Uses variables, inferred from previous studies to be successful predictors, to determine if pairs matched on basis of such variables would be more compatible than those randomly assigned. Indications are that variables, although possibly necessary for compatibility, are not sufficient determinants. (Author/CJ)