Probabilistic Classification Learning in Amnesia

Amnesic patients and control subjects participated in a s tudy of probabilistic classification learning. In each of three tasks, four different cues were each probabilist ically associated with one of two outcomes. On each trial, the cues could appear a lone or in combina t ion with other cues and subjects selected the ou tcome they though t was correct. Feedback was provided after each trial. In each task, the amnesic patients l ea rned gradual ly to associate the cues with the appropriate ou tcome at the same rate as control subjects, improving from 50% correct to 65% correct. Presumably because the cue-outcome associations were probabilistic, declarative m e m o r y for the ou tcomes of specific trials was not as useful for pe r fo rmance as the informat ion gradual ly accrued across trials. Nevertheless, declarative m e m o r y does appear to make a cont r ibu t ion to per formance w h e n training is extended beyond 5 0 trials, because with fur ther t raining control subjects eventual ly ou tpe r fo rmed the amnesic patients. It was also demons t ra ted that pe r fo rmance on the probabilistic classification task was not the resul t of ho ld ing knowledge of cueou tcome associations in shor t term memory , because both control subjects and amnesic patients demons t ra ted significant 5Corresponding author. savings w h e n testing was in te r rup ted by a 5-rain delay (exper iment 2). Probabilistic classification learning appears to provide an analog in h u m a n subjects for the habi t learning tasks that can be acqui red normal ly by animals with h ippocampal lesions.