| |
Abstract:
There is ample evidence for dissociations between declarative
and nondeclarative memory, but little is known about how these
systems work together in normal individuals. Evidence from animal
studies has shown that caudate-based nondeclarative memory and
medial temporal lobe-based declarative memory may compete with one
another, such that performance of MTL-lesioned animals can exceed
controls. Using fMRI at 1.5T, we examined neural activity during
learning of a task that is known to rely upon the striatum,
probabilstic classification. Twelve subjects received four scans
that alternated between a "weather prediction" task (36
trials/scan) and a simple baseline task. Across scans, subjects
acquired the classification skill (mean accuracy 85% in scan 4).
Performance of the classification task was associated with
activation in bilateral prefrontal, parietal, and occipital
cortices, and bilateral caudate nuclei, replicating previous
findings. Region-of-interest (ROI) analyses showed that caudate
activation decreased slightly with learning. The medial temporal
lobe (MTL) was deactivated during classification compared to
baseline, and ROI analyses showed that this deactivation persisted
beyond 100 trials. Connectivity analyses demonstrated a negative
correlation between the medial temporal lobe and the caudate,
consistent with competition between these structures. These
findings converge with results from animal studies to suggest that
striatum and MTL compete with one another during learning.
|
