Capacity limits for the detection and identification of change: Implications for models of visual short-term memory

Wilken, Patrick (2001) Capacity limits for the detection and identification of change: Implications for models of visual short-term memory. PhD thesis, University of Melbourne.

Full text available as:

PDF - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader 990 Kb

Abstract

The issue of capacity limits in visual short-term memory (VSTM) has been an area of active research since the 19th Century (Cattell, 1886). A common metaphor suggests that VSTM is akin to a limited capacity urn, able to hold only three-to-six items (Cowan, 2001). The 11 experiments in this thesis explore implications of this metaphor. Experiments 1-4 suggest that items in VSTM are not stored as coherent objects, contrary to recent suggestions (Luck & Vogel, 1997; Vogel, Woodman, & Luck, 2001). Experiments 5-6 contrast two distinct experimental techniques used to probe the structure of information encoded in VSTM. The "single-shot" technique measures sensitivity to differences across two sequential arrays (Luck & Vogel, 1997); the "flicker" technique measures response times to the detection of change, across multiple successive presentations of two alternate displays (Rensink, O'Regan, & Clark, 1997). It is shown that observers' underlying sensitivity to change, and by implication the structure of information encoded, is the same for these two techniques. Experiments 7-9 examine the relationship between observers' performance for detection and identification of change between two successive visual arrays. It is demonstrated that this relationship fails to meet basic assumptions inherent in the urn metaphor. An alternate model is proposed, which suggests that the capacity limit in VSTM is not determined by the number of items stored, but by the amount of information able to be extracted from a visual display. In Experiments 10-11, this alternative model is used to examine the short-term storage of visual information. Overall, the findings of the thesis are inconsistent with a model in which limits in the detection and identification of change are the result of a single process, which operates on a limited number of coherent objects held in a high-level memory store. Instead, it is suggested that detection and identification of change are separate processes, which share a common informational bottleneck at a low-level in the visual processing hierarchy.

Comments/Discussion

This thesis puts forward a series of arguments directed against the notion that visual short-term memory is limited in storage to four items, rather it argues that more items appear to be stored, but that increasing numbers of items lead to loss of information regarding each item. Put simply the bottleneck is increasing noise, not an absolute limit on how many things can be stored. This result can be used to argue against the view that change blindness is evidence of a radically impoverished phenomenology. A better methodology for analyzing these results can be found in Wilken, Patrick and Ma, Wei Ji (2004) A detection theory account of change detection. Journal of Vision, 4. 1120.

Repository Staff Only: edit this item