Structural ambiguity in serial pattern learning

Abstract Four experiments on serial anticipation learning of lights by college students show that (a) the operations of transposing and taking the mirror image are used, (b) temporal inversion is not used, (c) a sequence made of two contrasting halves may be as easy to learn as a sequence having a single homogeneous hierarchical structure. Data revealed intrasequence confusion errors made by importing an operation from one part of the sequence to another. In homogeneous sequences, subjects appear to lose track of where they are in a sequence, and therefore repeat an operation already performed. The evidence supports the idea of an hierarchical cognitive structure, but shows that learning is not a simple function of that structure. Errors and difficulties arise from structural ambiguities, substrings of events that can be reconciled with more than one possible cognitive structure.