Abstract: Rationale: Both the GABAB receptor positive modulator GS39783 and the GABAB receptor antagonist CGP46381 exhibit anxiolytic-like properties in animal models. In the present studies, the effects of GS39783 and CGP46381 in the modified Geller–Seifter task were assessed. First, the predictive validity of the task was confirmed by assessing the effects of multiple anxiolytic and non-anxiolytic compounds on punished and unpunished responding. Methods: Rats were trained in the modified Geller–Seifter task. After successful acquisition of the task, chlordiazepoxide, diazepam, MPEP, haloperidol, GS39783, 8-OH-DPAT, alprazolam and CGP46381 were tested consecutively. For each test compound, doses were administered in a randomized, counter-balanced, within-subjects design. Drug tests were performed only when rats exhibited baseline performance (the punished and time-out response rates were less than 10% of the unpunished response rate). Results: Chlordiazepoxide, diazepam, alprazolam and MPEP released punished responding with variable effects on unpunished responding. Haloperidol had a small but significant effect on punished responding at an intermediate dose, and decreased unpunished responding at the highest dose tested. In contrast, administration of the GABAB receptor positive modulator GS398783 or the GABAB receptor antagonist CGP46381 at doses up to 30mg/kg had no effects on either punished or unpunished responding. The 5-HT1A agonist 8-OH-DPAT did not release punished responding, but significantly decreased unpunished responding at the highest dose tested. Summary: The modified Geller–Seifter task generally exhibits good predictive validity for anxiolytic-like compounds. Neither GABAB receptor positive allosteric modulation nor blockade exhibited anxiolytic-like properties in the modified Geller–Seifter task. The 5-HT1A partial agonist buspirone was similarly ineffective. [Copyright &y& Elsevier]