Extinction of a food reinforced habit results in an increase in the variability of the response learned in acquisition and in the appearance of previously suppressed competing responses. The purpose of the present study was to examine the effects of chronically administered diazepam (0.0, 1.5, 3.0, or 6.0 mg/kg, IP, -30 min) or 10% ethanol (0.0, 1.0, 1.5, or 2.0 g/kg, IP, -15 min) on such behavioral variability in the extinction of radial maze performance. Eight groups of food deprived rats (n = 6) were given one of the forementioned doses for 2 sessions of baseline, 18 sessions of acquisition, and 5 sessions of extinction. In acquisition, eight rewards of two food pellets were obtained on each of three trials in each session. The food well at the end of each arm was rebaited when emptied by the animal, consequently an entry into any arm was reinforced. In baseline and extinction no food was available in the maze. Each session consisted of three 10-min trials. In extinction, compared to treatment with vehicle, both diazepam and ethanol treatments decreased the rate of the instrumental response, arm entry, and increased the variability of the instrumental response and of competing responses. Only the effects of the drugs on the competing responses in extinction were greater than those observed in acquisition. It was concluded that the interference-reduction model of drug action best described the magnitude of the drug effects and the variability-reduction model best predicted the direction of the effects.