In this article, we explore the nature of the conceptual knowledge retrieved when people use words to think about objects. If conceptual knowledge is used to simulate and guide action in the world, then how one can interact with an object should be reflected in the speed of retrieval and the content that is retrieved. This prediction was tested in three experiments in which a part verification procedure was used. Experiments 1 and 2 demonstrated that speed of part verification varied with the perspective imposed on the object by the language used to name the object (e.g., "You are driving a car" or "You are fueling a car"). In Experiment 3, parts were chosen so that actions directed toward them (on the real object) require movement upward (e.g., the roof of a car) or downward (e.g., the wheels of a car). Orthogonally, responding "yes" required an upward or a downward movement to a response button. Responding in a direction incompatible with the part's location (e.g., responding downward to verify that a car has a roof) was slow relative to responding in a direction compatible with the part's location. These results provide a strong link between concept knowledge and situated action.