Fig. 3. CA3 place fields exhibit relative lap-by-lap stability while CA1 place fields shift with experience in novel environments.
a Pearson’s correlation coefficient of all neurons with place fields within N (average activity map in first 10 laps versus next 10 laps of session: CA1 n = 1238, CA3 n = 255 neurons). Bootstrapped mean difference (∆) between CA1 and CA3 shown on the right. b Example place cells with stable or shifting place fields (PFs) in N (lap-by-lap activity for the first 25 laps). White arrows depict the direction of shifting. c, Linear regression analysis on the center of mass (COM) of individual PFs (CA1 in N) shows a high number of significantly shifting PFs (blue and red dots and bars). The slope measures the direction and amplitude of the PF shift: for significantly shifting PFs, backward is blue, forward is red. CA3 has a lower proportion of significantly shifting PFs than CA1. d Histograms of the COM difference between last and onset laps (five-lap weighted average, see Methods) normalized by the number of laps for each PF in N for CA1 and CA3. Shuffling the lap numbers (gray distributions) reveals an overrepresentation of backward shifting PFs, as confirmed by the bootstrapped mean difference (∆, insets). e Population shift of COM. Left, mean ±SEM, over all PFs, of five-lap sliding average COM difference relative to lap 12 (see Methods), in N. CA1 (green) and CA3 (orange). Linear regression analysis (on all data points, not means) shows that the population of PFs shifts backwards significantly, both in CA1 and CA3. Linear regression, F test, ***P < 0.001, CA1, P < 1 × 10−100, CA3, P = 1.3 × 10−4. Right, The CA1 data set was resampled 1000x using n = 175 PFs to match the number of CA3 PFs and the slope and intercept of the regression line were measured each time (green dots). CA1 slopes are always steeper than the CA3 slope indicating that the CA1 population shifts significantly faster than CA3.