1. Spatiotemporal Origin of Bursts and Reliable Spikes Generated by Neurons in V1 Jonathan D. Victor, Ferenc Mechler, Daniel S. Reich, Keith P. Purpura Department of Neurology and Neuroscience Weill Medical College of Cornell University and Laboratory of Biophysics The Rockefeller University Support: EY09314, GM07739, EY07138, NS01677, NS36699

2. Reliable spikes (and spikes in bursts) are disproportionately effective in signaling the visual input, as filtered by the receptive field. For most neurons, reliable spikes are preferentially generated by particular receptive field components. Receptive fields have distinct subregions that differ in spike generation characteristics. Spatiotemporal origin of reliable spikes and spikes in bursts

3. Experiment cross-correlate t X y receptive field structure spike generation m-sequence neural response (multiple identical trials) vs. all spikes vs. bursts (3 spikes within 11 msec) vs. reliable spikes (spikes in at least 4 of 16 trials, within same 3.7-msec bin)

4. Spatiotemporal receptive field map constructed from all spikes simple cell 35/1

5. Spatial Profiles (Per-event normalization) all spikes reliable spikes average gain change (n=38): 1.49 +/- 0.37 (1 SD) bursts simple cell 35/1 63 to 81 msec 1.8

6. Spatiotemporal receptive field map constructed from all spikes complex cell 34/20

7. Spatial Profiles (Per-event normalization) all spikes reliable spikes complex cell 34/20 44 to 63 msec 67 to 85 msec 2

8. Spatial Profiles (Per-event normalization) all spikes reliable spikes average gain change (n=38): 1.49 +/- 0.37 (1 SD) bursts complex cell 35/9 44 to 63 msec 2

9. Spatial Profiles (Best-fit normalization) all spikes reliable spikes bursts complex cell 33/1 59 to 92 msec average shape change (fraction of unexplained variance): 0.43 +/- 0.32 (1 SD) units with significant shape changes: 30/38 0.8

10. Relationship of Spike Probability to Reliable-spike Probability average across trials coincidences across trials p(t) t spike probability r(t) t reliable-spike probability

11. Size of Cross-correlation Depends on Best Linear Fit to the Behavior of the Spike Generator Sensitivity on a per-event basis = (slope of best-fit line)/(mean probability of events) mean probability of spikes drive distribution best-fit line behavior of spike generator

12. Effect of Spike Selection on Gain gain ratio =1.65 gain ratio =1.25 high threshold gain ratio =1.99 low threshold all spikesreliable spikes

13. average across trials coincidences across trials spike probability p(t) reliable-spike probability r(t) Looking at reliable spikes is the same as appending a nonlinearity. For linear-> static nonlinear systems driven by (nearly) Gaussian noise, this can produce an apparent change in overall gain, but not a change in receptive field profile. receptive field structure spike generation drive d(t)

14. change in per-event sensitivity of RF but no change in shape change in per-event sensitivity of RF and change in shape receptive field subregion 1 receptive field subregion 2 drive d 1 (t) drive d 2 (t) receptive field subregion 1 receptive field subregion 2 drive d 1 (t) drive d 2 (t)

15. Summary Reliable spikes (and spikes in bursts) are disproportionately effective in signaling the visual input, as filtered by the receptive field. For most neurons, reliable spikes are preferentially generated by particular receptive field components. Receptive fields have distinct subregions that differ in how threshold relates to gain and noise. Spatial selectivity is supported by subregions with a high threshold relative to their intrinsic gain.

16. Two Phenomena in Spatiotemporal Maps of Cortical Receptive Fields: Does spiking confer special properties on kernels? Are some spikes more special than others? Jonathan D. Victor, Ferenc Mechler, Daniel S. Reich, Keith P. Purpura Department of Neurology and Neuroscience Weill Medical College of Cornell University and Laboratory of Biophysics The Rockefeller University Support: EY09314, GM07739, EY07138, NS01677, NS36699

17. Spatiotemporal Origin of Reliable Spikes and Spikes in Bursts Do reliable spikes (or spikes in bursts) carry a special message? Are these spikes generated by specific areas within the receptive field? What does this have to do with the statistics of spike generation?

18. Factoring the Problem: “Drive” and “Spike Probability” receptive field structure spike generation drive d(t) average across trials coincidences across trials p(t) t spike probability r(t) t reliable-spike probability

19. Summary Reliable spikes (and spikes in bursts) are disproportionately effective in signaling the visual input, as filtered by the receptive field. For most neurons, reliable spikes are preferentially generated by particular receptive field components. To account for this observation, receptive fields have distinct subregions that differ in spike generation characteristics. These observations are compatible with both Poisson and non-Poisson spike train statistics.