Williams syndrome is a rare genetic disorder in which, it is claimed, language abilities are relatively strong despite mild to moderate mental retardation. Such claims have, in turn, been interpreted as evidence either for modular preservation of language or for atypical constraints on cognitive development. However, this review demonstrates that there is, in fact, little evidence that syntax, morphology, phonology, or pragmatics are any better than predicted by nonverbal ability, although performance on receptive vocabulary tests is relatively good. Similarly, claims of an imbalance between good phonology and impaired or atypical lexical semantics are without strong support. There is, nevertheless, consistent evidence for specific deficits in spatial language that mirror difficulties in nonverbal spatial cognition, as well as some tentative evidence that early language acquisition proceeds atypically. Implications for modular and neuroconstructivist accounts of language development are discussed.