Table 1.
Expected value of the high-uncertainty choice as a function of the probability of food for a high-uncertainty choice for each group in Experiment 1.
Fig 1.
Mean (±SEM) log odds of high-uncertainty choices for each group in Experiment 1 as a function of the probability of high-uncertainty food delivery (A) and the expected value of the high-uncertainty outcome (B).
The horizontal line indicates neutral preference of the two outcomes. Values below this line reflect risk aversion (more low-uncertainty than high-uncertainty choices), and values above it reflect risk proneness (more high-uncertainty than low-uncertainty choices). A second ordinate showing the percentages corresponding to the log odds values was included to aid interpretation.
Fig 2.
Mean (±SEM) log odds of high-uncertainty choices for each group in Experiment 1 as a function of the outcome of the previous choice, collapsed across the probability of high-uncertainty food delivery.
A second ordinate showing percentages corresponding to the log odds values was included to aid interpretation. L-S = low-uncertainty-small; L-L = low-uncertainty-large; H-Z = high-uncertainty-zero; H-S = high-uncertainty-small; H-L = high-uncertainty-large.
Fig 3.
Mean (±SEM) difference score between the log odds of high-uncertainty choices following high-uncertainty-large (H-L) outcomes and the log odds of high-uncertainty choices following high-uncertainty-small (H-S) outcomes for each group in Experiment 1 as a function of the difference between the magnitude of the H-S outcome and the expected value of the high-uncertainty choice.
A negative/positive difference on the abscissa indicates that the H-S outcome magnitude was greater than/less than the expected value of the high-uncertainty choice (gain/loss, respectively). A negative or positive difference on the ordinate reflects a greater likelihood to make high-uncertainty choices following H-S or H-L outcomes, respectively.
Table 2.
Expected value of the high-uncertainty choice as a function of the probability of food for a high-uncertainty choice for each group in Experiment 2.
Fig 4.
Mean (±SEM) log odds of high-uncertainty choices for each group in Experiment 2 as a function of the probability of high-uncertainty food delivery (A) and the expected value of a high-uncertainty outcome (B).
A second ordinate showing percentages corresponding to the log odds values is included to aid interpretation.
Fig 5.
Mean (±SEM) log odds of high-uncertainty choices for each group in Experiment 2 as a function of the outcome of the previous choice, collapsed across the probability of high-uncertainty food delivery.
A second ordinate showing percentages corresponding to the log odds values was included to aid interpretation. L-S = low-uncertainty-small; L-L = low-uncertainty-large; H-Z = high-uncertainty-zero; H-S = high-uncertainty-small; H-L = high-uncertainty-large.
Fig 6.
Mean (±SEM) difference score between the log odds of high-uncertainty choices following high-uncertainty-large (H-L) outcomes and the log odds of high-uncertainty choices following high-uncertainty-small (H-S) outcomes for each group in Experiment 2 as a function of the difference between the magnitude of the H-S outcome and the expected value of the high-uncertainty choice.
A negative/positive difference on the abscissa indicates that the H-S outcome magnitude was greater than/less than the expected value of the high-uncertainty choice (gain/loss, respectively). A negative or positive difference on the ordinate reflects a greater likelihood to make high-uncertainty choices following H-S or H-L outcomes, respectively.
Table 3.
Expected value of the high-uncertainty choice as a function of the probability of receiving 0 pellets (P[0]) or 1 pellet (P[1]) following a high-uncertainty choice in the high-uncertainty probability training procedure of Experiment 3A.
Fig 7.
Mean (±SEM) log odds of high-uncertainty choices for each group in the high-uncertainty probability training procedure of Experiment 3A as a function of the probability of the 0-pellet outcome (P[0]) or the 1-pellet outcome (P[1]).
Note that high-uncertainty choices are plotted against the probability of the manipulated outcome rather than as a function of overall probability of food, as was the case in previous figures. A second ordinate showing percentages corresponding to the log odds values was included to aid interpretation.
Fig 8.
Mean (±SEM) log odds of high-uncertainty choices collapsed across groups in the high-uncertainty probability training procedure of Experiment 3A as a function of the outcome of the previous choice, collapsed across probability.
A second ordinate showing percentages corresponding to the log odds values was included to aid interpretation. L-S = low-uncertainty-small; L-L = low-uncertainty-large; H-Z = high-uncertainty-zero; H-S = high-uncertainty-small; H-L = high-uncertainty-large; P[0] = probability of 0 pellets; p[1] = probability of 1 pellet.
Fig 9.
A: Zero-order correlation between the mean log odds of high-uncertainty choices across probabilities of the P[0] condition (abscissa) and the difference score in the log odds of high-uncertainty choices following high-uncertainty-zero (H-Z) outcomes relative to high-uncertainty-small (H-S) outcomes (ordinate).
Here, post H-Z high-uncertainty choice behavior was subtracted from post H-S high-uncertainty choice behavior. B: Partial correlation between the two variables in Panel A, controlling for mean post H-Z choice behavior. The abscissa and ordinate show the residuals derived from the correlations of these variables with mean post H-Z choice behavior. C: Zero-order correlation between the mean log odds of high-uncertainty choices across probabilities of the p[1] condition (abscissa) and the difference score in the log odds of high-uncertainty choices following H-S outcomes relative to H-Z outcomes (ordinate). Here, post H-S high-uncertainty choice behavior was subtracted from post H-Z high-uncertainty choice behavior. D: Partial correlation between these two variables, controlling for mean post H-S choice behavior. The abscissa and ordinate show the residuals derived from the correlations of these variables with mean post H-S choice behavior. The best fitting regression line and variance accounted for (R2) is shown in each panel.
Fig 10.
Mean (±SEM) log odds of high-uncertainty choices in the low-uncertainty magnitude training procedure of Experiment 3B as a function of the outcome magnitude of a low-uncertainty choice.
A second ordinate showing percentages corresponding to the log odds values is included to aid interpretation.
Fig 11.
Mean (±SEM) log odds of high-uncertainty choices in the low-uncertainty magnitude training procedure of Experiment 3B as a function of the outcome of the previous choice and the outcome magnitude (pellets, p) of a low-uncertainty outcome choice.
A second ordinate showing percentages corresponding to the log odds values is included to aid interpretation.
Table 4.
Summary of the primary results from Experiments 1–3.
Fig 12.
A proposed mechanism to account for the asymmetric effects of previous high-uncertainty choice outcomes on subsequent choice behavior.
The abscissa is the magnitude of the previous high-uncertainty outcome. The ordinate is the subjective value of the previous high-uncertainty outcome relative to the value of the low-uncertainty outcome choice (e.g., 3 food pellets). The low-uncertainty outcome reference (XRP) differentiates gains and losses for the high-uncertainty outcomes.