Nature and precision of temporal coding in visual cortex: a metric-space analysis

• DOI: 10.1152/jn.1996.76.2.1310
• Publication Date: 2017-12-25T00:30:34Z
• Authors: J. D. Victor; K. P. Purpura
• Keywords: Stimulus (psychology); Neural coding

1. We recorded single-unit and multi-unit activity in response to transient presentation of texture grating patterns at 25 sites within the parafoveal representation V1, V2, V3 two awake monkeys trained perform a fixation task. In experiments, stimuli varied orientation, spatial frequency, or both. contrast, check size, type, pairs these attributes. 2. To examine nature precision temporal coding, we compared individual responses elicited by each set terms families metrics. One family metrics, D(spike), was sensitive absolute spike time (following stimulus onset). The second D(interval), pattern interspike intervals. family, metrics depend on parameter q, which expresses coding. For q = 0, both collapse into "spike count" metric D(Count), is number impulses but insensitive their position time. 3. Each with values ranging from 0 512/s, used calculate distance between all trains dataset. extent stimulus-specific clustering manifest pairwise distances quantified an information measure. Chance estimated applying same procedure synthetic data sets were assigned randomly input stimuli. 4. Of 352 sets, 170 showed evidence tuning via count (q 0) metric, 294 272 interval attribute (contrast, type) under study. Across entire dataset, not attributable chance averaged 0.042 bits for 0.171 optimal 0.107 metric. 5. reciprocal cost serves as measure V1 highest contrast (ca. 10-30 ms) lowest type 100 ms). This systematic dependence provides possible mechanism simultaneous multiple attributes one train. 6. Our findings are inconsistent Poisson models trains. Synthetic firing rate governed time-dependent process matched observed poststimulus histogram (PSTH) overestimated induced D(count) and, low D(spike)[q] D(intervals)[q]. constructed modified process, preserved only PSTH also statistics accounted underestimated D(interval)[q].