1. Introduction
Being able to distinguish between safe and risky options is paramount to make functional choices in our daily lives. However, especially in a highly complex and uncertain world like the one we are living in, this is not always an easy task.
Research has increasingly highlighted the role of emotional reactions in helping people to generate an adaptive response through a quicker and easier way to navigate complexity and uncertainty [
1,
2]. In the risk studies domain, for example, approaches such as the “
risk as feelings” [
3] assume that people base their judgments not only on what they think or know about an event/activity but also on what they feel about it (“
I like it/I dislike it”). Following this theoretical approach, events/activities are marked in people’s minds with positive or negative affective reactions which occur rapidly and automatically and can be experienced, consciously or not, as a feeling state. Thus, according to the so-called “affect heuristic” [
4,
5], affective reactions come prior to and are used as an orienting mechanism in the decisional process and the perception of risks and benefits, especially in complex and uncertain situations.
Similarly, the “somatic marker hypothesis” (SMH [
6,
7,
8,
9]) proposed that, in particular situations, bodily responses and corresponding central nervous representations, can generate emotional responses (“
somatic markers”) that, in turn, can guide and assist the decisional process. This somatic mechanism is supposed to precede overt reasoning on declarative knowledge [
10]. Indeed, somatic markers mark mental images of future outcomes with positive or negative emotional tags and are then used as a factor of efficiency and accuracy in decision making, even before conscious knowledge of the situation is available. The generation of these visceral tags can be detected from physiological signals (e.g., skin conductance response (SCR), heart rate variability) in terms of autonomic arousal. The ventromedial prefrontal cortex (vmPFC), orbital frontal cortex (OFC), and amygdala have been indicated as the neural basis and core brain regions of the somatic marker system [
11,
12,
13]. Clinical observations on patients with damage to those areas have been indeed associated with decision deficits. Notably, this kind of patient performed poorly on the Iowa Gambling Task (IGT), an experimental paradigm used to simulate real-life decision-making under uncertainty [
7,
12,
14]. The IGT is used as a tool to investigate the role of emotional reactions in the decisional process, as well as to test the SMH, in multiple fields [
7,
12,
14,
15]. The goal of the task is to maximize a
$2000 virtual loan by picking cards from four decks. Two decks (A and B) are disadvantageous because they yield high immediate gains (
$100) but high future unpredictable losses, i.e., overall net loss. We refer to them as “Bad decks”. The other two decks (C and D) are advantageous because they yield lower immediate gains (50
$), but lower future unpredictable losses, i.e., overall net gain. We refer to them as “Good decks”. Participants can not guess the internal gambling schedule, and they do not know how many cards they will have to pick (the game is stopped after 100 selections), leaving them in an uncertain situation. Good performance at the IGT is typically associated with a higher proportion of selections from Good decks.
Patients with brain damages, indeed, usually selected more cards from Bad decks rather than from Good ones and failed to produce anticipatory autonomic somatic response, measured as anticipatory skin conductance response, before selecting a card from a Bad deck. On the other hand, non-impaired control subjects gradually tended to avoid Bad decks and started to generate pronounced anticipatory SCR before selecting a card from those decks [
7,
14]. Specifically, “they begin to generate anticipatory SCR whenever they ponder a choice that turned out to be risky before they knew explicitly that it was a risky choice. The results suggest that, in normal individuals, nonconscious biases guide behavior before conscious knowledge does” [
14]. Differently said, this affective reaction, experienced as a somatic state associated with the monetary outcome (deterministic gains and probable losses) of the decks, ultimately leads people with an intact somatic marker system to perform better, by helping them to distinguish between potentially risky choices (that are indeed avoided) and safer ones (that are indeed approached), even when their knowledge about the situation is not yet available or at least uncertain (“pre-hunch” period) [
6,
8,
16].
In these terms, affective reactions can be considered beneficial to decision-making [
6]. However, reliance on emotional reactions can also be detrimental. For example, induced strong and negative emotional states before the IGT have been associated with poorer performances [
6,
17,
18,
19].
