This item is only available as the following downloads:
EFFECT OF AROUSAL ON DECISION MAKING IN THE IOWA GAMBLING TASK By Erica Herzig A Thesis Submitted to the Division of Social Sciences New College of Florida In partial fulfillment of the requirements for a degree of Bachelor of Arts in Biological Psychology Under the sponsorship of Dr. Gordon Bauer Sarasota, Florida May 2013
iii Acknowledgements I would like to sincerely thank my family, friends and mentors, who have supported and evaluated me decisions over the past four years. Such support a nd redresses have resulted in an invaluable collegiate experience. Profound gratitude must be given to all of the professors and staff at New College. A special thanks goes to Dr. Bauer for ensuring that my time at New College would result in a gain nto the wondrous field of cognition and for her dedication to his students and to the field of neurobiology. Without him this research would not have been possible. Finally, thank you Mom, Dad, Nikki and Nathan for your love and support. To all of those who fuel my motivation ; Jordan, Morgan, Elliot and David.
1 Table of Contents Acknowledgements i i i Ta ble of Contents 1 L ist of Abbreviations 2 Abstract 3 Rational Emotions 5 The Iowa Gambling Task 7 What Determines Performance in the IGT? 9 Use of the IGT a s a Neuropsychological Tool 1 4 Validity & Relia bility of the IGT 1 7 Dissociable Processes I nvolved in Performance of the IGT 31 Neural Correlates of the Iowa Gambling Task 34 Skin Conductance Res ponses 4 6 The Present Study 50 Method 52 Results 5 5 Discussion 5 6 References 6 2 Figures & Tables 6 8
2 List of Abbreviations ( S): Microsiemen ACC: Anterior Cingulate Cortex aSCR: Anticipatory Skin Conductance Responses BN: B ulimia N ervosa dlPFC: Dorsolateral Prefrontal Cortex fMRI: Functional Magnetic Resonance Imaging GDT: Game of Dice Task ICD: Impulse Control Disorder IGT: Iowa Gambling Task OCD: O bsessive C ompulsive D isorder PAG: P eriaqueductal G ray PD: RSD : Reobsorptive Sweat D uct SCR: Skin Conductance Response SMA: Somatosensory Area SMH: Somatic Marker Hypothesis SN: Substantia Nigra VMPFC: Ventromedial prefrontal Cortex WCST: Wisconsin Card Sorting Task
3 EFFECT OF AROUSAL ON DECISION MAKING IN THE IOWA GAMBLING TASK Erica Herzig New College of Florida, 2013 ABSTRACT The Iowa Gambling Task (IGT) is a tool used to measure decision making, yet the cognitive components which determine performance on this task have not been established Two separate cognitive components may work together. O ne constituent may detect net est imated value, and the other fr equency of gains and losses. Loading estimated value while leaving the ability to determine frequency of losses intact Pr evious research has found that emotional arousal aids in performance on the IGT The goal of the current study was to verify the existence of two distinct components and to discover which componen t reflects emotional arousal Two groups of participants performed the IGT while skin conductance was recorded as a measure of emotional arousal. One group performed the IGT while executive functions were blocked and another group performed the task while listening to white noise as a control Results showed choices were determ ined by frequency of losses regardless of condition, and low frequency losses were preferred over high net gain. Controls were more emotionally aroused than those with blocked executive functions, indicating
4 executive functions were necessary to generate emotional responses. Skin conductance for controls was highest prior to high net estimated value choices. Low initial gains associated with high net estimated value choices may have appeared aversive to participants leading to high er emotional arousal pri or to making these choices Gordon Bauer Division of Social Sciences
5 Rational Emotions Elliot When observing small children and college students it is blatantly apparent when they make decisions without c onsidering the long term consequences. In his book (1994), neuropsychologist Antonio Damasio described his patient, Elliot, dedicated family Elliot began suffering from severe headaches and was diagnosed with a tumor in the region of his brain behind the eyes. The surgery to remove the tumor was collecting odd objects and fill ed his house w ith clutter. His work ethic began to falter and he had difficulty completing the many tasks his job demanded of him. Elliot sat for hours persevering over inane tasks, like choosing a proper pen, and eventually lost his job Without the distraction of wor k Elliot began making poor investments with his own finances which drove him to bankruptcy repeatedly took sex workers as mistresses. At this point Antonio Damasio entered lost character. Computerized tomography (CT) and magnetic resonance imaging (MRI) studies greater damage on the right than the left. The external surface of the left frontal lobe was intact, and damage was within the orbital and medial sectors. Both the left and right motor and premotor regions were intact, allowing Elliot full functional movement.
6 The basal forebrain, a region highly involved in learning and memory, was also intact. Though one might argue that much of the brain was intact, even phrenologists in the understood that extent of damage is a factor of location rather than size of the affected area. Elliot was issued a battery of neuropsycholog ical tests, with results indicating an above average Intelligence Quotient and normal performance on all subsets of the Wechsler Adult Intelligence Scale. Damasio reported, memory, short term memory, new learning, languag e and ability to do arithmetic were mind. Damasio began testing Elliot on tasks requiring frontal lobe process ing including the Wisconsin C ard Sorting Task. This task requires that subjects alter their categorization schemes to follow new rules set by the researcher. Previous neuropsychologists had shown that patients with frontal lobe damage were unable to perform this task (Milner, 1963). H owever Elliot completed the task without difficulty. Elliot performed the Minnesota Multiphasic Personality Inventory, and generated a valid, genuine, normal profile. To address the question of whether Elliot still possessed knowledge of appropriate social behavior, Damasio presented him with situations pertaining to social and financial dilemmas to which Elliot was asked to give a reasonable solution. to apply social conventions to problem s, and that stealing was wrong. However, this knowledge did not reflect his life choices. One test required Eliot to view emotionally
7 charged images and report his emotions. Elliot reported not feeling as he was ed, but could not generate the appropriate emotional response. In response t 206). perform these cognitive tasks to their simplicity. While performing the tasks, Elliot was not forced to weigh multiple aspects of decisions and alter his choices as new information was presented, as was the case in his real life. open ended, uncertain evolution of real was lacking which e experiences with emotions and to recall these directed emotions at a later time. At the time, Damasio did not have a task to measure emotional aspects of cognition. The Iowa Gambling Task The research performed by Bechara, Damasio, Damasio, & Anderson, (1994) was preceded by multiple observations of patients like Elliot, who suffered from ventromedial prefrontal cortex (VMPFC) damage. Patients with VMPFC damage were reported to make bad judgment calls in real life, but passed all batteries of neuropsychological tests, including those which test memory, perseverance, self ordering, and cognitive estimations. Bechara et al. (1994) devised a task which they
8 believed simulated real life choices through its incorporation of uncertainty of outcomes Participants included 44 healthy control subjects and 6 patients with damage specific to the VMPFC region; both groups consisted of men and women. Researchers initially present ed participants with $2000 of play money and four decks of cards from which they could choose. Trials consisted of choosing a card from any deck and participants completed five blocks of twenty trials each. Participants were told that their goal was to gai n as much money as possible by choosing from the four decks. Unbeknownst to the participants, two decks (A and B) were always associated with an initial payout of $100, while the other two decks (C and D) were associated with $50 of initial payout. However the future yield of each deck varied due to lo sses which occasionally occurred along with the initial gain. The net gain every ten draws from decks A and B was $ 250, whereas the net gain from the low initial payout decks (C and D) was $250 per 10 cards. Thus, the participant would benefit from choosing more often from the low initial payout decks (C and D) (Table 1) Performance on the task was determined by taking the number of choices from the advantageous decks (C and D) and subtracting the number of choices from the disadvantageous decks (A and B), i.e, (C+D) (A+B). Results showed that initially, both groups chose from the high initial payout decks (A and B) significantly more often than the low initial payout decks (C and D). Normal controls switche d their preference to decks C and D by block two, whereas
9 VMPFC patients continued to choose from decks A and B significantly more than decks C and D throughout the entire game. The current thesis will aim to determine how normal healthy participants make decisions while they perform this task. First, a brief overview of the early Somatic Marker Hypothesis will be presented which proposes arousal dictates their responses in performance of the task. Then, issues with this early hypothesis will be addressed, and will be followed by a recent hypothesis regarding components. Finally, a proposal to establish what factors influence performance of th e Iowa Gambling Task will be proposed and executed. What Determines Performance on the Iowa Gambling Task? The Somatic Marker Hypothesis. Researchers set out to uncover why patients with ventromedial prefrontal cortex damage could not learn to choose from the more advantageous decks in the Iowa Gambling Task over time (Bechara, Tranel, Damasio, and Damasio, 1996).The y proposed patients were either incapable of holding representations of future outcomes in their working memory, or unable to emotionally mark this representation of the future with positive or negative values. A final hypothesis combined the first two, stating that somatic (emotional) marking of outcomes help drive and maintain attention and working memory of outcomes and that VMPFC damaged p atients were incapable of using somatic information to predict future outcomes. This final proposal was believed to best reflect behavior observed in VMPFC patients. If somatic state activation was indeed necessary for distinction
1 0 between good and bad choi ces, researchers believed healthy participants performing the Iowa Gambling task should elicit emotionally charged responses when choosing between good and bad decks. Experimenters gathered 7 target subjects with ventromedial prefrontal cortex (VMPC) dama ge, and 12 control subjects with no cognitive deficits. Electrodermal skin conductance responses (SCRs) were measured for both groups of participants while they performed the Iowa Gambling Task. Skin conductance reflects arousal of the autonomic nervous sy stem and is linked to emotional arousal. Part icipants were told to select a card from one of four decks of cards which were laid out before them. Participants were not told to choose another card until skin conductance returned to baseline. Both groups of participants generated SCRs in response to reward and punishment with no significant differences between groups. However, over time, controls began generating SCRs prior to choosing the decks associated with greater losses, whereas the targe t subjects failed to generate these anticipatory responses to the disadvantageous decks. Bechara and his team concluded patients fail to elicit these anticipatory SCRs because 1) a connection was lost between categorization of previous experiences and a bi ological emotional response; or 2) patients had lost the ability to activate or inhibit bioregulatory mechanisms which respond to environmental stimuli. The scientists also proposed somatic signaling non of disadvantageo us decks and helps to bring about conscious reasoning for making this decision.
