Subtraction inhibition combined with a spiking threshold accounts for cortical direction selectivity.

We have modeled simple-cell direction selectivity by a nonlinearity consisting of a subtraction inhibition followed by half-wave rectification and compared the performance of this model to that of different versions of the elaborated Reichardt detector for similar inputs and parameter settings. Not only does the subtraction model fit the experimental data more closely than the elaborated Reichardt detector, but the subtraction model also is more plausible from a physiological and anatomical point of view. Moreover, the subtraction model operates optimally at plausible spatiotemporal parameter settings. Therefore, we conclude that there is no need to invoke specific synaptic interactions, such as implied in the Reichardt detector, to account for simple-cell direction selectivity.