Abstract
Computerized working-memory training (WM), despite typically yielding large practice effects in the training task, transfers reliably almost only to similar tasks and barely transfers to Fluid Intelligence (Gf). We hypothesized that WM training tasks gradually become less related to Gf due to the development of task-specific skills that reduce reliance on WM. As a result, what is being trained in ...
Abstract
Computerized working-memory training (WM), despite typically yielding large practice effects in the training task, transfers reliably almost only to similar tasks and barely transfers to Fluid Intelligence (Gf). We hypothesized that WM training tasks gradually become less related to Gf due to the development of task-specific skills that reduce reliance on WM. As a result, what is being trained in the advanced stages of training is weakly related to WM and Gf. This hypothesis leads to predicting that with training progression, there would be a gradual change in the rank-ordering of individuals (quasi-simplex) in the per-session scores of the training task coupled with a trend in reduction in Gf loadings of these scores. We reanalyzed individual differences in persession scores in the training task from two moderately large-scale published studies. Results show that, as predicted, the correlations between per-session scores decreased with increasing temporal distance between sessions, suggesting a quasi-simplex pattern indicative of a gradual change in the rank-ordering of individuals. However, contrary to the prediction, the training tasks maintained or even tended to increase their Gf loading with training progression. We provide post-hoc accounts for these results, some which challenge prevalent assumptions beyond the attempt to improve Gf through improving WM.
Altmetric