Impulsivity and risk-taking behavior in focal frontal lobe lesions

Transcription

1 Neuropsychologia 46 (2008) Impulsivity and risk-taking behavior in focal frontal lobe lesions D. Floden a,b, M.P. Alexander a,c,d,e,f, C.S. Kubu g, D. Katz h,i, D.T. Stuss a,b, a Rotman Research Institute, Baycrest, Toronto, Ont., Canada b Departments of Psychology and/or Medicine, University of Toronto, Toronto, Ont., Canada c Harvard Medical School, Boston, MA, United States d Behavioral Neurology Unit, Beth Israel Deaconess Medical Center, Department of Neurology, Boston, MA, United States e Youville Hospital, Cambridge, MA, United States f Memory Disorders Research Center, Boston University, Boston, MA, United States g The Cleveland Clinic, Department of Psychiatry and Psychology, Section of Neuropsychology, Cleveland, OH, United States h Department of Neurology, Boston University School of Medicine, Boston, MA, United States i Brain Injury Program, HealthSouth Braintree Rehabilitation Hospital, Braintree, MA, United States Received 26 January 2007; received in revised form 24 July 2007; accepted 25 July 2007 Available online 3 August 2007 Abstract Frontal lobe dysfunction may underlie excessively impulsive and risky behavior observed in a range of neurological disorders. We devised a gambling task to examine these behavior tendencies in a sample of patients who had sustained focal damage to the frontal lobes or nonfrontal cortical regions as well as in a matched sample of healthy control subjects. The main objectives of the study were: (1) to behaviorally dissociate impulsivity and risk-taking; (2) to examine potential associations between specific frontal lesion sites and impulsivity or risk-taking; (3) to investigate the influence of reinforcement and trial timing on both behaviors. Our results indicated that patients and controls were equally likely to perform impulsively. Risk-taking performance strategies, however, were related to left ventrolateral and orbital lesion sites. Moreover, risk-taking was also associated with blunted response alteration following a nonrewarded trial. Patients and control subjects showed identical responses to rewardtiming manipulations consistent with formal decision-making theory. These findings suggest that ventrolateral and orbital lesions are related to the reward-based aspects of decision-making (risk-taking) rather than to simple response disinhibition (impulsivity). Reduced reaction to the negative consequences of one s actions may underlie this behavior pattern Elsevier Ltd. All rights reserved. Keywords: Reward processing; Response disinhibition; Decision-making; Orbitofrontal cortex; Common Difference Effect 1. Introduction Impaired behavioral regulation, such as impulsivity (IMP) and risk-taking behavior (RTB), is commonly observed after injury to the frontal lobes. Attempts to objectively evaluate these behaviors are complicated by the interrelated character of these behaviors. The operational definitions of IMP and RTB are not straightforward, and it is sometimes unclear if a particular behavior represents impulsivity or risk-taking or both (i.e., a spur of the moment leap from a cliff into an unfamiliar lake). For patients with brain injury who exhibit poor behavioral control, Corresponding author at: Rotman Research Institute, Baycrest, 3560 Bathurst Street, Toronto, Ont., Canada M6A 2E1. Tel.: x2938; fax: address: dstuss@rotman-baycrest.on.ca (D.T. Stuss). it may be important to distinguish between IMP, which may have one set of potential causes (i.e., stimulus-bound responses, poor response inhibition, etc.) and RTB, which may have another set of potential causes (i.e., poor computation of risk, blunted concern about risk, etc.). Single case studies of patients with frontal lobe damage have provided striking descriptions of both behavioral types (Eslinger & Damasio, 1985; Shallice, Burgess, Schon, & Baxter, 1989). Systematic investigations of IMP and RTB in frontal lobe patients have employed gambling paradigms (Bechara, Damasio, Damasio, & Anderson, 1994; Miller, 1992; Rogers et al., 1999). In particular, the Iowa Gambling Task (Bechara et al., 1994) is sensitive to the behavioral problems that frontal patients may exhibit in their everyday lives. This task does not, however, permit unambiguous behavioral separation of IMP and RTB (although some excellent modeling work has been done to fractionate the contributing mechanisms post hoc; Busemeyer /$ see front matter 2007 Elsevier Ltd. All rights reserved. doi: /j.neuropsychologia

