There are many established reading strategy training programmes, which explicitly teach strategic and meta-cognitive knowledge to improve reading comprehension. Although instruction in strategy knowledge leads to improvements in meta-cognitive skills, the effects do not always transfer to reading comprehension. Therefore, to investigate preconditions for knowledge transfer, two different strategy training programmes were implemented in nine classes of Grade 6 students (N = 148) over the course of one school year. One programme involved teacher-directed instruction of declarative meta-cognitive knowledge (Reading Detectives; Rühl & Souvignier, 2006). The other aimed at improving executive meta-cognition by guided practice: students worked with a computer program based on latent semantic analysis (LSA) (conText) and received immediate feedback on written summaries. Although both groups improved their strategy knowledge to the same extent, the conText group showed a greater improvement in reading comprehension. These fndings suggest that guided practice, which is characterised by intensive practice and individualised corrective feedback, is superior to explicitly teaching strategy knowledge.