Three experiments were conducted to determine the relative merits of forward and backward chaining in the learning of sequential (serial) tasks. Previous research with animals has indicated the superiority of backward chaining and this principle frequently has been proposed for human learning. In all experiments the materials consisted of lists formed from familiar items (numbers, letters, words) arranged in arbitrary sequences. In the forward-chaining technique the subject begins by practicing the first item in the sequence. Next he practices the first and second and third items, and so on until he is practicing the entire sequence. In the backward-chaining technique the subject begins by practicing the last item in the sequence. He then practices the next-to-the-last and last items, then the third-from-last, next-to-last, and last items, and so on until he is practicing the entire sequence. In all three experiments, the forward-chaining technique was superior to the backward-chaining technique. In the first experiment, this difference was not reliable, but in each of the remaining experiments, it was. (Author).