Experimental Procedures and Resulting Value Functions.

<p><b>(a)</b> The PT-based experiment used two “gambles”, schematized by two spinners. One spinner showed two-thirds of its area as gains (+$10) and one-third as losses (-$8), leading to an expected outcome (i.e., referred to as actuarial outcome in Breiter et al. [<a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0135216#pone.0135216.ref038" target="_blank">38</a>]) of +$4. The second spinner showed one-thirds of its area as gains (+$10) and two-thirds as losses (-$8), leading to an expected outcome of-$2. Each trial lasted 20 seconds, with 10s focused on the arrow spinning (anticipation phase) and 10s focused on the arrow stopping, and the win/loss flickering (outcome phase). Order of presentation between the PT-based experiment and RPT-based experiment was counterbalanced across subjects. <b>(b)</b> The RPT-based experiment used a keypress procedure [<a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0135216#pone.0135216.ref021" target="_blank">21</a>]: a picture would appear for 200ms, then be replaced by a fixation point for 1800ms. After 2000ms, the face would reappear, and if subjects did nothing, the face would stay up another 6000ms (e.g., default condition). Subjects could increase viewing time via a scalloping resistive function, getting close to maximum 1400ms. Alternatively, they could decrease viewing time with the same resistive function close to a minimum of 2000ms. The scalloping resistive function incrementally reduced the viewing time alteration achieved by each keypress, so subjects needed to exert effort to increase or reduce viewing times. Its mathematical formulation can be found in Kim et al. [<a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0135216#pone.0135216.ref021" target="_blank">21</a>], along with multiple control analyses about its impact on subject behavior. <b>(c)</b> The value function for the PT-based experiment mapped subjective ratings made during the anticipation phase of the experiment on the y-axis, and the actuarial amount of the spinner on the x-axis. For the outcome phase of this experiment, the value function mapped the subjective ratings made when the arrow stopped spinning against the gain or loss. <b>(d)</b> The RPT-based graph showed the mean intensity of keypressing to increase viewtime (K₊) or decrease viewtime (K_-) calibrated against the Shannon entropy of keypress patterns to increase (H₊) or decrease (H_-) viewtime. Solid and empty triangles stand for individual data points for the five categories of facial expressions.