The History of Cognitive Science

Last reviewed on October 2, 2026.

Cognitive science is a young discipline with very old questions. Philosophers argued about knowledge, memory and perception for more than two thousand years, but it was only in the mid-20th century that psychologists, linguists, neuroscientists and the first computer scientists found a shared way to study the mind scientifically. This page sets out a dated timeline of how that happened, who was involved and how the field has changed since.

Timeline of Cognitive Science

The table below lists the milestones most often cited by historians of the field. Dates refer to publication or founding; some ideas were circulating earlier.

Ancient and Philosophical Foundations

The oldest questions in cognitive science come from philosophy. Plato argued that some knowledge is innate and that learning is a kind of recollection; Aristotle’s De Anima (On the Soul) analysed perception, imagination and memory, and described the mind as receiving the “form” of objects without their matter. Philosophers in India and China developed their own sophisticated accounts of perception, attention and self-knowledge, and medieval scholars such as Avicenna (Ibn Sīnā) proposed “internal senses” that they tentatively located in different parts of the brain.

In the 17th century René Descartes separated a thinking mind from a mechanical body, while Thomas Hobbes claimed that reasoning is “nothing but reckoning” – adding and subtracting ideas. The long debate between rationalists (who stressed innate structure) and empiricists such as John Locke and David Hume (who stressed learning from experience) is still alive in modern arguments about language acquisition and machine learning.

The Birth of Experimental Psychology

In the 19th century the mind became something to measure. In 1868 the Dutch physiologist Franciscus Donders used differences in reaction times to estimate how long a mental decision takes – a logic still used in experiments today. In 1879 Wilhelm Wundt opened his laboratory in Leipzig, studying sensation, attention and reaction time with trained introspection. Hermann Ebbinghaus (1885) memorised lists of nonsense syllables to produce the first forgetting curves, and William James’s Principles of Psychology (1890) laid out ideas about attention, habit and the “stream of consciousness” that cognitive psychologists later returned to.

Behaviourism and Its Limits

Introspection proved unreliable: different laboratories reported different “contents of consciousness” and had no way to settle disputes. In 1913 John B. Watson proposed that psychology should study only observable behaviour, and for the next four decades behaviourism, later led by B. F. Skinner, dominated American psychology. It gave the field rigorous experimental methods but treated the mind as a black box.

Cracks appeared early. Edward Tolman showed that rats learn “cognitive maps” of a maze even without reward (his best-known summary appeared in 1948). In Britain, Frederic Bartlett’s Remembering (1932) showed that people reconstruct memories using schemas. At the 1948 Hixon Symposium, Karl Lashley argued that fast, ordered sequences of behaviour such as speech and piano playing cannot be chains of stimulus and response – they must be planned in advance by the brain.

Which Intellectual Revolution Transformed Psychology?

The answer is the cognitive revolution of the 1950s and 1960s. Instead of asking only how stimuli are linked to responses, psychologists began asking how information is taken in, stored, transformed and used – and they borrowed the digital computer as a model of how that could happen. George Miller dated the moment to 11 September 1956, the second day of the MIT Symposium on Information Theory, when papers by Newell and Simon, Chomsky and Miller himself appeared side by side. That summer the Dartmouth workshop had already launched artificial intelligence as a field.

Our in-depth article The Cognitive Revolution in Psychology: History, Focus and Key Figures covers the key experiments, the arguments with behaviourism and the legacy of this turning point. Below we focus on the people from computing who made it possible and on what happened next.

Early Cognitive Scientists from Computer Science

Many of the founders of cognitive science were not psychologists at all. They were mathematicians, engineers and computer scientists who realised that the same formal tools used to design machines could describe thinking.

• Alan Turing defined what it means for a procedure to be computable (1936) and, in “Computing Machinery and Intelligence” (1950), proposed the imitation game as a test of machine intelligence. His work gave cognitive science its central hypothesis: that thinking might be a kind of computation.
• Warren McCulloch and Walter Pitts (1943) showed that simplified neurons can implement logic, linking brains and computers in a single formal framework.
• John von Neumann designed the stored-program architecture used by most computers and compared computers and brains explicitly in The Computer and the Brain (published 1958).
• Claude Shannon created information theory (1948), giving psychologists a way to quantify how much information a person can process; he also co-organised the Dartmouth workshop.
• Allen Newell and Herbert Simon, with programmer Cliff Shaw, built the Logic Theorist (1956), which proved theorems from Principia Mathematica, and the General Problem Solver (from 1957), which used means–ends analysis. They compared their programs to people’s think-aloud reports, effectively inventing cognitive modelling, and later formulated the physical symbol system hypothesis. They shared the 1975 Turing Award, and Simon received the 1978 Nobel Prize in economics for his work on bounded rationality.
• John McCarthy coined the term “artificial intelligence” in the 1955 proposal for the Dartmouth workshop and created the LISP programming language (1958), which became the standard language of AI research for decades.
• Marvin Minsky built an early learning neural network machine (SNARC, 1951), co-founded the MIT AI project with McCarthy in 1959, and later proposed “frames” for representing knowledge and the Society of Mind theory (1986).
• Frank Rosenblatt introduced the perceptron (1958), a neural network that learned from examples – a direct ancestor of modern deep learning.

