Current cognitive multimedia design theories provide several guidelines on how to integrate verbal and pictorial information. However, the recommendations for the design of auditory texts (narrations) are still fragmentary, especially with regard to the characteristics of the voices used. In the current paper, a fundamental question is addressed, namely, whether to use a male or a female speaker. In two experiments, learners studied dynamic visualizations on probability theory that were accompanied by narrations. The learner's gender and the speaker's gender served as between-subjects variables. In the first study, learners were randomly assigned to speakers of different gender. In the second study, learners could choose among different speakers. The results show that learners achieved better learning outcomes when the narration was presented by a female speaker rather than a male speaker irrespective of the learner's gender (speaker/gender effect). Being given the choice, learners preferred female speakers, but this individual preference had no impact on learning outcomes. The results suggest augmenting purely cognitive approaches to multimedia design by social-motivational assumptions.