The fear-potentiated startle (PSR) paradigm is a putative behavioral model for the determination of anxiolytic properties of drugs. The present study further investigated the predictive validity of the model. Predictive validity is high, when only drugs clinically used as anxiolytics attenuate PSR dose dependently. Results showed that startle potentiation decreased dose dependently after the administration of the anxiolytics CDP (2.5-10 mg/kg, IP) and alprazolam (1-3 mg/kg, IP). After administration of the clinically non-anxiolytic drugs amitriptyline (2.5-10 mg/kg, IP), carbamazepine (5-20 mg/kg, IP), fentanyl (0.0025-0.04 mg/kg, SC), naloxone (2.5-10 mg/kg, IP), nicotine (0.4-1.6 mg/kg, IP), alcohol (500-2000 mg/kg, IP), and d-amphetamine (0.6-2.4 mg/kg, IP), a dose-dependent decrease in startle potentiation was not found. The PSR correctly discriminated most of the drugs tested in clinically anxiolytic and clinically non-anxiolytic drugs. However, haloperidol behaved as a false positive, and results of nicotine and alcohol were at variance with results reported by others.