In the reported calculations testing maximum-likelihood molecular-replacement methods (Read, 2001) on the structure of Streptomyces griseus trypsin (SGT), I inadvertently used simulated data computed from the final refined structure of SGT (Read & James, 1986). The simulated data had been constructed to test isomorphous replacement methods in the presence of known errors. The test calculations have been repeated using the measured data, and the results are given in new Tables 1 and 2. The original conclusions are upheld, except that in a translation search with an orientation in error by 6.9°, the correct translation no longer has the highest likelihood score. However, when this translation search is repeated, increasing the assumed r.m.s. error of the molecular-replacement model from the default value of 1.4 to 2.0 Å to compensate for the effect of orientation error, the correct translation again has the highest likelihood score. As before, the discrimination from incorrect translations is poor with the most inaccurate orientation.