We investigate differences between three body-wave magnitude (mb) scales for 2009 earthquakes from 1996 to 1999 listed in the Preliminary Determination of Epicenter (pde) bulletin having mb between 5.0 and 5.5 and that also have moment tensor solutions available from the Harvard Centroid Moment Tensor (cmt) catalog. A total of 31,280 broadband seismograms are analyzed, for an average of 15 stations per event. Both the pde and Reviewed Event Bulletin (reb) procedures for determining an automated mb are reproduced, thereby eliminating any discrepancies that result from using different networks of stations. We compare the reproduced pde and reb magnitudes to another magnitude measurement, mb(P), that is based on the Worldwide Standard Seismographic Network (wwssn) short-period instrument. We find that differences between mb(P), mb(pde), and mb(reb) arise from four factors: response function, length of time window, and corrections for event depth and epicentral distance. Reproduced mb(pde) and mb(P) are strongly correlated, and we expect that magnitudes assigned from wwssn short-period instruments during the 1970s and 1980s are consistent with those assigned by the automated procedure used since 1991 by the U.S. Geological Survey (usgs), providing stability in mb measurements over several decades. The difference between mb(reb) and mb(P) is much greater because of the significantly shorter reb window length of 5.5 sec and the high-frequency passband of the reb displacement response.