AbstractHeterosis refers to the phenomenon in which an F₁ hybrid exhibits enhanced growth or agronomic performance. However, previous theoretical studies on heterosis have been based on bi-parental segregating populations instead of F₁ hybrids. To understand the genetic basis of heterosis, here we used a subset of F₁ hybrids, named a partial North Carolina II design, to perform association mapping for dependent variables: original trait value, general combining ability (GCA), specific combining ability (SCA) and mid-parental heterosis (MPH). Our models jointly fitted all the additive, dominance and epistatic effects. The analyses resulted in several important findings: 1) Main components are additive and additive-by-additive effects for GCA and dominance-related effects for SCA and MPH and additive-by-dominant effect for MPH was partly identified as additive effect; 2) the ranking of factors affecting heterosis was dominance > dominance-by-dominance > over-dominance > complete dominance; and 3) increasing the proportion of F₁ hybrids in the population could significantly increase the power to detect dominance-related effects and slightly reduce the power to detect additive and additive-by-additive effects. Analyses of cotton and rapeseed datasets showed that more additive-by-additive QTL were detected from GCA than from trait phenotype and fewer QTL were from MPH than from other dependent variables.