Great controversy exists on whether associative learning occurs without awareness. In Experiment 1, 31 participants received discrimination training by repeated presentations of two stimulus sequences (S1[subscript A] right arrow S2[subscript A], and S1[subscript B] right arrow S2[subscript B]), S1 being a masked stimulus. S2 were imperative stimuli for a reaction time (RT) task. After the acquisition phase, all participants were tested with 20 presentations of congruent (S1[subscript A] right arrow S2[subscript A] and S1[subscript B] right arrow S2[subscript B]) and incongruent (S1[subscript A] right arrow S2[subscript B] and S1[subscript B] right arrow S2[subscript A]) stimulus sequences. The RT in the testing phase was faster in congruent than in incongruent stimulus sequences. These results are considered strong evidence of associative learning without awareness of the contingency between the stimuli. A second experiment was designed with SOA varied between three groups (23, 58, and 117 ms). The results showed that the participants responded more quickly to congruent stimulus sequences and that the SOA did not affect RT. The SOA did not modify the effect of congruence either, although the interaction was near significance.