This paper by Neil Risch explores linkage detection strategies for genetically complex traits, focusing on multilocus models of inheritance. Two types of multilocus models are described: a multiplicative model representing epistasis (interaction) among loci and an additive model, which approximates genetic heterogeneity with no interlocus interaction. The author defines a ratio \( \lambda_R \) of risk for type R relatives compared to population prevalence. For a single-locus model, \( \lambda_R \) decreases by a factor of two with each degree of relationship. For an additive multilocus model, the same holds true. However, for a multiplicative model, \( \lambda_R \) decreases more rapidly with each degree of relationship. Examination of \( \lambda_R \) values for various relatives can suggest the presence of multiple loci and epistasis. The paper provides detailed mathematical derivations and examples, including an analysis of schizophrenia data, which suggests multiple loci in interaction. The second paper in this series will show that \( \lambda_R \) is crucial for determining the power to detect linkage using affected relative pairs.This paper by Neil Risch explores linkage detection strategies for genetically complex traits, focusing on multilocus models of inheritance. Two types of multilocus models are described: a multiplicative model representing epistasis (interaction) among loci and an additive model, which approximates genetic heterogeneity with no interlocus interaction. The author defines a ratio \( \lambda_R \) of risk for type R relatives compared to population prevalence. For a single-locus model, \( \lambda_R \) decreases by a factor of two with each degree of relationship. For an additive multilocus model, the same holds true. However, for a multiplicative model, \( \lambda_R \) decreases more rapidly with each degree of relationship. Examination of \( \lambda_R \) values for various relatives can suggest the presence of multiple loci and epistasis. The paper provides detailed mathematical derivations and examples, including an analysis of schizophrenia data, which suggests multiple loci in interaction. The second paper in this series will show that \( \lambda_R \) is crucial for determining the power to detect linkage using affected relative pairs.