An item response theory model for dealing with test speededness is proposed. The model consists
of two random processes, a problem solving process and a random guessing process, with the random
guessing gradually taking over from the problem solving process. The involved change point and change
rate are considered random parameters in order to model examinee differences in both respects. The
proposed model is evaluated on simulated data and in a case study.