Acquiring new skills is not a set-in-stone process and students often take various paths to the goal; the acquisition of the required skill. To assess this learning process, previous studies used hidden Markov models to separate the cognitive stages of a problem solving task similar to solving algebraic equations (e.g. Anderson et al., 2012; Anderson, 2012). Because of the slow nature of fMRI recordings, this method can only discriminate between relatively long states in the process. This study extends the approach by including eye movements as a predictor of state, in an attempt to increase temporal resolution of the method. The results show that eye movements can be used to trace the characteristics of the problem the subject is working on. Because tracking eye movements is a non-invasive measure that can be used outside experimental settings, this can benefit the discovery of problems students encounter while solving algebraic problems.