Mechanistic reasoning is an epistemic practice central within science, technology, engineering, and mathematics disciplines. Although there has been some work on mechanistic reasoning in the research literature and standards documents, much of this work targets domaingeneral characterizations of mechanistic reasoning; this study provides domain-specific illustrations of mechanistic reasoning. The data in this study comes from the Assessment of Mechanistic Reasoning Project (AMRP) (Weinberg, 2012), designed using item response theory modeling to diagnose individuals' mechanistic reasoning about systems of levers. Such a characterization of mechanistic reasoning illuminates what is easy and difficult about this form of reasoning, within the subdomain of simple machines. Moreover, this work indicates how domain-general principles may be limited. The study participants included elementary, middle, and high school students as well as college undergraduates and adults without higher education. Although the majority of participants responded to the AMRP by diagnosing at least one mechanistic element (elements inherent to the working of systems of levers) as they predicted its motion, such reasoning was not trivial. In fact, the diverse reasoning by participants shows how systems of levers support elements of mechanistic reasoning. Moreover, this study provides evidence that the development of mechanistic reasoning is dependent on domainspecific experience.