How anthropomorphic reasoning functions in scientific thinking has been a controversial topic. There is evidence it is problematic as well as evidence it can play productive roles, for scientists and for students. In science education, however, the prevailing view remains that it is an impediment. For this study, we have chosen examples of what we claim are productive instances in elementary students' reasoning, and we analyze them to understand how anthropomorphisms functioned to support scientific thinking. We argue that one productive role is to support temporary shifts from mechanistic reasoning to more general storytelling, in particular to fill gaps as students work to explain phenomena. That is, we propose that children may come to mechanistic explanation as a form of storytelling. Part of their value is in allowing students to "invent science" based on their existing knowledge, supporting them to understand science as sensemaking.