This paper explores theoretically and empirically the view that Taylor rules are often nonlinear due to asymmetric central bank preferences, and that the nature of these asymmetries changes across different policy regimes. The theoretical model uses a standard new Keynesian framework to establish equivalence relations between the shape of nonlinearities in Taylor rules and asymmetries in monetary policy objectives. These relations are estimated and tested for the United Kingdom (UK) and the United States (US) over various subperiods by means of smooth transition regressions.There is often evidence in favor of nonlinear rules in both countries, and their character changes substantially over subperiods. The period preceding inflation targeting in the UK is characterized by a concave rule supporting dominant recession avoidance preferences, while the inflation targeting period is characterized by a convex rule supporting dominant inflation avoidance preferences on the part of policymakers. Dominant inflation avoidance appears during the Vietnam War in the US while, during the Burns/Miller and the Greenspan periods, recession avoidance dominates. Under Volcker the Taylor rule is linear. This is consistent with an offset by inflation avoidance of the more prevalent recession avoidance of the Fed.Findings from both countries support the view that reaction functions and the asymmetry properties of the underlying loss functions change in line with the regime and the main macroeconomic problem of the day.