2 214 D. Floden et al. / Neuropsychologia 46 (2008) & Stout, 2002). Other gambling tasks designed to separate these behavioral influences have not produced consistent results in frontal lobe patients. Results of one study suggested increased IMP (Miller, 1992) whereas work with a related paradigm suggested increased RTB (Clark, Manes, Antoun, Sahakian, & Robbins, 2003; Manes et al., 2002; Mavaddat, Kirkpatrick, Rogers, & Sahakian, 2000; Rogers et al., 1999). The conflicting data may be due to procedural differences in these two tasks: skilled processing (Miller, 1992) or probabilistic decision-making (Rogers et al., 1999); inverse (Miller, 1992) or independent (Rogers et al., 1999) relations between reward values and probabilities. Substantial differences in lesion distributions and in methodologies for analyzing and reporting lesion site also exist between studies. With these issues unresolved, the relative contribution of impulsivity and risk-taking to behavioral problems in patients with frontal lobe damage remains unclear. To address these questions, we developed a gambling task which represents a compromise between the techniques used in other studies. The current procedure removed all elements of skilled performance and carried an explicit, consistent, inverse relationship between success probability and reward values (high probability-low reward, and vice versa). The conflict between reward size and probability of obtaining that reward ensures that preference for high reward values reflects actual risk-taking; preferences for large rewards where there is a high probability of obtaining them would reflect an adaptive decision-making process rather than true risk-taking. Moreover, previous decision-making research has demonstrated that decision options phrased in terms of possible gains promote risk-aversive responding whereas options phrased in terms of possible losses result in risk-seeking response tendencies a phenomenon referred to as the Framing Effect (Kahneman & Tversky, 1984). To maximize our ability to detect risk-taking and minimize additional influences on decisions, the task did not involve losses. This study had three main objectives. The first was to determine whether patients with frontal lobe lesions were more likely than control subjects to exhibit IMP or RTB. To best address this question within the context of prior work, we designed a gambling task to separate IMP, which we defined as a failure to suppress an immediate reaction to a stimulus (i.e., a lack of control over behavior), from RTB, which we defined as a preference for responses associated with a low probability of obtaining a large reward (i.e., a type of poorly calibrated control over behavior). The underlying assumption is that RTB is a strategic response which is selected whereas IMP reflects reduced control over behavior and is evoked. We recognize that IMP, in particular, is a complex construct and many different types of impulsivity may exist. As a check on the ecological validity of our impulsivity measure, subjects also completed the Barratt Impulsiveness Scale (Patton, Stanford, & Barratt, 1995). The second objective, closely allied with the first, was to examine whether specific lesion locations in our small sample were associated with IMP or RTB. We identified individual subjects with excessive IMP or RTB and compared their lesion locations to identify critical lesion sites. We anticipated that RTB would be more frequent with lesions encroaching on ventral frontal regions that constitute part of the reward circuitry (Rolls, 2000). IMP, on the other hand, was hypothesized to be related to more dorsal and medial regions of the frontal lobes that play a role in attentional and motor control (Floden & Stuss, 2006; Ullsperger, 2006). The third objective was to characterize response patterns that might provide insights into the processes underlying IMP or RTB. Specifically, we investigated how subjects modified their behavior following positive or negative outcomes. Positive and negative feedback are crucial elements in acquiring and altering stimulus-reward associations. In fact, people tend to use feedback to guide behavior even in contexts that do not require stimulus-reward learning (i.e., the Gambler s Fallacy). There is growing evidence that some patients with frontal lobe damage fail to make use of this feedback (Bechara et al., 1994; Fellows & Farah, 2005a). We also evaluated how the