These researchers gave cognitive science two competing traditions that still shape it: the symbolic tradition of rules and representations (Newell, Simon, McCarthy) and the neural network tradition of learning from connections (McCulloch and Pitts, Rosenblatt). For how the second tradition evolved, see connectionism and neural networks.

From Revolution to Discipline (1960–1979)

In 1960 George Miller, Eugene Galanter and Karl Pribram published Plans and the Structure of Behavior, replacing the reflex arc with the TOTE unit (Test–Operate–Test–Exit), a feedback loop borrowed from cybernetics. The same year Miller and Jerome Bruner founded the Harvard Center for Cognitive Studies. Ulric Neisser’s Cognitive Psychology (1967) gave the field its name and its first textbook, and models such as Atkinson and Shiffrin’s multi-store model of memory (1968) set the agenda for experimental work.

During the 1970s the separate strands were drawn together. The Alfred P. Sloan Foundation funded cognitive science programmes at several universities, and its 1978 state-of-the-art report pictured the field as a hexagon linking psychology, linguistics, computer science, neuroscience, philosophy and anthropology (see the disciplines of cognitive science). The journal Cognitive Science appeared in 1977, and the Cognitive Science Society was founded in 1979, holding its first conference at the University of California, San Diego.

Symbols, Networks and Levels (1980s)

David Marr’s Vision (1982) argued that any cognitive process must be understood at three levels: what is computed and why, the algorithm used, and how it is physically implemented. Jerry Fodor’s The Modularity of Mind (1983) proposed that perception and language rely on specialised, fast and encapsulated modules. Cognitive architectures such as John Anderson’s ACT (later ACT-R) and Newell, Laird and Rosenbloom’s Soar attempted unified theories of cognition.

In 1986 David Rumelhart, James McClelland and the PDP Research Group published Parallel Distributed Processing, and Rumelhart, Geoffrey Hinton and Ronald Williams popularised the backpropagation learning algorithm. Connectionism showed that networks could learn rules, such as English past tenses, without anyone writing the rules in – sparking a long debate with defenders of symbolic models.

The Brain Comes In: Cognitive Neuroscience (1990s)

The phrase “cognitive neuroscience” was coined in the late 1970s by Michael Gazzaniga and George Miller, but the field took off when imaging allowed researchers to watch healthy brains at work. PET studies of language in the late 1980s were followed by functional MRI, based on the BOLD signal described by Seiji Ogawa in 1990 and first used to map human brain activity in 1991–92. The 1990s were designated the “Decade of the Brain” in the United States. Neuroscience moved from the edge of cognitive science to its centre; see cognitive science vs neuroscience and our guide to research methods.

Embodied, Bayesian and Predictive Approaches

From the 1990s many researchers argued that the classical picture was too disembodied. The Embodied Mind (Varela, Thompson and Rosch, 1991), Rodney Brooks’s behaviour-based robots and Andy Clark and David Chalmers’s “extended mind” (1998) all stressed that cognition depends on bodies acting in environments – see embodied cognition. At the same time, Bayesian models treated perception and learning as probabilistic inference, and predictive processing (developed by Rao and Ballard, Karl Friston, Andy Clark and others) proposed that the brain is constantly predicting its own sensory input. The scientific study of consciousness also became respectable in this period.

Deep Learning and Large Language Models (2012–Today)

The neural network tradition returned in force when deep convolutional networks won the ImageNet image-recognition competition in 2012. The transformer architecture (2017) led to large language models, which reached the public with ChatGPT in late 2022. These systems are now both tools and objects of study for cognitive scientists: researchers test them with classic psychology experiments, use them as models of language processing in the brain, and debate whether learning from huge amounts of text resembles how children learn. The old arguments – innate structure versus learning, symbols versus networks, understanding versus imitation – are as live as ever. For more, see how AI is shaping cognitive science.

Looking back, each era of cognitive science was shaped by the technology of its day: clockwork and hydraulics, the telephone switchboard, the digital computer, the brain scanner and now the neural network. The history of the field is a reminder that our metaphors for the mind are tools – useful, but never the whole story.