In this study we report two experiments that investigate the acquisition of affective reactions. In Experiment 1, unfamiliar melodies were played to Korsakoff's syndrome patients and alcoholic and nonalcoholic control subjects who were matched with them according to age and education. Following a retention interval of 5 min, subjects received a preference test on old and new melodies. Korsakoff's syndrome patients showed the same increase in preference for old melodies as a consequence of prior exposures as control subjects did, but their recognition of melodies was significantly impaired in comparison with controls. In Experiment 2, the same subjects saw photographs of two men. Fictional biographical information depicted one as a "good guy" and the other as a "bad guy." After a retention interval of approximately 20 days, Korsakoffs recalled virtually none of the biographical information; however, 78% preferred the good guy, and impression ratings were less favorable for the bad guy. Korsakoff patients developed preferences and impressions even though they did not have voluntary access to the information on which the preferences were based. However, their impression ratings were less extreme than those of controls. The pattern of results of the two studies is discussed in terms of Johnson's (1983) MEM model of memory.