Visualization and memorization

Abstract

Previous investigations have suggested that visual memory may involve short-term (STVM) and long-term (LTVM) components. Evidence for this comes from the functional differences between visual memory tested after short, unfilled retention intervals (STVM conditions), and performance measured after any interpolated task with a high mental load (LTVM conditions). The suggestion is that stimulus information is maintained over short, unfilled intervals by visualization, an active, voluntary control process utilizing central resources. Under LTVM conditions interference prevents active maintenance, and the item must be memorized. The aim of this thesis was to provide further evidence on the functional distinction, and the nature of the underlying processes. A number of experiments were conducted using novel matrix patterns as stimulus materials, and on-line control to allow precise manipulation of timing and other display parameters. The dissociation of STVM and LTVM was reflected in several results: STVM and LTVM (a) have different requirements for display time (b) differ in the consistency of performance over trials (c) they involve different coding processes at acquisition and (d) they show quite different relations between accuracy of performance and mean response time. In contrast to this, varying the exposure of a recognition test probe did not dissociate STVM and LTVM performance, and the provision of feedback and retrieval cues durin recall had no clearly interpretable effect. Visualization is a limited capacity process, insofar as it is restricted to one item or presentation at a time, and can maintain information up to a certain level of complexity. Visualized descriptions are constructed rapidly from short display times, and have general application to this class of novel visual patterns. With other evidence, this suggests that visualization is based on low-level 'figural' descriptions, specifying stimuli as a spatial arrangement of shapes formed by groupings of the pattern elements. LTVM performance increases slowly and irregularly with display time and there is a wide variation in performance over trials. Higher-level, 'semantic' descriptions contribute to memorization, and these cannot be applied rapidly and consistently to randomly generated abstract patterns. The results have widespread implications for theories of visual memory. Single-process theories which deny any distinction between short- and long-term memory are ruled out by the data. Other models which (a) consider STVM as an 'activated' part of LTVM or (b) claim the dichotomy arises from simple distinctions in coding or storage or retrieval do not give a complete account of the results. The 'modal' model is also rejected since prolonged visualization of an item after stimulus offset does not lead to an increase in LTVM. To account for this latter finding, it is proposed that visualization and: the elaborate encoding processes required for memorization compete for-central processing resources.