No other receptive field property characterizes the neurons of the visual cortex like orientation selectivity. The great majority of neurons in the primary visual cortex of many carnivores and primates are exquisitely sensitive to the orientation of a stimulus. Yet the relay cells of the lateral geniculate nucleus (LGN), which provide the cortex with most of its information about the visual image, respond equally well to a stimulus at any orientation. In at least some species, including cats, this remarkable and quintessentially cortical property emerges fully formed at a single synapse, between thalamic axons and their targets in the cortical layers.
Because it is such a striking phenomenon, because it is relatively easy to measure, and because it is so strongly linked with the function of the visual cortex, orientation selectivity and the mechanisms that give rise to it have been subjected to intense study and debate. Much of the cerebral cortex performs tasks that are dauntingly complex, difficult to characterize, and only just becoming experimentally approachable. Although a complex spatial transformation, extracting the orientation of an image element is still relatively straightforward and tractable. No wonder orientation selectivity has become one of the standard models for how the synaptic circuitry of the cortex performs a complex computation.
The roots of the longstanding controversy over the synaptic mechanisms underlying orientation selectivity lie in the complexity of the cortical circuit. It is easy to say that orientation selectivity in cats emerges at a single synapse between the terminals of geniculate relay cell axons and the cortical cells that they excite. But these same cortical cells receive thousands of synapses altogether, and from many different sources. Determining which of these broad categories of inputs -- thalamic excitatory, intracortical excitatory, intracortical inhibitory, or some combination of all three -- gives rise to orientation selectivity and how they do so has proven to be a surprisingly difficult task. At issue is not just which of the various pathways contribute, but the entire nature of the cortical computation, whether orientation selectivity arises from a feedforward filtering of the thalamic inputs to the cortex, or from a more dynamic, feedback process that encompasses the entire cortical circuit.
We shall focus our discussion on cat V1, for two reasons. First, the vast majority of cells in layer 4, the cortical layer that receives the dominant LGN input, are orientation selective in cat, though the same is not true in many other species e.g. monkeys (Blasdel and Fitzpatrick, 1984, Hawken and Parker, 1984), ferret (Chapman and Stryker, 1993), and tree shrew (Humphrey and Norton, 1980). Second, the synaptic physiology underlying orientation selectivity is by far best studied in cats.