|
Skaggs and McNaughton (1998) reported that rats shuttling between two
visually identical boxes connected by a corridor showed similar place
codes in each box. The degree of "partial remapping" between boxes
varied somewhat from day to day, but with no discernable trend. This
poses a problem for place cell theories in which the hippocampal map
is a crossproduct of extrinsic (local view) and intrinsic (path
integrator, or PI) variables, and failure to properly reset the PI
upen reentry into an environment leads to remapping. If the PI is
reset to an identical value upon each box entry, why is there any
remapping at all? One possibility is that the rat remembers whether
it entered the current box via a left turn (north box) or a right turn
(south box), but it would have to retain that information throughout
its visit. On the other hand, if the PI is not reset, why isn't
remapping complete?
We suggest that the PI is not reset upon box entry, but exerts a
relatively weak influence on map selection when visual cues are
identical in two environments. The primary contribution of the PI is
to constrain the nascent place code to a proper 2D manifold, but the
discrepancy in PI coordinates in the two boxes does induce some
differences, hence partial remapping. Hebbian learning would then
cause cells with fields in both boxes to become bound to two
sets of path integrator coordinates. The PI thus both causes partial
remapping and preserves it. We explore this hypothesis with computer
simulations.
In Knierim's (2002) double cue rotation experiment, place code
dissociation, individual place field splitting, and partial remapping
are attributed to weak attractor dynamics. But they also imply a weak
but nonzero PI influence. When some cells appear to follow local
landmarks whle others follow distal, the weak PI may remain bound to the
distal reference frame and compete with local landmarks for control of
place fields. Support Contributed By: NIH MH59932
|