Common synaptic input is thought to give rise to both correlated trial-to-trial response variability and synchronous spiking in nearby cortical neurons. As a way to determine the source of this common input, we measured how correlation between neurons depends on the similarity of their tuning properties, the cortical distance between them, and the orientation of the stimulus.

We recorded from the primary visual cortex (V1) of anesthetized, paralyzed macaque monkeys using two techniques: (1) pairs of electrodes spaced 300 to 500 microns apart, and (2) A 100-electrode array. The array consisted of a 10 x 10 grid with 400 micron spacing between adjacent electrodes. Electrode length was 1.0 mm and the array was implanted 0.6 mm into cortex, resulting in superficial layer recordings. We presented sinusoidal grating stimuli drifting in different directions to elicit neuronal responses. From each electrode, we recorded all responses which exceeded a threshold and sorted these waveforms offline. Simultaneously, we recorded the local field potential (LFP) from each electrode, a signal considered to be related to the sum of all synaptic input to a region of tissue.

We measured correlation on two different time scales from the spiking data. Spike count correlation (rsc, or "noise" correlation) was calculated over the course of the entire stimulus presentation (1.28 seconds). Synchrony (precise temporal correlation) was measured by examining the spike-train cross-correlogram (CCG) peak height near zero time lag. We determined the coherence of the raw LFP in seven frequency bands, ranging from delta (1-4 Hz) to very high gamma (100-150 Hz). We found that spike count correlation did not change with stimulus orientation. Although rsc was smaller for widely separated neurons, it remained significantly above zero even at separations greater than 4 mm. The peak height of the CCG was strongest for the orientation that produced the best response for the pair of neurons. Similar to spike count correlation, CCG peak height decreased with distance, but at a faster rate. In addition we found that the coherence of high frequency LFP components falls off more precipitously with distance than that of low frequency LFP components (similar to a previous study in awake behaving macaques [1]). We also calculated the time scale of spiking correlation and compared it to the LFP. Pairs of single units which had their strongest correlation on a time scale of roughly 200 ms tended to have spiking activity that was tightly locked to the LFP.

Together, our results suggest that sharp synchrony and spike count correlation on long timescales arise from different mechanisms. The dependence of synchrony on distance and stimulus orientation suggests it is primarily (but not exclusively) driven by common input arising from short-range cortical circuitry. The spatial extent of spike count correlation and its lack of dependence on stimulus orientation suggest it is mediated by connections which cover large regions of cortex. Furthermore, different frequency components of the LFP are related to measures of spiking correlation on different time scales. Pairs of neurons which are correlated on a broad time scale appear to be the most influenced by the LFP.

[1] D.A. Leopold, Y. Murayama, N.K. Logothetis Cereb. Cortex. 13:422-433 (2003).