A new statistical approach to the study of conservation of amino acid and nucleotide sequences based on kernel density analysis is described that enables analysis of both conserved and highly variable HIV-1 protein sequences. The amino acid sequences of HIV-1 env proteins in 63 isolates were analysed to determine, first, whether the designations of regions identified in 1987 as conserved (C1-C6) or variable (V1-V5) were still valid. Even though the data base used was nine times larger, the designations that were based on seven isolates from five patients remain correct. Second, the new approach enabled the quantifications of the degree of conservation in reported B or T cell epitopes. Using this approach, highly conserved epitopes located in both gp41 and gp120 were identified.