Despite being widely used in both biology and psychology as if it were a single notion, heritability is not a unified concept. This is also true in evolutionary theory, in which the word ‘heritability’ has at least two technical definitions that only partly overlap. These yield two approaches to heritability: the ‘variance approach’ and the ‘regression approach.’ In this paper, I aim to unify these two approaches. After presenting them, I argue that a general notion of heritability ought to satisfy two desiderata—‘general applicability’ and ‘separability of the causes of resemblance.’ I argue that neither the variance nor the regression approach satisfies these two desiderata concomitantly. From there, I develop a general definition of heritability that relies on the distinction between intrinsic and extrinsic properties. I show that this general definition satisfies the two desiderata. I then illustrate the potential usefulness of this general definition in the context of microbiome research.