Studies of the frequencies of different alleles in young adults and aged individuals have implicated several genes, such as ApoE and ACE, in longevity. However such association studies can easily give rise to spurious results through unsuspected population subdivision, and an approach making use of genetic relationships among relatives is desirable. We have studied the effectiveness of non-parametric genetic analysis to detect different types of loci affecting longevity. The non-parametric method has high statistical power to detect infrequent recessive alleles that are required for, or significantly increase the probability of, survival to advanced age. Statistical power is reduced if a proportion of carriers of the alternative allele is allowed to survive. The method is least effective in detecting alleles that occur at low frequency in young individuals and that subsequently experience high mortality, as is the case for carriers of the epsilon4 allele of ApoE. Genotyping errors will also reduce the value of the NPL statistic in a linear fashion with the error rate and the number of loci genotyped. We have also used the method to analyse genotypes of seven highly polymorphic markers near the ApoE gene in a sample of 188 sibships of nonagenarians and centenarians (n=434) and their children (n=124), however no excess sharing of alleles was detected.