry As Cognition In The Natural Sciences. PSYCOLOQUY 4 (65) scientific-cognition.2.bookstein.

Brown, J.R. (1991) The Laboratory of the Mind: Thought Experiments in the Natural Sciences. London and New York: Routledge.

Fetzer, J.H. (1993) Philosophy Unframed. PSYCOLOQUY 4(33) frame-problem.10.fetzer.

Giere, R.N. (1993) Precis of Cognitive Models of Science. PSYCOLOQUY 4(56) scientific-cognition.1.giere.

Giere, R.N. (1992) Cognitive Models of Science. Minnesota Studies in the Philosophy of Science, volume 15. Minneapolis: University of Minnesota Press.

Gooding, D., Pinch, T. and Schaffer, S. (1989) The Uses of Experiment. Cambridge: Cambridge University Press.

Hacking, I. (1983) Representing and Intervening. Cambridge: Cambridge University Press.

Harnad, S. (1993) Problems, Problems: The Frame Problem As A Symptom Of The Symbol Grounding Problem. PSYCOLOQUY 4(34) frame-problem.11.harnad.

Hayes, P.J. and Ford, K.M. (1993) Effective Descriptions Need Not Be Complete. PSYCOLOQUY 4(21) frame-problem.5.ford+hayes.

Hesse, M.B. (1966) Models and Analogies in Science, Notre Dame IN: University of Notre Dame Press.

Horowitz, T. and Massey, G.J. (eds.) (1991) Thought Experiments in Science and Philosophy. Rowman & Littlefield.

Pickering, A. (1984) Constructing Quarks: A Sociological History of Particle Physics. Chicago: The University of Chicago Press.

Radder, H. (1988) The Material Realization of Science. Assen: van Gorcum.

Sellars, W. (1963) Science, Perception and Reality. London: Routledge.

Sorensen, R.A. (1992) Thought Experiments. Oxford: Oxford University Press.

van Brakel, J. (1992) The Complete Description of the Frame Problem. PSYCOLOQUY 3(60) frame-problem.2.vanbrakel.

van Brakel, J. (1993) Peirce's Pragmatisch Realisme. Tekenen van Waarheid: C.S. Peirce en de Hedendaagse Wetenschapsfilosofie (M. Hulswit and H.C.D.G. de Regt, eds.), Tilburg: Tilburg University Press, pp. 175-206 [English translation in preparation].

van Brakel, J. (1994) Emotions as the Fabric of Forms of Life: A Cross-Cultural Perspective. Social Perspectives on Emotion (W.M. Wentworth and J. Ryan, eds.), Vol. II, Greenwich USA: JAI Press, in press.

Volume: 5 (next, prev) Issue: 07 (next, prev) Article: 3 (next prev first) Alternate versions: ASCII Summary