Syntactic language features of word problems have often been identified by assessment studies and interview studies as potential obstacles in word problem solving. However, students' processes of noticing and interpreting these features have only rarely been investigated in depth because noticing is hard to investigate. Eye-tracking methodology has already been shown to be a suitable methodology for analyzing noticing other word problem features, but not yet for syntactic features. The longitudinal eye-tracking study presented here analyzes fifth graders' processes of noticing and interpreting of syntactic word problems before and after an intervention on word problem solving. Based on the eye-mind hypothesis, the study captures students' noticing through eyetracking, whereas interpreting can be captured by their solutions. The analysis of pre-intervention data reveals that long fixations cannot be identified with successful noticing and vice versa. But when comparing the net dwell time (i.e., the time that students look at the words indicating syntactic structures) before and after the intervention, an interesting decrease is revealed. This decrease occurs for both expected and less expected syntactic structures, regardless of the accuracy of mathematization. Meanwhile, analyzing the revisits of students indicates differences between items with more or less expected structures and correctly or incorrectly mathematized items. Methodologically, the article contributes to understanding which eye-tracking measures are relevant for capturing changes in students' processes of noticing syntactic language features: The revisits and scan paths reveal more insights than the net dwell time.