Introduction: Because of the preponderance of visual images, many mathematics texts are wholly or largely inaccessible to students who are blind. This study investigated how much description is sufficient to communicate math content in different types of images. Methods: Representative math textbooks from grades five, eight, and 11, aligned to the Common Core, were selected. Mutually exclusive and exhaustive image categories were identified. Clear examples of each image category were chosen, and digital files were created containing the examples and surrounding textual material. Files replicated the printed page as closely as possible, and all elements (words, math expressions, and descriptions of images) were readable by JAWS. Forty-four students who are visually impaired (that is, those who are blind or had low vision) listened to the math pages and answered questions related to the content contained in the images. Results: Students answered content-related questions better with more description, but across four description conditions with varying amounts of description the highest correct rate was low (29%). In looking at individual image categories, students had the most correct answers for number lines (41.0%). Rates of 20% to 33% correct were demonstrated for image categories of shapes, tables, line graphs, bar graphs, and ray diagrams. Correct rates for equations, pie charts, and maps were inconsistent or lower than 15%. Students were positive about math and did not indicate many problems with math texts. Discussion: Descriptions of visual images can communicate important math information, but there are images for which no level of description is sufficient. Many students in the study were not aware of how much visual math content to which they were not provided access. Implications for practitioners: Math texts need to be more accessible for students who are visually impaired. Although describing visual images can improve access to content, that may not be sufficient. Materials should be provided in several formats simultaneously so students can approach material in the mode that fits their needs in a variety of contexts.