We have developed a process model that learns while finding faults in a simple control panel device. The model accounts very well for measures such as problem solving strategy, the relative difficulty of faults, average fault-finding time, and the speed up due to learning. However, subjects tended to take longer to find a fault the second time they completed a task than the model predicted. To further test the model, we compared the model's sequential predictions-the order and relative speed that it examined interface objects with a subject solving five tasks. Both verbal protocols and mouse movements were compared with the model's performance. We found that (a) the model's operators were applied in basically the same order as the subject's actions; (b) during the initial learning phase there was greater variation in the time taken to apply operators than the model predicted; (c) the subject spent time reflecting on or checking their work after completing the task (which the model did not). The sequential analysis reminds us that though aggregate measures can be well matched by a model, the underlying processes that generate these predictions can differ. The failure to match can be accounted for by the time that the subject spent reflecting on or checking whilst they learn to solve the fault-finding problems.