The present study examines the existing published research about the effectiveness of learner-generated highlighting and instructor-provided highlighting on learning from text. A meta-analysis was conducted of scientifically rigorous experiments comparing the learning outcomes (i.e., performance on memory and/or comprehension tests) of students (i.e., college students and/or K-12 students) who read an academic text with or without being asked to highlight important material (i.e., with or without learner-generated highlighting) or who read an academic text with or without the important material already being highlighted (i.e., with or without instructor-provided highlighting). We found 36 published articles that met these criteria ranging from the years 1938 to 2019, which generated 85 effect sizes. The results showed that learner-generated highlighting improved memory but not comprehension, with average effect sizes of 0.36 and 0.20, respectively; and instructor-provided highlighting improved both memory and comprehension, both with an average effect size of 0.44. Learner-generated highlighting improved learning for college students but not for school students, with average effect sizes of 0.39 and 0.24, respectively; and instructor-provided highlighting improved learning for both college and school students, with average effect sizes of 0.41 and 0.48, respectively. We discuss the theoretical and practical implications of these findings.