PATTERN COMPLETION AND PATTERN SEPARATION IN CA3 DURING MORPHING OF TWO ENVIRONMENTS

Jill K Leutgeb¹*; Stefan Leutgeb¹; Alessandro Treves^1,2;
Marianne Fyhn¹; Retsina Meyer³;
Carol A. Barnes^1,3; Bruce L. McNaughton^1,3;
May-Britt Moser¹; Edvard I. Moser¹

1. Ctr Biol Memory, NTNU, Trondheim, Norway

2. SISSA Intl Sch Adv Studies, Trieste, Italy

3. NSMA, Univ Arizona, Tucson AZ

Computational models suggest that the hippocampus can perform both pattern completion and pattern separation, depending on the degree of similarity between sensory inputs. To test this idea, we recorded simultaneously from CA3 and CA1 while rats ran in a square environment that was gradually morphed into a circle, or vice versa, with all tests occurring in the same location in the same room. In phase 1, the rats were trained in randomly presented circle and square enclosures for 8 sessions x 10 min. The overlap of activity in the two shapes was significantly less in CA3 than in CA1 after correction for differences in sparsity, suggesting more efficient pattern separation in CA3 (Leutgeb et al., SfN 2003). In phase 2, one shape was gradually morphed into the other, using 5 intermediate shapes. CA3 cells remained close to their initial firing rates during the early parts of the morph sequence, suggesting that pattern completion occurs as the shape of the box starts to change. In the later parts of the morphing sequence, the majority of active CA3 cells stopped firing, with new cells becoming active (pattern separation). More heterogeneous responses were seen in CA1, where firing rates were modulated more continuously throughout the sequence. CA1 cells often had the highest rates in one of the intermediate shapes. These results demonstrate pattern completion and pattern separation in CA3, consistent with the idea that CA3 operates as an autoassociative network. CA1, in contrast, seems to be influenced to a larger extent by the current sensory inputs. The results also imply that the selection of which cells fire in CA1 for a given episode of experience may be determined more by the direct inputs from EC layer III than by the Schaffer collaterals.

Supp. By Norw. Res. Council Ctr. Excellence Grant.