Cornell University Medical College
We introduce several families of metrics between spike trains as a tool to study the nature and precision of temporal coding. Each metric defines the distance between two spike trains as the minimal "cost" required to transform one spike train into the other via a sequence of allowed elementary steps, such as inserting or deleting a spike, shifting a spike in time, or changing an interspike interval length. For many of these metrics, efficient computational algorithms (essentially modelled on the Sellers (1974) algorithm for genetic sequences) exist.
The geometries corresponding to these metrics are in general not Euclidean, and distinct families of cost-based metrics typically correspond to distinct topologies on the space of spike trains. Each metric, in essence, represents a candidate temporal code in which similar stimuli produce responses which are close and dissimilar stimuli produce responses which are more distant.
We applied this technique to single-unit and multi-unit activity recorded in the parafoveal representation of V1 and V2 in two awake monkeys trained to perform a fixation task. Metrics based on spike times provided a higher degree of stimulus-dependent clustering than metrics based on spike intervals or metrics which ignored temporal structure. The precision of temporal coding varied systematically with the attribute being encoded: temporal precision was highest for contrast and lowest for texture type. This systematic dependence provides a possible mechanism for the simultaneous representation of multiple stimulus attributes in one spike train. We also used these metrics to investigate the statistical structure of spike trains. We found systematic deviations from time-dependent Poisson processes, and determined that these deviations are capable of carrying stimulus-specific information.