We investigated the effects of instruction method and cross‐language similarity during second language (L2) grammar learning. English speakers learned a subset of Swedish using contrast and color highlighting (Salience Group), contrast and highlighting with grammatical explanations (Rule & Salience Group), or neither (Control Group with exposure only). Comprehension of grammatical features corresponding to three levels of L1–L2 similarity (similar, dissimilar, unique to L2) were contrasted in three posttests. Grammaticality judgments on L2 sentences: (a) improved across tests for all three training groups, (b) were least accurate for dissimilar features in all groups, and (c) were most accurate for cross‐linguistically similar features in both the Control and Salience groups. Only a trend was found for high language‐learning aptitude (as measured by the Words‐In‐Sentences instrument of the Modern Languages Aptitude Test) possibly facilitating the learning of grammatical features that are distinct from L1. The findings suggest that ideal instructional conditions for L2 morphosyntax may vary with cross‐language similarity.