Role of the amygdala in decisions under ambiguity and decisions under risk: Evidence from patients with Urbach-Wiethe disease

Abstract

Various neuropsychological studies have shown that decision-making deficits can occur in a wide range of patients with brain damage or dysfunctions. Decisions under ambiguity, as measured with the Iowa Gambling Task, primarily depend on the integrity of the ventromedial prefrontal cortex and the amygdala, as well as on further brain regions such as the somatosensory cortex. However, little is known about the specific role of these structures in decisions under risk measured with tasks that offer explicit rules for gains and losses and winning probabilities, for example, the Game of Dice Task. We aimed to investigate the potential role of the amygdala for decisions under risk. For this purpose, we examined three patients with Urbach-Wiethe disease—a rare syndrome associated with selective bilateral mineralisation of the amygdalae. Neuropsychological performance was assessed with the Iowa Gambling Task (decisions under ambiguity), the Game of Dice Task (decisions under risk), and an extensive neuropsychological test battery focussing on executive functions. Furthermore, previous studies found relationships between generating skin conductance responses and deciding advantageously in the Iowa Gambling Task. Accordingly, we recorded skin conductance responses during both decision tasks as a measure of emotional reactivity. Results indicate that patients with selective amygdala damage have lower scores in both decisions under ambiguity and decisions under risk. Decisions under risk are especially compromised in patients who also demonstrate deficits in executive functioning. In both gambling tasks, patients showed reduced skin conductance responses compared to healthy comparison subjects. The results suggest that deciding advantageously under risk conditions involves both the use of feedback from previous trials, as required by decisions under ambiguity, and in addition, executive functions.

Introduction

Deciding between different options is a very important function in everyday life and disturbances of decision-making can result in severe social, financial and health problems. In real life, decision situations differ in their degree and probability of associated reward and punishment. In decisions under ambiguity individuals have to decide between different options, but the outcome of their choices is uncertain and not defined by clear/obvious probabilities. These types of decisions – at least in patients with brain lesions – primarily depend on the integrity of the ventromedial prefrontal cortex and the amygdala as well as on structures involved in limbic circuits (e.g., Bechara et al., 2000a, Bechara et al., 2003; Bechara & Van Der Linden, 2005). According to the somatic marker hypothesis (e.g., Damasio, 1996), subjects have to follow their own feelings and hunches for optimal decision-making (Bechara, 2001). Although discussed controversially (see Maia and McClelland, 2004, Maia and McClelland, 2005), in the process of decision-making, the amygdala is suggested to be a critical structure as being involved in processing primary inducers and emotional arousal associated with anticipation of rewards and punishments (e.g., Kringelbach, 2005, Phelps, 2006; Phelps & LeDoux, 2005). Decisions under ambiguity are often tested using the Iowa Gambling Task (abbreviated with “gambling task” in the following) (Bechara, Damasio, Damasio, & Anderson, 1994; Bechara, Tranel, & Damasio, 2000). In this task, subjects are asked to select a card from one of four card decks. After card selection, a fictitious gain is displayed, irregularly accompanied by a loss. Two of the four decks are coupled with high gains but even higher losses. Choosing from these disadvantageous decks will lead to a negative net-balance in the long run. The other two decks will result in small gains but even smaller losses (advantageous decks). Thus, choosing from these advantageous decks will lead to a positive overall balance. Although participants are told that some card decks are “better” than others, they are not explicitly informed about the reward/punishment contingencies. Thus, the possible choices are full of ambiguity and participants have to learn to avoid the disadvantageous card decks using the feedback from previous trials.

Using this task, decision-making deficits have been revealed in a wide range of neurological patients suffering from frontal lobe damage (especially when lesion foci centred around the orbitofrontal/ventromedial section of the prefrontal cortex; e.g., Bechara et al., 1994, Bechara et al., 2000b). Furthermore, psychological patient populations were found to be deficient relative to healthy individuals in gambling task performance, such as patients suffering from substance dependencies (e.g., Bechara, 2005; Bechara & Damasio, 2002; Bechara et al., 2001; Bechara, Dolan, & Hindes, 2002; Bolla et al., 2003; Bolla, Eldreth, Matochik, & Cadet, 2005; Fishbein et al., 2005; Verdejo, Aguilar de Arcos, & Perez Garcia, 2004; Whitlow et al., 2004), or patients with obsessive compulsive disorder, schizophrenia, pathological gambling, anorexia nervosa, suicide attempters, and other patients with neuropsychiatric symptoms (e.g., Bark, Dieckmann, Bogerts, & Northoff, 2005; Cavedini et al., 2004; Cavedini, Riboldi, D’Annucci et al., 2002; Cavedini, Riboldi, Keller, D’Annucci, & Bellodi, 2002; Goudriaan, Oosterlaan, de Beurs, & van den Brink, 2005; Jollant et al., 2005) (an excellent review of studies with the gambling task is found in the article by Dunn, Dalgleish, & Lawrence, 2006).

