OBJECTIVE: To quantify the contributions of child, family, and environmental predictors to language ability at 4 years. METHODS: A longitudinal study was performed with a sample of 1910 infants recruited at 8 months in Melbourne, Australia. Predictors were child gender, prematurity, birth weight and order, multiple birth, socioeconomic status, maternal mental health, vocabulary, education, and age at child's birth, non-English-speaking background, and family history of speech/language difficulties. Outcomes were Clinical Evaluation of Language Fundamentals-Preschool, language scores, low language status (scores >1.25 SDs below the mean), and specific language impairment (SLI) (scores >1.25 SDs below the mean for children with normal nonverbal performance). RESULTS: A total of 1596 children provided outcome data. Twelve baseline predictors explained 18.9% and 20.9% of the variation in receptive and expressive scores, respectively, increasing to 23.6% and 30.4% with the addition of late talking status at age 2. A total of 20.6% of children (324 of 1573 children) met the criteria for low language status and 17.2% (251 of 1462 children) for SLI. Family history of speech/language problems and low maternal education levels and socioeconomic status predicted adverse language outcomes. The combined predictors discriminated only moderately between children with and without low language levels or SLIs (area under the curve: 0.72-0.76); this improved with the addition of late talking status (area under the curve: 0.78-0.84). CONCLUSIONS: Measures of social disadvantage helped explain more variation in outcomes at 4 years than at 2 years, but ability to predict low language status and SLI status remained limited.