Testing the Boundaries of "Paradoxical" Predictions: Pigeons Do Disregard Bad News

Several studies have shown that, when offered a choice between an option followed by stimuli indicating whether or not reward is forthcoming and an option followed by noninformative stimuli, animals strongly prefer the former even when the latter is more profitable. Though this paradoxical preference appears to question the principles of optimal foraging theory, Vasconcelos, Monteiro, and Kacelnik (2015) proposed an optimality model that shows how such preference maximizes gains under certain conditions. In this paper, we tested the model's core assumption that a stimulus signaling the absence of food should not influence choice independently of its other properties, such as probability or duration. In 2 experiments, pigeons chose between 2 options: the "informative option" delivered food on 20% of the trials after a 10-s delay, signaled by a "good-news" stimulus, and delivered no food on the remaining 80% of the trials, signaled by a "bad-news" stimulus. The "noninformative option" delivered food after 10 s on 50% of the trials, regardless of the signal shown. In Experiment 1, the probability of the bad-news stimulus was manipulated from 0.80 to 1.00; in Experiment 2, the duration of the bad-news stimulus was increased every time pigeons preferred the informative option, reaching at least 200 s. Consistent with the model's predictions, pigeons clearly preferred the informative option even when the noninformative option delivered 9 (Experiment 1) and 35 (Experiment 2) times more food.- This study was conducted at the Psychology Research Centre, University of Minho, Portugal, and was supported by the Portuguese Foundation for Science and Technology (FCT) and the Portuguese Ministry of Education and Science through national funds cofinanced by FEDER under the PT2020 Partnership Agreement (Grant UID/PSI/01662/2013). This work was also supported by FCT Grant PTDC/MHC-PCN/3540/2012 to Armando Machado. Ines Fortes and Marco Vasconcelos were supported by an FCT Doctoral Grant SFRH/BD/77061/2011 and an FCT Investigator Grant IF/01624/2013, respectively. We are grateful to Thomas Zentall and the members of the Animal Learning and Behavior Laboratory of University of Minho for their helpful comments on a prior version of this paper.