11 Bechara, Tranel and Damasio (2005) considered three possibilities which could lead to the poor performance on the Iowa Gambling Task observed in VMPFC damaged patients. Resear chers proposed the VMPFC patients could be hypersensitive to reward, insensitive to punishment, or insensitive to future consequences. The last possibility would lead the participant to choose decks according to immediate outcomes, rather than long term re wards or losses. To determine which of these factors led to poor performance of the IGT in VMPFC damaged patients, the researchers designed a variant version of the IGT which reversed the order of reward and punishment such that punishment became immediat e and reward became delayed. In the reversed game, decks were labeled E, F, G and H. Decks E and G had high immediate punishments, but were followed by higher future rewards, whereas the disadvantageous decks (F and H) had low immediate punishments with lo w future rewards (Refer to Appendix A). Control and VMPFC damaged participants played a computerized version of the normal and variant IGT. Skin conductance responses (SCRs) were measured after punishments and rewards five seconds after participants chose a deck. Bechara et al. (2005) reasoned that hypersensitivity to reward would be associated with higher reward SCRs of VMPFC patients when compared to controls, and that insensitivity to punishment would be associated with lower punishment SCRs when compar ed to controls in both versions of the task. Predictions were made to account for each of the possible causes of poor IGT performance observed in VMPFC damaged patients. First, in the variant task, preference
12 for advantageous decks (with high immediate pu nishments) combined with abnormally low punishment SCRs compared to healthy controls would be consistent with insensitivity to punishment as an explanation. Second, preference for advantageous decks in the normal IGT task (with low immediate rewards) combi ned with abnormally high reward SCRs compared to controls would be consistent with hypersensitivity to reward as an explanation. Lastly, preference for disadvantageous decks in the variant version of the IGT (with low immediate punishments) combined with n ormal reward and punishment SCRs would be consistent with insensitivity to future consequences as an explanation. Researchers also predicted that increasing consequences associated with disadvantageous decks would not result in improvement of performance i n patients with VMPFC damage, whereas controls should show marked improvement in the task given these conditions. Results showed that in the normal IGT, healthy controls shifted their preferences from disadvantageous to advantageous decks as they progress ed through the game; however VMPFC patients never shifted their preferences towards advantageous decks. In the variant task, normal controls increased their preference for the good decks, whereas, again, the VMPFC patients failed to shift away from choosin g the bad decks. Thus, VMPFC damaged patients were capable of shifting choices between decks but never acquired the most advantageous preference. In both cases, VMPFC patients made significantly fewer advantageous decisions than controls. Controls and VM PFC patients both generated SCRs after receiving rewards and punishments in both the normal IGT and in the variant task. Bechara et al. (2005) concluded that VMPFC patients
13 are insensitive to future consequences as indicated by their poor performance in bo th not affected by increasing the consequences of disadvantageous decks, and led increasing consequences. Aside from the reversed task, the study performed in Bechara et al. (2005) was strikingly similar to that of Bechara et al. (1996). In both cases VMPFC damaged patients elicited similar reward and punishment SCRs to controls, but unlike controls, the target patients never altered preferences toward advantageous decks. The team of researchers concluded that anticipatory skin conductances reflected an emotional labeling of options as either good or bad while information of rewards an d punishments was acquired. Patients with VMPFC damage were unable to use reward and punishment information to assign emotional labels to the various decks, thus their decisions were based on short term rather than long term consequences. The Iowa Gamblin g Task is currently used to measure decision making in numerous populations (Boeka & Lokken, 2006; Crone & Van Der Molen, 2007; Garcia Molina et al., 2007). The task requires that perceptions toward each deck are re evaluated each time a card is drawn. Thi s reflects decisions in real life which require re assessment of choices as information is acquired. For example, choosing a career requires that people base their choice on every new piece of information acquired pertaining to each option. The IGT also re quires that decks are somatically labeled according to their advantageous or disadvantageous properties. Referring back to the
14 previous example, people assess career options by distributing emotional value to each trait of the multiple careers. Use of the Iowa Gambling Task as a Neuropsychological Tool The Iowa Gambling Task has been used to evaluate decision making in people with eating disorders. Researchers Boeka and Lokken (2006) gathered 20 undergraduate students diagnosed with bulimia nervosa (BN) to perform the Iowa Gambling Task along with 20 healthy controls. IGT scores were obtained by subtracting the number of disadvantageous choices from the number of advantageous c hoices (C+D) (A+B). Results indicated that the females with BN performed significantly worse on the IGT when compared to the control group (Boeka & Lokken, 2006). Performance on the IGT was found to be negatively correlated with bulimic symptomology. This effect was present even after controlling for demographic and depressive symptoms. Immediate gratification seeking behavior observed in BN patients was proposed to be a result of an inability or unwillingness to assess future outcomes (Boeka & Lokken, 2006 ). Patients with moderate to severe traumatic brain injury (TBI) were also tested for decision making abilities using the Iowa Gambling Task (Garcia Molina et al. 2007). Forty patients with moderate to severe TBI were compared to 30 healthy controls in p explicit knowledge of the task was examined using a questionnaire. The questionnaire you disadvantageous choices from the number of advantageous choices. Results showed the
15 TBI group performed significantly worse than controls. Among TBI patients, explicit knowledge of the task was positively correlated with performance in the IGT. This was not true for the control group. R odents have been assessed for decision making capabilities using the I GT Lack of animal models for maladaptive decision making, along with the diffe rences in tasks between rodents and humans have made it difficult to determine the pathogenesis of poor decision making. Most animal studies measure decision processes after long periods of training, whereas the IGT examines decision making in a single ses sion and requires complex thought. In an attempt to resolve this issue, a team of researchers designed a single session rat gambling task (RGT) to allow comparisons between rodent and human decision making (Rivalan, Ahmed & Dellu Hagedorn, 2009). Figure 1 displays the human and rat version of the IGT (Rivalan et al., 2009). In this task, rats were trained to nose poke into one of four holes located at one end of the chamber. After two nose pokes, the selected hole remained illuminated, and the rat would co llect two food pellets from a dispenser at the opposite and of the chamber. Rats would receive two pellets simultaneously for nose poking in holes A or B, and would receive one pellet for nose poking into holes C or D. Nose poking in A or B resulted in lon poking in holes C or D. These rats could gain five times as many rewards by choosing from holes C or D than from A or B because timeout lengths for C and D were shorter.
16 F igure 1 Rat Gambling Task designed by Rivalan et al. (2009). In a second part of the experiment, Rivalan et al. (2009) te sted rats for risk taking using the light dark emergence task. In this task, rats were allowed to pass from one compartment exposed to light to another compartment enclosed in the dark. Rats tend to avoid bright areas, which leave them exposed to predators. The amount of time elapsed be tween placement in the dark compartment and emergence into the bright compartment was measured during a 10 minute interval. Rivalan et al. ascertained that the majority of rats (58%) initially selected equally from all options before eventually developing a significant preference towards the advantageous options (A and B). The rest of the rats fell into two groups, one which partially preferred advantageous options, and another which never learned to discriminate advantageous from disadvantageous options. Good and bad decision makers remained in the same categories when tested 6.5 months after the initial test.
17 Rats with poor scores Iowa Gambling Task scores were found to have faster emergence times in the light dark emergence task. Thus, higher levels of r isk taking were found to be correlated with poor decisions in the IGT. As evidenced by the literature, the IGT has been used as a neuropsychological tool to measure decision making in many populations. The task requires that information is updated constant ly, and that old information is re evaluated. The Somatic Marker Hypothesis posits that people rely on anticipatory skin conductance responses to guide them away from the disadvantageous decks. However, many issues exist with determining validity and relia bility of the IGT as a tool to measure decision making. Validity and Reliability of the IGT Two traits of an effective research tool include reliability and validity. Reliability is the ability of a tool to repeatedly display the same results. Construct validity is the degree to which a task measures the construct it aims to measure. The manual f or the Therefore the IGT aims to measure decision making ability and to determine cognitive impairments. Reliability of the IGT is hard to assess due to learning effects observed in most populations (Buelow & Sur, 2009). This means the test may not be applicable for assessing performance over multiple administrations ( e.g., pre post intervention studies). Furthermore, r ecent studies have found varying results in deck preference s from the early Bechara, et al 1996 & 1994 studies (Crone & Van Der Molen, 2007; 2004
18 Fum, Stocco & Napoli, 2009). In these recent studies, decks C and D are not always preferred over the disadvantageous decks A and B. These studies have utilized college aged participants and those under 18years old. The age disparity between these partic ipants and those in early studies (middle aged adults) may explain the differences in choice preference. Certain brain areas including the medial prefrontal cortex are reliably activated and required for performance in the IGT. However, anticipatory skin c onductance responses do not always reflect IGT performance. These issues of validity and reliability of the Iowa Gambling Task must be addressed if a more accurate tool to measure decision making is to be developed. Validity The Iowa Gambling Task was des igned to measure decisions under uncertainty, however recent evidence has shown that uncertainty only dictates decisions early on in the IGT (Brand, Recknor, Grabenhorst, & Bechara, 2007; Upton et al. 2011). Furthermore, it is currently unclear whether the IGT measures implicit or explicit knowledge. Uncertaint y vs. Risk Decisions are made under ambiguity ( uncertainty ) when outcomes are impossible to predict. For example, a patient can be completely unaware of whether a medication will improve their health or make their condition worse. Decisions under ambiguity differ from decisions under risk, which involve a certain known probability of outcomes. As an example, there could be a 95% chance that the encounter complications due to treatment. Autho rs of the Iowa Gambling Task
19 pro posed that knowledge was not available about the possible outcomes of choosing from each deck (Bechara et al., 1994). Thus the decisions in the game were believed to be made under complete uncertainty, or ambiguity (Bechara et al., 1994). Consequences of decisions ma de under uncertainty are unknown; therefore such decisions must be reevaluated as new information is acquired (Brand et al., 2007). Once a repertoire of information has been accessed, the decision can be made under risk, and detailed analysis of incoming d ata can be avoided (Brand et al., 2007). Brand et al. (2007) found that participants performing the Iowa Gambling Task made decisions under both uncertainty and risk. The researchers initially proposed that early decisions in the IGT were made under uncer tainty due to lack of knowledge of outcome probability. However, over time participants were predicted to generate outcome probabilities for each deck, at which point decisions would shift from uncertain to risky. To test their hypothesis, Brand et al (20 07) on a game known to involve risk, The Game of Dice Task (GDT), with performance of the IGT. In the GDT, participants bet on numbers they expect to see on a die. In this task a bet on more numbers results in a smaller reward if correct and a smaller punishment if incorrect. A bet on fewer numbers results in a large reward if correct, and a large punishment if incorrect. High scores indicate large risk and result from betting on fewer numbers. Scores between the two tas ks were found to be correlated only in the last the blocks (3,4,5) of the IGT ( Brand et al, 2007) participants were observed to shift their decision types from uncertain to risky midway
20 through performance of t he IGT. The researchers thus concluded participants deduced the probability of obtaining a certain outcome for each deck by the latter blocks of the IGT. Researchers Upton et al. (2011) aimed to determine whether performance in the Iowa Gambling Task was r the Balloon Analogue Risk Task (BART). In each trial of the BART participants decided to either pump a balloon or collect money and finish the game. Participants were aware that each pump added money to a money bag. It was also understood that the balloon could pop if it reached maximum capacity, which would result in a loss of any earnings and termination of the game. Risk for each participant was measured by the number of balloon pumps, where a high number of pumps indicated higher risk. Upton, et al. (2011) found high risk propensity according to the BART was significantly correlated with disadvantageous choices in blocks four, five, and six of the IGT. Upton et al. (2011) also found that choi ces made in the IGT by people with high BART scores did not significantly differ between the first and second half of the IGT. Thus, participants with low reward decks in the IGT. A tool used to measure executive function, the Wisconsin Card Sorting Task (WCST), requires that participants sort cards according to color, shape, or number of stimuli on each card. As rules of sorting change participants must override old rules and apply the new one. The WCST establishes a known set of rules for participants, and
21 Petre, Worsley, & Dagher, 2001). Conversely, in the IGT participants are not given rules as to which choices are good and which are bad. Rather, they must establish rules themselves during the game. Brand et al. (2007) found that performance on the WCST was correlated with performance on the IGT in the later, but not earlier block s of the IGT. This indicates rules (as to which decks are advantageous) are discerned by the end of the IGT, but are not present at the beginning of the game. If decisions in the earlier parts of the IGT are made under uncertainty, while decisions in the l ater portions of the IGT are made under risk, then the best alternative to retaining validity would be to score each block of the IGT separately, rather than compiling all blocks as one final score. This may explain why some populations (e.g., schizophren ics and smokers) perform poorly on the IGT and concurrently perform poorly on tasks which measure executive functions (Yip, Sacco, George, & Potenza, 2009). While populations including VMPFC damaged patients, OCD patients, and chronic gamblers do not perfo rm poorly compared to controls on the executive function tasks (Yip et al., 2009). Indeed, damage to anterior regions of the orbitofrontal cortex (OFC) and ventromedial prefrontal (VMPFC) region is associated with impaired IGT performance, but not impaired WCST performance (Yip et al., 2009). However, patients with damage to posterior regions of the OFC and VMPFC and dorsolateral prefrontal cortex (DLPFC) seem to be associated with impairments of performance on both the IGT and the WCST (Yip et al., 2009). It is possible that the DLPFC elicits rule following behavior, while the VMPFC and OFC allow the development of rules from rewards and punishments.
