Readers with autism (ASD), poor comprehension (PC), and typical development (TD) took part in three reading experiments requiring the production of inferences. In Experiments 1 and 2 reading times for target phrases--placed immediately after text implicitly indicating the emotion of a protagonist or after a number of filler sentences, respectively--were used as measures of inferencing. In Experiment 3, participants were explicitly asked to identify the protagonist's emotion. There were no significant differences among groups in Experiment 1. Compared to TD readers, the PC group performed poorly in Experiments 2 and 3. ASD readers performed worse than PC participants only in the explicit-question task. Although ASD readers can produce inferences, they respond to questions about them with difficulty.