Norman (1980) called for cognitive science to extend its domain beyond the investigation of pure cognition, suggesting, for example, a return to some of the topics that were central to cybernetics, such as feedback between agents and environments. This was not the first time that such a suggestion had been made.
Twenty years earlier, in Plans and the Structure of Behavior (Miller et al., 1960), cognitive psychologist George Miller, mathematical psychologist Eugene Galanter, and neuropsychologist Karl Pribram argued for cognitivism to revisit the contributions of cybernetics. The reason for this was that Miller, Galanter, and Pribram, like Norman, were worried that if cognitivism focused exclusively on mental representations, then it would be incomplete. Such a perspective “left an organism more in the role of a spectator than of a participant in the drama of living. Unless you can use your Image to do something, you are like a man who collects maps but never makes a trip” (p. 2).
A related perspective was the theme of another key work that preceded Norman (1980), Ulric Neisser’s (1976) Cognition and Reality. Neisser, an eminent pioneer of cognitivism (Neisser, 1967), argued that the relevance of cognitive psychology required it to be concerned with factors that lay beyond mental representations. “Perception and cognition are usually not just operations in the head, but transactions with the world. These transactions do not merely inform the perceiver, they also transform him” (Neisser, 1976, p. 11). Rather than being inspired by cybernetics, Neisser was interested in reformulating cognitivism in the context of Gibson’s (1966, 1979) theory of ecological perception. “Because perception and action take place in continuous dependence on the environment, they cannot be understood without an understanding of that environment itself” (Neisser, 1976, p. 183).
It would appear, then, that there is an extended history of important cognitivists calling for cognitive science to extend itself beyond the study of what Norman (1980) called the pure cognitive system. It is equally clear that this message has not had the desired impact. For instance, had the main theme of Miller, Galanter, and Pribram (1960) been widely accepted, then there would have been no need for similar proposals to appear decades later, as with Neisser (1976) and Norman (1980).
Why has cognitive science stubbornly held firm to the classical approach, emphasizing the study of pure cognition? One possible answer to this question is that the development of cognitivism in one of cognitive science’s key contributors, psychology, occurred in a combative context that revealed thesis and antithesis but was not conducive to synthesis. This answer is considered in more detail below.
It is often claimed that cognitive science is chiefly concerned with the human cognitive capacities (Gardner, 1984; von Eckardt, 1995). Ironically, the one discipline that would be expected to have the most to say about human mental phenomena—experimental psychology—was one of the last to accept cognitivism. This was because around the time cognitive science emerged, experimental psychology was dominated by behaviorism.
Behaviorists argued that a scientific psychology must restrict itself to the study of observable behavior and avoid invoking theoretical constructs that could not be directly observed, such as mental representation.
So long as behaviorism held sway—that is, during the 1920s, 1930s, and 1940s— questions about the nature of human language, planning, problem solving, imagination and the like could only be approached stealthily and with difficulty, if they were tolerated at all. (Gardner, 1984, p. 12)
Other disciplines were quicker to endorse cognitivism and to draw upon the insights of diverse fields of study because they were not restricted by the behaviorist yoke. For instance, mathematician Norbert Wiener (1948) created the field of cybernetics after realizing that problems involving communication, feedback, and information were general enough to span many disciplines. He held “the conviction that the most fruitful areas for the growth of the sciences were those which had been neglected as a no-man’s land between the various fields” (p. 8).
Wiener realized that progress in cybernetics required interaction between researchers trained in different disciplines. He was a key organizer of the first joint meeting concerning cybernetics, held at Princeton in 1944, which included engineers, physiologists, and mathematicians. This in turn led to the Macy conferences on cybernetics that occurred regularly from 1946 through 1953 (Conway & Siegelman, 2005). The Macy conferences broadened the range of participants who attended the 1944 Princeton meeting to include psychologists, sociologists, and anthropologists.
The success of the Macy meetings prepared the way for a variety of similar interdisciplinary conferences that in turn set the stage for cognitive science. One of these was a 1956 conference organized by MIT’s Special Interest Group in Information Theory. This conference included presentations by Newell and Simon on their logic machine, and by Chomsky on generative grammar (Miller, 2003). Thus conference participant George Miller, trained in the behaviorist tradition, would have heard computer scientists and linguists freely using representational terms to great effect.
The success of cognitivism in other disciplines, communicated to psychologists who participated in these interdisciplinary conferences, led to a reaction against behaviourism in psychology. “No longer were psychologists restricted in their explanatory accounts to events that could either be imposed on a subject or observed in one’s behavior; psychologists were now willing to consider the representation of information in the mind” (Gardner, 1984, p. 95).
