Frequency estimation in verbal learning

A multicopy theory of frequency estimation and discrimination is applied to some data found in the literature on verbal learning. The main theoretical assumption adopted is that ‘relevant’ copies generated during study have to be discriminated from ‘irrelevant’ copies as are available in memory from other occasions than the study session proper. At retrieval, copies of either type are added and their sum determines the frequency estimates that are actually reported. A formalization of these ideas is presented and applied to Hinztman's (1969) and Underwood, Zimmerman, and Freund's (1971) data. Two specific models that emphasize either storage or retrieval of copy information are studied. The results show that the models provide a close approximation to the frequency discrimination data. With absolute frequency estimation, both the means and standard deviations of the judgements are predicted reasonably well. However, some reservations were dictated by the fact that, in one instance, the tails of the judgement distributions were not accounted for. The theory offers a theoretical framework for Underwood et al.'s method of cross-relating absolute frequency and forced choice judgements; the usefulness of two approaches to this problem is established. Also, possible extensions of the theory are noted.