This article investigates how a person with dementia is made up through intersubjective acts of recognition. Based on ethnographic fieldwork in a Danish memory clinic, we show that identification of disease requires patients to be substituted by their relatives in constructing believable medical narratives; yet during memory testing, patients are not allowed any substitution to clearly expose cognitive shortcomings. In combining works of theorists Ian Hacking and Paul Ricoeur, we argue that the clinical identification of dementia unmakes the knowing subject, a deconstruction that threatens to misrecognize and humiliate the person under examination. The article ends by proposing that dementia be the condition that forces us to rethink our ways of recognizing persons more generally. Thus, dementia diagnostics provide insights into different enactments of the person that invite us to explore practices of substitution and modes of interaction emerging when our fundamental dependency becomes unquestionable.