Figure 3. The intermittent structure of natural stimuli (A) leads to non-Gaussian statistics (B,C). Different patches of natural stimuli have different variances. As a result, even if the distribution at one moment in time or space can be described by a Gaussian function, the average distribution will be strongly non-Gaussian. This can be seen in the log-Plot (C) where the resultant distribution decreases linearly with the deviation from the mean, whereas Gaussian distributions correspond to parabolas in these coordinates.