ABSTRACT Purpose: To perform the construct validation of enunciative signs of language acquisition for children aged 3 to 12 months. Method: The study sample consisted of 94 mother-infant dyads for the analysis of Phase 1 (3-6 months) signs and 61 dyads for the analysis of Phase 2 (7-12 months) signs. Data collection was performed through analysis of the interaction between mothers and infants in play situation recorded in 15-min videos in the predicted phases, with attribution of value types of present or absent to each sign analyzed. The collected data were organized on a spreadsheet and then converted to computer applications for factor analysis. Results: Factor analysis indicated the existence of two factors named “mother” and “infant” both for Phase 1 signs (explaining 71.9% of the variation) - with three signs relevant for the “infant” factor and one sign relevant for the “mother” factor, and for Phase 2 signs (explaining 74.4% of the variation) - with one sign relevant for the “infant” factor and one sign relevant for the “mother” factor. Conclusion: Construct validation showed that one “mother” factor and one “infant” factor were able to distinguish between at-risk and not-at-risk groups in both phases analyzed, which suggests that the absence of these signs may pose risks to language acquisition.