Postscript: Rejoinder to Brandstätter, Gigerenzer, and Hertwig (2008).

Abstract

We appreciate that E. Brandstätter, G. Gigerenzer, and R. Hertwig (see record 2006-04733-008) agree that process models are indeed useful for advancing researchers understanding of choice processes. Their statement that choices represent adaptive process is also welcome, as is the emphasis on multiple measures. However, we do disagree on two empirical matters: Describing our research, Brandstätter et al posit that most tests are either null or supportive of the priority heuristic (PH) and that only three of the tests were significant in the opposite direction. This ‘scorekeeping heuristic’ implies that all tests are equally weighted in theory testing. However, we believe that the presence of mostly probability-payoff transitions in the data is a critical test for any falsification of the PH. Brandstätter et al also ignore empirically their own distinction between reading and choice phase. By stating that their reading phase now consists mainly of probability-payoff transitions, and then combining the two phases, they use those transitions to compensate for their absence in the choice phase, making the PH look artificially good, and remove one of its clearest predictions. In essence, we agree with the goals of the approach taken by Brandstätter et al 2008 but argue that if process-tracing data is to inform the development of choice models, then it is important to listen to what the data are saying.