Although specific language impairment (SLI) often runs in families, most pedigrees are not consistent with a single defective gene. Before progress can be made in molecular genetics, we need a better understanding of which aspects of SLI are heritable. Twin studies are useful in allowing us to distinguish genetic from environmental influences. This point is illustrated with a study in which twins were given tests of nonword repetition (regarded as an index of phonological short-term memory) and auditory processing. Children with SLI were impaired on both measures, but these deficits had different origins. Auditory processing problems showed no evidence of genetic influence, whereas the nonword repetition deficit was highly heritable. Future genetic studies of SLI may be most effective if they use measures of underlying cognitive processes, rather than relying on conventional psychometric definitions of disorder.