We investigated the effect of rest breaks on mental‐arithmetic performance, examining performance as a function of the factor rest, time‐on‐task, and demand. We asked the following questions: (i) Does rest (vs a continuous‐work condition) improve cognitive performance? (ii) Is active rest (taking a walk) better than passive rest (watching a video)? (iii) Do compensatory effects of rest increase with time‐at‐work? (iv) Are there differential effects of rest on automatic and controlled processes? (v) Are there differential effects of rest on performance speed versus variability? The results indicate that while rest is generally beneficial for performance, these benefits are similar for active and passive rest. The benefits increase with time‐on‐task and are larger for high (vs low) demand. Further, the effects on average response speed originated only partially from a reduction in the probability of attentional failure, as indicated by reaction‐time (ex‐Gaussian model) distributional and delta‐plot analysis.Copyright © 2016 John Wiley & Sons, Ltd.