We present two experimental studies with accompanying theory in which the spatial correlations of naturalistic stimuli induce temporal correlations in neural net activity. Both situations arise in the electric sense, which can be seen as a combination of vision and audition. Spatiotemporal variations in the electric field at the skin of weakly electric fish are a signature of specific stimuli, such as navigational cues, food, or inter-fish communication. The geometries of random stimuli which have long range spatial correlations are shown to induce regular oscillations in the network, which are carved out of the noisy activity of the cells. This relies on delayed feedback from higher brain centers. These results, manifested e.g. in single cell spike train spectra and cross-correlations between cells, can be explained using a slightly non-linear version of linear response theory applied to the network with global feedback. Extensions to localized feedback are also discussed.
In a second study, we study the combined effect of positive and negative delayed feedback on the information transfer about stimuli into spike trains. We further relate this transfer to network dynamics. We finally present experimental results and theoretical support on how feedback modifies the information transfer via a noise shaping effect.