George Miller (2003) has provided a personal account of this transition. His first book, Language and Communication (Miller, 1951), deliberately employed a behaviorist framework, a framework that he would completely abandon within a few years because of the influence of the cognitivist work of others. “In 1951, I apparently still hoped to gain scientific respectability by swearing allegiance to behaviorism. Five years later, inspired by such colleagues as Noam Chomsky and Jerry Bruner, I had stopped pretending to be a behaviorist” (Miller, 2003, p. 141).
However, because cognitivism arose as a reaction against behaviorism in North American experimental psychology, cognitive psychology developed by taking an antagonistic approach to almost all of the central behaviorist positions (Bruner, 1990; Sperry, 1993). “We were not out to ‘reform’ behaviorism, but to replace it” said Bruner (1990, p. 3). In psychology, the cognitive revolution,
was not one of finding new positives to support the important role of cognition, many of which were already long evident. Rather, the story is one of discovering an alternative logic by which to refute the seemingly incontestable reasoning that heretofore required science to ostracize mind and consciousness. (Sperry, 1993, p. 881)
Consider but one example that illustrates the tone within psychology during the cognitive revolution. Skinner’s (1957) account of language, Verbal Behavior, elicited a review by Noam Chomsky (1959b) that serves as one of the pioneering articles in cognitivism and is typically viewed as the turning point against psychological behaviourism (MacCorquodale, 1970; Schlinger, 2008). Some researchers, though, have objected to the tone of Chomsky’s review: “It is ungenerous to a fault; condescending, unforgiving, obtuse, and ill-humored” (MacCorquodale, 1970, p. 84).
On the other side of the antagonism, behaviorists have never accepted the impact of Chomsky’s review or the outcome of the cognitive revolution. Schlinger (2008, p. 335) argued that fifty years after its publication, Verbal Behavior (and behaviorism) was still vital because it worked: “It seems absurd to suggest that a book review could cause a paradigmatic revolution or wreak all the havoc that Chomsky’s review is said to have caused to Verbal Behavior or to behavioral psychology.”
The tone of the debate about Verbal Behavior is indicative of the tension and conflict that characterized cognitivism’s revolt against behaviorist psychology. As noted earlier, cognitivists such as Bruner viewed their goal as replacing, and not revising, behaviorist tenets: “It was not a revolution against behaviorism with the aim of transforming behaviorism into a better way of pursuing psychology by adding a little mentalism to it. Edward Tolman had done that, to little avail” (Bruner, 1990, p. 2).
One behaviorist position that was strongly reacted against by cognitivism “was the belief in the supremacy and the determining power of the environment” (Gardner, 1984, p. 11). Cognitive psychologists turned almost completely away from environmental determinism. Instead, humans were viewed as active information processors (Lindsay & Norman, 1972; Reynolds & Flagg, 1977). For instance, the New Look in perception was an argument that environmental stimulation could be overridden by the contents of beliefs, desires, and expectations (Bruner, 1957). In cognitivism, mind triumphed over environmental matter.
Cognitive psychology’s radical rejection of the role of the environment was a departure from the earlier cybernetic tradition, which placed a strong emphasis on the utility of feedback between an agent and its world. Cyberneticists had argued that,
for effective action on the outer world, it is not only essential that we possess good effectors, but that the performance of these effectors be properly monitored back to the central nervous system, and that the readings of these monitors be properly combined with the other information coming in from the sense organs to produce a properly proportioned output to the effectors. (Wiener, 1948, p. 114)
Some cognitivists still agreed with the view that the environment was an important contributor to the complexity of behavior, as shown by Simon’s parable of the ant (Simon, 1969; Vera & Simon, 1993). Miller, Galanter, and Pribram (1960) acknowledged that humans and other organisms employed internal representations of the world. However, they were also “disturbed by a theoretical vacuum between cognition and action” (p. 11). They attempted to fill this vacuum by exploring the relevance of key cybernetic ideas, particularly the notion of environmental feedback, to cognitive psychology.
However, it is clear that Miller, Galanter, and Pribram’s (1960) message about the environment had little substantive impact. Why else would Norman (1980) be conveying the same message twenty years later? It is less clear why this was the case. One possibility is that as cognitivism took root in experimental psychology, and as cognitive psychology in turn influenced empirical research within cognitive science, interest in the environment was a minority position. Cognitive psychology was clearly in a leading position to inform cognitive science about its prototypical domain (i.e., adult human cognition; see von Eckardt, 1995). Perhaps this informing included passing along antagonist views against core behaviorist ideas.
