Retrieval from memory sets that exceed the memory span.

Publisher Website
Google Scholar
Abstract
Groups [3] of 12 human subjects memorized sets of 2, 4, 6, 8, 12 and 16 items, and received series of 96 recognition probes, 48 positive and 48 negative, for each set. The sets were drawn from the letters of the alphabet in group 1, from an ensemble of 26 two-syllable words in group 2, and from an ensemble of 64 two-syllable words in group 3. Latency to identify the probes as members or non-members of the set was an increasing function of set-size, but the shape of the function depended principally on whether the items were letters or words. For group 1 the set-size function was linear with a slope of 34 ms per letter; for groups 2 and 3 it was flatter, with a slope of 15 ms per word, and for group 2, at least, it was better fit by a logarithmic than by a linear function. It was concluded, contrary to a recent suggestion by Burrows and Okada, that high-speed scanning may persist through set-sizes which exceed the memory span, at least when the items are letters. Scanning does appear to break down at some critical memory size, which may depend on the type of material, the degree of practice and response mapping.

This publication has 1 reference indexed in Scilit: