A free-recall procedure demonstrated state-dependent learning using alcohol. Tests of long-term memory showed that both high- and low-imagery words were less likely to be recalled if stored while intoxicated rather than under sober conditions. However, information encoded and stored while intoxicated was more effectively retrieved when later tests of recall were performed while intoxicated, as compared to recall accomplished in the sober state. This dissociative recall effect was far more robust with low-imagery than with high-imagery words.