Metacontrol of decision-making strategies in human aging

  1. Florian Bolenz  Is a corresponding author
  2. Wouter Kool
  3. Andrea MF Reiter
  4. Ben Eppinger
  1. Technische Universität Dresden, Germany
  2. Harvard University, United States
  3. Washington University in St. Louis, United States
  4. University College London, United Kingdom
  5. Max Planck UCL Centre for Computational Psychiatry and Ageing Research, United Kingdom
  6. Concordia University, Canada
6 figures, 1 table and 2 additional files

Figures

Design of the sequential decision-making task.

(A) State transition structure of the task. (B) Trial structure. At the beginning of each trial, a stakes cue signaled the stakes condition of the current trial (low-stakes trial vs. high-stakes …

https://doi.org/10.7554/eLife.49154.002
Figure 2 with 1 supplement
Analysis of complete sample.

(A) Reward per trial (baseline-corrected). Gray dots indicate values for individual participants and bars indicate group means with error bars representing standard error of the mean. (B) Degree of …

https://doi.org/10.7554/eLife.49154.003
Figure 2—figure supplement 1
Degree of model-based control under the assumption of perfect transition learning.

Model-based weights are depicted for the complete sample of younger and older adults as a function of stakes condition and transition variability condition. Gray dots indicate values for individual …

https://doi.org/10.7554/eLife.49154.004
Analysis of performance-matched subsample.

(A) Reward per trial (baseline-corrected) for participants in the performance-matched sample. Gray dots indicate values for individual participants and bars indicate group means. (B) Degree of …

https://doi.org/10.7554/eLife.49154.005
Analysis of second-stage reaction times.

(A) Reaction times at the second stage during variable-transitions blocks for trials in which a change of the task transition structure was observed for the first time (revaluation trials) and all …

https://doi.org/10.7554/eLife.49154.006
Appendix 1—figure 1
Structure of the task-switching task.
https://doi.org/10.7554/eLife.49154.011
Appendix 2—figure 1
Posterior predictive check.

Comparison of simulated vs. empirically observed values for baseline-corrected reward. The diagonal represents the identity line.

https://doi.org/10.7554/eLife.49154.013

Tables

Table 1
Psychometric assessment of complete sample.
https://doi.org/10.7554/eLife.49154.007
VariableN assessed
(younger/older adults)
Mean (SD) younger adultsMean (SD) older adultsBayes Factor*
Working memory62/6118.42 (3.80)15.95 (4.22)31.88
Verbal intelligence60/6069.07% (11.82)81.94% (5.71)>100
Processing speed
(reaction times)
61/602032.17 s (315.87)3319.93 s (516.42)>100
Processing speed (accuracy)61/6094.10% (6.02)94.84% (5.48)0.24
Need for cognition61/5982.18 (13.25)81.44 (13.24)0.20
  1. *Bayes Factors quantify the evidence in favor of one hypothesis (here: non-equal means) as opposed to a competing hypothesis (here: equal means). Commonly, Bayes Factors between 3 and 10 are considered as representing moderate evidence for the hypothesis and Bayes Factor above 10 as representing strong evidence for the hypothesis. Conversely, Bayes Factors between 1/3 and 1/10 represent moderate evidence for the competing hypothesis and Bayes Factors below 1/10 represent strong evidence for the competing hypothesis (Lee and Wagenmakers, 2013). We calculated Bayes Factors with the BayesFactors package (Morey et al., 2015).

Additional files

Supplementary file 1

Tables for model parameters, model comparison and statistical results.

https://doi.org/10.7554/eLife.49154.008
Transparent reporting form
https://doi.org/10.7554/eLife.49154.009

Download links