The hypothesis was examined that at least a part of backward (R-S) recall results from a type of forward practice termed intratrial repetition (ITR) in which S repeats each pair several times in succession before going on to the next pair (e.g., S-R-S-R-S-R). Part of the Ss received paired-associate training in which ITR was “built in” by the use of trials on which each pair appeared three times in succession before S advanced to the next pair. The procedure on each trial for the remaining Ss was identical except that the order of the pairs was changed so that the three repetitions of each pair were not successive. In both conditions, the rate of presentation was very rapid to reduce covert ITR. The Ss in the ITR condition displayed more R-S recall in spite of the fact that all Ss learned to the same S-R criterion.