Broad receptive fields in cortex facilitate efficient and robust population coding of sensory information
1 Max Planck Institute for Neurobiology of Behavior
2 Center for Neurogenomics and Cognitive Research, Vrije Universiteit Amsterdam
In conventional views of the cortical circuitry, sensory processing starts in layer 4 (L4), where inputs from primary thalamus evoke responses in L4 neurons that are selective to specific stimulus features. However, it has become increasingly clear that in parallel to this canonical cortical circuit, the same sensory input from thalamus also drives responses in L5. In contrast to selective responses and hence narrow receptive fields in L4, L5 pyramidal tract neurons (L5PTs) respond unselectively to virtually any stimulus, and hence have broad receptive fields. We recently reported the cellular and circuit mechanisms that underlie broad receptive fields of these major cortical output neurons. Based on this mechanistic insight, we now addressed the question: What stimulus information could L5PTs broadcast to downstream targets despite their broad receptive fields? We found that sensory responses of L5PTs constitute a population code that allows decoding the features of any stimulus tested. In fact, we show that sampling responses of any small subpopulation of L5PTs across a primary sensory cortex allows for such decoding. We demonstrate that broad receptive fields and large cell-to-cell variability thereof are necessary to constitute this population code. The broader the receptive fields, and the larger the cell-to-cell variability, the fewer L5PTs are required to encode the specific stimulus features. Thus, broad receptive fields of L5PTs enable downstream targets to efficiently and robustly decode sensory information. Our findings indicate that sensory input from thalamus is transformed in parallel into two complementary population codes, one that relies on selective responses in L4, and one that relies on unselective responses of L5PTs.