It is easy to see that texts do not tell the whole story. Rather, they can be seen as detailed instructions to readers about which background knowledge they should activate and combine into a coherent mental model. There fore, readers will have to add knowledge and link information from the text with relevant background knowledge. In many cases, the text does not exhibit any explicit signal that something is missing, and the reader has to make an inference to maintain the global coherence of the text. It has been shown that students with poor comprehension sometimes fail to make such knowledge-demanding inferences because they fail to activate the relevant knowledge. In a first study, 11-year-old students were taught to activate relevant background knowledge by means of graphic organisers: some boxes had to be filled with information from the text, while others were to be filled with information from the reader’s background knowledge. The teaching turned out to be highly effective not only for inference making in reading, but also for reading comprehension in general. Two limitations are discussed. First, teaching to activate background knowledge is probably much more important for comprehension of expository texts than for typical narrative texts. This is so because comprehension of expository texts depends much more on background knowledge that is abstract, not first hand experience and thus not immediately available. Second, an over-reliance on background knowledge may stand in the way for learning something new that is in conflict with prior knowledge or beliefs. The paper ends with a call for replication studies. It is further suggested that the teaching of knowledge activation for inference making may be integrated into the teaching of common text structures such as cause-consequence, compare-contrast, and problem-solution.