It is well known that metaphors can be a powerful conceptual tool for scientific progress. But many such metaphors become detrimental once they become overly entrenched or simply outlive their usefulness. In this paper, we anaylze the metaphor GENES ARE CODES, which colors nativist hypotheses about language acquisition. We suggest that such views of grammatical development are ultimately unhelpful because they invite inferences that are not supported by the results of molecular biology and genetics. In short: 1) genes do not code (in the sense of information theory) for phenotypical traits; rather, they only specify the primary structure of proteins; 2) the correlation between a gene and a given phenotypical effect can be used only as a measure of probabilistic differences; it is an analysis of variance, not an analysis of causes; 3) while different causes will have different effects, both the environment and genes (including the genetic background) are necessary for development. (Joseph Hilferty and Òscar Vilarroya)