Diazepam (0.3, 0.6, 1.2, or 2.5 mg/kg) produced a dose-dependent reduction of the potentiated startle effect where acoustic startle amplitude is normally increased in the presence of a light previously paired with a shock. Even the lowest dose tested (0.3 mg/kg) significantly attenuated potentiated startle. The effect was selective since the same doses did not depress baseline startle amplitude measured in the same animals in the same test session. A 2 X 2 design in which rats were trained and tested under the same or different drug condition (diazepam or saline) showed the results could not be explained by state-dependent learning. The primary effect of diazepam was to block expression of rather than acquisition of fear as measured by potentiated startle. Flurazepam (2.5, 10, or 20 mg/kg) also reduced potentiated startle selectively but was 6--8 times less potent than diazepam. These and other results suggest that the potentiated startle paradigm, as a measure of classical conditioning that involves no operant, might provide a useful adjunct to behavioral methods currently being used to analyze antianxiety compounds.