The majority of N-methyl-D-aspartate receptors (NMDA-R) in the adult forebrain are di- or triheteromers composed of NR1, NR2A and NR2B subunits. Subunit non-selective NMDA-R antagonists produce anxiolytic-like effects together with motor and sensory side-effects. The graded anxiety test (GAT), permits the within-task distinction of drug effects on anxiety from those on activity and perception. By testing NMDA-R subunit selective agents in the GAT it might be possible to determine whether their effects on anxiety, activity and perception are interrelated, and whether separate NMDA-R subtypes are involved. Dextromethorphan (weakly NR2A-selective) (10 and 30 mg/kg, i.p.) and ifenprodil (highly NR2B-selective) (1, 3 and 5 mg/kg, i.p.) were tested in the GAT. Both drugs failed to induce anxiolysis devoid of side-effects. However, the 10 mg/kg dose of dextromethorpan showed an anxiolytic, whereas the 30 mg/kg dose showed an anxiogenic, behavioral profile. Since the selective blockade of the NR2B subunit by ifenprodil had no clear anxiolytic effect, the anxiolytic potential of NMDA subunit non-selective agents might involve NR2A-containing receptors.