Memory Storage Fidelity in the Hippocampal Circuit: The Role of Subregions and Input Statistics
A: Recall performance in the model as proposed in . Different colors show mean correlation between reconstructed patterns and stored ones in different regions; dashed lines show performance in a simulation where the recurrent connections in CA3 were turned off. B: Histogram of CA1 cell firing during storage. When sparsity is 0.01 (magenta) each cell fires about one time. This grandmother-like coding is abandoned if sparsity is 0.1 (black). C: Recall performance in CA1 (red) and EC (blue) for sparsity 0.1 measured for the last 10 patterns stored (stars) and for the first 10 (diamonds). Abandoning the grandmother-like code leads to a breakdown in performance by forgetting previously stored patterns. D: Recall performance in EC when connectivity from CA1 to EC is not complete and sparsity in CA1 is 0.01 (diamonds). A grandmother-like code cannot reproduce the whole pattern if the connectivity is sparse. When CA3-CA1 is a hetero-association with sparsity 0.1 (stars) diluting the connectivity has a milder effect. E: Our model as described in text yields similar results as in A, but is biologically more plausible, we believe.