This paper examines three sources of knowledge in the acquisition of English articles by second language (L2) learners: L1-transfer, L2-input and Universal Grammar (UG). The paper examines patterns of English article (mis)use among L2-English learners whose L1 either has articles (Spanish) or lacks articles (Russian). It is shown that L1-Spanish learners of English transfer the semantics of Spanish articles onto English, and thus correctly categorize English articles on the basis of definiteness. In contrast, L1-Russian learners of English exhibit fluctuation between the concepts of definiteness and specificity, consistent with earlier findings by . It is shown that the L2-learners’ performance cannot be a result of domain-general statistical learning: despite a large hypothesis space for what English articles might mean, learners of English entertain only those hypotheses that involve semantic universals. It is argued that while L1-Spanish speakers rely on transfer, L1-Russian speakers, in the absence of transfer, have direct access to semantic universals, and fluctuate between them. The learners’ difficulty in choosing the appropriate semantic universal for English (definiteness) is argued to stem from problems generalizing from input triggers. Thus, all three factors – L1-transfer, access to semantic universals through UG, and input triggers – are shown to be relevant for the acquisition of English articles. [Copyright &y& Elsevier]