22 Does the IGT measure implicit or explicit knowledge? The somatic marker hypothesis posits that knowledge u sed in performance of the Iowa Gambling Task is implicit; however this view has been under great scrutiny as new evidence to the contrary is continuously uncovered. Bechara et al. (2005) proposed participants decide advantageously on the Iowa Gambling Task (IGT) before explicitly knowing the strategy to gaining the most money (Bechara et al., 1994). However, Maia and McClelland (2004) deduced that more sensitive measures would show participants are explicitly aware of properties of each deck while playing t he IGT. Verbal reports from participants revealed evidence of knowledge pertaining to the various decks of cards, indicating that participants have more access to knowledge than previously believed. Researchers Gutbrod et al. (2006) further investigated th e role of explicit memory in performance of the IGT. Contrary to Bechara et al. (1994), Gutbrod et al. (2006) proposed that conscious remembering was an essential component in performance of the Iowa Gambling Task (IGT). Gutbrod et al. (2006) recruited 11 participants suffering from amnesia due to brain damage along with 8 healthy controls to perform the IGT. Five of the amnesic participants possessed damage in their ventromedial prefrontal cortex (VMPFC), and the other six suffered from damage to the hippocampus or adjacent regions. Since the amnesic patients were unable to rely on conscious remembering while performing the IGT, they were expected to rely solely on implicit memory, and according to Bechara et al. (1994) perform at control levels. ing Task.
23 Results showed that healthy participants chose from the advantageous decks significantly more often at the end of the task than at the beginning, indicating learning. The amnesic participants did not show any improvement in performance of the IGT over time. Levels of anticipatory skin conductance responses did not differ between the amnesic and control groups. Since amnesic patients performed poorly in the IGT but had similar SCRs to controls, researchers proposed that performance in the Iowa Gamb ling Task must be attributed at least partially to conscious recollection. In order to determine the extent of explicit knowledge during performance of the IGT Persaud, McLeod & Cowey (2007) designed an objective measure for awareness in cognitive tasks. Persaud et al. (2007) used a post decision wagering method to assess conscious knowledge in 1) a blindsight localization task and 2) during performance of the Iowa Gambling Task. In the blindsight localization task, a participant with a scotoma in his ri ght visual field was to indicate whether an image was present or absent in his damaged field of correct, he would receive the amount of money wagered, however if he answered incorrectly the wagered amount of money would be withdrawn from his earnings. Results showed that although the participant corre ctly identified the presence or absence of the stimuli 70% of the time. Only 48% of these correct choices were followed by high wagers. The participant only made advantageous wagers (high wager after correct response/ low wager after incorrect response) 52% of the time. Although the
24 participant was abl e to perform the task, he lacked conscious knowledge of his ability to do so. In the Iowa Gambling Task, healthy participants chose between four decks of cards and were asked to make a wager of either $10 or $20 after choosing one of four decks and prior to observing the underlying reward and possible associated loss. The turning over of a card resulted in a gain of either 1x or 2x their wager, and some cards also had a loss associated with the win. The net outcome of choosing from deck A or deck B was a loss of 5x their average wager per ten cards. The net outcome of choosing from deck C or D was a net gain of 5x their average wager per ten cards. Three different questioning conditions were established to evaluate the effect of questioning type on conscio us awareness. In the first condition participants were only asked to wager after deciding on a deck. In the second condition participants were asked open ended questions about their knowledge of the decks after selecting from each one (as performed in Bech ara et al. 2004; 2006). In the last condition, participants were asked to evaluate their decisions quantitatively (as in Mia & McClelland, 2004) every 10th trial after the 20th trial. Participants in all conditions improved in performance and eventually chose from the advantageous decks more often than the disadvantageous decks. In the no question condition, participants did not increase their wager as they began choosing from the advantageous decks, indicating they lacked conscious awareness of their goo d performance. Participants in the open ended question condition did not differ from those in the no question condition, and wagers did not increase in response to
25 high ly correlated with their deck selections, indicating that more probing questions elicit a greater awareness of the quality of choices. This experiment performed by Persaud et al., (2007) displays participants are not explicitly aware of their choices unles s they are prompted to describe their feelings toward each deck. Two major threats to the validity of the IGT include the component of risk, of the decks traits. T o reduce the effects of decision type (uncertain vs. risky), performance on the IGT should be evaluated by block rather than by total score, so early and later portions of the task can be compared. To reduce effects of explicit vs. implicit awareness of th e properties of each deck, participants should either be prompted to consider the properties of each deck (explicit priming), or prompted to ignore properties of each deck (implicit priming). Reliability Reliability of the Iowa Gambling Task has been ch allenged by an occurrence to choose from the disadvantageous deck B more often than the two advantageous decks (C and D) (Crone & Van Der Molen, 2007; 2004). This occurren ce is rarely noted children are hypersensitive to the frequency of losses, and less sensitive to t he overall gain or loss associate with each deck ( Crone & Van Der Molen, 2007)
26 Frequency of Gains and Losses. Researchers Fum, Napoli, and Stocco (2008) hypothesized preference for deck B was due to the low frequency of losses within this deck. Decks A a nd C have frequent losses, whereas decks B and D have infrequent losses of higher magnitudes ( Appendix A ). Fum et al. (2008) devised a task to determine the influence of frequency of unexpected events and long term expected value on performance in the IGT. Researchers stated that magnitude of gains and losses should not be a determining factor in performance because magnitude depends on immediate gains and the frequency of losse s. Fum et al. (2008) manipulated the long term expected value for each deck to d were used including; a standard condition where immediate gains varied inversely with expected values, a null condition in which expected value was zero, and a frequency c ondition where expected values were contrasted with loss frequency (i.e., decks with the most frequent losses were the most profitable). Net estimated value did not significantly affect instead influenced by the frequency of losses. Immediate gains were shown to play a role in only the first few trials in all conditions. As predicted by the Somatic Marker Hypothesis, participants were influenced by high initial gains when they first began sampling from decks. However, par ticipants consistently showed a preference for decks with low frequency losses (B and D) regardless of which condition they were assigned.
