Clinical use of the middle latency response (MLR) has been limited by the variability of the response during sleep in young children. Theoretically, this variability can be explained by the differential maturation of the primary and nonprimary components of the MLR generating system. The model is supported by animal neurophysiological data. Applied to the human system, the model predicts that, in children, MLR generators are active only' during certain stages of sleep. From a clinical standpoint, this has led to a procedure for signaling the clinician when a child is in a sleep state favorable for recording the MLR.