Abstract
The present thesis examined the role of past choices on present choice, specifically, the sunk cost effect. The sunk cost effect is the tendency to persist in a losing course of action due to past investments in that particular option, when the optimal alternative would be to abandon that endeavor. These past investments can be either in terms of money, effort or time spent. This problem is of importance in the context of choice and decision-making because both economic and behavioral theories of choice predict that optimal choice is determined only by future benefits and not by past investments. The sunk cost effect is therefore seen as suboptimal choice.
In this thesis, investments in time and effort were examined, and the underlying theme was whether a lack of discrimination between the outcomes of a choice could account for the influence of the past choices on current choice. Experiments in Chapters 2 and 3 concerned the sunk time effect, i.e., investments of time into a course of action. In these experiments, pigeons (Columba livia) chose between two concurrent alternatives, one producing food according to two Fixed-Interval schedules whereas the other allowed pigeons to escape from the trial by terminating it and starting a new, randomly chosen one. Escaping from the long Fixed-Interval schedule was the optimal choice because the short schedule had a higher probability of being selected (Chapter 2) or equal probability (Chapter 3), which would guarantee a sooner reward. The sunk time effect was observed in various conditions, such as: in the absence of cues signaling the type of trial or in the absence of reward in the long interval; at different fixed-interval durations and intertrial interval durations. Importantly, a Signal Detection analysis confirmed that lack of discrimination between scheduled contingencies was not the source of the preference for persisting.
Experiments in Chapters 4 and 5 examined the sunk cost effect, i.e., investments in effort into a course of action. In modified versions of a standard concurrent-chains procedure, pigeons chose between two options, each ending in reward. In some trials (Chapter 4) or conditions (Chapter 5), the concurrent choice was preceded by a prior investment in one of the options, in the form of a fixed number of responses. In both chapters, the outcomes of the choice were independent of the prior investment, in that the prior investment in one of the options did not bind the subject to that option, nor did the prior investment change the amount of effort in either option in the outcome phase. Because of this, subjects were expected to show a preference in the choice phase, for the outcome that represented less effort to reward, regardless of being associated with prior investment. The sunk cost effect was observed with pigeons showing a preference for the option associated with prior investment, and this effect was stronger when more effort was spent in the prior investment phase (Chapter 5). Additionally, this preference could be so strong that pigeons were willing to expend twice as much effort in the alternative associated with the prior investment, when they could freely choose the other alternative (Chapter 4). Importantly, Matching Law analysis confirmed that this result was not due to an inability to discriminate between outcomes of the choice, and that it was indeed a bias created by the prior investment.
The overall conclusion of the present thesis is that the sunk cost and sunk time effects do not result from an inability to discriminate between conditions of reward or between the outcomes of choice. Additionally, the sunk cost and sunk time effects seem to be learned effects rather than the result of a momentum to continue responding. Several possible accounts for this effect were discussed, including within-trial contrast, and delay-reduction, among others. At the present stage, within-trial contrast seems the best able to explain this preference for continuing an endeavor once time or effort has been invested.