The ability of a peptide-based serotyping assay to differentiate human immunodeficiency virus (HIV) type 1 (HIV-1) subtype B infections from non-subtype B infections was investigated with 166 anti-HIV-1- and HIV RNA-positive (by PCR) serum or plasma specimens. The specimens were divided genetically into those infected with subtype B and non-subtype B by application of a screening heteroduplex mobility assay (HMA) that used plasmids for subtypes A and B alone. Specimens that were not clearly infected with HIV-1 subtype B by HMA or for which the two methods had discordant results in distinguishing those infected with subtype B from those infected with non-subtype B were then investigated with a full HMA plasmid panel and, for selected specimens, env sequencing. For the 141 genotyped and serotypically reactive specimens, the correlation between genotyping and serotyping (all subtypes) was 69%. Of the 67 specimens that reacted monotypically as serotype B, 64 were shown to be infected with genotype B (positive predictive value, 96%). Of the 82 specimens that contained genotype B nucleic acid, 64 reacted monotypically as serotype B (sensitivity, 78%), and 4 specimens reacted with a single non-subtype B peptide; the viruses in 14 specimens could not be assigned a serotype. Initial screening results had indicated that 12 samples had results discordant between restricted HMA and serotyping. The V3 loop amino acids of the infecting HIV strains from the seven specimens with discordant serology results were analyzed. For five specimens discordance occurred when the amino acid sequence of the infecting virus closely resembled those of more than one consensus peptide antigen or when the observed V3 crown motif of the strain was atypical for the genetic subtype present. For the other two specimens no explanation for the discordance was identified. Five specimens gave unclear or discordant results in the initial HMA screen, but the results were resolved when the full plasmid panel was used. Serotyping, although of limited sensitivity, distinguishes between subtype B and non-subtype B infections with a high degree of specificity. However, it poorly differentiates the major non-subtype B subtypes, particularly subtypes A and C. When HIV-1 subtype B predominates, serological typing and/or subtype-restricted HMA screening usefully distinguishes between subtype B and non-subtype B infections.