The intuition that affective reactions, even if unrelated to structural aspects of an option (e.g., its outcome and chances), might disrupt decision-making is not new. For example, in the marketing and advertising domains, affective reactions are usually exploited to influence consumer choices through “normatively irrelevant affective cues”. With this term, we refer to aesthetic features (e.g., pictures, colors, or symbols) used to generate positive or negative affective reactions towards a product, but that do not inform the consumer about normative characteristics of the consumer good such as its price or quantitative or qualitative aspects [
20,
21,
22,
23]. Thus, from a rational and economic perspective, these aesthetic affective cues should be irrelevant in the options’ weight, while the consumer should focus only on their economic value to make his choice.
Although the power of irrelevant affective cues is widely recognized in the consumer-behavior field, very few studies investigated their role in the domain of decision under uncertainty and risk-taking. A study from Gnambs and colleagues [
24], for example, investigated the role of “normatively irrelevant” task-cues in the Balloon Analogue Risk Task (BART) [
25]. The BART is typically used to study risk-taking in uncertain conditions. It has been found to correlate with everyday risky behavior such as drug consumption, smoking, and safety behaviors [
25,
26,
27]. Gnambs and colleagues [
24] found that the color of the balloon to be inflated (red vs. blue) affected participants’ risk-taking. Participants inflated the balloon less often when it was red than blue-colored (keeping constant the gambling schedule). The change in risk-taking behavior as a function of the color of the balloon was interpreted as due to the symbolic value of the red color, traditionally associated with danger.
Following this relatively uninvestigated research line, in our study, we aimed to demonstrate how the same normatively irrelevant affective cue could be both detrimental and beneficial to risk-taking behaviors in conditions of uncertainty. Specifically, we wanted to understand whether the affective reactions driven by a negative manipulation of the options (i.e., the hearing of an unpleasant sound associated with card selections) interfered with risk-taking in the IGT.
Thus, in our study, a computerized manipulated version of the IGT was used (see
Section 2.3.1). Specifically, the hearing of an unpleasant sound has been systematically associated with selections from the Bad decks in a first condition (Congruent condition) and to selections from Good decks in a second condition (Incongruent condition) to provoke a negatively valenced affective reaction (see
Section 2.3.2). A condition with no manipulation was also included to serve as a control. The negative affective manipulation has been designed to be unrelated to the gains and losses contingency of the decks. Indeed, the unpleasant sound was heard each time participants selected a Bad or a Good deck, according to the condition, regardless of the monetary outcome of that selection. In these terms, the negative affective reaction driven by the unpleasant sound was normatively irrelevant to the overall task goal (i.e., maximize the profit), as it did not inform participants about the economic value of the decks.
Therefore, in the experimental conditions, participants would have experienced two negative anticipatory affective reactions: the first affective reaction was driven by the deterministic hearing of the unpleasant sound associated with a certain type of deck and was irrelevant to the overall task goal, while the second affective reaction was driven by the possible negative monetary outcome of the decks and was thus relevant to the overall task goal.
In line with the SMH, the relevant anticipatory affective reactions associated with the disadvantageous risk contingency of the Bad decks should have assisted participants’ decisional process and guided them to avoid those decks. However, we believed that the presence of another irrelevant affective reaction, driven by the unpleasant manipulation, would have influenced participants’ risk-taking. Specifically, we expected that in the Congruent condition, the irrelevant negative affective reaction driven by the unpleasant sound would have helped participants to avoid Bad decks, as this first irrelevant affective reaction would have been added to the relevant affective reaction associated with the probable disadvantageous outcome of those decks. Differently said, we expected Bad decks to be “more intensively” somatically marked. We hypothesized therefore that participants in the Congruent condition would have shown a higher proportion of selection from the Good decks compared to the control condition (H
1a). From a physiological perspective, we expected to find higher autonomic activation (i.e., higher anticipatory SCR and higher heart rate) before selections from the Bad decks rather than before Good decks as suggested by the SMH. Indeed significant albeit small-to-medium associations between IGT performance and autonomic activation (measured as anticipatory SCR) have been found in a recent meta-analysis, alongside small differences preceding selections from Bad and Good decks [
28]. However, due to the summative effect of the irrelevant and relevant affective reactions, we expected that autonomic activation before Bad decks in the Congruent condition would have been more pronounced compared to the control condition in which only the relevant affective reaction could have been experienced (H
1b).
