Bruce and Young's (1986) theoretical framework was actually a synthesis of ideas contributed by several people. Some of its insights have stood the test of time – especially the importance of using converging evidence from as wide a range of methods of enquiry as possible, and an emphasis on understanding the demands that are made by particular face perception tasks. But there were also areas where Bruce and Young failed to obey their own edicts (emotion recognition), and some topics they simply omitted (gaze perception). We discuss these, and then look at how the field has been transformed by computing developments, finishing with a few thoughts about where things may go over the next few (25?) years.