Department of Psychological and Brain Sciences
Computational Memory and Perception Lab
University of Massachusetts Amherst
Banishing Ghosts of Process from the Neural Machinery of Human Memory: A Representational Hierarchical Account of Cognition
In cognitive neuroscience, mental processes or cognitive functions have often served as labels for characterizing the functional division of labor in cortex. For example, distinct neuroanatomical substrates have been proposed for declarative memory versus perceptual learning, short-term versus long-term memory, and familiarity versus recollection. I will present theoretical and empirical work, including connectionist simulations, behavioral studies and fMRI data, that challenges process-based accounts of memory and visual cognition. Instead, I will argue that the ventral visual stream and medial temporal lobe (MTL) can be understood in terms of a hierarchy of representations – from simple features in V1, through feature-conjunctions in inferior temporal cortex, to complex conjunctions of items and context in MTL. Under this view, the processes that operate upon the representations to produce a behavioral output – e.g., “pattern completion” to produce recall of a memory – can occur anywhere along the hierarchy.
The studies I will present demonstrate how a single brain region can support more than one cognitive process (visual discrimination and recognition memory); that a cognitive process (recollection) previously thought tied to one brain region (hippocampus) can also unfold elsewhere; and that this “Representational-Hierarchical” framework makes accurate predictions for the organization of representations within the ventral visual stream, and for the effects of aging on recognition memory. This suggests that it may be possible to banish the ghosts of process from the neural machinery of memory, and explain the neuroanatomical organization of cognition in terms of representations alone.
A pizza lunch will be served.