A number of A-ring-reduced metabolites of deoxycorticosterone and progesterone, known to exert agonist activity at the GABA(A) receptor complex, have been reported to reduce anxiety-related behavior in rodents. In the present study, the behavioral selectivity of these effects was assessed in an ethological version of the mouse elevated plus-maze paradigm. Anxiolytic-like profiles, characterised principally by reductions in open arm avoidance measures, were observed following systemic treatment with 5alpha-pregnan-3alpha, 21-diol-20-one (5alpha,3alpha-THDOC; 5.0 and 20.0 mg/kg), 5beta-pregnan-3,20-dione (5beta-DHP; 10-20 mg/kg), 5beta-pregnan-3alpha-ol-20-one (pregnanolone; 20 mg/kg), and 5alpha-pregnan-3alpha-ol-20-one (allopregnanolone; 10-20 mg/kg). In contrast, 5alpha-pregnan-3,20-dione (5alpha-DHP; 10.0-20.0 mg/kg) and 5alpha-pregnan-3beta-ol-20-one (2.5-20.0 mg/kg) were without effect under present test conditions. Detailed behavioral analysis further showed that the antianxiety effects of 5alpha,3alpha-THDOC, 5alpha-DHP, pregnanolone and allopregnanolone were not associated with changes in general activity levels. In addition, profile comparisons revealed that the anxiolytic steroids tend to produce a narrower range of behavioral effects than diazepam (1.0 mg/kg) and, in particular, do not reliably decrease measures of risk assessment. It is concluded that neuroactive steroids produce anxioselective effects in the mouse plus-maze and that their profile of action can at least partially be distinguished from that of a well-characterised benzodiazepine.