What is Cognitive Science?
Cognitive science comprises a cluster of disciplines including portions of psychology, linguistics, computer science, neuroscience, philosophy, sociology, and anthropology. Its roots lie in the 1950s; it acquired an academic identity in the 1970s, and it continues to thrive in the twenty-first century.
It seeks to explain mental activities such as reasoning, remembering, language use, and problem solving, and the explanations it advances commonly involve descriptions of the mechanisms responsible for these activities. Cognitive mechanisms are distinguished from the mechanisms invoked in other domains of biology by involving the processing of information. Many of the philosophical issues discussed in the context of cognitive science involve the nature of information processing, especially the central notion of representation. One of the distinctive features of cognitive science is that it is not a discipline but a multi-disciplinary research cluster. It draws upon the contributing disciplines for the problems it investigates, the tools it uses to investigate them, and the explanatory strategies invoked, but the results transcend what is typically achieved in any of the contributing disciplines. This gives rise to philosophical questions about the nature of interdisciplinary research.The term “cognitive science” was only coined in the mid-1970s. In 1975 it was employed in two books. Explorations in Cognition, the product of a collaborative research group at the University of California at San Diego (UCSD), concludes with the suggestion: the “concerted efforts of a number of people from... linguistics, artificial intelligence, and psychology may be creating a new field: cognitive science” (Norman and Rumelhart 1975, 409). Although situated in psychology, the group employed computational models of semantic networks to explain word recognition, analogy, memory, and semantic interpretation of verbs, sentences, and even brief stories.
The collaborative book by computer scientist Daniel Bobrow and cognitive psychologist Allan Collins, Representation and Understanding: Studies in Cognitive Science (1975), invoked the term in its subtitle. Two years later, the Alfred P. Sloan Foundation announced its Particular Program in Cognitive Science and, over 10 years, provided $17.4 million to establish and foster interdisciplinary centers at selected research universities. UCSD received one of the early Sloan grants and used a portion of its funding to sponsor the 1979 La Jolla Conference on Cognition, which became the first meeting of the now international Cognitive Science Society. In 1980 the Cognitive Science Society assumed ownership of the journal Cognitive Science, which itself had begun publication in 1977.While it was during the 1970s that cognitive science began to acquire an institutional identity, its roots go back to the middle of the century when a new intellectual perspective began to inspire researchers in psychology and linguistics to reject the strictures that behaviorism had placed on most research in these disciplines in North America since John Watson (1913) issued his manifesto “Psychology as the Behaviorist Views it” and urged a focus on behavior, not hypothetical mental activities. Although B.F. Skinner, who advocated a radical behaviorism that eschewed mental entities, is perhaps the best known behaviorist, the behaviorist tradition was relatively diverse. Not all behaviorists were opposed to positing events inside the head: Clark Hull (1943) appealed to intervening variables such as drive, but stopped short of overtly mentalistic concepts. Edward Tolman (1948) was exceptional among behaviorists in postulating cognitive maps to explain navigational abilities of rats. Leonard Bloomfield (1939) carried behaviorism to linguistics where he advanced a strongly empiricist approach to cataloging and analyzing linguistic forms and rejected mentalistic accounts of these forms.
While behaviorism cast a broad shadow over psychology and linguistics in North America, in Europe a variety of alternative perspectives more favorable to mentalistic characteristics of human beings prospered and would come to influence the development of cognitive science.
