Donald Hebb
Donald Olding Hebb (1904–1985) was a Canadian psychologist at McGill University whose 1949 book The Organization of Behavior proposed that thinking and learning arise from the growth of connections among nerve cells, an idea that later became both a standard hypothesis in neuroscience and the namesake of Hebbian learning in artificial neural networks. He was a psychologist first and not a computer scientist, but his theory supplied the vocabulary — connection strength, assemblies of cells, learning through activity — that early researchers in cybernetics and neural-network modeling reached for.
His path to the theory ran through the clinic and the laboratory rather than a straight academic line. Born in Chester, Nova Scotia, Hebb trained first at Dalhousie University and then at McGill, worked for a time as a schoolteacher and school principal, and went on to a doctorate at Harvard in 1936 under Karl Lashley, the physiologist who had spent years failing to find a single location for a memory trace in the rat brain. Hebb then worked with the neurosurgeon Wilder Penfield at the Montreal Neurological Institute, where he studied how people behaved after brain operations and injuries, and observed that large losses of tissue in adults sometimes spared intelligence while early damage in children could be devastating; he spent the 1940s at Queen's University and with Lashley at the Yerkes Laboratories of Primate Biology in Florida before returning to McGill in 1947, becoming chair of the psychology department in 1948.
The Organization of Behavior (Wiley, 1949) tried to bridge neurophysiology and psychology with a mechanism. Against the behaviorist habit of treating the brain as a black box, Hebb proposed that repeated stimulation builds a cell assembly, a group of neurons whose interconnections strengthen until activating part of the group tends to activate the whole, serving as the physical substance of a concept or percept. Sequences of assemblies, which he called phase sequences, explained trains of thought. Central to this was the growth hypothesis later called the Hebb synapse: when one cell persistently helps to fire another, the efficiency of that connection rises. He offered it as a plausible hypothesis, not an equation, and it was explicitly a statement meant to guide experiment.
The sensory deprivation experiments at McGill were the other half of his public reputation. In the early 1950s Hebb and his students, including Woodburn Heron, Thomas Scott and William Bexton, paid volunteers to lie for as long as they could bear it on a bed in a small cubicle, wearing translucent goggles, cuffs over the hands and a sound-muffling pillow, so that perception was reduced to unvarying noise. The report, "The Effects of Decreased Variation in the Sensory Environment," appeared in the Canadian Journal of Psychology in 1954. Participants did worse at simple thinking tasks and many experienced hallucinations and visual disturbances. The research, supported in part by Canada's Defence Research Board, was prompted by interest in how coerced confessions and isolation affect the mind, and Hebb argued that a brain needs a continuing stream of varied stimulation to function normally, a view that fitted his theory of activity-dependent organization.
His influence on computing was real but indirect. Nathaniel Rochester of IBM, one of the organizers of the 1955 Dartmouth proposal, led a team that simulated Hebbian cell assemblies on an IBM 701 in 1955–56, and Hebb's synapse was the template for the learning rules in Rosenblatt's perceptron era and for John Hopfield's 1982 associative memory. His Textbook of Psychology (1958) was widely used, and he was president of the American Psychological Association in 1960. A pitfall is to credit him with the famous slogan about cells that fire together wiring together, which came later from Carla Shatz, or to imagine that he specified an algorithm; he was cautious, wrote in terms of probable growth processes, and could not have known the molecular machinery of long-term potentiation, discovered by Bliss and Lømo in 1973, that gave his hypothesis experimental footing.
Later work refined and sometimes complicated his picture. Neuroscientists found that assemblies exist as coordinated ensembles in recordings from hippocampus and cortex, that connection strength depends on the precise timing of spikes as well as on correlation, and that inhibition and neuromodulators shape what is learned. Hebb's claim that brains build concepts from experience by strengthening connections is now such a standard background assumption that it can seem obvious; his achievement was to state it before the tools to test it existed.
Read Donald Hebb as the psychologist who turned the question of how experience is stored into a question about connections between cells, leaving behind a hypothesis modest enough to be tested and general enough to still shape both neuroscience and learning algorithms.