An eye-tracking study with 60 native Swedish speakers (18-30 years) was conducted to investigate the positive effects on learning outcomes predicted by the disfluency effect. Subtle low-pass filtering was used as a disfluency manipulation and compared with a control condition using regular text. The text was presented on four separate text presentation screens (TPSs), and eye movements were recorded. A free recall task was given 25 min later, and working memory capacity (WMC) was assessed to test if it would moderate learning outcomes. The disfluency manipulation had no effect on learning outcomes, total reading times on words or lines, first- or second-pass reading on lines, or average fixation durations. Moreover, the disfluency effect was not moderated by students' WMC or self-reported prior knowledge of the topic. However, an adaptation to the disfluency manipulation was found. Total reading times on both words and lines were shorter in TPS 1 and 2 in the disfluency condition compared with the control condition, whereas reading times were longer in TPS 3 and 4. It is discussed if failures to replicate the disfluency effect arise from material features, with positive adaptations to disfluency (i.e., higher effort investment) possibly requiring more comprehensive materials.