Abstract Bipolar rating scales find considerable use in educational research, whether in formal instruments or as ad hoc scales produced by respondents (for example, in elicited repertory grids), and the data generated are often subjected to statistical analysis. This article examines the concept of bipolarity from primarily a linguistic perspective, taking particular inspiration from Lyons's contributions to the field and from the much earlier work of Ogden. It is argued that the conceptual quality of bipolar scales may be open to serious question, thus rendering statistical and qualitative analyses more problematic than many would prefer.