The visual syntax of modelling languages can support (or impede) the intuitive understandability of a model. We observed the process of problem solving with two notation variants of i∗ diagrams by means of an eye-tracking device. The number of wrongly answered questions was significantly lower when the alternative i∗ notation suggested by Moody et al. was used. For the eye-tracking metrics "time to solve a task" and "number of eye fixations", no such significant result can be given. Furthermore, we identified a deficiency for the 'dependency' symbol in the alternative notation.