This paper demonstrates that information criteria (IC) model selection procedures can be applied to detect a possible changepoint of unknown timing in a linear regression model. Our aim is to find which criteria among existing IC has the best ability to detect a change in a particular regression coefficient when the timing of the changepoint is unknown. We use, as our measure of the ability of a criterion to detect a changepoint, the average mean probability of correct selection (AMPCS) when the model is being selected from a group of alternative models. The AMPCS criterion suggests Hocking’s criterion is the best IC procedure for small samples and BIC is best for large samples.