This full paper concerns an exploratory study that investigates students' reasoning about torsion. Mechanical reasoning is critical to engineering applications and yet students still struggle to accurately predict, analyze, and model mechanical systems using formal symbolic notations (i.e., formalizations). To understand the nature of students' reasoning, we analyzed students' discourse to explore two competing hypotheses: (H1) The "Formalisms First" (FF) hypothesis that students report their mechanical reasoning predominantly using mathematical formalisms that take on a disembodied, allocentric (observer) point-of-view; or (H2) the "Grounded and Embodied Cognition" (GEC) hypothesis that students predominantly use independent speech which includes analogy and imagery to simulate the physical structure and function of an object(s) using an embodied, egocentric (first-person) point-of-view in addition to an allocentric point-of-view. Qualitative results from discourse analysis of two student dyads showed that students' mechanical reasoning revealed through their speech included both analogy and imagery, as predicted by H2. Students generated analogies and imagery that described dynamic behaviors, such as how torque caused displacement, stored and released energy, and fractured. Usage of analogies and imagery supports that students' mechanical reasoning often drew upon simulations of torsion-related sensorimotor experiences. Students' egocentric and allocentric imagery invoked sensorial experiences in their speech, with allocentric viewpoints being more common, as predicted by H1 and H2. Student discourse included many references to formalisms, also consistent with the H1. Data from students' verbal discourse on mechanical reasoning suggests they employ both GEC and FF viewpoints of torsion, which has implications for designing effective learning experiences and for assessing students' knowledge.