Purpose: The authors of this article examined the etiology of developmental language impairment (LI) at 4 and 12 years of age, as well as the relationship between the 2. Method: Phenotypic and quantitative genetic analyses using longitudinal data from the Twins Early Development Study (Oliver & Plomin, 2007) were conducted. A total of 2,923 pairs of twins (1,075 monozygotic [MZ]; 975 dizygotic same sex [DZss]; and 873 dizygotic opposite sex [DZos]) provided data at 4 and 12 years. At 4 years, (a) psychometric LI was defined on the basis of a low parent-reported expressive vocabulary score (-1.25 SDs; 226 MZ and 115 DZss probands for genetic analysis); and (b) parent referral was defined as having seen a medical professional or speech-language pathologist following parental concern (112 MZ and 104 DZss probands). The 12-year language measure was a composite of 4 web-administered receptive language tests. Results: (a) Psychometric LI at 4 years is more predictive than parent referral of poor language performance at age 12 years, and (b) parent referral is substantially and significantly more heritable than psychometric LI. Conclusions: Parents' concern about their child's language development seems to be the marker of a more heritable disorder than poor expressive language skills alone. However, the language difficulties that arouse parental concern in preschool children, although more heritable, are not predictive of language difficulties in early adolescence. Rather, poor expressive language skills at age 4 years, psychometrically defined, are a better predictor than parent referral of continuing language difficulties at age 12 years.