Generalization of serial learning in the pigeon

Sequence learning in pigeons was studied in asimultaneous chaining paradigm: all stimuli and the opportunity to respond to each stimulus were available simultaneously. In contrast to the traditionalsuccessive chaining paradigm, a simultaneous chaining paradigm provides no differential feedback following each response (except the last). Subjects were first trained to perform on sequences of two (AB), then three (ABC), and then four colors (ABCD). Performance greatly exceeded that predicted by models of random choice. Generalization to novel arrays of three and four colors was complete. After training with a four-color sequence, the subjects were tested with subsequences consisting of all possible combinations of two and three of the four training colors (e.g., BD, AD, BC, ACD, BCD, etc.). The successful completion of these subsequences showed that the organization of the original sequence did not entail overt pecks to successive elements of that sequence. That subjects can respond accurately on nonadjacent subsets is not readily explained by a chaining theory, or by any theory that assumes that responding to element n provides a cue for responding to element n+1.