Disentangling temporal and rate codes in primate somatosensory cortex

Abstract

AO_SCPLOWBSTRACTC_SCPLOWMillisecond-scale temporal spiking patterns encode sensory information in the periphery, but their role in cortex remains controversial. The sense of touch provides a window into temporal coding because tactile neurons often exhibit precise, repeatable, and informative temporal spiking patterns. In somatosensory cortex (S1), for example, responses to skin vibrations exhibit phase-locking that faithfully carries information about vibratory frequency. However, the respective roles of spike timing and rate in frequency coding are confounded because vibratory frequency shapes both the timing and rates of S1 responses. To disentangle the contributions of these two neural features, we measured S1 responses as animals performed a frequency discrimination task, in which differences in frequency were accompanied by behaviorally irrelevant variations in amplitude. We then assessed the degree to which the strength and timing of S1 responses could account for the animals performance on the task. First, we showed that animals can discriminate frequency, but their performance is biased by amplitude. Second, rate-based representations of frequency are susceptible to changes in amplitude, but in ways that are inconsistent with the animals behavior, calling into question a rate-based code for frequency. In contrast, timing-based representations are impervious to changes in amplitude, also inconsistent with the animals behavior. We account for the animals behavior with a model wherein frequency coding relies on a temporal code, but frequency judgments are biased by perceived magnitude. Our results constitute further evidence for the role of millisecond-scale spike timing in cortex. SO_SCPLOWIGNIFICANCEC_SCPLOWO_SCPCAP C_SCPCAPO_SCPLOWSTATEMENTC_SCPLOWWhile neurons in the cerebral cortex are known to produce temporally patterned responses with single-digit millisecond precision, the role of this patterning remains controversial. The alternative hypothesis is that the critical feature of neuronal responses is the rate at which spikes are emitted, and the distribution of the spikes in time only matters at much slower time scales (tens or hundreds of milliseconds). To disentangle these two putative neural codes, we trained animals to discriminate the frequency of skin vibrations while recording from somatosensory cortex. We show that only a timing-based representation of frequency can account for the performance of the animals, though overall population firing rate biases their frequency judgments.