Statistics of visual responses in primate inferotemporal cortex to 3 object stimuli 4 5

33 We have characterized selectivity and sparseness in anterior inferotemporal 34 cortex using a large data set. Responses were collected from 674 monkey 35 inferotemporal cells, each stimulated by 806 object photographs. This 806×674 36 matrix was examined in two ways: column-wise, looking at responses of a 37 single neuron to all images (single-neuron selectivity), and row-wise, looking 38 at the responses of all neurons caused by a single image (population 39 sparseness). Selectivity and sparseness were measured as kurtosis of 40 probability distributions. Population sparseness exceeded single-neuron 41 selectivity, with specific values dependent on the size of the data sample. This 42 difference was principally caused by inclusion, within the population, of 43 neurons with a variety of dynamic ranges (standard deviations of responses 44 over all images). Statistics of large responses were examined by quantifying 45 how quickly the upper tail of the probability distribution decreased (tail 46 heaviness). This analysis demonstrated that population responses had heavier 47 tails than those of single-neuron responses, consistent with the difference 48 between sparseness and selectivity measurements. Population responses with 49 spontaneous activity subtracted had the heaviest tails, following a power law. 50 The very light tails of single-neuron responses indicate that the critical feature 51 for each neuron is simple enough to have a high probability of occurring 52 within a limited stimulus set. Heavy tails of population responses indicate 53 that there are a large number of different critical features to which different 54 neurons are tuned. These results are inconsistent with some structural models 55