LGN coding of natural movies

In recent experiments I have been studying how visual neurons in the LGN encode more natural visual stimuli, to test whether neurons really have special adaptations for encoding natural stimuli efficiently. A movie recorded in the real world is presented hundreds of times, to assess which aspects of a neuron's response are reproducible and which aspects are variable or 'noisy'. The figure shows one frame of such a movie (from the van Hateren archive), and the responses of an LGN neuron to several hundred repeats of a short sample of the movie. Each row is from one presentation of the movie, each blue mark shows the time of a neural response (spike). Total time shown is 4 sec.

Like responses to random flicker, there are epochs of "peaky" firing: intense firing separated by periods of silence (center of raster). In addition, the natural movie also evokes other epochs of lower, more gradually modulated firing (left or right side of raster). I am now exploring systematically how a neuron's coding properties vary with different stimulus characteristics, to find out whether (and in what sense) neural codes are optimized for natural and/or recent stimulus statistics.