Learning outcomes have become widely used in higher education, but also misused to the point of being controversial and a bureaucratic burden. This paper distinguishes three kinds of learning outcome found in current literature: (1) those used in individual teaching events; (2) those specified for modules or short courses; and (3) those specified for whole degree programmes. The nature of each is explored and their use in assessment and auditing is discussed, together with related notions such as the "corridor of tolerance", emergent outcomes, etc. It is concluded that learning outcomes used in individual teaching events (1) are the most useful kind if employed flexibly, but that they cannot be specified exactly or used for auditing performance, and their relationship with assessment is complex. Learning outcomes specified for modules or short courses (2) state little more than a list of contents; they cannot be stated precisely and have limitations in guiding assessment. Learning outcomes specified for whole degree programmes (3) is a misuse of the term 'learning outcome'.