A metric-based analysis of the contribution of spike timing to contrast and motion direction coding by single neurons in macaque area MT

Spike timing is thought to contribute to the coding of motion direction information by neurons in macaque area MT. Here, we examined whether spike timing also contributes to the coding of stimulus contrast. We applied a metric-based approach to spike trains fired by MT neurons in response to stimuli that varied in contrast, or direction. We assessed the performance of three metrics, Dspike and Dproduct (containing spike count and timing information), and the spike count metric Dcount. We analyzed responses elicited during the first 200 msec of stimulus presentation from 205 neurons. For both contrast and direction, the large majority of neurons showed the highest mutual information using Dspike, followed by Dproduct, and Dcount. This was corroborated by the performance of a theoretical observer model at discriminating contrast and direction using the three metrics. Our results demonstrate that spike timing can contribute to contrast coding in MT neurons, and support previous reports of its potential contribution to direction coding. Furthermore, they suggest that a combination of spike count with periodic and non-periodic spike timing information (contained in Dspike, but not in Dproduct and Dcount which are insensitive to spike counts and timing respectively) provides the largest coding advantage in spike trains fired by MT neurons during contrast and direction discrimination.

Research highlights
► We demonstrate that spike timing is a source of information — in addition to spike counts — that can potentially contribute to the coding of stimulus contrast in area MT neurons.
► We validate previous reports of the contribution of spike timing to coding of stimulus direction in MT neurons.
► We demonstrate that discrimination of motion direction and contrast by a theoretical observer model improves when using metrics that incorporate both, spike counts, and the information provided by spike timing relative to using spike counts only.
► Different from previous studies in MT, we apply and validate in our sample of neurons the metric-based approach proposed by Victor and Purpura (1996) to determine the potential role of spike timing in information coding.