A feature-extracting program for detecting and quantifying periodic oscillations in auto- and cross-correlograms is described. The program performs like an "expert system" for extracting and evaluating neuronal periodic oscillations. A series of parameters (oscillation frequency, number of repeated cycles, modulation depth, etc.) is calculated, and used for grading (0 to 10) the periodic activity. The program has been applied to single-unit records from the basal ganglia of normal and parkinsonian (MPTP treated) monkeys, and the results have confirmed its accuracy and advantages over other methods.