Five different models for matching judgments are described. Each one is applied to an experiment employing both matching and ordered discrimination tasks, three payoff matrices, and three levels of stimulus discriminability. Under the matching instructions, subjects indicated whether two patterns of lights had the same or a different number of lights. Under the ordered discrimination instructions, subjects indicated which one of two patterns had more lights. One of the models consistently gives a better account of the individual and the pooled group data than the others, but even for this model the goodness-of-fit chi-squares are generally too large.