Cortical neurons are notoriously variable in response to multiple presentations of identical sensory stimuli, raising questions about their fidelity as signal processors, and the possible coding schemes at work in the cortex. However, using cell-attached and whole-cell patch clamp recording techniques in the intact animal, we have found that both spiking and sub-threshold responses in auditory cortex can be remarkably reliable. Moreover, what variability is present in these neurons can be decomposed into a small component arising from sources of biophysical noise within the recorded neuron, and a potentially much larger component arising from variability in the presynaptic input to the neuron. We are now extending these techniques to the awake behaving animal in order to study the mechanisms underlying cognition. In particular, I am developing a rodent model of selective auditory attention.