It has been known for some years that specific language impairment (SLI), an unexpected failure to acquire age-appropriate language skills, is highly heritable. However, molecular genetic studies have been hampered by the heterogeneity of the disorder and the predominant lack of clear genotype-phenotype relationships. We review recent studies suggesting that a better understanding of the genetics of SLI might emerge if we move away from clinical criteria for diagnosis to look instead at a theoretically based quantitative and cognitive measure of the phenotype: a test of phonological short-term memory (STM). Deficient phonological STM has been linked to specific genetic loci, and might play a role in determining some types of reading impairment as well as SLI. Identifying those cognitive deficits that work best as indices of heritable phenotypes will help us to uncover the aetiology of developmental disorders.