The implications for optimal monetary policy of relaxing the normal assumption of a quadratic loss function are examined. Several alternative specifications are considered, but the results suggest that the convenient assumption of quadratic losses may not be that drastic.