As illustrated in Fig. 13C, we
propose that there is a relationship between excitatory–excitatory and excitatory–inhibitory correlations that is dependent upon levels of excitation and inhibition. Increased excitation will tend to increase Selleckchem JQ1 correlations and increased inhibition will tend to decrease correlations between excitatory–excitatory and excitatory–inhibitory pairs. Inhibition may be important for maintaining optimal levels of excitatory–excitatory correlation in visual cortex. This implies that increasing inhibition makes it more difficult for an excitatory input to push the network out of the optimal regime and into a higher excitatory–excitatory correlation state (Fig. 13C). ACh’s role in V1, then, might selleck inhibitor be to further activate inhibitory neurons so that they can absorb the increase in excitation that comes with top-down attention and BF activation of mAChRs on excitatory neurons without adding in excessive correlations. It has been suggested that low-frequency excitatory–excitatory
noise correlations originate from cortico-cortical connections (Mitchell et al., 2009). It is possible that we do not see attention and mAChR-dependent decreases in excitatory–excitatory correlations, then, due to the fact that our model does not incorporate these connections. Interestingly, mAChRs have been shown to also decrease lateral connectivity in the cortex (Hasselmo & McGaughy, 2004), which could potentially mediate the decrease in excitatory–excitatory correlations. It would be interesting to develop a model that incorporates cortico-cortical connections to see if mAChR-dependent reductions in their efficacy can decrease noise correlations between excitatory neurons. It is important
to point out that decreases in excitatory–excitatory correlations only improve encoding when two neurons have high signal correlations (Averbeck & Lee, 2006). Because neurons in each column receive the same Gabor-filtered input, we assume they all have high signals correlations, and thus decorrelating the signal would improve coding. Neurons that have low signal correlations, by contrast, such as neurons that encode for orthogonal stimulus orientations Etomidate within a single receptive field, may improve encoding by increasing noise correlations. mAChR influences on lateral connectivity strength may thus be crucial for facilitating this type of improvement in information processing. From a modeling and experimental standpoint, it will be interesting to see how mAChRs influence noise correlations when signal correlations differ. We demonstrated that both BF and top-down attentional signals lead to an increase in cortical reliability as a consequence of their projections to the TRN. The reliability of a neuron is related to the probability that it will fire at a particular time and rate given repeated presentation of the same stimulus.