A common method of evaluating the scope for demand expansion is to try to estimate a threshold rate of unemployment above which policies of demand stimulus are non-inflationary. However, this non-accelerating inflation rate of unemployment (NAIRU) seems to have been anything but stable over the past decade. In the absence of major shocks it is an increasingly popular argument that the NAIRU itself depends on the time-path of previous actual unemployment. This effect has been christened "hysteresis" and various channels through which hysteresis may work are discussed in this paper. Moreover, we estimate the NAIRU for the Federal Republic of Germany using categorical business survey data for inflationary expectations.