Although hailing from cognate analytical schools, the contributors to Hedwig te Molder and Jonathan Potter's edited volume Conversation and Cognition hold a remarkable diversity of views on the nature of "mental states" and their import for the purposes of analyzing naturally occurring interaction. I offer a critical analysis of some of the contributors' discussions of cognition in social interaction in an effort to clarify some obstinate issues with respect to the meanings of words in our cognitive vocabulary (e.g. "thought" and "realization") and their identification in analyses of conversation.