AN EXTENSION OF THE RELATIVE JUDGMENT MODEL
In the relative judgment model (RJM) of absolute identification, the difference between the current and previous stimulus is used, together with the feedback for the previous stimulus, to derive a response for the current stimulus. Thus the model is able to predict the large accuracy benefit when the previous and current stimuli are the same. However, a smaller but significant benefit is also seen when the stimulus two trials ago is the same as the current stimulus. Here, I show that although the original RJM cannot predict this effect, a modified version of the RJM can. In the modified RJM, participants sometimes retain the representation of the previous stimulus at the expense of storing the current stimulus. In this way, the stimulus two trials ago is sometimes used as the standard against which the current stimulus is compared. This modified model provided a significantly better fit for 72% of participants.