Inferences are important because not everything in the text is explicit. Therefore, the reader must generate inferences that fill in "missing" information. Various factors can influence inference processes, including those that are related to the text and reader. Moreover, these two factors are likely to interact in highly complex ways, although the nature of that interaction is not well understood. In this study, a novel application of multilevel modeling was used to examine joint (additive and interactive) contributions of text-based and reader-based factors to the prediction of specific inference processes, namely integrative and elaborative processes. A simple two-level model was constructed in which sentences (level-1) were nested within persons (level-2). Sentence-level predictor variables included argument overlap, implied causal relationships, new-argument nouns, and prior knowledge overlap. Reading skill served as the person-level predictor variable and was assessed using the Nelson-Denny Reading Skills Test. In Study 1, participants read science texts and typed their understanding after each sentence. The thoughts that readers produced were coded for integrative and elaborative processes. Results revealed significant additive effects at level-1, such that integrative processes were positively correlated with the presence of an implied causal relationship, whereas elaborative processes were positively correlated with the introduction of new-argument nouns. There were no significant additive effects at level-2, although reading skill did moderate the relationship between prior knowledge overlap and integrative processes. Specifically, more skilled readers exhibited a stronger negative relationship between these two variables. Reading skill also moderated the relationship between argument overlap and integrative processes, such that skilled readers exhibited a stronger negative relationship between these two variables. In Study 2, participants read science texts and sentence reading times were recorded. Results revealed significant additive effects at level-1, such that reading times were negatively correlated with argument overlap, implied causal relationships, and new-argument nouns. There were no significant additive effects at level-2, nor were any of the cross-level interactions significant. These results suggest that inference processes are influenced by the features of a text and that a reader's sensitivity to those features may be moderated by reading skill. Moreover, the use of multilevel modeling to study inference processes appears to be a viable alternative to more traditional statistical techniques, such as analysis of variance and multiple regression. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://bibliotheek.ehb.be:2222/en-US/products/dissertations/individuals.shtml.]