Reviewing the `language instinct' debate, the paper identifies generativist views with the baby's proverbial bathwater. As Sampson suggests, empirical evidence can lend no support to the claim that grammatical analysis illuminates the study of development, evolution, or the brain. Language instinct theory is coherent only if we adopt Pinker's (dubious) hypothesis that syntax possesses `inner' reality. To argue that grammar is purely `cultural' also proves unsatisfactory. In Sampson's terms, indeed, it leads to dualism and/or a belief in a `haunted universe'. Thus, it implausibly suggests that neither real-time events, development, nor evolution have grammatical consequencs. To avoid allowing the biological baby to disappear with the generativist bathwater, I propose we reject Pinker and Sampson's basic shared assumption. Instead of analyzing language into form-based units, we can treat it as an aspect of social life deriving from a capacity to contextualize experience. Rather than seek evidence supporting language instinct theory, we ask how the dualism of genotype and phenotype constrains an idividual's contextualizing activity.