Quantifying neural codes

Introduction Explorations of sensory processing are founded on the model that, if two sensory stimuli can be discriminated, their associated neural representations must also be in some way distinct. Thus, a fundamental challenge in neurobiology is to discover the essential differences in the neural representations of two perceptually discriminable stimuli. There is general agreement that, in mammals, the cortical representations of even the simplest stimuli consist of thousands of active neurons, meaning that the distinguishing features of responses might take the form of highly complex spatial–temporal codes. Our aim here is to compare candidate cortical population codes by identifying features of the neural response that might underlie stimulus discrimination. We then systematically quantify the contribution of these features to the cortical population code. We review recent work that addresses three key issues. First, the spatial organisation of neural coding — is the essential information that encodes each stimulus carried by a widespread network or by a restricted subset of neurons? Second, the role of spike timing — are stimuli encoded entirely by the number of spikes occurring over long time intervals or does the millisecond precision of spike times convey additional information? Third, the role of spike correlations — does each spike code the stimulus independently, or do correlated spike patterns convey additional information? As an experimental paradigm, we focus on how the cortical population response specifies the location of a punctate stimulus. Finally, we briefly discuss whether the conclusions from this work might generalise to other species and other sensory systems.