27 than long term estimated value of decks a finding which does not abide by the Somatic Marker Hypothesis. The researchers performed a second experiment in which conditions remained the same as the first, but the payoff matrix was reversed such that every card selection was associated with a loss and the unexpected event was a monetary gain (Figure 2 reversed I GT). Therefore the good decks in the standard condition switched to A and B, and in the frequency conditions decks A and C were associated with frequent wins but also led to negative expected values, while decks B and D were associated with a low number o f rewards but led to greater expected value outcome. Results showed no significant differences between the standard, control, and frequency conditions. Participants in each condition were influenced by the frequency of wins, while the immediate losses did not appear to influence their selection behavior. was preference for low frequency consequences in the Iowa Gambling Task. Fum et al. (2008) concluded participants initially based de cisions in the IGT on high net gains, but eventually switched to decisions based on loss gain frequency. The second of which was more influential than the first. Further studies replicated these results, finding that net estimated values play a small role compared to the frequency of losses or gains (in the variant version) when choosing between the decks (Chiu et al.,2008; Lin, Song, Lin & Chiu, 2012). Frequent lo sses appear to be a more salient punishment than future net gains to participants who
28 labeled bad because participants find frequent losses more aversive than not winning money. Thus to obtain more reliable results in the IGT, performance should be (A+B). SCRs According to the Som atic Marker Hypothesis people who perform poorly on the IGT lack the ability to produce emotional responses to available options. However, recent observations show some populations perform poorly on the IGT despite the ability to elicit anticipatory SCRs w hen choosing between decks (Starke, Tuschen Caffier, Markowitsch, & Brand, 2009). Fourteen patients diagnosed with obsessive compulsive disorder (OCD) by criteria set by the DSMIV and fifteen healthy control subjects were compared on performance of the Io wa Gambling Task, and the Game of Dice Task (Starke et al., 2009). The Iowa Gambling task was used as a means to assess decision making under ambiguity, while the Game of Dice Task was used to measure decision making under risk. Skin conductance was recor ded during performance of both the IGT and the GDT. Patients with OCD did not differ from controls in performance on the Game of Dice Task, and SCRs recorded during this task did not differ between groups. However, patients with OCD performed poorly on th e IGT compared to controls, and again, there were no differences in SCRs between groups. Results indicated that dissociation exists between decision making under ambiguity and under risk in patients with OCD as confirmed by psychological data, but the Som atic Marker Hypothesis cannot explain the poor performance of OCD patients on the IGT. These results indicate that OCD patients
29 who performed poorly on the IGT were able to elicit SCRs, but were missing another component necessary for good decision making in the IGT. Age Effects of age on Iowa Gambling Task performance were investigated in a study which used a modified version of the IGT where all decks were chosen from equally often (Cauffman et al., 2010). In this version, participants were given the opti on to either pass or choose from each deck one at a time. A total of 901 individuals with ages ranging from 10 30 years old were gathered to participate. Researchers investigated how choices changed over six blocks of the IGT by calculating the percentage of good and bad choices in each deck. Researchers observed a curvilinear relationship between age and rate of change in percentage of good choices. Thus, middle aged participants were more likely to play rather than pass advantageous decks as the game progressed. Age was observed to have a negative linear relationship to choices from disadvantageous decks between blocks. Thus, as the game progressed, older participants chose significantly less often from the disadvantageous decks than younger participan ts. A pattern of poor performance was observed in children who played the Iowa Gambling Task. Researchers wondered whether this poor performance was a result of insensitivity to loss or of inability to use outcome information to predict future consequence s. In order to investigate this issue, Crone and Molen (2007) asked three different age groups, ranging from 8 10 years, 12 14 years, and 16 18 years, to play a modified version of the Iowa Gambling Task called the Hungry Donkey Task while their skin condu ctance responses (SCRs) were measured. Researchers predicted that if young
30 children performed poorly on the IGT as a result of insensitivity to loss, then younger children would have lower SCRs after losses than older children. Alternatively, Crone and Mo len believed that if children were unable to predict future outcomes, younger children should show significantly lower SCRs compared to older children prior to choosing disadvantageous decks. Only 16 18 year olds showed significant improvement in the numb er of advantageous choices over task blocks (Crone & Molen, 2007). All subjects preferred to choose from decks with low punishment frequency over those with high punishment frequency. SCRs following reward or loss did not significantly differ between age g roups, thus autonomic nervous system responses to loss were the same across age groups. However, a significant difference was observed in the SCRs preceding the decision of a deck between age groups. The 16 18 year old group showed higher SCRs prior to cho ices which would lead to frequent punishment, compared to infrequent punishment. There were no differences of SCRs following loss between age groups, meaning all ages were equally sensitive to loss. Therefore, insensitivity to loss could not explain the yo older children showed larger SCRs prior to deck choices, the largest responses were predictors of high loss probability rather than of low net gain. The authors believed that yo ung adolescents were using frequency of punishment rather than magnitude of reward or punishment to reason in the IGT, leading to poor performance overall. These results also indicate that age is a factor which may influence validity in performance of
31 the Iowa Gambling Task. Older subjects may yield significantly different results than younger subjects. Performance during the Iowa Gambling Task does not always yield reliable results. Frequency of gains and losses may play a larger role than previously belie ved. Furthermore, anticipatory SCRs may not lead to choices from the overall advantageous decks as initially believed. Lastly, performance may vary significantly with age. Dissociable Processes Involved in Performance of the Iowa Gambling Task These dispa rate findings led researchers Stocco, Fum, and Napoli (2009) to evaluate the possibility that two different cognitive componen ts drive choices from the most rewarding decks. First, they proposed that initial learning of the advantageous decks required cogn itive control involving the assessment of magnitude of payoff. Second, they believed that after learning the pattern of payoff, people would rely on a less cognitively demanding strategy where they would assess frequencies of losses. Experimenters had 152 participants perform an ordinary Iowa Gambling task in the first phase of the experiment. In phase two participants were not given feedback about gains or losses causing them to rely on previously acquired information to make advantageous choices. Partici pants were asked to perform a distraction task during phase one, two, both, or neither of the IGT. The task was to label a presented number as even or odd. Decisions made based on net gain as determined by the classic deck formula (C+D) (A+B) were compared to decisions based on frequency of losses established using
32 the formula (B+D) (A+C). The original scoring method was based on net monetary gain, whereas the scoring method developed by Stocco, Fum and Napolli (2009) was based on frequency of losses. Res ults showed that in phase one, condition (distraction phase; first, second, both, or neither) had a significant effect on decisions made based on monetary gain. Participants who performed the normal IGT in both phases and those who performed the distractio n task in phase two made more advantageous choices than those who performed the distraction task in both phases, and those who performed the dual task in the first phase. There was no effect of condition on choices made based on frequency of loss, indicati ng that more cognitive control was necessary to establish payoff than to determine frequency of losses. Condition assignment had no significant effect on choices made in phase two of the experiment, indicating that performance could be determined from sole ly the first condition. complex patterns early on in the game (net estimated value), and resort to simplified patterns later in the game (frequency of losses). When pattern searc hing in the first phase was disrupted by performance of another task, participants never acquired the ability to follow a simplified pattern. These findings are evidence for two separate cognitive components involve d in performance of the IGT. Another scoring method was used to evaluate performance on the Iowa Gambling Task which took into account the progressive and negative aspects of each deck. Visagan, Xiang & Lamar (2012) used a trial by trial method of scoring performance in the IGT, accom panied by the traditional deck based conceptualization of determining
33 advantageous and disadvantageous decks. In the trial based scoring method, the net outcome for each deck up to that point was assessed for every trial. Thus, if the participant chose fro m the deck with the best net outcome up till that point, it was considered an advantageous choice. SCRs were measured and scored according to the trial based approach and the original deck based approach to investigate the impact of scoring method on physi ological response. When Visagan et al. compared the two scoring methods the trial based approach resulted in a more advantageous choice profile than the deck based approach. When using trial based approach, it was observed that deck B led to both advantag eous and disadvantageous choices in the majority of participants. Although deck B switched from an advantageous to disadvantageous status after Block 2, participants continued to sample from this deck more often or equal to other decks throughout the entir e experiment. Significantly greater numbers of SCRs were observed when participants were choosing from deck B in its disadvantageous state then when deck B was advantageous, though the amplitude differences between these two groups was not significant. No significant differences of anticipatory SCRs to disadvantageous decks were observed between deck based and trial based scoring methods. Results from this study indicate new evaluation techniques for the IGT may lead to different cognitive profiles (choice making evaluations), but not different physiological responses. The trial based scoring method designed by Visagan et al.(2009) combined all aspects of decision making into one scoring method, masking the elements responsible choices at a given point in time. If participants were more
34 influenced by estimated values towards earlier portions of the task, the trial based scoring method would reflect these results. To determine effects of frequency vs. estimated value on decisions both scoring methods should be used to determine performance in the Gambling Task and compared between each block (Fum et al., 2009). Neural Correlates of the Iowa Gambling Task To determine which specific brain regions were involved in performance of t he IGT, researchers Li et al. (2010) recorded brain activity while participants performed the task Li et al. (2010) proposed that according to the Somatic Marker Hypothesis, certain brain areas are involved in the production of somatic markers including a reas responsible for working memory and others involved in triggering emotional responses. They suggested that performance of the IGT was likely to require activation in areas associated with memory including the dorsolateral prefrontal cortex (dlPFC), for working memory and the hippocampus for longer retention of information (Figure 1). They also proposed that areas involved in triggering emotional responses such as brainstem nuclei including the ventral striatum (in accordance with the error prediction hy pothesis) periaqueductal gray (PAG), and the insula cortex, an area involved in analysis of internal viscera would be activated. They proposed that these areas, involved in eliciting a somatic state, are coupled through the orbitofrontal cortex (OFC) and t he ventromedial prefrontal cortex (VMPFC). This compiled information was proposed to be sent to areas involved in carrying out the decision including the anterior cingulate cortex (ACC) and
35 the somatosensory area (SMA), along with the striatum which activa tes the dopamine motivation pathway. Ten university students performed the Iowa Gambling Task along with a control task. In the control task people were told to pick sequentially from four decks which had the gain/ loss value visible. There were no advant ageous or disadvantageous decks in the control task. Each participant performed five blocks of 20 trials each of both the control and the actual Iowa Gambling Task. The tasks were mixed, thus a participant would make one draw from the real IGT and the foll owing draw would be from the control task. Li et al. (2010) found that activation of the dlPFC working memory area was more highly activated in performance of the IGT than the control task; however this was not the case for activation of the hippocampus. E motional areas which showed higher activation in the IGT than controls included the anterior insula, and posterior cingulate, however no significant activation was observed in the amygdala. The coupling systems which were proposed to be involved in unitin g emotion and memory areas including the OFC and VMPFC were both more highly activated in the IGT than in the control task. Areas proposed to elicit behavioral choices including the dorsal striatum and the SMA were also more highly activated while particip ants were performing the IGT than while performing the control task. Researchers also looked at differences in brain activation from the beginning to the end of the task in participants who were able to figure out the task and win money. In the last block participants showed greater activation in the insula and the SMA
36 compared to the first block which suggests that execution of decisions relies on these two components. Figure 2 Gambling Task. Lin et al. (2008) used fMRI to examine neural correlations of anticipation vs. outcome, wins vs losses, and different contingencies between decks in the Iowa Gambling Task. Analysis of IGT performance was conducted between choices based on referred to the time interval six seconds prior to a participant's decision and the interval after the button press was defined as the experience period. Results showed that brain activation during anticipation periods included areas such as the insular cortex (an area activated dur ing aversive stimulation and the experience of fear or disgust), and the lentiform nucleus (an area associated with expectation of reward). Other activated areas included the left inferior parietal lobule, and the ACC The areas activated during anticipati on of gains compared to those activated during anticipation of losses did not differ. However, areas activated following
