III.

Memory and Brain Structures

Until recently the brain processes on which memory depends could only be inferred from the study of patients with memory loss. Thus an operation in the 1950s to treat a Canadian patient known only as HM resulted in the destruction of his hippocampus and parts of his temporal lobes. The result was catastrophic; HM retains declarative memory only for events prior to the operation; the succeeding half century of his life he glimpses only fleetingly, remembering events for only a few seconds. The conclusion that the hippocampus is in some way crucial for the transfer of short- to long–term memory is generally accepted. Similarly, a brain region closely adjacent to the hippocampus, the amygdala, is necessary for the registration of the emotional content of memory. The advent of brain imaging (positron emission tomography, functional magnetic resonance imaging, magnetoencephalography) has confirmed the importance of these brain structures in undamaged brains too. In addition many cortical regions are engaged in working memory tasks, and the lower and medial regions of the frontal cortex in choice and decision-making based on memory.

A fundamental limitation to such studies is that even imaging does not allow one to discover the intimate physiological and biochemical mechanisms that might be required for memories to be made. At the beginning of the 20th century Ivan Pavlov, working in St Petersburg, brought learning and memory into the laboratory by showing that it was possible to teach dogs to associate the sound of a bell or a flash of light with the imminent arrival of food (so-called classical conditioning). By the 1930s the American psychologist B. F. Skinner invoked another form of learning when he showed that rats could be taught to press a lever or run a complex maze to obtain food (instrumental conditioning). And in 1948, Donald Hebb, in Montreal, proposed that such conditioning could result in alterations in the physical structure of synapses—the junction points between the nerve cells (neurons) in the brain—which would as a result generate new patterns of connections between the neurons. Granted that the hundred billion neurons in the human brain are connected by up to a hundred trillion synapses, such a model would allow ample possibility to encode the memories of a lifetime. Hebb’s hypothesis for this type of associative memory has opened the way for neural modellers to create computer-generated neural nets capable of showing “learning”. (See Cognitive Science.)

Hebb’s hypothesis also offered a route by which neuroscientists could begin to explore whether training animals on learning tasks actually did produce measurable synaptic changes. One early approach was to compare rats reared in “deprived” versus “enriched” environments; the latter, it could be assumed, would learn and remember more than the former. The “enriched” rats had a thicker cerebral cortex and increased numbers of synapses, in line with Hebb’s ideas, but to be sure that such changes were specifically related to learning and memory required a more precise experimental model and the development of analytical methods sensitive enough to measure the small changes in biochemical, morphological, or physiological properties that might be expected to result from such learning experiences. The study of memory has become the study of the brain events occurring when an animal learns and subsequently is called upon to show that it remembers a particular task—running a maze, finding food, escaping unpleasant stimuli, and so on.