| Even in V1, where neurons have well characterized receptive fields, it has been difficult to deduce which features of complex natural scene stimuli neurons respond to. This may be partly due to network activity that depends upon the stimulus in the whole visual field, not only that in the neuron's receptive field. Here we quantify the extent to which V1 neuronal activity in the behaving macaque monkey is modulated by both stimuli outside the classical receptive field, and by local field potentials (LFPs), a network activity surrogate. We simultaneously recorded spikes and LFPs in V1 and mapped the neurons receptive fields using moving bars. We then presented natural scene movies under two conditions. In the first (unmasked) condition we showed the unobscured movie. In the second (masked) condition only the portion corresponding to the neuron's receptive field was visible. To quantify changes the spiking response to the two stimuli we fit Generalized Linear Models (GLMs) of the spike probability to the data. The GLM included a spline based PSTH-like term accounting for the stimulus and an auto-regressive-type term accounting for the neuron's previous spiking history. For many neurons (85 of 102) the spiking responses to the masked and unmasked stimuli were substantially different (determined by 95 % confidence bounds on the fitted spike probabilities). This suggests that visual stimuli outside the classical receptive field (the context) strongly influence V1 spiking. We next determined the extent to which differences in the spiking response to the two stimuli were explainable by LFP changes. The LFP was decomposed into different time scales using a stationary Daubechies wavelet multi-resolution analysis and instantaneous power and phase of each scale added to the GLM. Averaged across neurons, approximately 25% of the difference could be explained, with the phase of the lower frequency (<20 Hz) scales most predictive. Using a similar analysis we also quantified the extent to which trial to trial variability in the LFP modulated V1 spiking. The trial to trial LFP induced variance of the spike probability iwas 40% (averaged across neurons) of the stimuli induced. In contrast to the trial averaged LFP, the phase of the high frequency (>40 Hz) trial to trial LFP was most predictive. Further, a clear locking of spikes to specific LFP scale phases was observed across neurons, possibly indicating a consistent network effect. Our study suggests that the global context of the natural scenes stimuli, possibly mediated by network activity is an essential component of V1 spiking. Supported by NIH grants K25 NS052422-02, DP1 OD003646-01, MH59733-07 and the Max Planck Society. |
|