How context alters value: The brain’s valuation and affective regulation system link price cues to experienced taste pleasantness

Informational cues such as the price of a wine can trigger expectations about its taste quality and thereby modulate the sensory experience on a reported and neural level. Yet it is unclear how the brain translates such expectations into sensory pleasantness. We used a whole-brain multilevel mediation approach with healthy participants who tasted identical wines cued with different prices while their brains were scanned using fMRI. We found that the brain’s valuation system (BVS) in concert with the anterior prefrontal cortex played a key role in implementing the effect of price cues on taste pleasantness ratings. The sensitivity of the BVS to monetary rewards outside the taste domain moderated the strength of these effects. These findings provide novel evidence for the fundamental role that neural pathways linked to motivation and affective regulation play for the effect of informational cues on sensory experiences.


Supplemental behavioural analyses and results
1.1. Experienced taste pleasantness ratings during fMRI scanning 1.2. Experienced taste pleasantness ratings during blind wine tasting 2. Supplementary fMRI analyses and results 2.1. Taste-responsive brain regions (contrast: wine > water-like control) 2.2. Price cue-responsive brain regions during wine tasting (contrast: €18 > €3) 2.3. Whole-brain mediation analysis for the linear price cue effect on experienced taste pleasantness ratings 2.4. Whole-brain mediation analysis across the full 8-second wine tasting period 2.5. Experienced value-related brain responses during the monetary decisionmaking task 2.6. Localizing brain mediators or price cue effects on experienced taste pleasantness within brain regions of interest activated under placebo analgesia 3. Additional tables S8 -S20

€3 €6 €18
Experienced pleasantness Supplementary Figure S1. Behavioural results from wine tasting task. (a) Bar graphs depict average experienced pleasantness ratings (z-scored across wine types) for each price and payment (pay/free) condition. Error bars correspond to the standard error of the mean. (b) Line graphs depict average experienced pleasantness ratings (z-scored) across the 6 trials of each price condition (€3, €6, €18) for trials during which participants needed to pay for the sample (left graph) and trials during which samples were free (right graph). Error bars correspond to within-subject SEM.

Experienced taste pleasantness ratings during fMRI scanning
In an additional analysis, we examined whether income predicts the effect of price cue on experienced taste pleasantness ratings. To get at this question we correlated the difference in experienced taste pleasantness ratings of wine cued with a high price (€18) versus a low price (€3) and income information given by 23 participants (coded 1 to 7 for < €15,000 to > €100,000). Seven participants were excluded because they declined to respond to the income questionnaire. Permutation tests revealed no significant correlations between income and the effect of price cue on experienced taste pleasantness (r = -0.10, p = 0.63, 95% CI [-0.36-0.34]).

Taste-responsive brain regions (contrast: wine > water-like control liquid)
As a sanity check, in order to localize brain regions that selectively activated in response to the taste of wine, correcting for effects of price, payment condition and wine type we used a general linear model (GLM) that assessed brain responses at time of wine tasting versus mouth rinse. Accordingly, GLM1 involved the following seven regressors: onset cue display (boxcar duration 2.5 seconds) parametrically modulated by pay/no-pay condition and price cue; onset ITI (duration 6-8 seconds) parametrically modulated by pay/no-pay condition and price cue; onset wine tasting (duration 3 seconds) parametrically modulated by pay/no-pay condition, price cue and wine type; onset swallow wine (duration 2 seconds); onset experienced pleasantness rating (duration reaction time); onset rinse mouth with water-like neutral liquid (duration 3 seconds); and onset swallow (duration 2 seconds). Delta functions of onset regressors and parametric modulators were convolved with the canonical hemodynamic response function and regressed against each subject's fMRI data. Linear contrasts were fit into a second-level random-effects analysis, which used paired t-tests to identify brain regions that activated more strongly to the taste of wine than to the water-like neutral liquid (contrast: onset wine tasting > rinse with waterlike neutral liquid). Findings are displayed in Supplementary Figure S3 and Table S2.
We found significant activations of the nucleus accumbens and the ventromedial/orbitofrontal cortex involving Brodmann areas 25 and 11, which activated more strongly in response to tasting wine than a water-like neutral liquid. At a more lenient threshold of p < 0.005 uncorrected (k = 50 voxels), activations extended to the dorsolateral and dorsomedial prefrontal cortex, areas of the motor cortex (M1, SMA, premotor cortex) and parts of the precuneus and occipital cortex.  Note: The table was obtained using the contrast between the onset of wine tasting versus onset of mouth rinse with a water-like control liquid. All regions listed survived a whole-brain threshold of p < 0.001 extent threshold k = 98 corresponding to cluster level p FWE < 0.05 (family-wise error). OFC: orbitofrontal cortex; vmPFC: ventromedial prefrontal cortex ; M1: primary motor cortex.

