Richard C. Atkinson

Introduction

Richard Chatham Atkinson, born 19 March 1929 in Oak Park, Illinois, is an American psychologist and cognitive scientist who did two careers in sequence, each of them consequential. In the first he built, with his graduate student Richard Shiffrin, the theory of memory that psychologists call the modal model: sensory register, short-term store and long-term store, with control processes that move information between them. Their 1968 paper is among the most cited works in the behavioural and cognitive sciences and still draws several hundred citations a year. In the second career he stopped doing the research and began administering it, as director of the National Science Foundation, chancellor of the University of California San Diego, and president of the University of California. The two halves are connected by a single conviction, that the way to understand a cognitive system is to model it formally and then test the model against behaviour, applied first to memory and then to institutions.

Early Life and Education

Atkinson took his bachelor's degree at the University of Chicago in 1948 and his doctorate at Indiana University in 1955, with a thesis on rote serial position effects written as a statistical model and supervised by Cletus J. Burke. The training was in mathematical psychology at the point when the field was being invented, and it gave him the toolkit he later applied to memory: specify the structure, write the equations, and let the data decide. After two years in the United States Army he joined Stanford in 1956, first in the Applied Mathematics and Statistics Laboratories.

Academic and Professional Career

He spent three years at UCLA as an assistant professor, then returned to Stanford in 1961, where he stayed until 1980 with appointments that accumulated as his interests widened: psychology, education, the Applied Mathematics and Statistics Laboratories, the Institute for Mathematical Studies in the Social Sciences, and the School of Engineering. He chaired the Stanford psychology department from 1968 to 1973 and was a founding editor of the Journal of Mathematical Psychology.

At Stanford he also worked with Patrick Suppes on computer-assisted instruction, building programs that taught mathematics and reading to young children and kept a running response history for each student, updated moment by moment and used to decide what that child should study next. The two later founded Computer Curriculum Corporation, the first company to put computers into classrooms. The Stanford Education Program for Gifted Youth descends from those experiments.

In 1975 he moved to Washington as deputy director of the National Science Foundation, becoming acting director and then, appointed by President Carter, director from 1977 to 1980. He negotiated the first memorandum of understanding between the United States and the People's Republic of China for the exchange of scientists and scholars, a precursor of the science and technology agreement signed in January 1979, and he established the NSF's industry-university cooperative research programme at a time when such collaboration was rare, elevated engineering to a full directorate, and initiated the policy work that led to the Bayh-Dole Act of 1980.

He was chancellor of UC San Diego from 1980 to 1995 and president of the University of California from 1995 to 2003. His honours include election to the National Academy of Sciences, the American Academy of Arts and Sciences, the National Academy of Education, the Institute of Medicine and the American Philosophical Society, the American Psychological Association's Distinguished Scientific Contribution Award in 1977, its E. L. Thorndike Award in 1980, the Revelle Medal in 1995, the APA Award for Lifetime Contributions to Psychology in 2002, and the Clark Kerr Award in 2018. A mountain in Antarctica bears his name.

Major Contributions to Psychology

The 1968 paper with Shiffrin proposed a memory system whose structure is fixed and whose control processes are not. Information enters a sensory register, passes under attention into a short-term store of limited capacity, and is transferred by rehearsal to a long-term store from which it can be retrieved. Encoding, storage, retrieval and the decision rules governing them are variable and under the subject's control. The point of the formalisation was that one could derive specific models for specific tasks, predicting performance on memory types and on procedures such as the Brown-Peterson task, rather than describing memory in prose. It merged mathematical psychology with the emerging computer metaphor, and for a decade it was the framework against which every other account was defined.

Mathematical models of learning and memory

Before and alongside the modal model, Atkinson wrote a series of formal models of paired-associate learning, rote serial learning and short-term retention, generally with colleagues and students, that established the style of work: a model precise enough to fail. Shiffrin went on to extend the tradition into the SAM retrieval models with Jeroen Raaijmakers.

