Integration of orthographic and semantic information in memory retrieval.

Our central goal was to determine how multiple sources of information are evaluated and integrated during memory retrieval. An expanded factorial design was used to vary two sources of information independently of one another and to present each source alone. Subjects solved crossword-like puzzles with varying numbers of orthographic and semantic cues. The results of the experiment indicated that (a) performance is better given two sources of information relative to just one; (b) evaluation of each source of information provides continuous rather than just categorical (all-or-none) information; and (c) the two sources are integrated multiplicatively rather than simply used independently of one another as claimed by nonintegration models. A fuzzy logical model of perception (FLMP)-taken from the pattern recognition domain--gave a good description of the memory results. A single channel model, an averaging model, and an adding model produced poor descriptions of the results.