Skip to Content

Search: {{$root.lsaSearchQuery.q}}, Page {{$}}

What is Cognitive Science

Given that the term “cognitive science” is relatively new and susceptible to distinct interpretations – some of which are quite narrow – it will be useful to specify what cognitive science does and does not mean.

We start by first setting aside some of the more narrow interpretations that do not form the basis of the major, because they serve as useful contrasts and points of departure for what we are. Contemporary cognitive science has its roots in the cognitive revolution of the 1950s and 60s, including the growth of computer science and artificial intelligence, and the accompanying articulation and exploration of the idea that thought is a kind of computation physically realized by the brain 1. This idea remains central to the brain and behavioral sciences, and its import for scientific practice continues to grow. It was the theoretical and methodological role of computation that most clearly marked cognitive science as an interdisciplinary venture at the outset, and distinguished it most sharply from much older traditions in psychology. However, as a result, for many scientists “cognitive science” came to represent something quite narrow: a commitment to computation and only computation, to the exclusion of explanations at other levels of abstraction – for example, at neural/biological or social levels. An early focus on “cold cognition” to the exclusion of putatively non-cognitive phenomena such as emotion also led to a perception of cognitive science as narrowly circumscribed. Some textbook treatments of cognitive science still implicitly endorse this approach by introducing the field as a catalog of types of computational representations. Subsequent scientific advances – especially in the last 10–15 years – strongly indicate that a more illuminating, useful and accurate characterization of cognitive science as it has matured is one that is radically broader in its conception, even while still embracing a foundational theoretical role for computation. This broader conception serves as the basis for the proposed major, and can be summarized as follows:

Contemporary cognitive science is the scientific study of mind, brain, and behavior that: (a) seeks to develop integrated explanations of the structure of brain and behavior at multiple theoretical levels of abstraction, ranging from neuro‐biological levels to abstract computational levels to evolutionary levels grounded in assumptions about adaptive pressures 2; (b) sharpens fundamental questions that arise within theoretical levels and within specific traditional disciplines by articulating them in ways that allow them to be addressed productively across theoretical levels or via different disciplinary approaches; and (c) draws on whatever methodological and theoretical resources can be exploited to address the scientific problems at hand – that is, cognitive science does not respect disciplinary (or institutional/departmental) boundaries.

Two examples of specific domains that exemplify these characteristics will serve to ground these abstract points. These domains also form the core of two of the proposed tracks in the major, to be described below, and each is a locus of significant recent progress in the field.

Example 1: The Nature of Choice

Understanding the nature of human choice has long been a core problem in psychology and economics (and of course philosophy), but over the last 10–15 years it has enjoyed particularly rapid theoretical and empirical advances. Crucially, these advances take the form of integration of levels of explanation. For example, the new field of neuro-economics is largely based on recent discoveries concerning the neural basis of the component quantities underlying the computation of expected value in classical normative approaches (such as value and probability). In psychology, mathematical and computational approaches to statistical decision theory – especially the optimal sequential integration of evidence – have been successfully reformulated as precise models of the information processes underlying rapid decisions (those taking approximately less than a second – for example, the choice of when and where to move the eyes next when searching for a visual target among a clutter of distractors). These models have in turn led to new understanding of patterns of neuronal firings in humans and other mammals making such decisions. Similarly, algorithms developed in computational reinforcement learning (a branch of artificial intelligence concerned with designing autonomous agents that learn how to act effectively in their environments) have now been mapped onto underlying cognitive and neural systems in ways that have revolutionized our understanding of how the mind and brain learns to control thought and action. In short, contemporary approaches to choice now integrate theoretical levels that range from formal descriptions of what is to be computed (e.g., grounded in expected utility theory and optimal control theory) to information processing accounts of how it is to be computed (e.g., grounded in reinforcement learning theory). And these “how” accounts are themselves specified at various levels ranging across abstract algorithms, brain-circuit systems, and coding schemes of neuronal firing rates. These advances would not be possible without the contributions of traditionally separate disciplines, including psychology, economics, and computer science.

Example 2: The Nature of Language

One of the major catalysts of the cognitive revolution in the 1950s was Chomsky’s (re)formulation of the scientific study of human language as the study of an internal biologically based cognitive system, rather than the systematization and description of external linguistic artifacts (e.g. sets of speech sounds or “sentences”) and their apparent patterns and regularities. The key idea in this reformulation was (and continues to be) the positing of a mental computational capacity for the generation and interpretation of an infinite set of linguistic expressions. In the linguistic sciences, theoretical accounts of this capacity – and crucially, its limits – take the form of formal grammars. One of the theoretical roles of such grammars in the cognitive science of language is analogous to the role of statistical decision theory in contemporary accounts of choice: the grammars constitute a specification of what must be known and computed by the organism for it to be able to engage in the (behavioral) processes of language generation and comprehension, whether speech, reading, or sign language. In the last 10–15 years, a new field of computational psycholinguistics has emerged that provides integrated accounts relating together the formal theories of human linguistic capacity with models of computational processing informed by recent advances in mathematical and cognitive psychology, and machine learning. These integrated theories are in turn reformulating many of the leading edge questions in the brain basis of language – for example, motivating the cognitive neuroscience of language to move beyond questions concerning “where syntax resides in the brain” to questions concerning how brain circuits might implement the dynamic processes posited in the independently motivated computational theories. And these integrations open the door to new perspectives on the even deeper “why” questions that continue to be raised in theoretical linguistics. 


  1. The University of Michigan played a key role in this revolution: major contributions with lasting influence include signal detection theory, which remains a foundation of modern perception research (Tanner & Swets, 1955); computational approaches to adaptation (Holland, 1975); reaction time approaches to human cognition and performance (Edwards, 1965; Meyer & Shvanaveldt, 1973); mathematical approaches to motor control (Fitts, 1967); and mathematical approaches to human decision making (Coombs, Dawes & Tversky, 1970).
  2. Thus the use of the terms “mind and brain” in cognitive science is not a regression to Cartesian dualism but rather reflects a commitment to multiple levels of theoretical explanation.