非表示:
キーワード:
-
要旨:
Vocalizations or speech constitute dynamic inputs that are represented in auditory cortices by precise time-varying activity patterns. Such response patterns are typically analyzed by aligning spikes and sensory events using the experimenter’s clock, a laboratory-based reference not available to the brain. In contrast, neural systems must interpret time-varying responses using only intrinsic reference frames, a particularly challenging task for stimuli appearing suddenly or unpredictably. One solution could be provided by encoding information in the relative timing of neural responses, thereby exploiting intrinsic temporal reference frames. But it remains unclear whether and how sensory cortices implement a neural reference suitable for relative coding schemes. We investigate the viability of such a relative coding scheme in primate auditory cortex using a paradigm where naturalistic sounds were presented at random (unexpected) times. Recording neural responses in macaque auditory cortex we found that neurons clustered in two subsets with different properties. A set of stereotypical neurons responded rapidly and unselectively to individual stimuli with minimally varying latency, while another set of stimulus-selective neurons responded slowly and selectively with high latency variability. We then tested the hypothesis that the latency of the stereotypical neurons can provide a reliable and intrinsic reference frame for relative coding schemes. Specifically, we calculated the stimulus information carried by the selective neurons in different neural codes based on the relative timing of their neural responses to either a stereotypical neuron or another selective neuron. Two codes were considered: the relative onset latency between neurons and the full spike train of the selective neuron aligned to the response onset of a reference neuron. Information in latency relative to stereotypical neurons reached 91 of the information in latency with respect to the stimulus onset. The spike trains of the selective neurons relative to the stereotypical neurons contains 84 of the information in spike trains aligned to the stimulus onset, but only 41 relative to another selective neuron. At the population level, an estimate of the latency based on 20 stereotypical neurons allows preserving 95 of the information as measured with the experimenter's clock. We thus demonstrate that information in response latencies and sustained time-varying responses may be decoded by measuring these relative to another neuron’s or a population response. Stereotypical neurons responding unselectively and rapidly to various complex stimuli may serve as an early saliency signal that provides a reliable temporal reference frame that can be used to extract information in the responses of more selective neurons.
