Being able to comprehend informal arguments in scientific texts is important for scientific literacy in higher education. Successful intervention studies demonstrating that these skills can be trained in university students have not yet provided evidence that gains of explicit training can be maintained beyond immediate post-training assessment. In this study, we tested whether the gains in argument structure comprehension achieved using a self-directed, web-based training intervention could be maintained over several weeks as an indication of sustained improvement in scientific literacy. We also explored characteristics of students and their engagement with the training intervention that resulted in significant and sustained improvements of their argument structure comprehension skills. One hundred students took part in a voluntary supplement to their university courses, completing an online pretest, a 45-minute training session, a posttest (n = 88), and a follow-up test (n = 31). Training effects at posttest were compared with an active control group. The results suggest that the training group exhibited significant gains in argument structure comprehension. These gains were maintained across a four-week period. Students with low starting ability profited the most from the training and gains in argument comprehension were greatest for complex arguments. Training results were positively related to student motivation and this effect was fully mediated by their engagement with the training exercises. The results demonstrate that training gains can be maintained after immediate post-training assessment and suggest that training is particularly effective for low-performing students, for complex arguments, and when students are motivated and engage with the training exercises.