Conversely, we expected that the irrelevant negative affective reaction would have been disruptive in the Incongruent condition. Indeed, in this condition, participants should have experienced an irrelevant negative affective reaction driven by the hearing of the unpleasant sound associated with the Good decks, and a relevant affective reaction associated with the disadvantageous contingency of the Bad decks. We believed that the presence of the irrelevant affective cue associated with the Good decks would have misled participants masking the advantageous probable outcome of those decks and leading them to choose a lower proportion of cards from the Good decks compared to the control condition (H2a). We also expected that this disruptive effect could have been detected also from a physiological point of view. Specifically, we expected to find a more pronounced autonomic activation before selections from the Good decks in the Incongruent condition rather than in the control condition (H2b). Indeed, when the Good decks were not manipulated as in the control condition, the generation of an alarming affective reaction was not necessary. However, in the Incongruent condition, the presence of the unpleasant sound associated with Good decks would have made those decks somatically marked, leading participants to generate a somatic anticipatory alarm reaction towards them that could have been used to guide their decisional process in the wrong direction.
Alongside classic anticipatory SCR, heart rate variability (HRV) was also included in the present study. The measurement of heart rate acceleration/deceleration is interesting because changes in HRV have been associated with autonomic activation [
29,
30,
31,
32,
33]. In particular, a faster heart rate has been associated with heightened sympathetic arousal, reflecting preparation for defensive action, while heart rate deceleration has been associated with parasympathetic responses reflecting attentional orienting in response to a potential threat [
34,
35,
36]. Neuroimaging studies also founded a link between HRV and brain regions (e.g., amygdala and vmPFC) involved in threat perception [
32,
33]. However, to the best of our knowledge, only few studies investigated both anticipatory SCR and HRV in relation to IGT performances, and results are mixed. A first study from Crone et al. [
37] found good performance to be related to higher anticipatory SCR and heart rate slowing preceding selections from Bad decks (versus Good decks), but those results were not supported by a recent study from Hayes et al. [
38]. Thus, our study aimed also to better investigate the effect of HRV in relation to IGT performance.
Finally, IGT performances have also been associated in the literature with some personality characteristics (impulsivity and sensation-seeking traits and thinking styles) and participants’ state mood at the moment of the task. Thus, possible effects of the aforementioned individual differences were controlled for also in the present study. Theoretical rationale, methods, and results can be found in the
Supplementary Materials (Section 1, Tables S1 and S2).
4. Discussion
In this study, we tested how the same normatively irrelevant affective cue could have been both beneficial and detrimental to risk-taking using a manipulated version of the IGT. Drawing on the SMH [
6,
7,
8,
9] and risk as feeling [
3] approach, we hypothesized that an irrelevant affective cue (i.e., the hearing of an unpleasant sound) could have helped participants to choose even more advantageously when it was associated with selections from the Bad decks, while it would have misled participants to choose less advantageously when it was associated with Good decks. The normatively irrelevant affective cue was designed to be irrelevant to the overall goal task (i.e., maximize a monetary loan), as it was not informative about the monetary outcome of the decks.
In the Congruent condition, we expected that the hearing of the unpleasant sound associated with selections from the Bad decks would have generated a normatively irrelevant affective reaction towards those decks that would have summed up with the relevant affective reaction driven by the disadvantageous risk contingency of the decks suggested by the SMH. We expected that this additive effect would have led participants to avoid even more selections from the Bad decks while preferring selections from the Good one in the Congruent condition compared to a control condition with no manipulation. However, no differences in the probability of selecting a Good deck were found between Congruent and Control condition. Furthermore, in both conditions, a significant increment in the probability to select from a Good deck as the task unfolded was detected. Said differently, in both conditions, participants initially sampled from all the decks and then increasingly tended to select more often from the Good ones, which is a typical finding with normal populations in the IGT [
7,
12,
14,
15].
