Fig. 2: Escalations of commitment on foraging tasks.

Once a subject is waiting out a delay, two escalations of commitment can be measured on these tasks. At a given point within the countdown in the wait zone, we ask what is the likelihood that the subject will wait out the delay to receive the reward. This is p(Earn). a As an agent approaches the reward, it becomes less likely to quit (p(Earn) increases). b The probability of waiting out the delay (p(Earn)) increases with the time spent. c This increase is not a simple exponential decrease of quitting, but increases non-linearly with time-spent. d Comparing the same future (10 s remaining before receiving the same reward) with different times already invested can also reveal a sensitivity to sunk costs. e The evidence for increased p(Earn) with time spent (time already waited) originally reported in ref. ¹⁹. The graph shows the probability of earning, p(Earn), as a function of the time remaining in the wait zone. Each colored dot and each line aligned to the colored dots corresponds to the p(Earn) given that the subject has waited a certain amount of time in the wait zone already. Notice that the slopes of the lines decrease, indicating that having waited longer, subjects are more likely to wait out the delay, even for a given countdown remaining. Thus, following the logic in d, we conclude that the subjects are behaviorally sensitive to sunk costs. f The full extent of the data from e, showing the continuous “bubble” of increased willingness to wait with time already waited. Models and future comparisons will be plotted in this form, without the linear slope assumption used in¹⁹. b, c Replotted from ref. ³⁰. d Reprinted from ref. ¹⁹ with permission from the publisher.