Discovering individual differences in the near transfer of cognitive training by learning curve analysis of naturally occurring data

A predominant assumption in cognitive training is that practice yields improvements that go beyond the practiced tasks. To date, the findings regarding individual differences in the transfer of learning are inconclusive. Leveraging naturalistic learning data from 22,252 users of Lumosity-an online brain training platform comprising games targeting various cognitive functions-this observational study used decomposition-based learning curve analysis to estimate the degree of transfer between similar tasks within the platform. The results of the learning curve analysis validated near transfer among games sharing overlapping cognitive processes. To improve our understanding of individual differences in this near transfer effect, we examined how individual parameters of transfer and learning rate vary among learners of different ages and initial performance levels. The results reveal that younger adults achieve greater transfer within similar tasks than middle-aged and older adults. Moreover, older adults who are poorer performers at the beginning of cognitive training tend to achieve greater transfer than older adults with greater initial performance. The findings of this article provide an alternative perspective and external validation for existing laboratory studies.