An essential step towardsunderstanding how the brain orchestrates information processing at the cellular and population levels is to simultaneously observe the spiking activity ofcortical neurons that mediate perception, learning, and motor processing. Inthis paper, we formulate an information theoretic approach to determine whethercooperation among neurons may constitute a governing mechanism of informationprocessing when encoding external covariates. Specifically, we show thatstatistical independence between neuronal outputs may not provide an optimalencoding strategy when the firing probability of a neuron depends on thehistory of firing of other neurons connected to it. Rather, cooperation amongneurons can provide a “message-passing” mechanism that preserves most of theinformation in the covariates under specific constraints governing theirconnectivity structure. Using a biologically plausible statistical learningmodel, we demonstrate the performance of the proposed approach insynergistically encoding a motor task using a subset of neurons drawn randomlyfrom a large population. We demonstrate its superiority in approximating thejoint density of the population from limited data compared to a statisticallyindependent model and a maximum entropy model.