This article investigates how students with mathematical difficulties (MD) differ from typically developing (TD) students in enumeration processes of small sets of objects (of 1 up to 9 dots). We present a study with 20 fifth-grade students of which ten were found to have MD in initial diagnostics. The students were supposed to exactly enumerate sets of dots, i.e., to say how many dots they saw. This took place in three conditions: (a) in random arrangements in the subitizing range (1-4 dots), (b) in random arrangements in the counting range (5-9 dots), and (c) in canonical (dice-like) arrangements (1-9 dots). We used eye tracking (ET) to analyze student enumeration processes derived from ET video data. Whereas we did not find significant group differences in students' error rates, we found differences in response times with longer response times for students in MD in the canonical arrangement condition. Further, we found significant group differences in students' enumeration processes in all three conditions (subitizing range, counting range, canonical): Students with MD tended to count all dots more often whereas TD students used more advantageous enumeration processes such as simultaneous enumeration or enumeration of groups of dots more often. Our results support the assumption of qualitatively different enumeration processes between students with and without MD.