37 outcome appearance of gains and losses could be segregated into different brain regions. The researchers concluded that activation of areas involved in positive and arousal preparing for any outcome. They also concluded that the observed differences between decks during the experience period indicated th at all decks elicited different responses from the participant, and that participants were not sensitive to the overall net gains of individual decks. In conclusion, this study established the experience period must guide decisions, since it was the only p hase which showed different activation areas between decks. A team of researchers from Kyoto University used fMRI to investigate brain activity during risky and safe decisions in the Iowa Gambling Task (Fukui et al., 2005). Data were obtained from 14 volu nteers who completed the Iowa Gambling Task while being monitored under fMRI. Participants were given a 3.5 second response time to pick a card from a deck, after which a computer would randomly generate a response. No participants had more than 5 decision decisions were defined as choices from the advantageous decks C and D. Data were analyzed for the period just prior to, and positively correlated with medial prefrontal activ ity. Thus, participants who were more
38 sensitive to risky choices were more likely to choose from advantageous decks. The authors propose that long term risk anticipation is controlled by the medial frontal gyrus and possibly the adjacent anterior cingualat e cortex. This study did not look at differences in medial prefrontal activity between the beginning and the later parts of the IGT in response to decisions of the various decks. Fukui et al. (2005) assumed that participants were sensitive to the risk of c hoosing from the bad decks thought the entire game. However, it is possible that participants were unaware of the riskiness of their choices at certain times during the game. Dopamine Agonists and Impulse Disorders Ann Klinestiver, a high school English te acher from West Virginia was diagnosed highly involved in control of movement. Th is disease also reduces dopamine producing neurons in the Substantia Nigra. Anne was prescribed Requip which acted to imitate the neurotransmitter dopamine in the synapse. She was restored to normal health until the a trip to Las Vegas and found herself, for the first (Lehrer, 2010). Upon her return to West Virginia, she reported entering the dog tracks daily from 7:30 am until 3:00 online computer game until the dog track reopened With maxed out credit cards and a diet of peanut butter, Anne began stealing quarters from her grandkids for the slots. After she left her husband to gamble she was alone and nearly $300,000 in debt. A
39 week after ceasing the drug Requip, Anne no longer had the compulsion to gamble, though eventually symptoms of PD returned. After a few rewarding slot trials dopamine neurons continuously pre dicted rewards in every circumstance near the slots. The drug the possibility that it could learn how to succeed at predicting the pattern of winning how to gamble. Anne was not the first or only person to experience an impulse dis order after being prescribed drugs which alter the dopamine pathway. Weintraub et al. (2006) investigated the prevalence of impulse control disorders (ICDs) in a population of types were predictors of ICDs. The researchers also investigated the prevalence of types of ICDs in this population, including compulsive gambling, compulsive sexual behavior, and compulsive spending. A sample of 272 patients with PD was screened for ICD s in accordance with the Minnesota Impulsive Disorders Interview ( Weintraub et al., 2006) Eighteen of the PD patients were found to meet the criteria for an ICD. Presence of an ICD was predicted by two factors including treatment of PD through dopamine ag onists, and history of ICD symptoms prior to PD onset. Patients with PD who also had an ICD were significantly more likely to have been administered daily doses of dopamine agonists when compared to PD patients who did not have an ICD. Prevalence of gambl ing ICDs were equal to the prevalence of compulsive sexual behavior ICDs in the population examined.
40 In order to determine whether decision making was impaired in people suffering ents not currently diagnosed with dementia and 22 normal controls to perform the Iowa Gambling Task (IGT). All patients suffering from PD were taking either a dopamine conduct ance responses were measured while they performed the IGT. Prior to performing the IGT, participants were given a noise mismatch test while their skin conductance was measured to establish a baseline SCR level. Results showed that PD patients gained signif icantly less money than normal controls, which was due to the significantly greater number of disadvantageous choices made by the PD patients. There was no differences of SCRs between groups on the noise mismatch task, indicating that both groups were able to produce SCRs. Normal controls generated higher SCRs when choosing from disadvantageous decks, whereas in PD patients, no differences in SCRs were observed when choosing between advantageous and disadvantageous decks. Normal controls had higher anticipa tory SCRs, reward SCRs, and punishment SCRs when compared to SCRs of PD patients. Kobayakawa et al. (2008) proposed that PD patients have amygdala dysfunction early in the disease, and thus are unable to produce an emotional reaction to decisions. In this study decks were analyzed by grouping the advantageous (C & D) and the disadvantageous (A & B) decks, not allowing for an assessment of individual deck choices.
41 The Reward Prediction Error Hypothesi s Three groups of midbrain dopamine neurons send axons al ong long distance routes to influence brain activity in many areas. A8 and A10 groups of neurons stem from the ventral tegmental area (VTA) of the inner basal ganglia. These neurons project to the ventral striatum and fronto cortical regions. A9 neurons pr oject from the substantia nigra pars compacta (SN) named for the dark melanin pigment found in the dopamine neurons, plainly apparent in the SN due to the density of these neurons here. These neurons innervate the caudate and putamen of the basal ganglia. The large size of cell bodies, electrical coupling of the cells, low firing rates of cells, and homogenous dopamine distribution throughout the cells imply that although the messages from these neurons are few, they are w idely heard (Glimcher, 2011). Dopam inergic targets include the frontal cortex and basal ganglia. Projections from the frontal cortex pass to the caudate and the putamen of the basal ganglia complex (Middleton 2002). Researchers Fiorillo, Tobler and Schultz (2003) aimed to determine differences of dopamine neuron activation between decisions made under risk, and those made under uncertainty. The researchers trained two monkeys in a Pavlonian procedure where visual stimuli indicated the probabili ty (P= 0, 0.25, 0.5, 0.75, and 1) of a liquid reward being delivered after a 2 s delay. Anticipatory licking increased as the probability of the reward increased, indicating that the animals were able to discriminate between the reward schedules. Dopamine neurons in the ventral midbrain showed no response to fully predicted reward (P=1.0). However, phasic activation of these neurons was observed when the probability of receiving the reward was less than one. The
42 magnitude of the neural responses increased a s the probability of receiving a reward decreased. Researchers also found that during uncertain circumstances (P=0.5), the dopamine neurons had a sustained increase in activity, which grew from the onset of the conditioned stimulus until the expected tim e of reward. This was not the case in more certain conditions where probability of receiving a reward was less than or greater than chance. Fiorillo et al. proposed that sustained activation of these neurons was coding uncertainty. An experiment was desig ned to verify sustained activity under uncertainty was a property of motivationally relevant stimuli, and not generalizable to all uncertain events. Monkeys viewed two stimuli in a series, with the second stimuli following the first in only half of the tri als (P=0.5). Neither phasic nor sustained responses were observed in dopamine neurons during this task. To further investigate the activation of these neurons during motivationally relevant uncertainty, Fiorillo et al. (2003) used visual stimuli to indica te the magnitude of the potential reward (P=0.5), and observed that sustained activation of the dopamine neurons increased with increasing reward magnitude. To determine whether this activation was a result of discrepancy in potential reward rather than a ctual reward magnitude, Fiorillo et al. (2003) designed yet another experiment. In this experiment reward was delivered on each trial, but varied between two magnitudes. Three stimuli indicated two possible rewards. One stimulus predicted a small or medium reward, another predicted a small or large reward, and a third
43 predicted either a medium or large reward. Sustained activation of the dopamine neurons was greatest after the stimulus preceding the largest variation (small or large reward). Results from Fi orillo et al. (2003) indicate one type of dopamine neuron carries information about reward probability, and another type of activation carries information about reward magnitude. If this is the case, it would be expected that choices from decks A and C on the Iowa Gambling task which have are uncertain (P=0.5) would result in highly phasic activations, whereas choices from decks B and D would be represented by tonic activations of the dopamine neurons. From this study it would be expected that dopamine neu rons would have sustained firing in response to choices from decks with more frequent losses (p=.5) It would also be expected that the sustained activation would be greatest for the disadvantageous deck which would have the largest discrepancy between gai ns and losses. symptoms of pathological gamblers (Weintraub et al., 2006).Researchers have shown that peripheral concentration of dopamine is elevated when healthy controls and chron ic gamblers gamble (Marazziti et al., 2008). Peripheral concentration of 5 HT were observed to decrease in pathological gamblers (Marazziti et al., 2008). However, both positive and negative effects have been observed when treating pathological gamblers wi th dopamine antagonists (Marazziti et al., 2008). Inconclusive results are also found when serotonin reuptake inhibitors are used to elevate levels of serotonin in the
44 synapses of pathological gamblers (Grant & Potenza, 2007). Researchers Zeeb, Robbins, an d Winstanley (2009) investigated the roles of dopamine and serotonin in performance on the Iowa Gambling task using rats as models for human gamblers. The researchers tested effects of agonists at the dopamine, D1 and D2 receptors, the serotonin, 5 HT1A re ceptor, and d amphetamine on performance of the Iowa Gambling Task. They also tested effects of a 5 HT1A receptor antagonist to mimic the low serotonin observed in pathological gamblers. Zeeb, et al. (2009) had rats nose poke into one of four holes in orde r to receive a reward. Each hole was associated with a reward of a certain number of pellets. Rats could receive either one, two, three or four pellets. Upon a nose poke, rats either out which they could not gain any rewards by nose poking. Holes with higher rewards were accompanied by longer time outs Holes with larger rewards also had higher probabilities of receiving a punishment (e.g., p=0.5 and p=0.6), compared with holes which had lower rewards (p=0.1, and p=0.2).Researchers calculated the percentage of choices from each hole to determine the preferred choice. A baseline value of choice preference was determined for each group. Rats then received an agonist or antagonist o f dopamine or serotonin, or a saline injection (control). Results showed that all baseline groups significantly preferred the two pellet reward option over the 1, 3, and 4 pellet option. Amphetamine was shown to switch the llet option over the two pellet option. Even when punishment time was held constant, rats in the amphetamine group still preferred the
45 lowest punishment, lowest reward option. sensitivity to loss frequency, leading the m to choose the lowest reward option. The dopamine agonist was not observed to alter choices in this version of the rat gambling task. A dopamine, D2 receptor antagonist significantly improved optimal choices, increasing the number of choices from the two pellet option and decreasing choices from the three and four pellet option. A Dopamine, D1 receptor antagonist had no effect on choices compared to controls. A serotonin agonist (8 OH DPAT) significantly impaired performance on the rat gambling task. Thoug h the 5 HT1A antagonist (WAY100635) did not significantly alter choice behavior in isolation, it was observed to block negative choice effects of the 5HT 1A agonist. Results reflect that rats choose decks according to two factors including the probability of loss and the magnitude of loss. Drugs such as amphetamine may have an effect on only one of these factors, indicating that different brain pathways control each individual factor. It is possible that pathological gamblers become either insensitive to lo ss frequency or insensitive to loss magnitude. Kang et al. (2010) found that differences in the genomic code can result in impaired performance on the IGT. Chromosome 11 houses the BDNF gene, responsible for production of brain derived neurotropic factor p roteins which have been shown to influence proliferation of serotonin and dopamine neurotransmitters. A common polymorphism of this gene occurs at the 66th codon and results in a substitution of the amino acid Valine with a different amino acid Methionine (Val66Met). This genomic polymorphism has been linked with poor short term memory and slow cognitive
46 function. The current researchers investigated the effects of this polymorphism on ypes were determined from a sample of white blood cells using genomic assays designed for the gene of interest. Results showed participants with the Val66Met polymorphism performed significantly worse than participants who did not possess the Val66Met poly morphism in the last three blocks of the task. No difference was observed between groups in the first two blocks. The researchers proposed that the Val66Met polymorphism may be associated with IGT performance through impaired reward processing capabilities Skin Conductance Responses Physiology Skin conductance is one aspect of electrodermal activity which measures how well the skin conducts electricity when a constant voltage (using DC current) is applied. The unit of measurement for skin conductance is the conductance are related to the activity of eccrine sweat glands innervated by sympathetic nerves. Sweat is an electrolytic solution, meaning higher sweat sympathetic division of the autonomic nervous system (ANS) has unmylenated preganglionic neurons in the thoracic and lumbar spinal cord which descend from efferent nerves running down the spinal column. These preganglionic fibers release acetylcholine (ACh) onto long unmylenated C fiber postganglionic neurons which te rminate around the sweat glands.
