Barchi Library (140 John Morgan Building)
Princeton Neuroscience Institute
Error-robust modes of the retinal population code
Across the nervous system, activity in neural populations tends to be highly structured, with certain spiking patterns occurring far more frequently than others. A hypothesis about this structure is that these collective activity patterns function as population codewords – collective modes – carrying information which is distinct from that contained in any single cell’s response. I will describe an investigation of this phenomenon in large-scale recordings of ∼150 retinal ganglion cells, the spiking neurons forming the retina’s output to the rest of the brain. I will discuss a novel statistical model, explicitly incorporating the decomposition of the population response into modes, and show that it predicts retinal spiking activity with high fidelity. This model allows the full population’s state over time to be reduced into a sequence of modes, and I have investigated the information contained in this collective code. I will discuss the visual feature selectivity of the collective modes as well as their capacity to suppress noise fluctuations. The results suggest that correlations across retinal ganglion cells endow their collective signaling with a form of error-correcting code – a general principle that may have validity in brain areas beyond the retina.