A diffusion model analysis of age and individual differences in the retro-cue benefit

Abstract

The limited capacity of working memory (WM) constrains how well we can think and act. WM capacity is reduced in old age, with one explanation for this decline being a deficit in using attention to control WM contents. The retro-cue paradigm has been used to examine the ability to focus attention in WM. So far, there are conflicting findings regarding an aging deficit in the retro-cue effect. The present study evaluated age-related changes and individual differences in the retro-cue effect through a well-established computational model that combines speed and accuracy to extract underlying psychological parameters. We applied the drift-diffusion model to the data from a large sample of younger and older adults (total N = 346) that completed four retro-cue tasks. Retro-cues increased the quality of the evidence entering the decision process, reduced the time taken for memory retrieval, and changed response conservativeness for younger and older adults. An age-related decline was observed only in the retro-cue boost for evidence quality, and this was the only parameter capturing individual differences in focusing efficiency. Our results suggest that people differ in how well they can strengthen and protect a focused representation to boost evidence-quality accumulation, and this ability declines with aging.

Figure 1

Illustration of key features of retro-cue tasks used in the present study. These tasks varied the nature of the memory test (change detection, CD vs. delayed estimation, DE), the memoranda (orientation vs. color), and the type of retro-cue (central arrow vs. peripheral circle).

Figure 2

Diffusion model account of the responses in the change detection paradigm (A) and the delayed estimation paradigm (B). Both model versions have three main parameters: drift rate (ν), boundary separation (a), and nondecision time (T_er).

Figure 3

Posterior estimates (mean, 95% HDI, and full distribution) for the retro cue effect on drift rate (A), nondecision time (B), and boundary separation (C) for younger (black triangles) and older (grey circles) adults. CD = Change-detection; DE = Delayed estimation.

Figure 4

Posterior estimates (mean, 95% HDI, and full distribution) of the drift rate (A), nondecision time (B), and boundary separation (C) in the no-cue condition for younger (black triangles) and older (grey circles) adults in each task. Note. CD = Change-detection; DE = Delayed estimation.

Figure 5

Simplified path diagrams for the bayesian structural equation models isolating condition general variance from variance specific to the ability to focus attention (i.e., benefitting from retro-cues) for drift rate (A), nondecision time (B), and boundary separation (C). We report posterior means of standardized path coefficients for younger (black font) and older (grey font) adults. Additionally, posterior means of the error variances for each indicator are displayed in italics. Note v = drift rate, Ter = nondecision time, a = boundary separation. Col = color, Ori = orientation, CD = change detection, DE = delayed estimation, g = general.