The evil of banality: When choosing between the mundane feels like choosing between the worst

Abstract

Our most important decisions often provoke the greatest anxiety, whether we seek the better of two prizes or the lesser of two evils. Yet many of our choices are more mundane, such as selecting from a slate of mediocre but acceptable restaurants. Previous research suggests that choices of decreasing value should provoke decreasing anxiety. Here we show that this is not the case. Across three behavioral studies and one fMRI study, we find that anxiety and its neural correlates demonstrate a U-shaped function of choice set value, greatest when choosing between both the highest value and lowest value sets. Intermediate (moderate-value) choice sets provoke the least anxiety, even when they are just as difficult to select between as the choice sets at the two extremes. We show that these counterintuitive findings are accounted for by decision makers perceiving low-value items as aversive (i.e., negatively motivationally salient) rather than simply unrewarding. Importantly, though, neural signatures of these anxious reactions only appear when participants are required to choose one item from a set and not when simply appraising that set's overall value. Decision makers thus experience anxiety from competing avoidance motivations when forced to select among low-value options, comparable to the competing approach motivations they experience when choosing between high-value items. We further show that a common method of measuring subjective values (willingness to pay) can inadvertently censor a portion of this quadratic pattern, creating the misperception that anxiety simply increases linearly with set value. Collectively, these findings reveal the surprising costs of seemingly banal decisions. (PsycINFO Database Record (c) 2018 APA, all rights reserved).

Figure 1.

Task timeline. Participants rated products individually, then chose between sets of four products. After making these choices, participants viewed the choice sets again and (in separate blocks) rated how much they liked the set overall, how anxious they were about the choice, and how confident they were in their earlier choice. Studies 1–2 included a fourth phase, in which participants reevaluated each item using either a bipolar scale (Study 1) or a willingness-to-pay procedure (Study 2). For Study 4, Phase 2 was performed while undergoing fMRI, and interleaved with set liking trials. Product images were always labeled with 2– 4 word descriptions to clarify the product’s identity, and all phases were free response. See the online article for the color version of this figure.

Figure 2.

Left: In Shenhav and Buckner (2014), item values were determined by a WTP procedure and $0.00 items were excluded. Anxiety was primarily a linear function of average set value (online supplemental Table S1). Right: In Study 1, item values were determined by liking ratings and no items were excluded (see also Figure 3). Anxiety instead only exhibited a quadratic relationship with set value. Unless otherwise noted, plots reflect parameter fits based on trial-wise mixed-effects regressions. Shaded error bars reflect standard error of the mean (SEM). See the online article for the color version of this figure.

Figure 3.

In Study 2, we again observe a quadratic relationship between anxiety and set value (left) but this becomes a positive linear relationship when we exclude choice sets that include a zero-valued bid (based on a follow-up item WTP rating; right). Projected anxiety ratings are overlaid on histograms of item ratings based on the two rating procedures, demonstrating a pronounced peak at $0.00 for the WTP procedure. Shaded error bars reflect SEM. See the online article for the color version of this figure.

Figure 4.

Left: In Study 3, we again observe a quadratic relationship between set value and anxiety, but in this case a bipolar rating scale was used (−10 to 10) and we see that anxiety is minimal around the zero point. Right: Anxiety increases linearly with the distance between a given set value from the fixed zero midpoint on the rating scale. Shaded error bars reflect SEM. See the online article for the color version of this figure.

Figure 5.

BOLD activity in anxiety-related regions correlates with anxiety (left) and set salience (right) when participants are making a choice (pink) but not when they are just “browsing” the set (i.e., appraising the overall value of the set; burgundy). The a priori mask (defined by previous anxiety-related activations) is shown in the left inset. Shaded error bars reflect SEM.

Figure 6.

Left: Collapsing across Studies 1, 2, and 4, we find a negative association between set value and anxiety for lower-value choice sets (those below a participant’s mean set value; dashed line) and a positive association for higher-value choice sets (solid line), confirming that our data exhibit a true U-shaped effect. Right: Across these same studies, we find that the quadratic relationship between set value and anxiety is invariant to within-participant variability in item ratings. This quadratic curve did not differ significantly between participants who used a wider range of the Phase 1 rating scale (light red) and those who used a more limited range (dark red), suggesting that choice anxiety reflected an adaptation to the range of items observed by a given participant. Inset: distribution of interquartile ranges (IQRs) for Phase 1 ratings across subjects. For display purposes, we divide these based on a median split of IQR values, but analyses in the main text analyze this as a continuous variable. Shaded error bars reflect SEM. ^***p < .001. See the online article for the color version of this figure.