Compare problems are the most difficult type of word problems among primary school students, especially inconsistent problems, when the relational term is not consistent with the arithmetic operation required for the solution. The present study examines whether different methods help students successfully solve them and how these approaches influence the reading pattern of the text of the problems. Participants are 50 pupils studying in 4th grade (10-11 years old) divided into two treatment and one control group. In one treatment group the compare problems were solved using graphical representations. In the other treatment group, the focus was on the context of the problem. Eyetracking technology was used for studying the change in the reading pattern of the problem text, using a problem sheet with 6 inconsistent two-step compare problems for pre-test and a similar one for post-test. The findings suggest that the symbolic group, which prioritized focusing on the context, demonstrated the most substantial improvement in terms of performance, followed by the graphical representation group, and finally the control group, for which performance did not improve significantly by practicing these types of tasks. Meanwhile, there are no differences between groups in terms of attentional focus: the percentage of the reading time fixated on key elements (numbers, relational term and pronominal reference word) and context.