This article analyzes the instructed learning of the English article system by second language (L2) learners. The Competition Model (MacWhinney, 1987, 2012) was adopted as the theoretical framework for analyzing the cues to article usage and for designing effective computer-based article instruction. Study 1 found that article cues followed a Zipfian distribution for availability or frequency and that the cues had overall high reliabilities. Study 2 assessed the initial level of cue reliance in a group of intermediate-advanced L2 learners. As expected, the input variables of "cue availability" and "cue reliability" clearly influenced both the accuracy of learners' choices in a cloze test format and the subjects' response times. Study 3 demonstrated that the form-function mappings relevant for native-like article choice can be taught in two 1-hour sessions using the strategy of "cue focusing." The type of explicit instruction (analogical, i.e., by giving analogous examples without metalinguistic comments, vs. metalinguistic feedback) had an additional effect on response time, while both instruction types led to a similar increase in accuracy. These findings are novel and are highly relevant to both theory and pedagogy.