Artefacts offer opportunities for action, ‘affordances’, that can be described on various levels, from manipulations (‘pushing a button’) to social activities (‘dialling a friend’). However, research in design into affordances has not investigated what an ‘action’ is, nor has it distinguished those levels. This paper addresses the question of which kinds of descriptions can be applied to affordances. Its main claim is that different descriptions can apply to a single affordance. On this claim a descriptions-of-affordances-model is built that shows how these levels are connected, and that specifies what knowledge the artefact user would need in order to perceive affordances under each kind of description. The paper also shows several ways in which the descriptions-of-affordances-model can contribute to affordance-based design.