Skip to main content
. 2018 May 29;7:e31185. doi: 10.7554/eLife.31185

Figure 1. fMRI Paradigms and Choices.

(A) Food Task. Subjects chose between on-screen food items that varied in tastiness and healthiness and a neutral default food. Choices were made in ‘Natural’ [NC], ‘Focus on Health’ [HC], and ‘Focus on Taste’ Conditions [TC]. (B) Altruism Task. Subjects chose between on-screen proposals that affected the payoff of themselves ($Self) and an anonymous partner ($Other) and a default option ($20 for both). Choices were made in ‘Natural’ [NC], ‘Focus on Ethics’ [EC], and ‘Focus on Partner’ Conditions [PC]. (C) (D). Bar plots illustrate condition-wise percentages of healthy (C) and generous (D) choices (M ± SD), and subject-specific scores (circles). *p < 0.05, corrected, †p < 0.05, uncorrected. (E) Computational behavioral model (DDM). Choices (yes/no) are made when the sequential accumulation of noisy value information that unfolds over time crosses the predefined upper or lower threshold for choice. The relative decision value (RDV) at a point in time (t) is computed as the weighted sum of choice relevant attributes plus noise (ε) (i.e., RDVt = RDVt-1 + wTastiness * Tastiness + wHealthiness * Healthiness + εt). In the example displayed here, the value of a candy bar will tend to accumulate in a positive direction if the weight on Tastiness is high (blue line), yielding a choice in favor of a tasty but unhealthy item. However, the value of the food item is more likely to accumulate in a negative direction if the weight on Healthiness is high (brown line). Note that saying Yes can sometimes indicate a healthy choice, and sometimes an unhealthy choice. (RT = reaction times [sec]; figure adapted from [Hutcherson et al., 2015b; Adolphs and Tusche, 2017]).

Figure 1.

Figure 1—figure supplement 1. Drift diffusion model (DDM) fits to behavior in both choice tasks.

Figure 1—figure supplement 1.

(A) Correspondence in the altruism task between observed acceptance rates (top) and response times (bottom) for different proposal types (bars) and model predictions (blue circles, determined using best-fitting parameters for each subject). On average, subject-level correlation between observed and predicted acceptance rates across trial types was generally quite high. (B) Correspondence in the food task between observed and model-predicted acceptance rates (top) and response times (bottom) for foods of varying taste and healthiness (subject-specific ratings outside the scanner). For illustration purposes, for both tasks model fit to behavior is shown for eight bins created based on the displayed color scheme (right) for variations in choice-relevant attributes (increased attribute values from left to right). Thus, bar colors correspond to trials with specified combination of attributes.