From a physiological perspective, however, a higher trend in anticipatory SCR before selections from Bad decks was found in the Congruent condition compared to the control one as hypothesized. Thus, results suggest that the irrelevant affective reaction and the relevant somehow integrated to more intensively somatically mark Bad decks. However, as told above, this strengthened somatic marker did not affect participants’ behaviors towards safer performances. Moreover, no effects of anticipatory SCR were found in the Control condition. This result is in contrast with classic findings of SMH and with a recent metanalysis showing differences in anticipatory SCR between Bad and Good decks, even though those effects were small to medium [
28].
Nevertheless, a peculiar effect of heart rate variability was detected. Indeed, a growing trend in heart rate deceleration was found before selections from the Good decks rather than before Bad decks in both Control and Congruent conditions. Furthermore, this trend was higher before selections from Good decks in Congruent condition. Previous literature associated anticipatory heart rate deceleration with attentional orienting responses in preparation for a possible threat [
34,
35,
36]. Indeed, slower heart rate was found before selections from Bad decks and associated with better performance on the IGT [
37]. However, a slower heart rate has been also associated with lower stress reaction due to the parasympathetic influence on the heart through the vagus nerve [
29,
30,
31,
33]. We can therefore speculate that participants in both Control and Congruent conditions were less stressed while pondering an advantageous choice. This would be in line with Damasio’s hypothesis that “when a positive somatic marker is juxtaposed to a particular future outcome it becomes a beacon of incentive” [
7] to pursue that action. Indeed, our participants correctly chose more from Good decks in Control and Congruent condition and were less stressed before doing so (i.e., positive somatic marker) as shown by their cardiac activity. This might be the first empirical evidence of a positive somatic marker’s effect measured as heart rate variability, although more data are required to confirm this hypothesis.
Taken together, these results showed that the affective cue was effective in generating a different somatic response between Good and Bad decks when the latter were associated with it, but that it was ineffective from a behavioral point of view. We can speculate that the disadvantageous risk contingency of the Bad decks played a predominant role in guiding participants toward Good decks and that the effect of the affective cue has been actually irrelevant. These results are in line with some critiques raised to the SMH and IGT. For example, Maia and McCelland [
47] suggested that the gains and losses contingency of the IGT might be more cognitively penetrable than suggested by the authors and that it is performed through access to explicit knowledge. This implies that the generation of somatic markers might be subsequent to conscious knowledge and not necessary for adequate performance. Moreover, learning models of the IGT, such as reinforcement learning models and reversal learning models, suggested that other mechanisms might be involved in IGT performance (see [
48] for a review, see also [
38,
49]).
In the Incongruent condition instead, we expected participants to experience a negative irrelevant affective reaction towards Good deck when they were associated with the hearing of the unpleasant sound and that this affective reaction would have had a disrupting effect on their performance. Results confirmed that participants in this condition showed a systematically lower probability to select from a Good deck compared to both Control and Congruent conditions, thus showing a higher risky behavior performing more potentially disadvantageous choices. The disrupting effect of the affective cue was also detected from a physiological point of view. Indeed, even though no difference between the decks could be detected for anticipatory SCR, a higher heart rate deceleration before selections from a Bad deck was found. We can speculate that the presence of the affective cue associated with the Good decks misled participants’ somatic marker system to associate a positive somatic marker with the riskier, but not negatively manipulated, Bad decks, making participants less stressed when pondering a potentially disadvantageous choice. Nevertheless, the affective cue did not seem to completely disrupt participants’ performances. Indeed, a growing trend in the probability to select a Good deck as the task unfolded was found as in the Control and Congruent condition. Said differently, their performance increasingly improved as the task unfolded, even though participants in this condition chose a lower amount of cards from Good decks. It seems therefore that participants in this condition adequately understood the differences in decks’ risk contingency (i.e., Bad decks alluring large gains but potentially larger losses, resulting in an overall net loss, and Good decks alluring smaller gains but also potentially smaller losses, resulting in an overall net gain) and learned they had to endure the hearing of the unpleasant sound to reach a larger final overall gain.
