Adam M. Sillito and Helen E. Jones
Department of Visual Science, Institute of Ophthalmology,
University College London
Visual processing in the central visual system is conventionally presented as a sequential process. This involves the local filtering properties of receptive fields constructed by the convergent organisation of afferent inputs and in V1 is for example reflected by the characteristics associated with layer 4 simple cells. However, the retinal input relayed via the lateral geniculate nucleus to this layer constitutes less than 20% of the drive to these cells, the remainder originates from cortical circuitry. This suggests that response properties may be more subtly organised than the pattern of afferent convergence implies. Indeed, work from a range of laboratories show that cells in V1 and even the visual thalamus exhibit sensitivity to visual context that extends beyond the receptive field (Knierim and Van Essen, J. Neurophysiol. 67: 961-980, 1992; Sillito et al, Exp. Brain Res. 93:6-16, 1993; Sillito et al, Nature 378:492-496, 1995; Cudeiro and Sillito, J. Physiol. 490:481-492, 1996; Jones et al, IOVS 37:S1058, 1996; Sillito and Jones J. Physiol (Paris) 90:205-209, 1996; Lamme, J. Neurosci. 15:1605-1615, 1995). Sensory systems are as much characterised by extensive top down projections as they are by the organisation of the bottom up, feed forward systems and it seems some of the context effects link to feedback influences from higher levels in the system. For example there are direct feedback connections from visual area MT (concerned with the processing of visual motion) to layer 4 and 6 of V1.
Both orientation and direction of motion context will influence the responses of cells in V1 to stimulation of the classical receptive field (CRF) and the response to stimuli involving orientation or direction contrast can for a section of the population exceed the response to a single stimulus localised to the classical receptive field (Sillito et al, Nature 378:492-496). These effects can be observed within, on the edge of, and significantly beyond the apparent confines of the CRF (Jones and Sillito, Soc. for Neurosci. Abs. 22:632.7, 1996) and follow from stimulation with sinusoidal gratings or texture patterns. The nature of the effect though can be strongly influenced by the spatial relationships and scale of the stimuli driving the CRF and providing the context. Further to this the temporal characteristics of the response to the various stimulus configurations suggests the interplay of several mechanisms, with some effects resolving later than others in the response. Broadly responses that can be linked to stimulation of the CRF seem to emerge earlier than those that link to the context dependent modulation of this, but again the detailed stimulus configurations exert a strong influence on the pattern of effect. Fine dissection of the temporal characteristics of even apparently strongly consistent responses to a particular range of stimuli indicates a degree of ``stochastic jitter'' of the response to the stimuli, as if the system is ``hunting'' around some set of foci. In detail they seem to suggest that local processing is always rephrased in terms of the more global context of the image. All this is more reminiscent of the behaviour of an interlinked multilevel network than local filters.