Changes in Error Patterns during N-back Training Indicate Reliance on Subvocal Rehearsal

Contemporary cognitive training literature suggests that training on an adaptive task produces improvements only in the trained task or near transfer effects. No study has yet systematically explained the mechanism behind improved performance on the N-back. In this study, we first investigated how improvements in an N-back task using eight pairs of phonologically similar words as stimuli occurred by examining error distributions of the task over training sessions. Nineteen participants (non-native English speakers) trained for 20 sessions over 5 weeks. We observed a reduction in false alarms to non-target words and fewer missed target words. Though the absolute number of phonological-based errors reduced as training progressed, the proportion of this error type did not decrease over time suggesting participants increasingly relied on subvocal rehearsal in completing the N-back. In the second experiment, we evaluated if improvements developed during N-back training transferred to tasks that relied on serial order memory using simple span tasks (letter span with phonologically distinct letters, letter span with phonologically similar letters, digit span forward, and digit span backward). Twenty-nine participants trained on the N-back and 16 trained on the Operation Span (OSPAN) for 15 sessions over 4 weeks. Neither group of participants showed improvements on any of the simple span tasks. In the third experiment, 20 participants (16 native English speakers) trained on the N-back for 15 sessions over 4 weeks also showed increasing reliance on subvocal rehearsal as they progressed through training. Self-report strategy use did not predict improvements on the N-back.