47 The structure of the eccrine sweat gland is comprised of the reobsorptive sweat duct (RSD) and the secretory coil (Gibbons et al., 2009). The intraepidermal portion of ns et al., 2009). When Ach is released from postganglionic cholergininc nerves, cholinergic receptors in the secretory cytoplasm, resulting in sweat secretion (Gib bons et al., 2009). Brain activation Studies have shown that production of the skin conductance response is impaired after VMPFC, anterior cingulate and parietal lobe lesions. However, the lack of evidence for neuroanatomical locations associated with pro duction of spontaneous fluctuations of SCR, and generation of discrete SCR events led researchers Critchley et al. (2000) to perform a functional magnetic resonance imaging (fMRI) study in order to pinpoint activation sites during somatic arousal. The rese archers aimed to differentiate between activity relating to generation and afferent representation of skin conductance responses, or the feedback of the skin conductance response. While monitored under fMRI, participants were shown two decks of cards and w ere told that on half the draws choices from the red deck would be rewarded, but on the other half of the draws choices from the back deck would be correct. If they chose from the wrong deck participants were told they would lose money. The total amount of money gained or lost was displayed on the side of the stimuli presentation screen for participants to gauge their progress.
48 Researchers were able to apply a mask to the fMRI results to eliminate results showing areas of the brain directly related to rewar d components of the task, to better target areas of the brain involved in the skin conductance response and feedback. Results from fMRI showed that brain activity preceding SCRs occurred in the cerebellum, extrastriate visual cortices, and in the left medi al prefrontal lobe. Brain activity immediately following production of SCRs was observed in the right medial prefrontal cortex. Areas of the brain involved both early and late in the SCR response included the medial prefrontal cortex, lingual gyrus, right posterior cingulate, orbitofrontal cortex, and cerebellum. Though researchers observed SCR related activity in prefrontal and insula regions, there was not much activation observed in the anterior cingulate nor in amygdala regions during the production of the SCR. The anterior cingulate is associated with physical and cognitive effort, sympathetic cardiovascular arousal, and subjective emotional experience. The experimenters were able to differentiate activity involved in generation of the SCR from activity involved in cognitive representation of the SCR. Confounds of Skin Conductance as a Dependent Measure A main assumption of the SMH poses that SCRs are a result of an emotional response, however recent evidence points to SCRs as a result of other factors s uch as task complexity. Botvinick and Rosen (2009) sought to elucidate whether effortful thought would be considered a costly outcome by participants. Previous research had led the authors to believe that since people prefer less cognitively demanding task s, that
49 people would display a different physiological response to high cognitive demand tasks and low cognitive demand tasks. In this study, 12 participants were given the choice between two decks of cards, and were told to perform one of two tasks dep ending on the color of the bottom of the card which the participant chose. The participants performed either a magnitude judgment (indicating whether the number on the card was greater or less than five) or a parity judgment (indicating odd or even). Numbe rs of both colors were placed in each deck, but unannounced to participants, one deck tended to stay a constant color, and the other deck alternated between colors. Performance of this task was broken into 34 ten trial blocks. Once a participant chose from one deck on the first trial, they were told to choose from that deck for the following nine trials. Skin conductance was measured throughout the experiment to determine the magnitude of responsiveness toward each deck. Results showed that when choosing f rom the deck which switched tasks often (high cognitive demand), participants produced higher SCRs than when choosing from the deck which tended to stay constant (low cognitive demand, Botvink & Rosen, 2009). The researchers concluded that these skin SCRs were preparing for attention and task engagement, based on a prediction of upcoming cognitive demand. Authors also concluded that registration of expected cognitive demand was seen as an impending cost to the participant, information valuable in further ex periments on value based decision making.
50 This could be a potential confound to determining whether emotional responses respond to frequency as opposed to estimated values, since calculating estimated values should be cognitively demanding for participant s. The Current Study The Iowa Gambling Task (IGT) was designed to investigate how people make decisions under uncertainty Poor performance on this task is linked to ventromedial prefrontal cortex damage, obsessive compulsive disorder, depression, and eat ing disorders (Bechara, 2005). In this task which aims to mimic real life decision making participants choose from four decks of cards to gain the most money possible. Unannounced to the participant two decks are advantageous and lead to a net gain of mone y while the other two decks lead to a net loss of money. The Somatic Marker Hypothesis (SMH) proposes that a healthy body produces emotional responses (measured by skin conductance) which guide choices away from the disadvantageous decks (Damasio, 1994). A n alternative hypothesis proposes that people evaluate individual decks based on the frequency of wins and losses in each deck and are not aware of the net winnings of the individual decks. According to this theory, skin conductance responses (SCRs) are sa id to result from high frequency gains or losses associated with certain decks. However, no empirical evidence currently exists to support the claim that decisions based on frequency of gains and losses can be physiologically separated from decisions made based on calculated estimated value. played a computerized version of the Iowa Gambling Task while their skin conductance was recorded One group played the
51 IGT while listening to numbers played on headphones an d were told to indicate whether participants played the IGT while listening to white noise. The Somatic Marker Hypothesis assumes no differences of anticipatory SCRs exist b etween decisions made based on frequency of wins and losses and those made based on net value of the decks (Bechara et al., 1994). However if emotional responses predict choices based on gain loss frequency larger anticipatory SCRs should be observed when participants choose from decks with high frequency losses (decks A and D). A main effect of frequency should be observed. The present research should help to uncover the evaluative processes involved in performance of the IGT, and will therefore help to pinpoint compromised cognitive components in populations whom perform poorly on the IGT. Aim 1: Replicate findings that pa rticipants prefer decks with low frequency losses and with high net gains M ain effect s of frequency of loss and estimated value should be observed. Implications : If both components display a significant effect, then both play a role in decision making during IGT performance. Aim 2: Determine whether choices based on frequency increase and choices based on estimated value decrease, when executive function is blocked. This would be represented by a group by frequency and estimated value interaction. I mplications : Executive functions would be necessary for determining estimated value but not for determining frequency of losses
52 Aim 3: Determine whether decisions based on frequency of losses elicit higher emotional responses than those based on overall estimated value. Implications : If the frequency of losses is responsible for eliciting the anticipatory SCRs, then both groups should possess a main effect for frequency. If estimated value is resp onsible for eliciting the aSCRs, then a main effect of estimated value should be observed. Aim 4: Determine whether an interaction exists between the estimated value and frequency of losses. Implications: An interaction would indicate a relative effect of skin conductance, such that combined effects of frequency and estimated value produce the observed effects. Method Participants Participants included 36 students from the New College of Florida who volunteered to participate in the p resent study. All participants were between 18 and 23 years old. Partici pants responded to verbal and e mail methods of recruitment. Materials Participants were presented with a modified computerized version of the IGT on a Toshiba Laptop computer ( Bechara et al., 1994; Mueller, 2009; Herzig, 2013). Punishment and reward schedules are described on Table 1. The original computerized version of the IGT was modified so that any monetary loss (even those that led to an overall $0 gain) appeared in red. T he game was also modified from no inter trial interval period to allow six seconds to elapse between each trial Skin conductance was
53 measured for each participant using GSR electrodes (MLT116F, ADInstruments ) attached to the medial portion of the index and middle finger of the non dominant hand. A Powerlab GSR unit recorded skin conductance and data were converted into units of skin conductance, microsiemens. Protocol Prior to participating, participants w ere told they would be involved in a study measuring how choices are made. Participants were given two informed consent forms, one to sign and return to the experimenter and another to keep. They sat in front of a laptop computer running a modified version of the computerized IGT. The experimenter read instructions out loud to participants while a baseline skin conductance was (Figure 3 ) A minimum of six seconds elapsed between each card choice to allow skin conductance to return to baseline after receiving feedback from a deck choice Participants performed five blocks of 20 trials, thus 100 ca rds were chosen before the game was over. One group of participants listened to numbers on headphones while they performed the IGT and indicated whether the number was even or odd to block executive functions The other group of participants listened to wh ite noise on headphones while they performed the IGT (simplynoise.com). Skin conductance was recorded while participants performed the Iowa Gambling Task (AD Instruments, Powerlab GSR). Anticipatory skin conductance responses were defined as SCRs in the s econd preceding a deck selection. Anticipatory SCRs were
54 Chart). The six seconds between choices allowed SCRs to return to baseline before another decision was made. Prev ious studies have reported that the increased task length does not affect performance (Bowman, Evans & Turnbull, 2005). Analysis T wo separate scoring methods were compared. The first score, currently used as the dependent measure in most studies, wa s found by the deck formula (C+D ) (A+B). This score was based on estimated value of each deck since the low estimated value decks were subtracted from the high estimated value decks. From now on this score will be termed the estimated value score or EV score. The second score was obtained by subtracting the high frequency loss decks from the low frequency loss decks (B+D) (A+C). This score will be referred to as the frequency value score or FV score. Both scores were gathered for each of the five blocks of 20 choi ces to determine whether the two scoring methods yielded different results This analysis was confounded by dualistic properties of each deck, thus a second analysis was performed to determine the individual effects of estimated value and frequency of loss es. The second analysis investigated combined effects of loss frequency and estimated value on choices made during the IGT. For example, dualistic properties of deck B result in low loss frequency value and low estimated value, whereas deck C possesses hig h loss frequency and a high estimated value. The dual properties of each deck were therefore separated into two categories including estimated value and frequency value, which could either be low or high ( Table 1 ) To determine effect of loss
