When insulin interacts with its target cells, it binds to a specific membrane receptor and triggers a chain of events which stimulate many physiological responses1,2. Although the mechanism of insulin action remains uncertain2-4, electron microscopy has suggested that unoccupied insulin receptors appear in clusters of two to six molecules and that insulin becomes internalized after binding to them5,6. Insulin receptors on 3T3 fibroblasts labelled with rhodamine–insulin (R-insulin) are diffusely distributed and mobile on the cell surface, with a lateral diffusion coefficient D∼(3–5)×10−10cm2s−1. At 37°C, hormone–receptor complexes aggregate into immobile patches which are soon endocytosed7,8. A new tool for such studies has been provided by the discovery of autoantibodies to the insulin receptor in the sera of certain patients with insulin-resistant diabetes9. These antibodies stimulate various physiological responses normally mediated by insulin10-14. Both the bivalent antibodies (Fab′)2 and the monovalent Fab′ block insulin binding, but only the bivalent antibodies are biologically active14, Fab′ acting as a competitive antagonist14. Similarly, antibodies raised against purified insulin receptors can mimic insulin15, suggesting that microaggregation of receptors could be involved in the action of this hormone14,16. We now report that at 37°C both R-insulin and the rhodamine-labelled antibodies against insulin receptors form a single cap on one pole of the cell. Cells which were first treated at 37°C with fluorescein–insulin (F-insulin, at partial receptor occupancy), fixed with formaldehyde and then treated with either R–insulin or rhodamine-labelled anti-receptor antibodies, showed overlapping caps of the two markers. These experiments demonstrate directly the aggregation of insulin receptors and suggest either that insulin receptors are multivalent towards the hormone or that the unoccupied receptors migrate together with the occupied ones in the formation of caps.