S. P. Strong, William Bialek Rob R. de Ruyter van Steveninck and Geoffrey D. Lewen,
Institute for Advanced Study,
Olden Lane, Princeton, New Jersey 08540
NEC Research Institute, 4 Independence Way, Princeton, New Jersey 08540
Frequently, direct estimates of the information contained in neural spike trains are constrained by limits on the number of possible repetitions of the stimulus without degradation of the preparation. These limits imply that the conditional entropy rate must be estimated with inadequate sampling. To combat this, we have obtained upper and lower bounds on this rate that can be used for as few as two presentations of the stimulus. In experiments on the H1 neuron of the blowfly, Calliphora vicina, the utility of the bounds was verified by a comparison with the directly estimated conditional entropy rate.