Learning from texts is an important part of STEM education, but linguistic text features can affect comprehension and learning. We synthesized 45 experimental studies that investigated the effect of linguistic modifications on STEM text comprehension and learning, with a total of N = 6477 learners. In a meta-regression, we distinguished between eight categories of linguistic modifications and several other moderators, including prior knowledge. We found a small positive mean effect of linguistic modifications (g = 0.15). In line with previous research, the results showed that learners with low content prior knowledge benefited more from linguistic modifications. Personalization and increasing clarity and elaboration had positive effects, while reducing complexity and increasing cohesion did not have significant effects. This study highlights the potential and limits of linguistic modifications in STEM texts and provides implications for educational practice.