Abstract
Synchronized neural responses, which often are accompanied by oscillations in the gamma frequency band, exist extensively in visual cortex and are proposed as supporting perceptual mechanisms. Neural synchrony and oscillation are normally studied with cross-correlation analysis and coherence analysis respectively. We studied responses from cat visual cortex to explore the relationship between synchrony and coherency. With a Cyberkinetics 10×10 microelectrode array, we recorded 66 complex cells from areas 17 and 18 in two paralyzed and anesthetized cats. Drifting sinewave gratings (SF = 0.5 cycle/°, TF = 2Hz, Contrast = 50%) were used as visual stimuli. We identified 694 pairs that showed significant synchrony using the JPSTH representation of correlation. We also studied frequency dependence in these synchronized pairs with multi-taper coherence analysis (Chronux 1.0) and found that 98.4% data samples showed coherence values higher than the 95% confidence interval at certain frequency band(s). Linear regression analysis reveals strong correlation between neural synchrony and the corresponding coherence (R2 = 0.63), which validates the normalization implemented in the JPSTH. To test dependence of synchrony/coherence on fine response structure, we randomly jittered the neural spike trains over different time ranges (±5ms, ±10ms, ±20ms) to deconstruct the timing accuracy. The strength of synchrony and coherence systematically decreased with increase of jitter range. We defined the average coherence with spike trains jittered ±20ms as the “baseline” coherence (independent of fine structure), and derived modulation functions. The coherence of the unjittered spike trains in the gamma frequency band showed the greatest losses after spike trains were jittered. We believe that the fine temporal structures in spike trains are important in maintaining the temporal and frequency dependence between neurons. Our results suggest that cross-correlation analysis and coherence analysis are internally related, though these two methods study neural connectivity from different perspectives.
Supported by NIH EY014680-3, USARL W91INF-05-2-0019