In order to investigate the influence of learner-controlled pacing in educational animation on instructional efficiency, three versions of an audio-visual computer animation and a narration-only presentation were used to teach primary school students the determinants of day and night. The animations were either system-paced using a continuous animation, learner-paced using discrete segments or learner paced using ‘stop’ and ‘play’ buttons. The two learner-paced groups showed higher test performance with relatively lower cognitive load compared to the two system-paced groups, despite the fact that the ‘stop’ and ‘play’ buttons were rarely used. The significant group differences regarding test performance were obtained only for more difficult, high element interactivity questions but not for low element interactivity questions. Copyright # 2007 John Wiley & Sons, Ltd.