Certain semantic categories, such as the polysemous senses of English prepositions, present specific problems for adult second language (L2) learners, whether they attempt to acquire these meanings through implicit learning mechanisms or through explicit mechanisms associated with incidental learning or instruction. This study examined research on categorization and practice, along with results of learner corpus analyses, to arrive at a characterization of the learning problem posed by English prepositions. An experiment then assessed the effectiveness of a novel pedagogical intervention called semantic highlighting (SH), which employed an inductive, integrative approach to the acquisition of procedural knowledge while accounting for some of the distinctive features of the learning problem posed by polysemy and semantic complexity. A between-subject comparison examined the performance of a control group and four treatment groups. One treatment group (D-P) received explicit explanations of the senses of various prepositions, followed by practice with immediate feedback. Another group (SH) received only a practice session in which cues, referred to here as "semantic highlighting" (SH), were used to draw participants' attention to concrete form-meaning mapping as it applied to the target sentences. The other two treatment groups received hybrid instruction with explicit explanations preceding SH practice (D-SH) or with SH practice preceding explicit explanations (SH-D). Acquisition was measured using a fill-in-the-blanks (FB) test and a written sentence-elicitation (SE) test that was scored using a target-language use analysis (Pica, 1984). Two ANCOVAs, using pretest scores as a covariate, showed significant differences between groups on the FB measure (p < .001) and SE measure (p < .001) at an alpha level of .025. On the FB test, results indicated an advantage for the SH (p < .001) group relative to the SH-D group. On the SE measure, the SH group outperformed the D-P (p = .010), SH-D (p = .013), and D-SH (p = .002) groups. The results suggested that the SH treatment, and possibly the D-SH treatment, as well, constitute viable alternatives to a conventional presentation-and-practice approach when teaching complex semantic targets. The results were further discussed in terms of implications for theoretical accounts of explicit instruction and categorization.