For example, Jean Piaget proposed cognitive operations in his genetic epistemology; Frederick Bartlett introduced schemas (organizing structures) to account for memory distortions; Gestalt psychology recast perception in terms of self-organizing forms; and Lev Vygotsky and Alexander Luria initiated studies demonstrating cultural influences on language and thought. Even in North America, psychophysics and parts of developmental and social psychology were pursued outside of behaviorism’s shadow. But for much of psychology, a revolution was required to reverse behaviorism’s proscription on appeals to mental phenomena in explaining behavior.The cognitivism that emerged in the 1950s maintained behaviorism’s emphasis on explaining behavioral phenomena and invoking only behavioral evidence. Hence, unlike earlier mentalistic psychology, it rejected introspection as the avenue to the mind. What it required was a way of conceptualizing internal events that construed them as causal processes contributing to the generation of behaviors. This conceptualization was provided by information theory, which developed from formal engineering analyses of communication channels, such as telephones, conducted at Bell Laboratories between the 1920s and 1960s. This endeavor attempted to quantify the capacity to transmit information across channels that are subject to capacity and rate limits and to noise. Construing information as a reduction of uncertainty at the end of the channel about the message at the beginning of the channel, Claude Shannon (1948) introduced the bit as the unit of information: the unit of information required to differentiate between two equally likely messages. Shannon also showed that one could determine the redundancy in a message in terms of the reduction of uncertainty. George Miller drew upon this analysis in his Harvard PhD dissertation which used redundancy in a message to explain how messages in spoken English could be understood in noisy environments.
In perhaps his best known research, Miller (1956b) identified comparable capacity limitations in a number of cognitive domains, including short-term memory: humans can hold up to seven, plus or minus two, separate items in memory over a period of minutes (unless they are interrupted earlier). In this research, information is construed as the commodity the mind utilizes and the various tasks it performs (remembering, planning, problem solving) are construed as the processing of information. Donald Broadbent (1958) advanced a model in which information about sensory events is held briefly in a short-term store, and an attentional filter restricts which gets transmitted along a single, limited-capacity channel for further processing. These ideas were incorporated into a general framework for understanding cognitive activity by Ulric Neisser (1967) in his pathbreaking textbook Cognitive Psychology.
The idea of the mind as an information processor was further promoted with the introduction of the digital computer as itself an information processor. Shortly after its creation following World War II, some researchers in the new field of computer science began to explore the possibility of programming a computer to behave intelligently (e.g., perform activities that would be judged intelligent if performed by humans). A pivotal conference at Dartmouth College in the summer of 1956 introduced the name artificial intelligence and witnessed the first presentation of a program performing intelligently: Alan Newell and Herbert Simon's Logic Theorist, which developed proofs of theorems in symbolic logic.
These contributions were brought together on September 11, 1956, the second day of the second Symposium on Information Theory. Newell and Simon (1956) again reported on Logic Theorist, and George Miller (1956a) presented his research on capacity limitations of short-term memory. In between, a young linguist, Noam Chomsky (1956), presented a paper, “Three Models of Language,” in which he argued that various computational systems, such as finite-state automata, were inadequate to model the grammar of human languages and introduced arguments for what he called transformational grammars.
These employed procedures for generating core linguistic structures (trees) and transformations to modify these structures. For Miller, on that day, “cognitive science burst from the womb of cybernetics and became a recognizable, interdisciplinary adventure in its own right” (Miller 1979, 4). The interdisciplinary interaction that day between psychology, artificial intelligence, and linguistics became characteristic of cognitive science, although, as discussed above, 20 years were to pass before the name was introduced and the field became institutionalized.From these beginnings, research in cognitive science has burgeoned. We can here note just a few landmarks that provide an indication of the breadth of the field. Newell and Simon (1972) introduced the idea of a production system, consisting of a working memory and operations (productions) designed to respond and alter the contents of that memory, and employed it to model the strategies humans use to solve problems. Chomsky (1957) developed the first of several grammatical theories (minimalism is the most recent; see Chomsky 1995). Chomsky elicited an opposition movement which rejected the autonomous status Chomsky claimed for syntax and interlaced syntax with semantics (resulting in what Harris (1993) characterizes as the linguistic wars of the 1960s and 1970s). More recently, cognitive linguistics has emphasized how other cognitive processes such as spatial representation (Fauconnier 1994) or metaphors grounded in the body (Lakoff and Johnson 1999) serve as the basis for linguistic structures. Research in the field of memory that started with the distinction between short- and longterm memory has expanded as researchers have distinguished different forms of long-term memory and distinctive features of how each is processed (Schacter 1996, Tulving 2002).
2.