55 frequency and net estimated value on choices made during the IGT, a 2 (group; blocked Executive function, control) x 5(block) x 2(estimated value; High, Low) x 2(Frequency of Loss; High, low) mixed ANOVA was performed using SAS 9.2. Anticipatory skin condu ctance responses we re analyzed using a 2 (group; blocked Executive function, control) x 5(block) x 2(estimated value; High, Low) x 2(Frequency of Loss; High, low) mixed ANOVA. Results Choice Data There was a significant effect of scoring method (EV or FV) on scores in the IGT, F (1, 33) = 8.58, p =.004, 2 =.076, where scores according to EV formula (M=19.9) were greater than scores according to FV formula (M=13.8). Scores significantly differed between block during performance of the IGT F (1, 33) = 5.68, p <.001, 2 =.076 (Figure 4 ). sensitive to frequency of losses, estimated value, or both equally. An analysis was performed to determine the effect of both the fre quency (high and low) and estimated value (high and low) compone nt of each deck on choices made A significant main effect of frequency value showed preference for low frequency losses (M=5.62, SD= 3.36) over high frequency losses (M=4.3, SD=2.65). This ef fect was constant throughout the entire task. A significant interaction was found between estimated value and frequency value, F (1, 31) = 9.50, p= .004 (Figure 7 ). To further aled that participants chose significantly more often from deck B with low estimated value
56 coupled with low frequency losses than any other deck (Figure 5). This finding could be as present. Deck A with low net gains coupled with high frequency loss was chosen from least compared with any other deck. A significant interaction between estimated value and block was observed, F(4, 124) = 5.6, p=.0004, such that lower estimated value decks (with higher initial gains) were favored in the first two blocks of the Iowa Gambling Task and over time higher estimated value decks were preferred (Figure 8 ). SCR Data When anticipatory SCRs were analyzed according to deck properties a main effe ct of group was observed, F (3, 32) =72.53, p <.0001 where the non distracted group had significantly high aSCRs than the distracted group. A significant interaction was found between estimated value and frequency value F (1, 7) = 6.97, p = .033(Figure 10 ). These results indicated aSCRs were higher for low estimated values than low frequency values, and were also higher for high frequency values than for high estimated values. A significant interaction was also observed between group, estimated value, and f requency of losses, F (1, 7) = 6.7, p =.033 (Figure 11 ). Discussion Results from the deck cho ice data indicate a prominent B deck phenomenon in both groups as Crone and Van Der Molen (2007) found in their population of children under the age of 18 Deck B was chosen significantly more than both advantageous decks C and D. The young student population sampled here may not have attended to the net amount of money as consistently as adult populations in previous studies They
57 may have been more influenced by the high initial gain paired with the low frequency of losses (Bechara et al. 1994, 1996, 2005). Deck A was chosen least of all indicating participants were sensitive to net estimated value (Bechara et al. 1994) The interaction of choices by block shows that participants developed an affinity towards the advantageous decks C and D over time which was also observed in Bechara et al. (1994) Scores obtained by the commonly used estimated value scoring method differed significantly from the loss frequency scoring method developed for this study In other evaluations of frequency. It was therefore imperative to determine which of these two or if both methods were essential for performance of the Iowa Gambling Task. Evaluation of deck properties (net estimated value and frequency of losses) showed preference for high estimated value grew over time and exceeded preference for low estimated value. Preference for low frequency loss was found to be constant throughout the task. Such results would be expecte d based on work presented by Stucco, Fum and Napolli (2008); (2009); (2010) which showed preference for the low frequency loss, low net gain deck B over the high net gain, high frequency loss deck C. The present finding that high estimated value was favored over high frequency losses and that low frequency value was favored over low estimated value indicate that both components play a significant role in determin IGT. Preference for level of estimated value was shown to vary by block during performance of the Iowa Gambling task. As Figure 8 displays, participants initially
58 favored low estimated value, and over time preferred hig h estimated value. This trend was likely due to the high initial rewards (+$100) associated with the low estimated value decks. Over time, participants preferred high estimated value choices based on previous experience with the decks. Though both groups elicited larger aSCRs over blocks, this response was not affected by levels of net estimated value nor by the frequency of losses within each deck. The interaction observed between estimated value and frequency of losses with skin conductance as the depend ent variable revealed that lower estimated value was accompanied by higher skin conductance than lower frequency losses. This interaction also revealed high frequency losses were accompanied by larger anticipatory SCRs than high estimated value. This interaction was present in both groups (distracted and non distracted), though the non distracted group had higher anticipatory skin conductance responses overall. Results show no effect of level of estimated value of frequency value on skin conducta nce; however i t is possible that skin conductance re fl ects both frequency and estimated value component of the Iowa Gambling Task throughout the entire task. The cognitive components which guide decision making may not be influenced by skin conductance as previously believed. Rather, cognitive activation may elicit the skin conductance responses observed. Li at al (2010) proposed that the OFC and VMPFC unite emotion and memory and are highly activate during performance of the IGT. It is possible that these coupling areas were very active in the particip ants in this study, and
59 lead to the skin conductance interaction between estimated value and frequency of losses. Though previous research found that a distraction task could shift choices toward preferring low frequency losses, such results were not revealed in the present study ( Stucco, Fum & Napoli 2009) Participant s in the present study were between the ages of 18 and 23, whereas participants in the previous study using the distraction task were all und er 16. It is possible that the task was not salient enough to load executive functions of the older participants in the present task. A more challenging distraction task may have loaded executive functions more effectively. Thus a future study may use an n back test to distract participants while performing the Iowa Gambling Task. It is also possible that participants in the present study were not relying on executive functions to determine their choices. However, the interaction between block and estimated value indicated participants did switch preference toward high estimated value with experience Thus this process of determining estimated value may not have taken place in the ventromedial PFC of participants in th e present study. Preference for Another limitation includes the short (6 second) interval between choices. This short interval was not long enough for skin conductance to return to baseline after each trial for every participant. Thus anticipatory responses for some participants may have been mixed with reward or punishment responses. Future studies would increase this length of time to verify aSCRs were indeed the only signal being recorded. Lastly, reward
60 frequency (in contrast with loss frequency) was not evaluated as a variable in the current study, however this may have had a larger effect on skin conductance than loss frequency according to the reward prediction error hypothesis. In conclusion, both distracted and non distracted groups significantly preferred low frequency losses over high frequency losses throughout the entire task. Both groups also learned t o prefer high estimated value cards over low estimated value cards. Thus, this population of participants may not have utilized prefrontal regions to perform the task at hand. It is possible both groups relied on primary inducers to predict outcomes. Anti cipatory skin conductance was not a direct reflection of choices made in this study since the two groups possessed significantly different aSCRs, but similar choices. It is possible that older participants rely more on areas involved with execut ive functi ons such as the vmPFC, to incorporate emotions with decisions. This youthful population may not have relied on the vmPFC to perform this task, and were incapable of pairing emotions with decisions during the task. The increase in skin conductance over blo cks for both groups may have resulted from an increase in preference for the high estimated value decks. In this case, it was likely that skin conductance was not leading choices in the IGT, but rather may have been elicited by the choices made. In other w ords, suppression of preference toward low estimated value decks may have generated higher aSCRs of participants. Results support the proposal that future studies using the Iowa Gambling Task as a neuropsychological tool look at the five blocks separately and to evaluate each deck
61 according to its in the task.
62 References Bechara, A., Damasio, A. (2005). The somatic marker hypothesis: A neural theory of economic decision. Games and Economic Behavior 52(2), 336 372. doi: 10.1016/j.geb.2004.06.010 Bechara A., Damasio A. R., Damasio H., Anderson S. W. (1994). Insensitivity to future consequences following da mage to human prefrontal cortex. Cognition 50 (1 3), 7 15. Bechara, A., Tranel, D., Damasio, H., Damasio, A.R. (1996). Failure to Respond Autonomically to Anticipated Future Outcomes Following Damage to Prefrontal Cortex. Cerebral Cortex 6 (2), 215 225. Behcara, A., Tranel, D., Damasio, H. (2005). Characterization of the decision making deficit of patients with ventromedial prefrontal cortex lesions. Brain 123 (11). 2189 2202. Boeka, A.G., Lokken K.L. (2006). The Iowa Gambling Task as a measure of decision making in women with bulimia nervosa. Journal of International Neuropsychology Society. 12 (5), 741 5. Botvinick, M.M., Rosen, Z.B. (2009). Anticipation of Cognitive Demand during Decision Makin g. Psychological Research (73) 835 842. Bowman,C.H., Evans, C.E.Y., Turnbull, O.H. (2005). Artificial time constraints on the Iowa Gambling Task: The effects on behavioural performance and subjective experience. Brain and Cognition 57(1), 21 25.
63 Brand, M., Recknor, E.C., Grabenhorst, F., Bechara, A. (2007). Decisions under ambiguity and decisions under risk: Correlations with executive functions and comparisons of two different gambling tasks with implicit and explicit rules Journal of Clinical and Expe rimental Neuropsychology. 29 (1)86 99. Buelow, M., Suhr, J. (2009). Construct Validity in the Iowa Gambling Task. Neuropsychological Review 19 102 114. Cauffman, E., Shulman, E., Steinberg, L., Claus, E., Banich, M., Grham, S., Woolard, J. (2010). Age diff erences in affective decision making as indexed by performance on the iowa gambling task. Developmental Psychology. 46 (1), 193 207. Critchley, H.D., Elliot, R., Mathias, C.J., Dolan, R.J. (2000). Neural Activity Relating to Generation and Representation o f Galvonic Skin Responses: A Functional Magnetic Resonance Imaging Study. The Journal of Neuroscience 20(8), 3033 3040. Gibbons, C. H., et al. Quantification of sweat gland innervations: a clinical pathologic correlation. Neurology 72, 1479 1486 (2009). C rone E.A., Van Der Molen M.W. (2007). Development of Decision Making in School Aged Children and Adolescents: Evidence From Heart Rate and Skin Conductance Analysis. Child Development 78(4), 1288 1301. Damasio, A., Antonio R. (1994). Descartes' error : e motion, reason, and the human brain New York : G.P. Putnam Fiorillo, C., Tobler, P., Schultz. (2003). Discrete coding of reward probability and uncertainty by dopamine neurons. Science, 299 (5614), 1898 1902.
