A new nonparametric method of linkage analysis is described based on identity by descent relationships between all pairs of affected relatives within a pedigree. This approach is an extension of ESPA, which only uses information from pairs of affected siblings. The new method, called ERPA, uses the risk calculation facilities of the LINKAGE programs to obtain the necessary information in a fashion which is simple to implement and which automatically generalizes to allow for marker loci which may be multiple, non-codominant and sex-linked. We have investigated the relative performance of ERPA, ESPA and the lod score method on simulated data. ERPA appears to be more sensitive than ESPA for detecting linkage in pedigrees with small sibships, though both nonparametric methods are inferior to the lod score method when the true mode of transmission can be specified.