Of course, cognitive psychology’s antagonism towards behaviorism and the behaviorist view of the environment is not the only reason for cognitive science’s rise as a classical science. Another reason is that cognitive science was not so much inspired by cybernetics, but was instead inspired by computer science and the implications of the digital computer. Furthermore, the digital computer that inspired cognitive science—the von Neumann architecture, or the stored-program computer (von Neumann, 1993)—was a device that was primarily concerned with the manipulation of internal representations.
Finally, the early successes in developing classical models of a variety of highlevel cognitive phenomena such as problem solving (Newell et al., 1958; Newell & Simon, 1961, 1972), and of robots that used internal models to plan before executing actions on the world (Nilsson, 1984), were successes achieved without worrying much about the relationship between world and agent. Sense-think-act processing, particularly the sort that heavily emphasized thinking or planning, was promising new horizons for the understanding of human cognition. Alternative approaches, rooted in older traditions of cybernetics or behaviorism, seemed to have been completely replaced.
One consequence of this situation was that cognitive science came to be defined in a manner that explicitly excluded non-classical perspectives. For example, consider von Eckardt’s (1995) attempt to characterize cognitive science. Von Eckardt argued that this can be done by identifying a set of domain-specifying assumptions, basic research questions, substantive assumptions, and methodological assumptions. Importantly, the specific members of these sets that von Eckardt identified reflect a prototypical classical cognitive science and seem to exclude both connectionist and embodied varieties.
Consider just one feature of von Eckardt’s (1995) project. She began by specifying the identification assumption for cognitive science—its assumed domain of study. According to von Eckardt, the best statement of this assumption is to say that cognitive science’s domain is human cognitive capacities. Furthermore, her discussion of this assumption—and of possible alternatives to it—rejects non-classical variants of cognitive science.
For instance, Simon’s (1969) early consideration of the sciences of the artificial cast intelligence as being the ability to adapt behaviour to changing demands of the environment. Von Eckardt (1995) considered this idea as being a plausible alternative to her preferred identification assumption. However, her analysis of Simon’s proposal can be dismissed because it is too broad: “for there are cases of adaptive behavior (in Simon’s sense) mediated by fairly low-level biological mechanisms that are not in the least bit cognitive and, hence, do not belong within the domain of cognitive science” (p. 62). This view would appear to reject connectionism as being cognitive science, in the sense that it works upward from low-level biological mechanisms (Dawson, 2004) and that connectionism rejects the classical use of an explanatory cognitive vocabulary (Fodor & Pylyshyn, 1988; Smolensky, 1988).
Similarly, von Eckardt (1995) also rejected an alternative to her definition of cognitive science’s identification assumption, which would include in cognitive science the study of core embodied issues, such as cognitive scaffolding.
Human beings represent and use their ‘knowledge’ in many ways, only some of which involve the human mind. What we know is represented in books, pictures, computer databases, and so forth. Clearly, cognitive science does not study the representation and the use of knowledge in all these forms. (von Eckardt, 1995, p. 67)
If cognitive science does not study external representations, then by von Eckardt’s definition the embodied approach does not belong to cognitive science.
The performance of classical simulations of human cognitive processes led researchers to propose in the late 1950s that within a decade most psychological theories would be expressed as computer programs (Simon & Newell, 1958). The classical approach’s failure to deliver on such promises led to pessimism (Dreyfus, 1992), which resulted in critical assessments of the classical assumptions that inspired alternative approaches (Rumelhart & McClelland, 1986c; Winograd & Flores, 1987b). The preoccupation of classical cognitivism with the manipulation of internal models of the world may have prevented it from solving problems that depend on other factors, such as a cybernetic view of the environment.
As my colleague George Miller put it some years later, ‘We nailed our new credo to the door, and waited to see what would happen. All went very well, so well, in fact, that in the end we may have been the victims of our success.’ (Bruner, 1990, pp. 2–3)
How has classical cognitivism been a victim of its success? Perhaps its success caused it to be unreceptive to completing the cognitive dialectic. With the rise of the connectionist and embodied alternatives, cognitive science seems to have been in the midst of conflict between thesis and antithesis, with no attempt at synthesis. Fortunately there are pockets of research within cognitive science that can illustrate a path towards synthesis, a path which requires realizing that each of the schools of thought we have considered here has its own limits, and that none of these schools of thought should be excluded from cognitive science by definition. One example domain in which synthesis is courted is computational vision.