Psychological approaches to depression emphasize the importance of cognitive dysfunction in the disorder, and the Diagnostic and Statistical Manual of Mental Disorders-V (DSM-V; American Psychiatric Association, 2013) considers decision-making impairment a possible symptom for major depressive disorder (MDD). MDD is associated with aberrant reward and punishment learning (e.g., Depue and Iacono, 1989; Must et al., 2006; Steele et al., 2007; Cella et al., 2010; Eshel and Roiser, 2010), and altered sensitivity to reward and punishment may contribute to the maintenance of depressive symptoms (Martin-Soelch, 2009). For example, depressed patients have an enhanced sensitivity to negative feedback and are highly influenced by punishments and losses (Elliott et al., 1996). Punishment sensitivity can be observed during risk-based decision-making tasks, where the individual has to select between multiple options associated with uncertain consequences: when people engage in risk decision making, they pursue some form of reward while exposing themselves to potential punishment (Wallach et al., 1962).

The Balloon Analogue Risk Task (BART; Lejuez et al., 2002) requires repeated decisions for financial gain, and is sensitive to real-world behaviors (Bishara et al., 2009). However, in contrast to tasks such as Iowa Gambling Task (IGT; Bechara et al., 1994) that require participants to learn the reward and punishment contingencies relating to their choic of cards, the BART offers a direct measure of risk-taking. The BART requires participants to make successive pumps within each trial to inflate a balloon. The participant can stop pumping at any time and will win an amount of money depending up on how many pumps of air were put in the balloon: the more pumps you made, the more you could win. However, putting more pumps of air into the balloon increases the risk that the balloon will burst and you will be punished by losing that amount of money. Increased pumping on the BART was significantly related to alcohol and drug use, cigarette smoking, gambling, theft, aggression, and unprotected sexual intercourse in both adolescent (Lejuez et al., 2003; Aklin et al., 2005) and adult samples (Lejuez et al., 2002, 2004). It was also associated with psychopathy and impulsivity among young adults (Hunt et al., 2005). The BART is a reliable (White et al., 2008) simple laboratory task that captures the defining characteristic of risk taking in the real world—when participants pump the balloon, they pursue the reward of monetary gain while exposing themselves to potential punishment with the loss of money.

Of specific interest for the present study, Pleskac et al. (2008) modified the BART wherein the instructions given to participants were altered to inform them explicitly about the best decision strategy to use during the course of the BART (i.e., select 64 pumps) trials to produce the best long-term outcome. However, participants are told that this strategy might not always work and that if 64 is entered, the balloon may still explode before or at that number of pumps. Participants are informed that the balloons may explode at any given pump though they are unaware of the actual explosion point for each balloon. The task was further modified so that instead of making the sequential pumps, the participant now simply entered the number of pumps they wanted to inflate the balloon by and they watched the balloon inflate on screen. If the number the participant enters at the start does not exceed the explosion point, the balloon on the screen inflated by the amount entered and the money they earned was deposited into their “bank account”. However, if the participant typed in a number that exceeds the balloon’s explosion point, the balloon on the screen pops and the participant loses the amount entered.

To date, there has been less research using the automatic BART than the original BART. Research examining the effect of these modifications indicates that they produce unbiased BART statistics (Pleskac et al., 2008). As noted by the developers of the original BART, analysis of only unexploded balloons produced biased results: scores were biased toward low scores, because the more times respondents choose a risky option, the more likely it is that a trial will end in failure. Consequently the authors proposed an adjusted score, which filters out longer response sequences and biases scores toward a lower number of pumps (Pleskac et al., 2008). Furthermore, whereas participants were largely risk averse on the original BART and routinely stop pumping much earlier than is optimal (Lejuez et al., 2002; Wallsten et al., 2005), the modifications have been associated increased risky pump selections by participants (Pleskac et al., 2008). The automatic BART retains similar relationships to substance use indices as the standard version of the task. For example, the automatic BART was associated with increased familial risk of alcohol use (Gorka et al., 2015), and drinking quantity, self-efficacy to control drinking and drinking acceptability (DeMartini et al., 2014). It has also been associated with levels of daily smoking (Larsen et al., 2014), risky driving behaviors and attitudes among adolescents (Vaca et al., 2013).

We compared the performance of participants with MDD with that of a matched comparison group on the automatic BART. We used the automatic BART instead of the original BART as we were primarily interested in how people responded to a loss in the context of being explicitly told the optimal long-term strategy. We were interested to see if participants actually used the optimal strategy consistently (i.e., always enter 64 pumps) or if they used a more risky (i.e., enter higher than 64 pumps) or conservative strategy (i.e., enter lower than 64 pumps) and how their responses changed after reward or punishment. As using the optimal strategy on the BART will occasionally result in losses, we were particularly interested in how the MDD group responded to such losses. In the automatic BART the participants could not avoid losses as there was no way to predict the outcome on individual trials. Based on an assumption of depressed patients’ increased sensitivity to negative feedback, we hypothesized that after a loss, the MDD patients will decrease their next selection significantly more than the comparison group will do after a loss. We did not predict any difference between the MDD and comparison group after a gain.

