225 undergraduates learned 6-choice, repeating sequences of lights by the method of anticipation. 4 forms of each of 2 patterns were tested: initial, transposed, inverted, and transposed and inverted. Learning depended on the pattern, but not on its specific form, both in terms of mean errors at each serial position of the pattern and the frequency of particular errors. Tests were made of 5 stimulus-response explanations of serial learning: (a) simple associative chain, (b) serial position as cue, (c) sequences of events as stimuli, (d) transfer of specific sequences, and (e) a fixed tendency to continue runs. All were inadequate to account for the results. Instead, Ss seemed to detect the rules that generate runs (2-3-4) and trills (3-2-3) and to generate their responses by applying those rules. (15 ref.) (PsycINFO Database Record (c) 2006 APA, all rights reserved)