Knowledge about the role of the amygdala in decisions under ambiguity comes from studies which examined patients with selective damage of the amygdalae with the gambling task. For instance, Bechara, Damasio, Damasio, and Lee (1999) studied gambling task performance of five patients suffering from Urbach-Wiethe disease (UWD) (n = 1), encephalitis during childhood (n = 2) or herpes simplex encephalitis during adulthood (n = 2), all having lesions comprising bilateral medial temporal lobes and being more or less restricted to the amygdalae. The authors compared gambling task performance of these patients with performance of healthy subjects as well as with that of patients with selective damage to the ventromedial section of the prefrontal cortex. The behavioural data of both patient groups appeared very similar indicating poor gambling task performance compared to the healthy participants. The authors also recorded skin conductance responses (SCRs), as a measure of emotional reactivity, while the subjects performed the task. Both, amygdala-damaged patients and patients with ventromedial lesions had significantly lower anticipatory SCRs (that is SCR during the time period before a choice) than the normal comparison subjects. SCRs following each choice differed between patient groups: Amygdala-damaged patients generated significantly lower SCRs after both types of choices, selections of decks that led to reward as well as after selections of cards that were followed by punishments, whereas ventromedial PFC patients showed similar responses compared to healthy individuals. Four of five patients with ventromedial lesions generated SCRs within the range of the healthy comparison subjects’ scores. The authors concluded that the lack of SCR generation after choice in amygdala-damaged patients may reflect that gaining or losing money in the gambling task does not evoke a somatic state in individuals with amygdala damage. This, as the authors stated, is in accordance with real life decision-making deficits of those patients (as described in Tranel & Hyman, 1990). SCRs are seen as reflecting general arousal associated with processing emotional stimuli (Boucsein, 1992; Venables & Christie, 1980). Recording SCRs during gambling task performance was done by a few studies and most of them replicated the finding of the studies by Bechara and colleagues (e.g., Bechara & Damasio, 2002; Bechara et al., 2002; Bechara, Tranel, Damasio, & Damasio, 1996) who revealed higher anticipatory SCRs before choosing a disadvantageous option relative to those before choosing an advantageous card deck in healthy subjects (e.g., Crone, Somsen, Van Beek, & Van Der Molen, 2004; Hinson, Jameson, & Whitney, 2002). The results were interpreted as the higher anticipatory SCRs preceding disadvantageous decisions act as warning signals from the periphery that gradually lead to avoiding the disadvantageous alternatives and preferring the advantageous options. Other studies found that generating SCRs after the feedback is delivered (potentially indicating appraisal of the decision) may be more important to perform well in the gambling task than the anticipatory SCRs. For example, Suzuki, Hirota, Takasawa, and Shigemasu (2003) found that feedback SCRs were higher after choosing a disadvantageous option than an advantageous and were higher for punishments than for rewards. However, it is still a topic of debate whether or not SCRs, as measures of emotional arousal, guide future decisions and whether anticipatory and/or feedback SCRs are primarily important for successful gambling task performance (see the critical review by Dunn et al., 2006).

