Sheila Nirenberg   Professor of Computational Neuroscience in Computational Biomedicine

  • +1 212 746 6372

Information Processing in Neurons

Our lab works on the general question, "How do networks of neurons process information?", and we use a combined experimental and computational approach.The projects fall into three main areas. The first focuses on how neural circuits in the visual system carry out computations. Our approach is to dissect the circuits using a method we developed for targeted cell class ablation (a genetic, inducible method). Currently, we are focusing on the retina. We stimulate the retinal input cells, the photoreceptors, with computer-generated images while recording the responses of the retinal output cells, the ganglion cells. We then ablate specific classes of interneurons as a way to perturb the transfer of information from input to output and to test computational models for how the output is generated. Recently, we have expanded this work to higher brain areas, specifically visual cortex.The second focuses on how populations of neurons in the visual system represent information, and also uses the retina as the model system. Our aim to understand how the retinal output cells collectively encode visual scenes. Can we look at a set of spike trains coming out of the retina and know what the animal is seeing? Our approach involves three general steps: The first is to determine which aspects of the spike trains carry visual information, the second is to build a decoder that translates spike trains into visual scenes, and the third is to test the decoder against the animal's behavior.The last project focuses on how neural networks recognize and classify patterned inputs. Currently, we are addressing this question in the context of attractor networks, and are using networks of dissociated CNS neurons as our model system. The project involves two parts. The first is to examine the intrinsic behavior of the cultured networks. The second is to examine their behavior when they receive patterned input (i.e, patterned current pulses). We can then determine whether the network can learn to distinguish among the inputs by representing them as different patterns of stable activity, i.e., as attractors.e-mail: shn2010@med.cornell.eduFor further information please visit:


Sort by

Selected publications


Sort by

Grants awarded



Primary Email