Testing for linkage: phase known/unknown.

It has been well established in human genetics that when one is testing for linkage between two loci, the significance test must be one-sided for the purpose of reflecting the theoretical limits placed on the estimated recombination fraction. However, the issue becomes obscure when phase-known/-unknown data are considered. This paper presents the subtle difference in hypotheses for testing linkage between two loci when data from which one estimates recombination fractions is phase known or phase unknown. It is shown that a one-sided significance test is appropriate for phase-known data, while a two-sided significance test is correct for phase-unknown data.