Taken together, these results seem to resemble real-life behaviors. Imagine being told by a doctor to avoid a particularly bad-tasting food you always disliked to get better health. In this situation, you probably will not have any trouble following the diet and improve your health. Similarly, participants in our study did not have trouble avoiding the Bad decks when they were associated with the negative affective cue. Imagine instead the opposite situation, in which the doctor recommends you to always eat the same bad-tasting food, as it is very beneficial for your health. In this situation, it is probable that you will stick with the diet as you want to have better health and will endure the unpleasant food to reach your goal. However, it is also probable that from time to time, you will not comply with the diet, preferring a delicious and not healthy food instead of the bad-tasting and healthy one. Similarly, participants in the Incongruent condition endured the unpleasant sound associated with the Good decks, as they recognized it was beneficial to their overall performance but had more trouble in doing so.
Limits and Future Directions
In this study, one negative auditory manipulation was used as an affective cue to make the decks unpleasant and elicit a normatively irrelevant affective reaction. Considering that the sound used was the same throughout the unfolding of the IGT, it is possible that participants got used to it, thus being less influenced by the manipulation, especially in the last portions of the game. Future studies could use two or more sounds to control for a possible habituation effect. Moreover, the manipulation used in this study was meant to be highly generically unpleasant. We did not focus on more specific and potentially powerful negative emotions, such as fear or disgust. It is possible that a manipulation able to elicit those specific categorial emotions could have led to stronger results. Indeed, fewer advantageous choices have been found in previous studies when Good decks were associated with fear-relevant stimuli (i.e., pictures of spiders and angry faces, respectively) in subjects with spider phobia or social anxiety [
50,
51,
52]. In those studies, however, the stimuli used to manipulate the decks were emotionally relevant and meaningful for the participants. In our study instead, the affective cue was designed to elicit a general negative affective reaction of unpleasantness, not emotion-relevant or meaningful for participants. Future studies should investigate possible effects of affective cues able to generate categorial negative emotions such as fear or disgust while keeping the stimuli not meaningful to the participants.
Additionally, we did not test for the effect of pleasant affective cues. Nevertheless, it could be interesting to investigate if such cues could lead to similar results, but in the opposite direction, i.e., increased risky choices when they are affectively pleasant.
Moreover, our negative manipulation represented a deterministic aversive outcome associated with the action of choosing a certain deck, but not with its monetary outcome. In these terms, the manipulation was irrelevant to the overall goal of the task (i.e., maximization of the initial monetary loan). However, it is possible that participants perceived the unpleasant sound as a punishment for their choice. This eventuality, in a reinforcement learning perspective [
38,
49], could have made the negative manipulation relevant to the task. Future studies should investigate this possibility.
In this study, we focused mainly on anticipatory physiological response according to the SMH, using SCR and HRV. However, some studies suggested that post-selection SCR (i.e., physiological activation in response to the choice’s outcome) might be more influential than anticipatory SCR on IGT performance [
37,
38,
49,
53,
54,
55]. Thus, the effect of post-selection arousal should be taken into consideration in future studies. Moreover, greater levels of HRV at rest have been related to better emotional regulation [
32,
56,
57,
58]. Future studies should investigate if the effect of the manipulation could be mediated by emotion regulation abilities. It could be that people who better regulate their emotional responses can be less influenced by the manipulation. Alongside resting HRV, emotional regulation abilities could also be investigated as a personality trait.
Finally, other types of physiological indexes such as pupil dilatation, respiration, or brain activity should be included in future studies to reach a clearer understanding of autonomic activation in response to anticipatory emotional reactions.