The role of Pavlovian contingencies in human skilled motor behavior was investigated in three experiments by means of a new conditioning preparation. In Experiment 1 the present method was shown to be appropriate for the study of associative learning. Subjects who experienced a standard delay configuration performed significantly more conditioned responses than subjects who received either backward conditioning or random pairings. Stimulus generalization was shown to be slight in two additional groups. Subsequent experiments examined conditioning with multiple conditioned stimuli (CSs). In particular, in Experiment 2 some reciprocal overshadowing was demonstrated when two conditional stimuli (tone and vibration) were compounded. Experiment 3 investigated blocking. Blocking was less than expected, however. Subjects' perceptions of the stimuli and reaction time data suggest that a certain proportion had shifted their attention to the added element of the CS compound. Results are discussed in relation to other studies on Pavlovian learning in humans and animals, which are concerned with "stimulus selection."