The MDD group showed an increased sensitivity to punishment on the BART: after a trial where they were punished by losing money, the next number of balloon pumps they entered was much lower than their previous number. Whilst such a strategy may reduce the risk of subsequent punishment, it also entails a loss of potential reward. Even when presented with the optimal value to enter, the MDD group on average was significantly below this value. The negative correlation between the BDI and the BART target score indicates that those participants with more depression symptoms entered lower pump values on the BART, which may reflect a strategy to avoid the possible punishment associated with higher values. Of note, a study examining the effects of alcohol on performance on the automated BART revealed that in comparison to those drinking placebo, the alcohol group did not approach the optimal level of 64 pumps, but kept on using a less effective strategy of entering lower pumps per balloon (Euser et al., 2011)—this mirrors the pattern observed in the present study by the MDD group.

The pattern of pump value entries by the MDD group is consistent with a punishment avoidant strategy. Consequently, this impairs their ability to focus on the rewards and to develop adaptive rewarding strategies. It has been noted that positive events such as winning money may fail to adequately reinforce behavior in people with MDD (Henriques and Davidson, 2000), as depressed patients fail to perceive rewards as reinforcing due to low hedonic capacity (Meehl, 1975). MDD has been characterized by significant changes in both motivational and affective processing (Leppänen, 2006). Dominated by persistent dysphoric emotions and thoughts (anhedonia) such patients can exhibit a decreased motivation to seek and a reduced ability to experience reward (Drevets, 2001). Those with depression may be sensitive to the saliency of punishing stimuli, and they may act to minimize exposure to such punishing outcomes.

Although avoiding punishments may be adaptive in some circumstances, the continual use of a strategy in a rigid or context-insensitive manner is likely to reduce the probability of being exposed to rewarding environments, which, in turn, may exacerbate depressive symptoms (Smoski et al., 2008). Thus, avoiding risk often will lead to missed opportunities for rewards. Engaging in an increasing number of rewarding experiences has been shown to be beneficial for improving depression; furthermore, behavioral activation as part of cognitive-behavioral therapy requires people to increase behaviors that lead to a sense of mastery and self-efficacy, as well as those that result in pleasurable consequences (Jacobson et al., 1996). However, behavioral activation involves the potential for reward as well as punishment. For example, going out and getting involved in social activities carries the possibility of both positive (e.g., making friends, strengthening relationships, engaging in enjoyable activities) and negative consequences (e.g., rejection, social disapproval, criticism). Depressed individuals may be particularly attuned to and avoidant of the negative consequences of these activities, and consequently, may choose not to take the risk. Thus, if generalized, this punishment avoidance strategy may be an important factor in the maintenance of depression (Chapman et al., 2007). Punishment sensitivity has been associated with depression levels (Eshel and Roiser, 2010); it has been argued that avoiding decisions that have potential negative consequences reflects an underlying process impacting on depression and psychological flexibility (Leahy et al., 2012). Additional research examining such processes is warranted.

The neurobiological basis of punishment sensitivity remains to be determined (Kim et al., 2015): punishment-based learning is associated with greater activation in the insula and lateral orbitofrontal cortex (Wächter et al., 2009; Elliott et al., 2010; Xue et al., 2013). Furthermore, dopamine plays a role in the formation of behavioral learning strategies aimed at avoiding aversive stimuli (Ilango et al., 2012).

The MDD group self-selected to participate in the study, were predominantly inpatients, and were on antidepressant medications, which can affect sensitivity to reward and punishment; antidepressants may enhance sensitivity to negative outcomes so that they can be perceived and avoided (Dalgleish et al., 2004). Consequently the effect of the medication on performance is unclear as we did not record the dosage at the time of the study. The strength of drug may impact on performance and future research should examine the nature of this relationship. In addition, the study did not use self-report psychometric instruments to assess loss aversion, impulsive personality traits and risk tendencies in the two groups; this limits the capacity to rule out the possible confounding effects of these traits on BART performance. Given the relationship between impulsivity and depression among clinical populations (Jakubczyk et al., 2012; Dervic et al., 2014) future research should include such measures.

In line with previous decision making studies (e.g., Whitlow et al., 2004), a performance-related monetary reward was offered to enhance participant motivation to perform successfully; however, this may not have proved sufficient to engage the participants, as depressed patients are characterized by lack of appropriate responses to rewards (e.g., Layne, 1980). Although the comparison group all scored in the normal range on the BDI-II and did not self-report any history of psychiatric and medical conditions that could impact on performance, a psychiatric diagnostic interview (e.g., SCID-V) was not used to confirm their mental health status. In addition, the comparison group were a self-selected group of community dwelling adults who were interested in taking part in psychological research.

Individuals with MDD engaged in punishment-sensitive decision-making on the automatic BART: although this strategy can minimize potential negative outcomes it comes at the cost of minimizing potential rewarding outcomes, which may contribute to the maintenance of depression.

DH wrote the paper, analyzed the data, and contributed to the design of the study. KT, KL, and RB contributed to the paper and the design of the study. SL-S contributed to the paper, collected the data, and the design of the study.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.