Embodied cognition is an idea in high fashion. Concepts are postulated to entail simulations of sensory experiences or reenactments of motor behavior with instances of these concepts. I will argue that strong versions of embodied cognition are untenable. The choice between a purely symbolic and a fully embodied cognition obscures a more realistic question: what does it mean for cognition to be “grounded” in perception and action and still permit abstraction? This kind of question has a rich history in neurology dating back to the 19th century. Spatial perception, thought, and language offers a continuum within which to consider the disembodying of cognition. Using fMRI and patient studies I will discuss ways in which actions are extracted from actors, thinking about object properties is shifted to object relations, and richly textured perceptions are reduced to schematic analog representations. I will also present data to suggest that analog behavioral effects need not arise from reenactments as typically conceived.