Despite the increased use of video for data collection, most research using assessment interviews in early childhood education relies solely upon the analysis of linguistic data, ignoring children's bodies. This trend is particularly troubling in studies of marginalized children because transcripts limited to language can make it difficult to analyze embodied power relations between majority researchers and minority children. This article responds to this problem by outlining a theoretical position on power and bodies, describing multimodal analysis strategies, and using these strategies to analyze the subject positions available during a mathematical assessment interview for three African American preschool child-participants and the European American adult researcher. This study draws attention to the complexity of human interactions during assessment interviews by describing the ways children positioned themselves as willing (or not), attentive (or not), and competent (or not) as well as describing the ways the researcher sought to position herself.