The widespread use of restriction endonucleases and DNA sequencing provides a wealth of data on the genetic structure of natural populations. From such data, detailed phylogenies can be constructed and qualitatively different kinds of mutational and substitutional processes can be studied. A neutral model can be constructed to describe the frequencies of sequence haplotypes according to the haplotypes from which they arose and the types of substitution that distinguish them. One feature of such a model is that it examines the ancestors of various sequences. Deleterious selection against a character has a distinct effect on descendant sequences. Individuals containing many deleterious characters leave few or no descendants because these individuals are quickly eliminated by selection. Hence, such a model lends itself to the study of deleterious selection. It is possible to determine if selection is required by searching for any set of mutation rates that can explain an observed set of data. Simulations of artificial populations without selection suggest that this method seldom indicates selection when none is present. Furthermore, recent recombination events between the sequences do not induce false indications of deleterious selection. The method may, however, require relatively large simple sizes in order to accurately reflect the true nature of populations. The method is often very conservative and may not indicate selection when it is, in fact, present.