Gary G. Blasdel and Lawrence C. Sincich
Dept. of Neurobiology and Program in Neuroscience,
Harvard Medical School
The orientation selectivity of neurons in primary visual cortex is one of the more celebrated response properties in the mammalian nervous system, partly because it achieves something intuitively useful: the abstraction of edges from visual scenes. Despite its popularity, it has never been explained by any generally accepted mechanism. Hubel and Wiesel (1962) suggested originally that it arises from LGN afferents-specifically, from an oriented alignment of their receptive fields. And though this appears plausible in cat striate cortex, where it recently has been supported (Chapman et al., 1991; Reid and Alonso, 1995; Ferster et al., 1996), this explanation encounters difficulty in primates where the vast majority of LGN afferents contact neurons that are not selective for orientation. Alternative models have argued that orientation selectivity, along with other cortical response properties, arise through recurrent excitation-especially if laterally projecting axon systems are as abundant and as structured as they seem (Douglas and Martin, 1991; Somers et al., 1995). The problem (until recently) was that no one had linked orientation selectivity to local circuitry, as would appear to be required by this mechanism. All this changed with the discovery of aligned axon systems in tree shrews (Fitzpatrick et al., 1994, 1995), however, where pyramidal cells in superficial laminae were observed to terminate in discrete clusters at patterned intervals, that extended farther along axes representing their preferred orientations in visual space. Since the axon collaterals of these neurons provide monosynaptic, short-latency excitation onto other, laterally displaced pyramidal cells, with reciprocal excitatory connections, they establish a clear substrate for positive feedback (through recurrent excitation) that is biased along one axis. For visual signals arriving along a line of similar orientation, therefore, the amount of positive feedback should be greater since the associated axon collaterals spread farther in the aligned directions.
Accordingly, this circuitry could provide the 'tuned excitation' required by Douglas and Martin's model. In primates we have found similar correlations between oriented axon systems and the orientation preferences of neurons. Not surprisingly, they are not seen in layer 4c but in layers 2/3 and 4B, where the strongly orientationally selective responses are first observed. Since comparable oriented axon systems appear to have been found in cats (Lowell and Singer, personal communication), they could be a general feature of visual cortex in mammals, in which case they might also provide a cortical substrate for orientation selectivity that operates through recurrent, 'tuned' excitation.
Funded by NIH and Human Frontiers Science Program.