Natural Stimulus Statistics Alter the Receptive Field Structure of V1 Neurons

Abstract

Studies of the primary visual cortex (V1) have produced models that account for neuronal responses to synthetic stimuli such as sinusoidal gratings. Little is known about how these models generalize to activity during natural vision. We recorded neural responses in area V1 of awake macaques to a stimulus with natural spatiotemporal statistics and to a dynamic grating sequence stimulus. We fit nonlinear receptive field models using each of these data sets and compared how well they predicted time-varying responses to a novel natural visual stimulus. On average, the model fit using the natural stimulus predicted natural visual responses more than twice as accurately as the model fit to the synthetic stimulus. The natural vision model produced better predictions in Ͼ75% of the neurons studied. This large difference in predictive power suggests that natural spatiotemporal stimulus statistics activate nonlinear response properties in a different manner than the grating stimulus. To characterize this modulation, we compared the temporal and spatial response properties of the model fits. During natural stimulation, temporal responses often showed a stronger late inhibitory component, indicating an effect of nonlinear temporal summation during natural vision. In addition, spatial tuning underwent complex shifts, primarily in the inhibitory, rather than excitatory, elements of the response profile. These differences in late and spatially tuned inhibition accounted fully for the difference in predictive power between the two models. Both the spatial and temporal statistics of the natural stimulus contributed to the modulatory effects.

Predicting neuronal responses during natural vision

“…These stimuli mimicked the stimulation occurring in an area three to four times the CRF diameter during inspection of a natural scene with voluntary eye movements (Vinje & Gallant 2000;David et al 2004). A brief segment taken from a natural vision movie appears at the top of Figure 1A.…”

“…At the same time, functional models of visual neurons have been developed largely using simple synthetic stimuli, such as bars, sine wave gratings and white noise (Hubel & Wiesel 1959;DeValois et al 1982;Jones et al 1987). The question of how well models developed using synthetic stimuli generalize to natural vision has received little attention (Vinje & Gallant 1998;David et al 2004). This is an issue of fundamental importance for understanding visual processing.…”

The Embedded Neuron, the Enactive Field?

“…If V1 neurons were linear filters, this would not be the case; RF models would generalise between classes of stimuli. So the stimulus specificity of RF predictions that David et al (2004) report, which seem to be due to crucial nonlinearities in V1, again points to a more dynamic notion of RF than was originally conceived.…”

“…Two studies (Ringach, Hawken & Shapley 2002, Smyth, Willmore, Baker, Thompson & Tolhurst 2003 have shown that receptive field maps generated from responses to natural images resemble the elongated, oriented fields derived from the classic grating experiments. In contrast, David, Vinje and Gallant (2004) have made the case that RF models generated by stimulation with natural stimuli give significantly better predictions of responses to novel natural stimuli than do RF models generated by stimulation with gratings. Likewise, the predictions of responses to novel grating stimuli were superior if the RF model had been constructed from the correlations of responses to gratings.…”

“…It is sometimes suggested (e.g. David et al 2004) that the RFs revealed by natural stimuli are so complex that a natural image derived model is only good at predicting responses to other natural stimuli. Adopting natural images as canonical stimuli thus entails giving up on the idea that once one fully characterises a neuron's RF, the characterization can then be used to predict responses to any stimuli.…”