Vision: in the brain of the beholder

It is commonly thought that neural activity in the visual cortex reflects retinal input. Recent studies, however, suggest that patterns of cortical activity are mostly intrinsically generated, and that visual input exerts but a modulatory influence. At first glance, it seems obvious that what the visual cortex does is an analysis of the information conveyed by the retinal images in the two eyes. One would therefore expect that there should be a rather strict relationship between retinal input and the output of the visual cortex – the spiking activity of its neurons. But researchers in the field know that many neurons display a high variability of response when the same stimuli are shown repeatedly. This variability is attributed to the spontaneous activity of cortical neurons, which in turn is usually dismissed as ‘noise’. A recent study by Fiser et al. [1], however, provides evidence that spontaneous activity may shape the response to sensory stimuli, which systematically improves during early postnatal development. That spontaneous activity might not be as random as previously thought has been demonstrated in a number of studies of its spatio-temporal pattern at various levels of the visual pathway. Patterns of locally correlated bursting activity were originally discovered in the developing retina, where they became known as retinal waves [2]. Later they were also observed in the lateral geniculate nucleus [3] and the primary visual cortex, V1 4. and 5.. Interestingly, correlated spontaneous activity in V1 exhibits a similar kind of patchy organization as exhibited by horizontal intracortical connections [5], and periodic fluctuations in correlated activity match with periodic changes in ocular dominance [6]. Even more surprising than the mere presence of spatio-temporal patterns in spontaneous activity was the finding that those patterns often match those observed in response to actual visual stimuli. Using voltage-sensitive dye imaging, Kenet et al. [7] recorded and compared patterns of spontaneous activity and those elicited with full-field gratings of various orientations ( Figure 1). Spontaneous activity patterns were highly dynamic, but at any moment in time they often corresponded quite closely to an orientation map in response to visual stimulation. This finding raises the question, addressed by Fiser et al. [1], whether spontaneous activity is not just ‘noise’ but plays an important role in how neurons respond during natural viewing