First language learners are “morphology machines,” but second language learners are not. This phenomenon is at the heart of those cases of exaptation (in the sense of Lass 1990) where a loss in morphology due to language contact (second language acquisition) triggers new interpretations of morphological relics in a new generation of speakers (first language acquisition). Exaptation is, however, not restricted to morphology. This paper presents two case studies, one from phonology to morphology and syntax (the Celtic mutations), and one from syntax to discourse (verb-second in Early Modern English). The paper argues that a central notion in exaptation, and possibly the key fact that distinguishes it from the notion of reanalysis, is breakdown in transmission. Breakdown makes it more challenging for learners to recover the interpretation of a feature. They will often succeed nevertheless by fine-tuning hypotheses until they have a reasonable fit. Exaptation, then, is an accidental by-product of the (first and second language) acquisition toolkit: learners’ hypotheses may occasionally differ spectacularly from their “model”. It is then that we see how powerful the toolkit really is.