PROJECT SUMMARY (See instructions): Varying cognitive processes like attention can be a powerful tool for investigating the function and flexibility of cortical circuits. Attention has dramatic effects on behavior, and deficits in selective attention are a hallmark of neuropsychiatric disorders ranging from attention deficit hyperactivity disorder to autism to Alzheimer’s disease. Knowing which of the concomitant changes in the brain are responsible for those shifts in behavior can reveal what is important about circuit function in the first place. One of the most notable aspects of how any cognitive process affects cortical circuits is its heterogeneity: attention has vastly different effects on different neurons, having different magnitude and often sign effects on the rates and shared variability of even nearby, simultaneously recorded neurons. This heterogeneity is not unstructured, however. For example, there are neuron-by-neuron correlations between the magnitude and sign of modulation by spatial attention and by having multiple visual stimuli in the receptive field, between modulation by feature attention and the strength of a neuron's tuning for the relevant feature, and between modulation by attention and neuron-behavior relationships. This structured heterogeneity has the potential to provide unique insight into cortical function because it could be caused by either network or cellular mechanisms. Through our long running collaboration, the Cohen and Doiron groups proposed a circuit mechanism that accounts for the effects of attention on single neurons and on the response variability that is shared within and between neuronal populations. However, our network model only captures the neuron-averaged correlates of attention. We will study the role of cell-intrinsic and network properties in the flexibility of circuits in visual cortex. We will measure, manipulate, and model the activity of large groups of visual and premotor neurons while modulating multiple types of attention that pose distinct challenges for mechanistic models: spatial attention (Aim 1) and feature attention (Aims 2 and 3). This ambitious project will combine multi-neuron, multi-area recordings, causal manipulations, flexible and rigorously controlled behavioral paradigms, and circuit modeling. Our efforts will produce rich data sets and models that provide a platform for studying the mechanisms underlying cognition and complex behavior in many systems.