The behaviors of some heritability estimators in the complete absence of genetic factors.

Heritability is an important concept in quantitative genetics and is widely used in human genetics. A high or even a moderate value of heritability estimate is usually taken as evidence for a genetic component for a quantitative trait. In this paper, the behaviors of some correlation-based heritability estimators are reexamined under the assumption of complete absence of any genetic factors. It turns out that when monozygotic (MZ) twins (or full sibs) are environmentally more similar than dizygotic twins (or half sibs), or when there is placement bias in MZ twins reared apart, those correlation-based heritability estimates can lead to nonnegligible or even high heritability values, even when genetic factors are completely absent. These alarming results suggest that extreme care should be exercised when using these heritability estimators.