The aim of this study is to investigate the neural mechanisms underlying the drawing of inferences based on a reader s knowledge during reading. Previous research studies have investigated this topic by using different types of text materials varying in coherence (e.g. Kuperberg et al., 2006), complexity (e.g. Xu et al., 2005), comprehensibility (e.g. Vandenberghe et al., 2002) or acceptability (e.g. Hagoort et al., 2004). Instead of using different types of text materials, we used a less explored method that manipulated the reader s reading goals to vary the level of engagement of inference processes. Cognitive psychologists have shown that the reader s reading goals have considerable influence on the cognitive processes of comprehension and on the content of the resulting representation of the text (Graesser et al., 1994). Here, two functional magnetic resonance imaging (fMRI) experiments were conducted to investigate the neural mechanisms of drawing strategic and routine inferences. The experimental data were analysed using two complementary approaches, namely conventional fMRI data analysis and effective connectivity analysis. Combined with an anatomical model developed in this study, the latter approach enabled us to quantify the interregional interactions modulated by the experimental conditions and to discriminate between several plausible hypotheses regarding how inferences are drawn. The results of both fMRI experiments show that the left inferior frontal gyrus (IFG) in Brodmann area (BA) 45/47 is involved in inference processes, regardless of whether inferences are drawn strategically or routinely, while the left anterior prefrontal cortex in BA 9/10 is only involved in retrieving strategic inferences. The effective connectivity analyses show that the retrieval of strategic and routine inferences consistently enhances the connectivity between the left posterior superior temporal sulcus and the left dorsal lateral IFG.