During instruction, students are typically presented with new information through several modalities, such as through language and images. Students need to attend to these different modalities and integrate the information in both in order to learn and generalize from instruction. Many studies have shown that the features of each modality, such as the use of generic noun phrases or perceptually bland visualizations, influence how much students generalize. However, few studies have manipulated both the linguistic and visual information to examine how students integrate the two modalities and how they generalize when the modalities cue to different levels of generalization. Study 1 examines what combinations of linguistic and visual information are common in elementary school science books. Studies 2-6 show that undergraduate students rely primarily on the linguistic information when generalizing. Study 7 reduced the possible split of visual attention by reading out the text for participants and shows that undergraduate students generalize more broadly when the information in either modality promotes generalization, but their effect does not compound. Study 8 shows that elementary school students generalize more broadly when the linguistic information is broad, but the visual information is rich. These results suggest that students across ages use linguistic features similarly to guide their generalizations, but how they integrate the linguistic and visual information changes with age. Based on these findings, I propose the cues to generality hypothesis, as an account of how students use information in lessons to determine how far to generalize.