2.2.
Price cue-responsive brain regions during wine tasting (univariate contrast: €18 > €3). As another sanity check, we used two additional GLMs to localize brain responses that varied as a function of price cue. To this aim GLM2a was designed in order to match the mediation path a regression. The design matrix consisted of a categorical model with the following onset regressors modelling each event within one trial: cue display (duration 2.5 sec), ITI (duration 6-8 sec), wine tasting (duration 3 sec), swallow wine (duration 0 sec), experienced pleasantness rating (0 sec), rinse (0 sec) and swallow (0 sec). Onsets cue, tasting and swallow wine were further broken down into €3, €6 and €18 price conditions. GLM2b applied the same regressors, but the wine tasting onset was modelled across the whole tasting duration (8 seconds), and similar boxcar functions were used for the wine swallow (2 seconds), the experienced pleasantness rating (response times), and the rinse (3 seconds) and swallow (2 seconds) onsets. Taken together both GLMs contained a total of 13 onset regressors convolved with the canonical hemodynamic response function and regressed against each subject's fMRI data.
Linear contrasts for each regressor were fit into a second-level random-effects analysis, which used paired t-tests to identify brain activity in responses to highpriced (€18) versus low-priced (€3) wines. One-sampled t-tests additionally estimated brain responses at onsets cue display, wine tasting and swallow for each price category (€3, €6, €18), respectively. We restricted our search to value-encoding brain regions using a region of interest (ROI) mask that included brain regions reported by an independent study to respond positively to stimulus value 1 . Moreover, we applied small-volume correction using a more restricted region of interest that consisted of a 10-mm-radius sphere centred around the [x = -4, y = 33, z = -13] MNI coordinates, reported by Plassmann et al. (2008) 2 , for the contrast €18 versus €3 at time of tasting with a lenient initial whole-brain threshold of p < 0.05 uncorrected.
As depicted in Supplementary Figure S4, we found stronger brain activations in response to the tasting of a high-priced wine compared to a low-priced wine (i.e., €18 > €3) in the ventral striatum and the ventromedial prefrontal cortex for GLM 2a (3second wine tasting) (SVC p FWE < 0.05). These activations were observed during the first 3 seconds of wine tasting. GLM2b revealed that across the full 8 seconds of wine tasting value-related activations shifted to a more dorsal part of the vmPFC (Supplementary Figure S5), with a prominent cluster located in the anterior cingulate cortex at p < 0.001 uncorrected. Supplementary Figure S4: Univariate analysis of price cue effects on brain activations at time of wine tasting across the first 3 seconds of the tasting period. Significant voxels for the contrast high versus low price (€18 > €3) are for visualization purposes displayed in yellow at a threshold of p < 0.001 uncorrected, masked by an ROI mask comprising brain regions reported by Bartra et al. (2003) to respond positively to stimulus value. SPMs are superimposed on the average T1 anatomical scan.
The [x, y, z] coordinates correspond to Montreal Neurological Institute (MNI) coordinates and are taken at maxima of interest. Line graphs depict time courses across seconds of activation in the ventromedial prefrontal cortex (vmPFC) and ventral striatum peak, respectively. Black dotted lines denote the 8-second wine tasting period. Shaded errors represent confidence intervals (means ± intersubject SEM).

Y=56 Z=10 X=2
Supplementary Figure S5: Univariate analysis of price cue effects on brain activations at time of wine tasting across the full 8-second tasting period. Significant voxels for the contrast high versus low price (€18 > €3) are displayed in yellow at a threshold of p < 0.001 uncorrected. SPMs are superimposed on the average T1 anatomical scan. The [x, y, z] coordinates correspond to Montreal Neurological Institute (MNI) coordinates and are taken at maxima of interest.

