It is a widely held assumption that social cognition is wholly the result of natural selection and learning, debates arising over how much was naturally selected versus how much is learned. I argue here, however, for there being a third factor, namely physics, specifically symmetries and symmetry breakings in neural dynamics. These symmetries manifest themselves in social judgments in a fairly direct way as descending chains of subgroup types in mental social schemata. These schemata are the four models of Alan Page Fiske's relational-models typology. Descending chains of subgroup types are a phenomenon widely observed in nature; their presence in social cognition is consistent with there being a relevant neural network, the activity of which can undergo symmetry breakings. This would be analogous to the neural activity that has been computer modeled in an attempt to explain animal locomotion. This should encourage work towards specifying the particular symmetry groups in social cognition as a step towards devising computer models of the relevant neural mechanism. Approaches to animal locomotion suggest at least the broad outlines of how to proceed. Evidence of symmetry groups in social schemata also supports the view that the innate aspects of social cognition are at least partly structured by dynamics without being encoded in genes, just as the shape of the protective shell of some viruses results from dynamics without being genetically encoded. (C) 2007 Elsevier Ltd. All rights reserved.