Six families of HLA-A alleles have been previously proposed on the basis of nucleotide sequence and phylogenetic analysis. Here, sequence polymorphism has been examined at both the protein and DNA levels in a family specific manner and new minimal signatures for each of the families have been delineated. The DNA and protein sites that constitute these signatures are distributed throughout the length of the sequence and generally do not appear to act to promote structural or functional features of the molecules. This is explained by the fact that traditional signatures suffer biases where, for example, recombination products of low frequency can obscure one family's trend by introducing 'impurities' intrinsic to another family. In the absence of complete frequency data, a closer approximation of family signatures can be defined by sites that show strong correlation with the family groups. Using this description, the amino acid positions 62, 97 and 114, localized in the antigen-binding cleft are, in combination, sufficient to discriminate between the six families. Thus, while the composition of the whole cleft defines the details of antigen specificity, these sites in particular, play a key role in modulating supertype peptide specificity and T cell recognition.