Extended Data Fig. 2: Mouse movement behavior and CA2 and CA1 neuron spatial decoding properties in various conditions.

a–b, There was no significant difference between (a) mean velocity (ANOVA, p = 0.36, n = 6,3,5,5 mice), or (b) average path length (ANOVA, p = 0.71, n = 6,3,5,5 mice) in the different experimental groups of mice. Data is presented as mean ± SEM. c, The relationship between the number of cells used to decode position by an SVM linear classifier and the accuracy of decoding for CA1 (left) and CA2 (right) cells. Note: CA2 spatial decoding accuracy was not significantly greater than chance levels in any of our recordings for any number of cells. CA1 decoding accuracy became greater than chance as additional cells were added to the decoder. Even with 30 CA2 cells, spatial decoding accuracy was less than chance and less than decoding accuracy with half as many CA1 cells. d, Example plots of real (black) versus predicted position by the model (red). e, CA1 population activity decoded position significantly better than chance in all sessions of the 3-chamber task. Data shown for two mice (CA1_1, n = 21 neurons; CA1_2, n = 27 neurons). f, CA2 population activity did not decode position better than chance in any 3-chamber task session (left graphs) or during the four individual ten-minute sessions of the pre-habituation session (right graphs). Data shown for two mice (CA2_2, n = 25 neurons; CA2_3, n = 31 neurons). Box plots display the center line as the mean; box limits are upper and lower quartiles; whiskers show min to max values in data sets. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.