In contrast to the decisions measured in the gambling task, many decisions in real life can be made on the basis of probabilities and explicit knowledge about options and their associated rewards and punishments (e.g., the amount of gains and losses). Such decisions are termed “decisions under risk” (e.g., Bechara, 2004; Brand, Labudda, & Markowitsch, 2006). We have recently developed a gambling task, the Game of Dice Task (abbreviated with “dice task” in the following) that is believed to simulate those decisions under risk. Here, subjects are asked to predict the outcome of a dice throw (see detailed description of the task in Section 2). This task offers explicit rules for gains and losses and obvious winning probabilities. Thus, participants are allowed the possibility of utilising explicit clues to optimise their performance (e.g., by calculating the risk associated with each option). In a series of studies with patients suffering from neurological diseases (e.g., Morbus Parkinson, Korsakoff's syndrome) or psychological disorders (e.g., pathological gambling), we found that performance in decisions under risk was associated with both specific executive functions (e.g., categorisation, set-shifting) and processing feedback from previous trials within the task (Brand, Fujiwara et al., 2005; Brand, Kalbe et al., 2005; Brand, Labudda et al., 2004). In accordance with these behavioural studies and results obtained from neuroimaging investigations (recent review in Krain, Wilson, Arbuckle, Castellanos, & Milham, 2006), we suggested that decisions under risk, as measured by the dice task, rely on intact fronto-striatal loops connecting mesencephalic dopaminergic structures (substantia nigra, ventral tegmental area) with both the dorsal striatum and the dorsolateral section of the prefrontal cortex (the so-called cognitive loop; Alexander & Crutcher, 1990; Alexander, Crutcher, & DeLong, 1990) as well as limbic structures such as the amygdala, the nucleus accumbens as well as the ventromedial and the orbitofrontal part of the prefrontal cortex (the so-called limbic loop according to Alexander & Crutcher, 1990; Alexander et al., 1990). The relevance of the dorsolateral prefrontal cortex for dice task performance is indirectly confirmed by a case study of a female patient who suffered from severe decision-making deficits in real life following a foramen of Monro cyst removal (Brand, Kalbe et al., 2004). She was severely impaired relative to control subjects in the dice task and executive tests while other cognitive functions were intact. A positron emission tomography (PET) examination revealed hypometabolic zones within the dorsolateral prefrontal cortex (bilateral) as well as the fusiform and the cingulate gyri. We proposed that decisions under risk can be made by two different but interacting ways (Brand et al., 2006). A cognitive way to select an option relies on explicit knowledge about probabilities and consequences as well as knowledge about the risk of an option, which is the combination of the probability and amount of gain/loss. In addition to this cognitive way, we suggest that decisions under risk – as other kinds of decisions, such as decisions under ambiguity – can also be made on the basis of feedback from previous decisions. (The role of feedback use in decisions under ambiguity according to the somatic marker hypothesis is described in Bechara, 2001, Bechara, 2005; Bechara, Damasio, Tranel, & Damasio, 1997; Bechara, Damasio, Tranel, & Damasio, 2005; but see also the statements on SCRs measures and decision-making above.) In decisions under risk, feedback from previous trials may as well guide the decision-making process: If someone chooses a risky option and receives punishment, additional to his or her prior knowledge about probable negative consequences, the feedback can be used to reconsider the current strategy or to more explicitly observe the rules. Under risk conditions, advantageous performance, i.e. decisions with maximal rewards, is believed to rely on both logical strategies based on explicit knowledge and the use of feedbacks from previous decisions. Therefore, structures of both fronto-striatal loops, the cognitive and the limbic loop (see description above), are thought to be involved in decision-making under risk conditions.

So far, the potential role of the amygdala for decisions under risk is not clear. To our knowledge, this is the first study that aims to reveal differential contributions of the amygdala to decisions under ambiguity and decisions under risk in patients with selective amygdala damage and recording SCRs during performing both tasks. As mentioned above, gambling task performance is reported to co-vary with SCRs as a marker of emotional reactivity. As the amygdala is crucial for processing emotional stimuli and initiating arousal changes (Phelps, 2006; Phelps & LeDoux, 2005), we hypothesise – in accordance with findings of Bechara et al. (1999) – that in our amygdala-damaged patients generating SCRs is reduced during gambling task performance. While SCRs have not previously been measured during dice task performance, behavioural studies revealed that processing feedbacks from previous trials contribute importantly to successful dice task performance. As feedback processing seems to rely on amygdaloid activation and corresponding generation of SCRs (Bechara et al., 1999), amygdala-damaged patients are hypothesised to demonstrate lower reactions on emotional feedback and reduced SCRs during dice task performance.

The healthy subjects should generate higher SCRs preceding disadvantageous relative to advantageous decisions in the gambling task, as revealed in most of the previous studies which recorded SCRs. In the dice task, healthy subjects are hypothesised to primarily show higher SCRs for risky relative to the non-risky decisions in the feedback phase, when gains or losses are displayed. Anticipatory SCRs should not differ between risky and non-risky choices, because we believe that the upcoming decision is based on reflecting about the probabilities and the amount of rewards/punishments more than on anticipatory reactions that “alert” the individual for the potential punishment.

On the behavioural level, we hypothesise that in the patients with UWD, performance in the dice task is less severely affected than in the gambling task because in the dice task it is possible to decide on the basis of cognitive strategies from the very beginning of the task, as mentioned above. More precise, we hypothesise that in UWD patients one of the suggested ways to solve the dice task is affected (feedback use and generating SCRs as markers for emotional processing of the feedbacks) while the other way (evaluating the options on the basis of their probabilities and amounts of gains and losses) is intact, resulting in an overall moderately reduced performance compared with healthy subjects, who should use both suggested ways to solve the task.