This article examines how ‘trauma’ has been conceptualised in the unresolved state of mind classification in the Adult Attachment Interview, introduced by Main and Hesse in 1990. The unresolved state of mind construct has been influential for three decades of research in developmental psychology. However, not much is known about how this measure of unresolved trauma was developed, and how it relates to other conceptualisations of trauma. We draw on previously unavailable manuscripts from Main and Hesse's personal archive, including various editions of unpublished coding manuals, and on Main–Bowlby correspondence from the John Bowlby Archive at the Wellcome Trust in London. This article traces the emergence of the unresolved state of mind classification, and examines the assumptions about trauma embedded in the construct. These assumptions are situated both in the immediate context of the work of Main and Hesse and in terms of wider discourses about trauma in the period. Our analysis considers how a particular form of trauma discourse entered into attachment research, and in doing so partly lost contact with wider disciplinary study of trauma.