Article Figures & Data
Figures
- Figure 1.
Social food choice task. A, Experimental session. In Part 1, hungry participants with unrestricted diets were introduced to 2 partners, and provided taste and health ratings for the foods used in the experiment. Then, in Part 2, participants made food choices for themselves and the partners while their brain activity was measured with EEG. Participants knew that their choices mattered because a single trial was randomly selected and implemented for each recipient in Part 3. B, Sample statements from the Similar (top) and Dissimilar (bottom) partners in the experiment. C, Sample stimuli and screens from the decision task.
- Figure 2.
Behavioral results. A, Percentage of healthy choices for each recipient. B, Average observed RT by recipient. C–E, Observed RT for trials separated on the basis of whether, given the same food option, the decision maker chose the same for (C) Own versus Similar, (D) Own versus Dissimilar, and (E) Dissimilar versus Similar. Thus, Own-Similar:Same Response reflects RT for trials in which the choice was for Own self, and the response was the same as for the Similar other, whereas Own-Similar:Different Response reflects RTs on those trials where the response was different from the comparable choice for the Similar other; and likewise for the other comparisons. *p < 0.05, **p < 0.01.
- Figure 3.
ERP analysis of stimulus value integration by recipient. A, Heat map of significant p values associated with stimulus value, showing significant effects from 500 to 650 ms (green box) and 700 to 850 ms (blue box) after stimulus onset. B, Source reconstructions from 500 to 650 ms (green) and 700 to 850 ms (blue) overlaid on a representative brain image. Particularly during the earlier 500–650 ms window, stimulus value activity was localized to regions including VMPFC (circled). Inset, Top, Spherical masks based on peak coordinates from three neuroimaging studies: blue (Plassmann et al., 2007), magenta (Hare et al., 2009), and cyan (Litt et al., 2011). Inset, Bottom, Source localization of stimulus value from ∼450 to 600 ms after stimulus onset in two ERP studies: red (Harris et al., 2013) and yellow (Harris and Lim, 2016). C, Stimulus value integration by recipient. Topographic scalp distribution (left) and average waveforms for the linear ordering of stimulus value (red represents Strong No; orange represents Weak No; cyan represents Weak Yes; green represents Strong Yes) in the 500–650 ms window (shaded green box) and 700–850 ms window (shaded blue box), plotted separately for each recipient. Solid line indicates Own. Dashed line indicates Similar. Dotted line indicates Dissimilar. D, Comparison of scalp topography at 550 ms (left) and 800 ms (right) after stimulus for the main effect of stimulus value (top), the interaction of stimulus value with Similar recipient (middle), and the interaction of stimulus value with Dissimilar recipient (bottom) revealed significant reductions in stimulus value signals for the 2 other recipients during the late window, in line with average waveform data in C.
- Figure 4.
ERP analysis of attribute-weighting by recipient. A, Scalp topography during time WOI from 500 to 650 ms after stimulus revealed differences in neural weighting of taste ratings (left) and health ratings (middle) by recipient. SOIs were identified by taking the conjunction of sensors showing significant ERP activity for Similar and Dissimilar in the Taste versus Health comparison (black box, right). B, Average waveforms associated with the linear ordering of taste (left) and health (right) plotted separately for each recipient. Solid line indicates Own. Dashed line indicates Similar. Dotted line indicates Dissimilar. Orange represents taste. Green represents health. C, D, Linear contrast weights for taste (C) and health (D) as a function of recipient.
- Figure 5.
Differences in average ERP waveforms by recipient identity. A, Heat map of significant p values showing stimulus-locked regression results for the main effects of Similar (left) and Dissimilar (right). Significant effects were visible for the Dissimilar recipient, but not the Similar recipient, at time windows from 350–400 ms after stimulus onset (green box). B, Scalp topography (left) and average waveforms (right) associated with choice recipient identity (cyan represents Own self; blue represents Similar other; orange represents Dissimilar other) 350–400 ms after stimulus onset. Inset, Source reconstruction from 350 to 400 ms after stimulus overlaid on a representative brain image localized this response to regions including the TPJ and STS. C, Heat map of significant p values showing response-locked regression results for the main effects of Similar (left) and Dissimilar (right). Significant effects were visible for the Similar recipient, but not Dissimilar recipient, ∼300 ms to 180 ms before the response (green box). Later effects in the Dissimilar condition coincide with motor activity related to initiation of the key press response. D, Scalp topography (left) and average waveforms (right) associated with choice recipient identity. Cyan represents Own self. Blue represents Similar other. Orange represents Dissimilar other. Inset, Source reconstruction from 300 to 180 ms before the response localized this differential response to regions including the IPS.
