Abstract
We explore several alternative formal models of working memory capacity limits and of the effect of ageing on these capacity limits. Three models test variations of resource accounts, one assumes a fixed number of free slots in working memory, one is based on decay and processing speed, one attributes capacity limits to interference, and one to crosstalk between associations of content and context representations. The models are evaluated by fitting them to time–accuracy functions of 16 young and 17 old adults working on a numerical memory-updating task under varied memory-load conditions. With increasing complexity (i.e., memory load), both asymptotic accuracy and the rate of approach to the asymptote decreased. Old adults reached lower asymptotes with the more complex tasks, and had generally slower rates. The interference model and the decay model fit the individual time–accuracy functions reasonably well, whereas the other models failed to account for the data. Within the interference model, age effects could be attributed to the older adults' higher susceptibility to interference. Within the decay model, old adults differed from young adults by a higher degree of variability in the activation of working memory contents.