Millions of U.S. adults lack the literacy skills needed for most living-wage jobs. We investigated one particular comprehension process for these adults: generating predictive inferences. If a sentence says that someone falls from a 14th-story roof, a reader should infer almost certain death. On any test of comprehension, there are two dependent variables: the speed of the response to a test item and accuracy. To address both simultaneously, we used a decision model that separates how much information an individual understands from a text and the individual's speed/accuracy trade-off settings. We found that adult literacy students do differentiate between predictive inference sentences and control sentences, a finding that illustrates how a decision-making model combined with tests of particular comprehension processes can lead to further understanding of low-literacy adults' reading skills.