Objective
Reversal learning (RL) tasks, known to be frontal-lobe dependent, are frequently used in experimental paradigms to assess components of executive function. Extant literature has historically focused on measures of accuracy as their primary outcome, but reaction time (RT) has not yet been well examined. The present study thoroughly examines RT in order to better characterize RL performance and how it changes with age.
Method
182 healthy aging participants completed a RL task, including 43 young (ages 18-30) and 139 community dwelling middle-aged adults (ages 40-61). In order to identify the best predictors of age among RL performance, an adaptive elastic net generalized regression with Poisson distribution modeling and Akakie information criteria (AIC) penalty was utilized. Variables were included from the Diffusion model (a, v, t0), ex-Gaussian distribution (mu, sigma, tau), Normal-3 Mixture distribution (location1-3, dispersion1-3, probability1-3), and RL trials-to-criterion.
Results
A Generalized R2=.83 demonstrated good prediction of age with 7 of 17 variables being significant predictors, including, in order of significance: Mu, Sigma, Dispersion1, Location1, Dispersion2, t0, and Tau. Only two of those predictors had Independent Resampled Variable Importance values greater than .1 (Mu=.645, Dispersion1=.221), indicating older age is associated with slower and more variable efficient responding.
Conclusions
RT adds unique variance to explaining age differences in reversal learning performance which suggests it’s important to measure RT as a fractionated construct. Older adults showed slower efficient mean RT and increased intra-individual variability, which has been linked to poorer frontal lobe processes and age-related cognitive decline.