Background: Maternal responsiveness has been shown to predict child language outcomes in clinical samples of children with language delay and non-representative samples of typically developing children. An effective and timely measure of maternal responsiveness for use at the population level has not yet been established. Aims: To determine whether a global rating of maternal responsiveness at age 2 years predicts language outcomes at ages 3 and 4 in a community sample of slow-to-talk toddlers. Methods & Procedures: In an Australian population-based study, at child age 1:6 years, 301 slow-to-talk toddlers (scoring =20th percentile on a parent-reported expressive vocabulary checklist) were invited to take part in a 15-min free-play video of mother-child interaction at 2:0 years. Each free-play video was rated for maternal responsiveness using a five-point global rating scale, where 1 is "very low" responsiveness and 5 is "very high" responsiveness. Language skills were measured at 3:0 years using PLS-4 and at 4:0 years using the CELF-P2. Outcomes & Results: In adjusted linear regression models (potential confounders: gender, maternal education, socioeconomic status) maternal responsiveness strongly predicted receptive, expressive and total language standard scores at ages 3 (coefficient = 5.9, p < 0.001; coefficient = 5.4, p < 0.001; coefficient = 6.2, p < 0.001, respectively) and 4 years (coefficient = 4.6, p < 0.001; coefficient = 3.1, p = 0.004; coefficient = 4.0, p < 0.001, respectively). Conclusions & Implications: Slow-to-talk toddlers of mothers with higher global ratings of responsiveness have higher language scores at 3 and 4 years of age. This global measure of maternal responsiveness could be further developed as a clinical tool for identifying which slow-to-talk toddlers are most in need of early intervention.