Inference ability is considered central to discourse processing and has been shown to be important across models of reading comprehension. To evaluate the impact of inference instruction, a meta-analysis of 25 inference studies in Grades K-12 was conducted. Results showed that inference instruction was effective for increasing students' general comprehension, d = 0.58, inferential comprehension, d = 0.68, and literal comprehension, d = 0.28. Although skilled and less skilled readers responded similarly on general and inference outcomes, less skilled readers benefited more on literal outcomes, d = 0.97, than skilled readers, d = 0.06. Findings suggest that students can increase their inference ability and that less skilled readers gain the extra benefit of increases in literal comprehension. Findings also suggest that instruction provided in small groups is beneficial for increasing readers' inferential understanding of text.