The ‘action effect’ according to which actions produce more regret than failures to act, has been shown to disappear in between-subject designs. This phenomenon is replicated in a first study. It is then argued that this disappearance is due to the inability of regret scales to capture differences in perceived regret when used in between-subject designs, a difficulty that is highlighted by a second study. A new method, the common reference method, is proposed in order to overcome this problem. This method is demonstrated in a third and fourth study, and its boundary conditions are discussed, together with its possible extensions.