Twenty-six participants, divided into three groups, learned to relate English words to Czech and Portuguese words in a matching-to-sample procedure. Half the word pairs were learned using English words as samples and foreign words as corresponding comparisons and the other half were learned with the foreign words serving as samples and English words as corresponding comparisons. The only difference in training across the three groups involved a programmed delay between the removal of the sample stimulus and the presentation of comparison stimuli. For Group 0, Group 2, and Group 8, the programmed delay values between sample offset and comparison onset were 0 s, 2 s, and 8 s, respectively. Test trials assessed the extent to which the conditional discriminations established during training had become reversible or the extent to which the effects of learning had transferred to a new situation. The results suggest that the likelihood of transfer was greatest for the group that learned the task with the largest delay (i.e., an 8 s delay between sample offset and comparison onset).