This paper begins by reviewing some of the different models of third-party observation of university teaching that can be found in the literature. Having analysed these, it argues that--if "peer" is taken to indicate equality of status--only one is genuinely a model of peer-observation. It proposes an alternative categorisation of third-party observations of teaching dependant on who controls the information generated by the process. A preferred six-dimensional model based on control by the person being observed of the data-flow, and other procedural aspects, is presented and explored. Evaluative comments, by university teachers who have undertaken the process, are presented to illustrate the benefits of adopting this model. (Contains 1 figure.)