In a study from four laboratories using two commercial vitamin B12 radioassays (impure hog intrinsic factor concentrate containing both intrinsic factor and R binders (IF + R) to measure total corrinoids and the same concentrate presaturated with cobinamide (Cbi) to block B12 binding sites on R binder (IF + R + Cbi) to measure only cobalamins), the rank order of results was generally the same. The concordance between the two tests for classifying sera as normal or deficient was 91% in 311 serum samples. Three percent of sera below the "true B12" (B12 binding to IF + R + Cbi) normal cut-off point were not below the cut-off point for normal "total B12" (B12 binding to IF + R); 6% of sera below the total B12 normal cut-off point were not below true B12 cut-off point. The correlations between Euglena gracilis and the radioassays were 0.80 and 0.83 in the 50 serum samples that also had E. gracilis serum vitamin B12 levels. Lactobacillus leichmannii serum vitamin B12 levels were determined in 49 of the 311 serum samples and results were comparable with results obtained by four radioassay binder systems: IF + R, IF + R + Cbi, highly purified hog IF, and saliva R binder. The closest correlate with L. leichmannii was radioassay using IF + R as binder (r = 0.93), then IF + R + Cbi (r = 0.92), pure IF (r = 0.80), and pure R (r = 0.73). The key to reliable results appears not to reside in a particular assay but rather in determining for each assay its own range of results in participants determined clinically and morphologically normal vs. participants with deficient vitamin B12 (with B12 deficiency defined independently of a serum B12 assay). When laboratory assay results differ from clinical judgment, further evaluation is the appropriate course. There is no "gold standard" for human serum vitamin B12 assay.