When and why do old adults outsource control to the environment?

Abstract

Old adults' tendency to rely on information present in the environment rather than internal representations has been frequently noted, but is not well understood. The fade-out paradigm provides a useful model situation to study this internal-to-external shift across the life span: Subjects need to transition from an initial, cued task-switching phase to a fade-out phase where only 1 task remains relevant. Old adults exhibit large response-time "fade-out costs," mainly because they continue to consult the task cues. Here we show that age differences in fade-out costs remain very large even when we insert between the task-switching and the fade-out phase 20 single-task trials without task cues (during which even old adults' performance becomes highly fluent; Experiment 1), but costs in old adults are eliminated when presenting an on-screen instruction to focus on the 1 remaining task at the transition point between the task-switching and fade-out phase (Experiment 2). Furthermore, old adults, but not young adults, also exhibited "fade-in costs" when they were instructed to perform an initial single-task phase that would be followed by the cued task-switching phase (Experiment 3). Combined, these results show that old adults' tendency to overutilize external support is not a problem of perseverating earlier-relevant control settings. Instead, old adults seem less likely to initiate the necessary reconfiguration process when transitioning from 1 phase to the next because they use underspecified task models that lack the higher-level distinction between those contexts that do and that do not require external support.

Figure 1

Sample displays used in Experiments 2 and 3. Experiment 1 displays were similar, only that vertically or horizontally aligned rectangles were used instead of squares versus circles as response-relevant stimuli and that grey patches were used as placeholders instead of the cues on cue-absent trials (display type C). Display type A was used for cued task-switching or corresponding single-task block segments (i.e., Exp. 1: trials 1–40; Exp. 2: trials 1–30; Exp. 3, fade-in blocks: trials 31–63; Exp. 3. Fade-out blocks: trials 1–30). Display types B and C were used during fade-out or fade-in segments or corresponding single-task control segments (i.e., Exp. 1: display type B for transition trials 41–43, display type C for trials 44–60 and display type B for trials 61–80; Exp. 2: display type B for transition trials 31–33 and display types B and C randomly intermixed for trials 34–63; Exp. 3, fade-in blocks: display types B and C randomly intermixed for trials 1–30; Exp. 3. Fade-out blocks: display type B for transition trials 31–33 and display types B and C randomly intermixed for trials 34–63). Note, that stimuli were always ambiguous, that is contained response-relevant values for both task dimension. Stimulus displays are not drawn to scale.

Figure 2

Fade-out costs in Experiment 1 as a function of fade-out conditions for young and old adults. Error bars reflect 95% within-subject confidence intervals testing for each segment separately fade-out costs against the single-task baseline condition.

Figure 3

Fade-out costs for the instruction and the no-instruction condition. Only old adults were tested in this experiment. Error bars reflect 95% within-subject confidence intervals testing for each segment separately the no-instruct condition against the instruction condition and the instruction condition against the single-task baseline condition.

Figure 4

Idealized structure of young and old adults’ hierarchical task models for fade-out blocks. Note, that the complexity of the task models is not affected by whether or not the cued switching phase comes first (as in fade-out blocks) or last (as in fade-in blocks).

Figure 5

Fade-in and fade-out costs for young and old adults. Error bars reflect 95% between-subject confidence intervals testing for each segment separately the age difference in fade-in/fade-out costs.