Fig. 2.

Example neurons. Each row depicts a single example neuron, where the responses of the top 3 neurons reflect relatively pure selectivity and the responses of the bottom 3 neurons reflect mixtures of different selectivity types. Left: the mean spike count responses computed within a window 50–250 ms after stimulus onset to each of the 16 conditions (the “response matrix”), averaged over 20 repeated trials, normalized to range from the minimum (black) to the maximum (white). Center: the orthonormal components with the 5 largest weights, plotted as shown in Fig. 1D but with intensity scaled by the weight applied to each component. The response matrix can be reconstructed as a weighted sum of these matrices (once the grand mean spike count is also factored in, which is not shown). Right: the temporal evolution of the closed-form bias-corrected signal modulation magnitudes for each type of signal, computed as the square root of the sum of squares of the bias-corrected weights normalized by the number of components for each signal type (Eq. 8). To perform this analysis, spikes were counted in 50-ms sliding windows shifted 1 ms for each successive time bin. The example neuron depicted in the 4th row was recorded in inferotemporal cortex (IT); the other neurons were recorded in perirhinal cortex (PRH).