64 Fukui, H., Murai, T., Fukuyama, H., Hayashi, T., & Hanakawa, T. (2005). Functional activity related to risk anticipation during performance of the iowa gambling task. NeuroImage, 24 (1), 253 259. doi: 10.1016/j.neuroimage.2004.08.028 Fum D, Napoli A, Stocco A. (2008). Somatic markers and frequency effects: Does emotion really play a role on decision making in the gambling task? In 30th Annual Conference of the Cognitive Science Society; Washington, DC Edited by Sloutsky V, Love B, McRae K. Lawrence Erlbaum; 1203 1208. Garcia Molina, A., Roig Rovira, T., Ensenat Cantallops A., Sanchez Carrion, R., Pico Azanza, N., Pena Casanova, J. (2007). Examination of decision making processes in patients with traumatic brain injury. Neurologia. 22 (4), 206 12. Glimcher, P.W. (2011) Understanding dopamine and reinforcement learning: the dopamine reward prediction error hypothesis. PNAS. (108) 15647 15654. Gutbrod, K., Krouzel, C., Hofer, H., Muri, R., Perig, W., Ptak, R. (2006). Decision making in amnesia: Do advantageous decisions require conscious knowledge of previous behavioural choices? Neuropsychologia. (44) 1315 1324. Kobayakawa, M., Koyama, S., Mimura, M., Kawamura, M. (2008). Decision making in sis of behavioral and Physiological Patterns in the Iowa Gambling Task. Movement Disorders(23) 4, 547 552. Li, X., Lu, Z., D'Argembeau, A., Ng, M., & Bechara, A. (2010). The Iowa gambling task in fMRI images. Human Brain Mapping, 31 (3), 410 423. doi: 10.1002/hbm.20875
65 Lin C H, Chiu, Y C, Cheng, C M, Hsieh, J C. (2008). Brain maps of Iowa Gambling Task. BMC Neuroscience. 9 (72).doi:10.1186/1471 2202 9 72 Maia, T.V., McClelland, J.L. (2004). A reexamination of the evidence for the somatic marker hypothesis: what participants really know in the Iowa gambling task. Proc Natl Acad Sci 101(45), 16075 80. Marazziti D, Golia F, Picchetti M, Pioli E, Mannari P, Lenzi F et al (2008). Decreased density of the p latelet serotonin transporter in pathological gamblers. Neuropsychobiology 57: 38 43. Middleton FA, Strick PL (2002) Basal primate. Cereb Cortex (12 ) 926 935. Monchi O, Petrides M, Petre V, Worsley K, & Dagher A., (2001). Wisconsin card sorting revisited: Distinct neural circuits participating in different stages of the task related functional magnetic resonance imaging. J Neurosci. 21 7733 7741. Napoli, A. & Fum, D. (2010 ). Rewards and punishments in iterated decision making: An explanation for the frequency of the contingent event effect. In D. D. Salvucci & G. Gunzelmann (Eds.), Proceedings of the 10th International Conference on Cognitive Modeling (pp. 175 180). Philade lphia, PA: Drexel University. http://iccm2010.cs.drexel.edu/proceedings/papers/Napoli.pdf Persaud, N., McLeod, P., & Cowey, A. (2007). Post decision wagering objectively measures awareness. Nature Neuroscience 10 (2), 257 261. doi:10.1038/nn1840
66 Rivalan, M., Ahmed, S. H., & Dellu Hagedorn, F. (2009). Risk prone individuals prefer the wrong options on a rat version of the iowa gambling task. Biological Psychiatry, 66 (8), 743 749. doi: 10.1016/j.biopsych.2009.04.008 Starcke, K., Tuschen Caffier, B., Markowitsch, H., and Brand, M. (2009). Skin Conductance responses during decisions in ambiguous and risky s ituations in obsessive compulsive disorder. Cognitive Neuropsychiatry 14(3), 199 216. Stocco, A., Fum, D., Napoli, A. (2009). Dissociable Processes underlying decisions in the Iowa Gambling Task: a new integrative framework. Behavioral and Brain Functions 5 (1), open access. Suhr, J.A., & Tsanadis, J. (2007). Affect and personality correlates of the Iowa Gambling task. Personality and Individual Differences 43, 27 36. Upton, D.J., Bishara, A.J., Ahn, W Y, Stout, J.C. (2011). Propensity for risk and trait impulsivity in the Iowa Gambling task. Pers Individ Dif. 50 (4):492 495. doi:10.1016/j.paid.2010.11.013. Visagan, R., Xiang A., & Lamar, M. (2012). Comparison of deck and trial based approaches to advantageous decision making on the Iowa Gambling Task. Psychological Assessment 24 (2), 455 463. doi: 10.1037/a0025932 Weintraub, D, Siderowf, A., Potenza, M, Goveas, J, M orales, K, Duda, J, Moberg, P, Stern, M. (2006). Association of dopamine agonist use with impulse control disorders in parkinson dis ease. Archives of Neurology, 63 (7), 969 973. Yip SW, Sacco KA, George TP, Potenza MN. Risk/reward decision making in schizophrenia: a preliminary examination of the influence of tobacco smoking
67 and relationship to Wisconsin Card Sorting Task performance. Schizophr Res. 2009;110:156 164. Zeeb, F., Robbins, T., & Winstanley, C. (2009). Serotonergic and dopaminergic modulation of gambling behavior as assessed using a novel rat gambling task. Neuropsychopharmacoogy. 34 2329 2343.
68 Figure 3 Screenshot of one trial of Iowa Gambling Task running on Toshiba Laptop computer using PEBL Psychological Battery. Participant chose deck A, gained $100, lost $150, and received a net gain of $ 50.
69 Table 1 Characteristics of Decks in the Iowa Gambling Task A B C D Initial Gain $100 $100 $50 $50 Net Estimated Value $250 $250 +$250 +$250 Loss Frequency 50% 10% 50% 10%
70 Figure 4 Scores by block during performance of the Iowa Gambling Task. The classic EV scoring method (shown in black) was determined for each block by subtracting the number of cards picked from disadvantageous decks from the number of cards drawn from advantageous decks, (C+D) (A+B)= EV. The newly established FV scoring met hod was determined for each block by subtracting decks with high frequency losses from those with low frequency losses, (B+D) (A+C). EV and FV scores significantly differed in block five of the IGT, p = .03, while difference in block three approached signif icance, p =.055. Bars represent +/ 1 SE of the mean. 0 5 10 15 20 25 30 1 2 3 4 5 Score Blocks of 20 Trials EV FV
71 Figure 5 Mean number of choices from each deck during performance of the IGT. All comparisons were significant, p<.001, except for comparison between deck C and D. Bars represent SE of the mean.
72 Figure 6 Mean number of cards drawn from each deck during the Iowa Gambling Task. Red lines indicate decks with low net gain, disadvantageous decks in the original publication, while black lines indicate decks with high net gain, advantageous decks in the original publication. Dashed lines identify decks with high loss frequency; solid lines are those with low loss frequency. Bars represent SE of the mean. 0 1 2 3 4 5 6 7 8 1 2 3 4 5 Choice Count Block A B C D Deck
73 Figure 7 Interaction between e stimated value and frequency value. A significant main effect of frequency was found indicating low frequency losses (M=5.62, SD= 3.35) were preferred over high frequency losses (M=4.38, SD= 2.65). No main effect of estimated value was observed. Red lines indicate choices made based on frequency of losses, while black lines indicate choices made based on net estimated value. 3 3.5 4 4.5 5 5.5 6 Hi Lo Choice Count Level of Estimated and Frequency Values EV FV
74 Figure 8 Choices made from decks with high and low net estimated values. Gray lines indicate the number of choices from decks with high estimated value, while black lines indicate the number of choices from decks with low estimated value. 0 1 2 3 4 5 6 7 1 2 3 4 5 Choice Count Block EV lo EV hi
75 Table 2 Mean Anticipatory SCRs by Estimated Value (EV), Frequency Value(FV), and Deck During IGT Performance A (Low EV, High FV) B (Low EV, Low FV) C (High EV, High FV) D (High EV, Low FV) Mean SE Mean SE Mean SE Mean SE 3.56 2. 99 3.59 2.60 3.94 3.27 4.83 4.39 Table 3 Mean Anticipatory SCRs for Estimated Value and Frequency Properties of Decks during IGT Performance Estimated Value Frequency Value Mean SE Mean SE High 4.39 2.11 3.79 1.67 Low 3.58 1.13 4.15 2.13
76 Note. indicating p p <.0001. Figure 9 Anticipatory skin conductance responses (aSCRs) for distracted and non distracted groups for each deck during performance of the IGT. SCRs for the two groups (distracted and non distracted) differed significantly for every choice (p<.05) except for choice B (p>.05).For the non distracted group, aSCRs in response to deck D were significantly greater than those from deck A ( p = .01) and B ( p <.0001). SCRs did not differ for choices in the distracted group. Bars indicate SE of the mean. 0 1 2 3 4 5 6 7 A B C D Anticipatory SCR (S) Deck Choice Non-Distracted Distracted *** ***
77 Figure 1 0. Interaction of anticipatory skin conductance responses during choices made based on net estimated value and those based on loss frequency. Red lines indicate aSCRs in response to choices based on frequency, while black lines indicate choices made based on estimated value. 3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 Lo Hi Anticipatory SCR (S ) Level of Estimated and Frequency Values FV EV
78 Figure 11 Interaction between group (Distracted and not distracted) estimated value (high and low) and Frequency of losses (high and low). Black lines indicate skin conductance in response to estimated values, whereas red lines indicate anticipatory skin conductance in response to frequency values. Solid lines indicate the distracted group, while dashed lin es indicate the non distracted group. 0 1 2 3 4 5 6 low high Anticipatory SCRs ( S) Distracted EV Distracted FV Not Distracted EV Not Distracted FV