Whole-brain mediation analysis for the linear price cue effect using all three price levels on experienced taste pleasantness ratings
Third, we checked if the results hold for a whole-brain mediation model across all three price conditions for variable x. To this end we conducted an analogous analysis as described in the methods section of the main text. The results hold both for wholebrain and ROI-based mediation analyses, and are reported in Supplementary Figure  S6 and Supplementary Table S3 below.
Supplementary Figure S6: Brain mediators and moderators of the linear price cue effect in N = 30 participants. BOLD activity in the ventral striatum, anterior prefrontal cortex and ventromedial prefrontal cortex mediated linear price cue effects on experienced pleasantness ratings (p FDR < 0.05). Significant voxels are displayed for visualization purposes in yellow (p < 0.001 uncorrected) and orange (p < 0.005 uncorrected), and are superimposed on the average anatomical brain image. These results imply that the ability of the €6 price tag to serve as a quality signal is much noisier than the two extreme prices. Some participants perceived the €6 wine as expensive and some as cheap. To further strengthen this point, we investigated in more depth the consumption behaviour of our participants. To understand what price levels they perceived as acceptable and would buy most frequently, we asked in the debriefing session a question about how much they spend on average for a bottle of wine. Supporting our point, we found that on average participants spend €6.50, with a standard error of the mean of €0.40 (see also SI Table S8 below). Looking at the distributions we indeed observed that there is a lot of heterogeneity about whether €6 is an acceptable bottle price for our participants (supplementary Fig. S7).

Supplementary Figure S7. Distribution of willingness to pay for a bottle of wine.
Bar graphs show the distribution of participants across different prices they would be willing to pay for a bottle of wine ("What is the price for a bottle of ordinary wine you would purchase?").
Taken together, these findings show that €6 was seen as an ambiguous quality signal for our participant pool. Because the price placebo effects described here depend crucially on the price quality intuition of participants, it makes sense that we could not find evidence for price placebo effects on the brain level between the two intermediary price levels (i.e., €18 versus €6 and €6 versus €3).

Whole-brain mediation analysis across the 8-second tasting period
Paralleling the univariate analysis of price cue responses reported in paragraph 2.2, we also conducted a whole-brain mediation analysis over the full 8-second wine tasting period. Analysis details are analogous to the main analysis described in the methods section of the main text. We considered whole-brain activations of continuous voxels at p < 0.001 uncorrected with a cluster extend threshold of k = 5 voxels, and also performed small volume corrections at p FWE < 0.05 using the same ROIs described in the methods section. ) and the correlation of a and b coefficients (cov) across participants denote the joint activation in paths a and b at *** p < 0.001, ** p < 0.01 or + p < 0.05 two-tailed. Note that multilevel mediation effects can be driven either by significant path a and path b co-activation or by covariance of path a and path b coefficients. Note: The table was obtained using a whole-brain multilevel mediation analysis at the time of wine tasting for the full 8-second wine tasting period. All regions listed survived a whole-brain threshold of p < 0.001 uncorrected. antPFC: anterior prefrontal cortex; MFG: medial frontal gyrus; ACC: anterior cingulate cortex.

Experienced value-related brain responses during the monetary decisionmaking task
To investigate experienced value-related brain responses, a GLM was fitted to brain responses at time of reward feedback. The GLM included the following regressors: two onset regressors at time of choice and reward feedback; parametric modulators involved at time of reward feedback: expected value (expressed as the probability of reward conditioned on the number of boxes to choose from in each trial) and experienced value (expressed as €0/€10 reward dummy coded 0 and 1); and at time of choice onset: hand used to respond (dummy coded 1 and -1 to control for motor responses). Linear contrasts for each regressor were fit into a second-level randomeffects analysis, which used one-sampled t-tests to identify brain activity for each regressor compared to a baseline. Figure S10, we found significant activations located in the bilateral ventral striatum and the ventromedial prefrontal cortex extending into the anterior cingulate cortex for experienced value at time of reward feedback (p FWE < 0.05, cluster level). Additional significant activations are reported in Supplementary  Table S6. Each trial started with the display of boxes, and participants had to make a choice within 2 seconds. The number of boxes indicated the expected reward (e.g., one chance of reward out of three). As soon as they chose one box their choice was highlighted for a jittered duration of 1.5 to 4.5 seconds. Next, the reward feedback was displayed on the screen for 2 seconds. Trials were separated by a jittered inter-trial interval (ITI) during which a fixation cross was displayed on the screen. (b) Statistical parametric maps (SPMs) show significant voxels in yellow and orange at time of monetary reward outcome in response to experienced value. SPMs are thresholded at p < 0.001 uncorrected for display and superimposed on the average anatomical brain image. MNI coordinates depict the maximum of interest in each contrast.

