The current study analysed article use in Spanish-speaking children with specific language impairment who are learning English as a second language. The surface hypothesis account of specific language impairment was evaluated in relation to the use of articles in these children. Language samples were obtained from 15 Spanish-speaking children with language impairment and 15 with normal language, ages 5;0 to 7;1, matched by age, gender, and school. The percentage of article errors was tested between groups with a nonparametric analysis and an analysis of covariance with mean length of terminable unit as the covariate. Results revealed significant differences between groups on percent of article errors with and without MLTU as the covariate. Nonparametric statistics on percent of article error types indicated that most errors consisted of omissions and gender agreement substitutions. As predicted by the Surface Hypothesis, article errors were likely to occur in unstressed definite articles, suggesting that perceptual and prosodic processes have some impact on children's production of articles.