Using a real-time data set for German GDP over the period from 1973 to 1998 we calculate various measures of real-time output gaps and use these to calibrate and estimate Taylor-type reaction functions for the Bundesbank. Most of the reaction functions we find fit the Bundesbank's actual policy, as represented by the short-run interest rate, quite well. In contrast to previous findings based on ex post revised data for the output gap, we find the reaction coefficients to resemble quite closely those originally proposed by Taylor for some of our realtime measures of the output gap. Broad monetary aggregates such as M3, in contrast, only played a small role for the Bundesbank's interest rate decisions. Given the good record of the Bundesbank in fighting inflation, the results give support to the use of the Taylor rule for monetary policy.