Localizing brain mediators or price cue effects on experienced taste pleasantness within brain regions of interest activated under placebo analgesia
To explore this idea we used an ROI mask that was previously reported by Wager and Atlas (2015). The ROI mask combined the vmPFC, dorsolateral prefrontal cortex, lateral orbitofrontal cortex, anterior prefrontal cortex, nucleus accumbens and ventral striatum, amygdala, hypothalamus, periaqueductal grey and rostroventral medulla. It was used to mask mediating brain regions activating in path a x b+cov(a,b). This analysis corresponds to a conjunction of the ROI mask with the mediating brain responses. Because of the exploratory purpose of this analysis and the fact that the sample size was relatively small, we applied a less conservative significance threshold of p < 0.001 uncorrected.
We found that brain mediators in the anterior PFC, the nucleus accumbens/ventral striatum and the vmPFC were indeed located within this ROI mask (Supplementary  Table S7). This overlap of activations may suggest the existence of a common neural signature for expectancy effects across sensory domains from pain to pleasure.

Supplementary Table S7: Mediating path a x b activations within brain regions of interest underpinning increased activation under placebo analgesia.
Note: The table was obtained using whole-brain multilevel mediation analysis at onset of wine tasting in N = 30 subjects. All regions listed survived a whole-brain threshold of p < .001 uncorrected.  Note: The table was obtained using whole-brain multilevel mediation analysis at onset of wine tasting in N = 30 subjects. Regions listed survived a whole-brain threshold of p < .001 uncorrected at the voxel level and *p FWE < 0.05. ACC: anterior cingulate cortex; dmPFC: dorsomedial prefrontal cortex; TPJ: temporal-parietal junction. Note: The table was obtained using whole-brain multilevel mediation analysis at onset of wine tasting in N = 30 subjects. Regions listed survived a whole-brain threshold of p < .001 uncorrected at the voxel level and *p FWE < 0.05 (small volume corrected). Note: The table was obtained using whole-brain multilevel, moderated mediation analysis at onset of wine tasting in a subset of N = 17 subjects. All regions listed were positively moderated in their responses to price cue by BVS responses to reward during the monetary decision-making task. They survived a whole-brain threshold of p < .001 uncorrected at the voxel level. *p FWE < 0.05. mOFC: medial orbitofrontal cortex; lOFC: lateral orbitofrontal; PAG: periaqueductal grey. Note: One week before the fMRI experiment, participants were asked to fill out a Qualtrics questionnaire designed to screen wine drinking habits. The aim of this procedure was to identify participants who appreciated drinking red wine but were not oenology experts. Participants rated liking of red wine and wine expertise on a seven-point Likert scale ranging from not at all to a lot for liking, and from no expert/novice to expert for wine expertise. Thus, higher scores indicate more liking for red wine and greater expertise in wine-related issues. The agreement to the statements listed in the table was given on a nine-point Likert scale (agree to disagree). Higher scores indicate greater disagreement. WTP = willingness to pay.  all to a lot). * Paired ttests revealed a significant difference between pre-and post-fMRI scanning in positive affect (t(29) = 3.36, p < .001), with higher positive affect ratings pre-scanning compared to post-scanning. No significant differences between pre-and post-fMRI scanning were found for negative affect (t(29) = -0.33, p > 0.25) and happiness ratings (t(29) = 0.62, p > 0.25).