Some Anomalies in Memory Theory
4:30 p.m. 24 May 2007
Given the importance of memory for both perception and cognition, it is surprising to find that current theories of learning in the nervous system appear to be in a state of discord. In this seminar, I will describe some experiments using brain imaging techniques and recordings of electrical activity, mostly in the visual and association cortices.
These experiments show that, after the onset of a stimulus, the initial activity in the primary visual cortex (V1) is followed by an efferent (descending) global signal, which modifies processing in V1. In memory experiments, these ‘top-down’ responses, which are modulated by attention, can be used to predict what will be recalled in subsequent tests. These findings support the view that perception is an active process.
However, neuroscientists investigating the molecular basis of memory do not always obtain their results from intact, awake animals. Instead, many researchers use isolated networks of cells, or anaesthetised animals. Further, computational modellers of neural networks generally ‘clamp’ inputs into their models, neglecting the role of attentional selection in learning.
In this respect, both groups of investigators ignore the dependence of learning on the motivational states of organisms. As such, results obtained from neuroscientific studies and computer simulations that neglect the ‘top-down’, global signals cannot be considered complete.
In this talk, I will also argue that the discovery of widespread synchronous neuronal firing (to within ~1 ms) in the neocortex without memory formation necessarily taking place casts doubt on Hebb’s rule. According to this rule, near-synchronous firings of pre- and post-synaptic neurons lead to the establishment of memory. The rule has been used extensively to model memory and to account for the development of cognitive functions.