Abstract
Correlated firing among sensory neurons limits the accuracy with which sensory information can be decoded from neuronal populations (Zohary et. al., 1994). Attention might improve sensory processing by reducing correlations in the responses evoked by an attended stimulus. To examine this we recorded from pairs of neurons in visual area V4 as monkeys performed an attention-demanding object tracking task that placed a preferred stimulus inside the overlapping region of the two neurons‘ receptive fields. We measured correlations in the neurons‘ responses when the stimulus was either attended or ignored. After controlling for correlations induced by the stimulus we find that trial-by-trial spike count correlations are reduced when attention is directed to the stimulus. The dominant source of noise contributing to the variability in spiking is best characterized as ‘1/f‘ low frequency noise, as seen in the spike power spectrum of single units. We also examined this low frequency noise reduction by computing the coherence of spiking with local field potentials (LFPs). Consistent with our pairwise spike count analysis, we observe a significant attention-dependent reduction in spike-LFP coherence at low frequencies. We also find that attention reduces the variability of individual neuron's responses, as measured by the Fano Factor. An analysis of the spectra of individual neuron's spike trains with and without attention reveals that this reduction in Fano Factor reflects a reduction in rate fluctuations over the same range of low frequencies where we see reductions in correlated activity. Therefore, we conclude that attention improves signal quality by reducing low frequency rate fluctuations, and this reduction contributes to our previously reported attention-dependent reductions in Fano Factor.