Web resources allow a learner to have more opportunities for study at any time and any place. It is still difficult, however, for learners to choose the right study materials to match their desired learning. A competence-based system for recommending study materials from the Web (COSREW) is proposed, based on the learner's competences. COSREW generates a list of learning paths, and extracts search terms from the competence statements on the chosen learning path. Three experiments were conducted to evaluate COSREW's recommendations. The first explored the differences between search engines and the qualities of the study material links in helping learners achieve their competences. The second experiment explored the differences between search keywords, and the third experiment compared COSREW with freely-browsing learning modes. The results showed that the Web is currently not a good resource for a pedagogically informed competence-based system, since Web pages predominantly comprise text-based subject matter content with little support for learning competence or capability.