For the analysis of quantitative traits in nuclear families, extreme discordant sib pairs proved to be more powerful than unselected sib pairs. Here, we present a test that makes use of selected pairs and, in addition, restricts the parameters of the identical-by-descent distribution analogously to the "possible triangle" for affected sib pairs. In the Problem 2A data, extreme discordant sib pairs are selected. The analysis allowed the detection of most simulated major genes.