The receptive field is one of the most familiar concepts in sensory physiology. For a visual neuron, it is commonly defined as the area of visual space in which stimulus presentation leads to a change in neuronal activity. But stimuli outside the 'classical' receptive field can modulate a neuron's response to stimuli within this area. A new tracing technique promises to shed light on the anatomical connections that mediate this effect.

A good example of the 'non-classical' receptive field is contour integration. A neuron in the primary visual cortex (V1) will respond more strongly to a bar in its receptive field if there is another bar just outside this area that is colinear with the first bar, so that if the two were joined they would make a line. This 'colinear facilitation' is strikingly similar to the psychophysical finding that observers can detect a faint bar more easily if another, colinear bar is presented alongside it.

There is healthy debate about many aspects of V1 function, often centring on whether certain properties are intrinsic to V1 or are imposed upon it by feedback from higher visual areas, such as V2. And colinear facilitation is no exception. Gilbert et al. have developed a very precise anterograde tracing technique, and used it to compare intrinsic connections in V1 with those from V2 back to V1. Their results support the idea that intrinsic connections in V1 are more likely to mediate colinear facilitation.

The new technique uses an adenovirus that contains the gene for green fluorescent protein (GFP). When small amounts of virus are injected into the cortex, the virus enters neurons and the gene is transcribed. The GFP that is produced acts as a tracer, labelling not only the cell bodies but also all of the dendrites and axons. Because the label is produced within the neurons, rather than injected directly, there is no saturation at the injection site and it is easy to identify the cell bodies of the labelled neurons. When Gilbert et al. injected the adenovirus into specific locations in V1, it labelled intrinsic axonal connections that extended for up to 3.5 mm from the cell bodies at the injection site, corresponding to around 4° of the visual field.

The intrinsic connections were not uniform. Rather, they formed periodic clusters or patches, about every 0.75 mm. The authors found that these clusters corresponded to columns with the same orientation selectivity as the neurons at the injection site, making these connections good candidates for mediating colinear facilitation.

Connections from a single injection site in V2 covered a similar expanse in V1, although the projections were much less dense. However, there was no orientation specificity in these projections. This makes it less likely that feedback from V2 is responsible for contour integration in V1. As the authors point out, feedback from other visual areas, such as V3 or MT, might show orientation specificity — and the adenoviral tracing technique, which allows precise identification of labelled cells and quantification of their projections and geometry, should make it easier to address this question, as well as others relating to the functional anatomy of cortical connections.