Consistent with previous studies, all participants learned to avoid the risky decks over the course of the task, with significant declines in the number of risky cards selected over time. Depressive participants chose fewer risky cards across the entire task compared to control participants and showed a trend toward winning more money overall. This result only partially supports our hypotheses, in that the depressive group showed better performance across the entire task on average, rather than showing a faster rate of learning than controls (as would be indicated by an interaction of group and block). Consistent with the study hypotheses, by the end of the task, depressive participants were more “successful” than controls at making decisions that avoided negative feedback (i.e., higher risk), winning an average of $140 compared to the average loss of $375 by controls in the final task block.
It is somewhat difficult to interpret a finding of better performance overall by the depressive group, as opposed to an interaction between trial block and group. An interaction would strongly suggest a learning effect, with both groups starting at the same level of performance but diverging by the end of the task. A main effect suggests that the depressive participants were less likely to choose risky decks throughout the task, but does not support the hypothesis of faster learning. We therefore interpret our finding as indicating that depressive subjects are more risk averse than controls, but that this effect is not necessarily manifested in faster learning of contingencies.
The structure of the IGT task places greater value on negative feedback than positive feedback, requiring participants to forego high-reward cards for less risky but less rewarding choices. The finding of differential IGT performance in depressive participants suggests that depressed individuals may have a heightened sensitivity to aversive contingencies. This is consistent with a bias towards negative stimuli or self-evaluation in depression, as well as the construct of harm avoidance. Other correlates of harm avoidance such as anxiety (Schmitt, Brinkley, & Newman, 1999
) and the personality trait of neuroticism (Carter & Smith Pasqualini, 2004
) also have been linked to enhanced IGT performance, though clinically significant anxiety can be associated with impaired performance (e.g., OCD; Cavedini et al., 2002
). It may be that depressed individuals behave as if negative or punishing consequences are more likely to occur than positive or rewarding consequences. For example, Elliott and colleagues found that depressed individuals did not change their behavior under conditions of absent versus negative feedback. In contrast, non-depressed controls did not demonstrate different behavior under conditions of absent versus positive
feedback (Elliott, Sahakian, Michael, Paykel, & Dolan, 1998
). They conclude that in the absence of feedback, “normal subjects behave as if they are expecting success while depressed patients behave as if they are expecting failure.” Our results are consistent with this conclusion.
Results from the present study also are consistent with Damasio's (1994)
somatic marker hypothesis, which attempts to explain how individuals make decisions between alternative courses of action characterized by varying potentials for positive and negative feedback. According to this hypothesis, a process of trial-and-error learning leads individuals to develop non-conscious learned associations between emotional states (called “somatic markers”) and particular courses of action. Over time, these emotional states (both negative and positive) begin to function as guides in effective decision-making. Essentially, when individuals are faced with different options, negative somatic markers will guide them away from unfavorable options, and positive somatic markers will guide them toward favorable options.
In support of the somatic marker hypothesis, several studies have found that individuals with frontal lobe damage (particularly, ventromedial prefrontal cortex damage), who do not experience anticipatory physiological reactions when facing decisions, also perform poorly on the IGT (Anderson, Bechara, Damasio, Tranel, & Damasio, 1999
; Bechara, Damasio, & Damasio, 2000
). Likewise, a lack of anticipatory response to potentially punishing situations has been linked to sociopathy; however, there has been mixed support for the presence of impaired IGT performance in individuals with sociopathic traits (Blair & Cipolotti, 2000
; Blair, Colledge, Murray, & Mitchell, 2001
; Lösel & Schmucker, 2004
; Mitchell, Colledge, Leonard, & Blair, 2002
). Enhanced performance by depressive participants on the IGT may reflect either stronger or more rapid acquisition of somatic markers in response to negative feedback, potentially as a result of a processing bias toward negative information in general. Likewise, depressed individuals may have weaker positive somatic markers, and as a result, demonstrate less appetitive, reward-seeking behaviors. These empirical questions should be addressed in future studies examining decision making and risk taking in depression.
The findings in this study are consistent with depressed individuals' tendency toward enhanced attention and responsivity to negative feedback, relative to positive feedback. It is rare to find a neuropsychological task in which psychopathology is associated with positive
performance. Indeed, our findings speak to the potentially adaptive nature of adopting a low-risk low-reward strategy when faced with a clearly differentiable high-risk high-reward alternative. Similar to the concept of depressive realism (e.g., Alloy & Abramson, 1979
), it may be that individuals with high depressive symptoms are more accurate at discriminating punishment from reward, and, accordingly, are differentially protected from punishment in the environment. However, when chronically used in a rigid or context insensitive manner, such a strategy likely reduces the probability of being exposed to rewarding environments, which, in turn, may exacerbate depressive symptoms. Such strategies may keep you from losing, but won't necessarily help you win.
It is important to note that it is uncommon in the natural environment to find punishment and reward as closely tied (hence the risk) as in the IGT. Thus, in the natural environment, unlike the IGT, avoiding risk often will lead to missed opportunities for rewards. There is value in the depressive style of minimizing losses; however, the flexibility to maximize rewards as well may be one of the keys to emotional health and overcoming depression. Along these lines, engaging in an increasing number of rewarding experiences has been shown to be sufficient in improving depression (Hopko, Lejuez, LePage, Hopko, & McNeil, 2003
; Jacobson et al., 1996
; Lejuez, Hopko, LePage, Hopko, & McNeil, 2001
). Thus, if generalized, this risk-averse strategy may be an important factor in the maintenance of depressive experiences (Chapman et al., 2007
), and successful treatments for unipolar depression may operate, at one level of analysis, by blocking avoidance of risk for punishment and providing opportunities for new learning to occur.
Several limitations should be considered when interpreting the findings of this study. First, depressive symptoms were assessed by a screening interview (HAM-D) as opposed to formal diagnostic testing. Although the modified HAM-D interview used in the study has strong validity and reliability (Miller et al, 1985
), formal diagnostic testing would help provide a more precise picture of the effects of depression on learning from feedback. Based on established cutoffs on the 17-item HAM-D (Miller et al., 1985
), our sample may be best characterized as mildly to moderately symptomatic. In addition, future studies should include an assessment of emotional state and task motivation on the day of testing to examine these factors as potential covariates of IGT performance. Another challenge is that our data are cross-sectional data cannot be used to discern whether these effects are related to state or trait characteristics. However, previous findings of no differences between individuals with a history of affective disorders and a lifetime-free control group (Jollant et al., 2005
) would suggest that our findings may be driven primarily by state effects. As mentioned earlier in this section, the lack of an interaction effect makes interpretation of the current results less clear. An additional concern is that the design of the present study makes it difficult to discern whether IGT performance among depressive individuals was superior due to a heightened response to punishment, a decreased response to reward, or both. This study represents a first demonstration of relative risk avoidance/lack of reward pursuit among depressive individuals in a well-validated task. Future studies are planned to use modified IGT tasks to measure the behavior of depressed individuals when feedback is neutral (i.e., risk equals reward) or weighted towards the positive (i.e., reward outweighs risk).