Simple macroeconomic models based on IS-LM have become unfashionable because of their lack of micro-foundations, and are in danger of being effectively forgotten by the profession. Yet while thinking about micro-foundations is a productive enterprise, complex models based on such foundations are not necessarily more accurate than simple, ad-hoc models. Three decades of attempts to base aggregate supply on rational behaviour have not displaced the Phillips curve; inter-temporal models of consumption do not offer reliable predictions about aggregate demand. Meanwhile, the ease of use of small models makes them superior for many practical applications. So we should not allow them to be driven out of circulation. Copyright 2000 by Oxford University Press.