Definitions of modeling competence in science education do not yet include noncognitive factors. However, noncognitive factors are central to competence and might thus substantially improve our understanding of modeling competence. In this article, we analyze volition during preservice science teachers' engagement with a black-box modeling task and its relation to established aspects of modeling competence: metamodeling knowledge, modeling process, and modeling product. A cluster analysis of the occurrence of volition categories resulted in three clusters of volitional behavior. The clusters describe three different volition types: one action-oriented type applying a self-regulative strategy and two state-oriented types applying self-controlling strategies. Correlation analyses between clusters, volition categories and modeling process variables indicate benefits of the self-regulative strategy.