The nature of memory

Randy Gallistel

Rutgers University

In the behaviorist framework within which most neuroscientific discussions of memory occur, memory is conceived of as the rewiring (changes in synaptic conductance) produced by experience, which rewiring adapts the animal's behavior to its circumstances. In this conception, the problem of coding does not arise. From an information processing perspective, on the other hand, memory is the mechanism that carries information forward in time. Formulated this way, the coding question is central to our understanding of the mechanism of memory. Yet, the coding question is seldom raised in discussion of the neurobiological mechanisms of memory and the most popular suggestions regarding this mechanism (e.g., LTP) do not offer any very plausible solutions to it. Moreover, current autoassociative memory proposals based on something like this mechanism have shockingly poor quantitative properties. It is argued that we should shift our focus to the coding problem