timing of reward opportunities may influence performance on this task. Our task manipulation was based on the Common Difference Effect (Loewenstein & Prelec, 1992) from formal decision-making theory. Research in both humans and non-human animals has demonstrated that imposing a delay between reinforcement opportunities biases decisions towards larger rewards. This is best illustrated by contrasting two decisions: given a choice between one dollar today and two dollars tomorrow, most people will chose the small but immediate reward. However, if a time constant is added to both options, one dollar in 50 days or two dollars in 51 days, preferences switch and most people will choose the larger delayed reward. This effect is observed in decision contexts involving choice between two delayed options (in contrast to simple temporal discounting contexts involving choice between an immediate and a delayed option) but has not, to our knowledge, been investigated in the decision contexts similar to the current type of gambling task involving serial presentation of multiple options. We manipulated the intertrial interval to evaluate the influence of delayed reinforcement opportunities on performance in the patient and control groups. Our procedure is idiosyncratic in that we use probabilistic options rather than surety of rewards. The probabilistic nature of the choices means that, in effect, choice of any gamble is a selection of reward delay. Extending the intertrial adds a temporal constant to each choice, although it does not change the expected value of each choice. 2. Methods The Research Ethics Board of Baycrest Centre for Geriatric Care and University of Toronto approved the study. All participants gave written consent in accordance with the Declaration of Helsinki Subjects Eleven patients with chronic focal frontal lobe damage, six brain-damaged controls with cortical lesions outside the frontal lobes, and 11 age- and educationmatched neurologically normal control subjects participated in the study (see Table 1). Exclusion criteria included history of neurological or psychological

3 D. Floden et al. / Neuropsychologia 46 (2008) Table 1 Sample characteristics mean (S.D.) Group N Age Education NART-IQ BIS a Pre Post Frontal (9.0) 13.6 (2.0) (8.0) 66.0 (13.8) 69.8 (17.5) Nonfrontal (8.4) 14.7 (3.0) (10.6) 64.3 (9.5) 58.5 (12.5) Control (14.2) 14.7 (1.7) (7.0) 63.2 (11.9) a Pre-injury and post-injury total scores on the Barratt Impulsiveness Scale, 11th ed. disorder unrelated to lesion, estimated premorbid IQ < 90, history of alcohol or drug abuse, and impaired and uncorrected hearing or vision. All patients were in the chronic stage of recovery (4+ months post-event). Lesion location (documented from structural scans obtained for clinical purposes), etiology, chronicity, and size are displayed in Table 2. Focal lesions in patients diagnosed with epilepsy were a result of surgical intervention for treatment of seizures. All frontal patients diagnosed with epilepsy had late onset seizures (in adulthood). Of the nonfrontal patients diagnosed with epilepsy, two patients had early onset while three had late onset of seizures. All patients surgically treated for seizures, as well as two patients with frontal tumour resections and two patients with frontal lesions due to cerebrovascular events, were on standard dosages of anticonvulsant medications. One patient with epilepsy was taking medication thought to produce significant drug-related cognitive impairments (i.e., Topimax), although this was a patient with nonfrontal damage. All patients who underwent tumour resection had non-invasive/infiltrating masses and did not undergo chemotherapy or radiation treatment. A mixture of etiologies ensured lesion representation in all frontal regions. Prior work has demonstrated that etiology is less relevant for cognitive performance than lesion location (Burgess & Shallice, 1996; Stuss et al., 1994; Stuss, Floden, Alexander, Levine, & Katz, 2001). Two frontal patients had minor lesion extension into nonfrontal areas. Lesions were depicted on a standard anatomical template (Stuss et al., 2002) based on cytoarchitecture (Petrides & Pandya, 1994) and coded for the presence or absence of a lesion in each of seven lateralized frontal regions (see Table 2) Baseline neuropsychological measures All subjects completed a battery of clinically validated neuropsychological tests to establish baseline cognitive function. Tests assessed naming (Boston Naming Test; