To the Editor.— I believe the answer given by Dr Barnett to the question of inaccurate automated chemical analysis (SMA 12) results was insufficient (233:911,1975). The problem described concerned a proportion of values that fell outside normal ranges: when second specimens were repeated by hand methods, they were found to be normal. Whenever a normal range is defined as 95% of a subject population's values, one must expect that one value in 20 will be abnormal. Since there are 12 channels for each patient, most panels will have at least one value outside the normal range. If these tests are repeated on a second specimen, whether by a specific method or by the same method, the results will very likely fall within the normal range. This statistical phenomenon is called "regression toward the mean" and probably accounts for much of the discrepancy noted. The hospital's clinical pathologist should make this