- Figure 6.
Hierarchical Bayesian parameter estimates of the influence of neural response on model parameters. A–C, Trial-by-trial fluctuations in the ERP in SOIs for the value-related ERP from 500 to 650 ms (A) were linked with significantly greater influence on the drift rate v of the taste attribute when choosing for Own self and Similar other (B) and the health attribute when choosing for the Dissimilar other (C), suggesting that this response relates to evidence accumulation in the computational model. D–F, Trial-by-trial fluctuations in the response-locked ERP component (D) differentiated choices for the Similar other to a greater extent than Dissimilar or Own choices, consistent with computational model-fitting of the behavioral barrier parameter (E), and likewise demonstrated a significant influence on the barrier for response (F).
- Figure 7.
DDM simulations. A, Simulations (N = 10,000) of RT data for same versus different choice using best-fitting specification M2. Trials are separated based on the same criteria described for observed data in Figure 2, with observed data (left) plotted next to predicted model RTs (right) for each combination of recipient. B, C, Predicted (B) and observed (C) relationship between drift rates and percentage of different choices when comparing choices for 2 recipients, simulated separately for each of the 16 possible combinations of health and taste rating. Proximity of the dots to the diagonal reflects similarity of food values for 2 recipients, and size and color of dots are coded according to the percentage of different choices.
Tables
Parameter with trial type Model no. DIC MSE choice MSE RT yes MSE RT no v M1 46384.2 0.0137 0.0537 0.0694 v, a M2 46318.6 0.0135 0.0532 0.0689 v, z M3 46522.1 0.0137 0.0559 0.0655 v, a, z M4 46415.6 0.0137 0.0538 0.0682 ↵av, Drift rate; a, threshold; z, starting point.
Parameter Mean SD Drift (v), Intercept −1.982 0.055 Drift (v), HealthOwn 0.062 0.011 Drift (v), HealthSimilar 0.183 0.010 Drift (v), HealthDissimilar 0.660 0.011 Drift (v), TasteOwn 0.722 0.011 Drift (v), TasteSimilar 0.665 0.010 Drift (v), TasteDissimilar 0.007 0.011 Barrier (a), Intercept 2.006 0.047 Barrier (a), Similar 0.125 0.015 Barrier (a), Dissimilar 0.049 0.015 Starting point (z) 0.493 0.006 NonDecisionTime 0.357 0.026 Convergence R̂ 1.0002 0.0002 ↵aStatistics for parameter estimates reflect the mean ± SD of the N = 10,000 posterior samples of group-level parameters.
- Table 3.
DDM posterior estimates and convergence for Model M2 with stimulus-locked ERP dataa
Parameter Mean SD Drift (v), Intercept −2.004 0.057 Drift (v), ERP −0.049 0.010 Drift (v), ERP HealthOwn 0.000 0.005 Drift (v), ERP HealthSimilar −0.003 0.004 Drift (v), ERP HealthDissimilar 0.017 0.004 Drift (v), ERP TasteOwn 0.028 0.004 Drift (v), ERP TasteSimilar 0.023 0.004 Drift (v), ERP TasteDissimilar −0.001 0.004 Drift (v), HealthOwn 0.054 0.011 Drift (v), HealthSimilar 0.179 0.011 Drift (v), HealthDissimilar 0.672 0.011 Drift (v), TasteOwn 0.734 0.011 Drift (v), TasteSimilar 0.675 0.010 Drift (v), TasteDissimilar −0.006 0.010 Barrier (a), Intercept 2.000 0.043 Barrier (a), Similar 0.118 0.016 Barrier (a), Dissimilar 0.043 0.016 Starting point (z) 0.494 0.006 NonDecisionTime 0.360 0.025 Convergence R̂ 1.0006 0.0010 ↵aStatistics for parameter estimates reflect the mean ± SD of the N = 10,000 posterior samples of group-level parameters.
