EFFECTS OF PHENTERMINE AND PENTOBARBITAL ON CHOICE PROCESSES DURING MULTIPLE PROBABILITY-LEARNING (MPL) AND DECISION-PROCESSES MANIPULATED BY PAY-OFF CONDITIONS

The primary research question in this investigation concerned whether arousal manipulation by a stimulant (phentermine 20 mg) and a depressa nt (pentobarbital 100 mg) will oppositely affect choice behaviour in a probability learning task and decision processes manipulated by pay-o ff. A 3-source probability learning task was used for this purpose, in that, upon the presentation of a pacing cue (every 2 s) which initiat es a sampling response required the subject to predict which source wi ll contain a signal. Three groups of 12 subjects received the drugs or placebo, double-blind and crossover. One group performed the task und er an all-or-none pay-off function (V1), wherein correct responses on the high (0.60), medium (0.30) or low (0.10) probability source were e qually paid; the second group performed the task under a pay-off funct ion, approximately inversely proportional to the signal probability ra tio (V2), wherein their gain was maximal on the low probability source when their predictions were correct on this event; while the third gr oup was paid according to a pay-off function (V3), closely related to the first, in that, the maximum gain could be reached when predictions were correct on the high or medium probability source. The results sh owed that pentobarbital significantly increased sampling responses on the low probability source in the V1 group, while the effects of phent ermine did not reach significance when compared to placebo. In the V2 group, no significant treatment effects were found on sampling respons es on the high, medium or low probability source a finding explained i n the context of arousal theory. Further, in the V3 group relative to placebo, both phentermine and pentobarbital marginally significantly i ncreased and decreased sampling responses on the medium probability so urce, respectively. These effects reached significance on the low prob ability source for both drugs when compared to placebo. It was conclud ed that: (1) the results generally confirm the expectations regarding the different pay-off conditions and the general finding that full max imization is hardly ever found in probability learning experiments; (2 ) the expected effects of the drugs were overall small but consistent and in the hypothesized direction; (3) significant drug effects were m ostly found on the low probability source; (4) pentobarbital produced more behavioural changes than phentermine. (C) 1997 John Wiley & Sons, Ltd.