Instruction as an engineering problem

The computer-assisted instruction work with Suppes treated teaching as a control system, with a model of the learner updated from response data and used to choose the next item. It produced the first commercial classroom computing company and, arguably, the intellectual ancestry of modern adaptive learning software.

Science policy

His NSF tenure changed how American universities relate to industry and to their own intellectual property, through the industry-university programme, the work behind Bayh-Dole, and the advocacy for research tax credits. It is not psychological research, but it shaped the conditions under which psychological research is funded.

Major Publications

Atkinson, R. C., & Shiffrin, R. M. (1968). Human memory: A proposed system and its control processes. In K. W. Spence & J. T. Spence (Eds.), The Psychology of Learning and Motivation (Vol. 2, pp. 89-195). New York: Academic Press.
Atkinson, R. C. (1968). Computerized instruction and the learning process. American Psychologist, 23, 225-239.
Atkinson, R. C., & Wilson, H. A. (Eds.). (1969). Computer-Assisted Instruction: A Book of Readings. New York: Academic Press.
Atkinson, R. C. (1967). Instruction in initial reading under computer control: The Stanford project. Educational Data Processing, 4, 175-192.

Impact and Legacy

The modal model is one of the small number of theories that organised a field. Fifty years after publication the journal Memory and Cognition devoted a special issue to the research it generated, and in 2023 the Journal of Memory and Language reprinted the 1968 paper with a historical commentary. Its vocabulary, sensory memory, short-term store, long-term store, rehearsal, is still how memory is taught.

It was also substantially dismantled, and the dismantling is the more interesting part of the story. The model's central empirical pillar was that maintenance rehearsal transfers material into long-term storage, and Craik and Watkins showed in 1973 that the amount of rehearsal did not predict later recall while the depth of processing did. That result, and the levels of processing framework of Craik and Lockhart in 1972, attacked the premise that memory is divided into structural stores at all, arguing instead that what matters is the kind of processing an item receives. Separately, Baddeley and Hitch's working memory model of 1974 replaced the single short-term store with a multi-component system having separate verbal and visuospatial slave systems under a central executive, because the unitary buffer could not accommodate the neuropsychological and dual-task evidence. What survives of Atkinson and Shiffrin is the architecture as a first approximation and the methodological standard it set, not the specific claims about rehearsal and capacity.

There is also a standing complaint about the theory's structure. Because the control processes were left variable, the model could be adapted to a wide range of findings by adjusting them, and critics argued that this made it difficult to falsify: a result that contradicted a prediction could be accommodated by changing an assumed strategy rather than by abandoning the theory. Defenders reply that this flexibility was the point, since the strategies genuinely do vary between subjects and tasks, and that the formalism made those assumptions explicit rather than hidden.

Atkinson's own relationship to the theory is worth noting. He left memory research for administration in 1975, seven years after the paper appeared, so the theory's subsequent development, defence and revision were carried out largely by others, Shiffrin above all. He is the author of a framework he did not spend his career defending.

The applied work has aged less comfortably. Computer-assisted instruction of the kind he built with Suppes produced a commercial company before there was convincing evidence that it improved learning outcomes, and the literature on its effectiveness has remained mixed ever since, with effect sizes that shrink as controls improve. That he and Suppes founded a company commercialising the approach while evaluating it as researchers is the obvious structural criticism, and it is the same criticism that attaches to much of educational technology since.

Learn More

Types of memory, the taxonomy the modal model was built to explain.
Levels of processing, the framework that challenged it.
Brown-Peterson task, the procedure used to test short-term retention.
Kenneth Spence, who edited the volume the 1968 paper appeared in.
All Psychologists — browse the full biographical collection.

References

Atkinson, R. C., & Shiffrin, R. M. (1968). Human memory: A proposed system and its control processes. In Psychology of Learning and Motivation (Vol. 2, pp. 89-195). Academic Press.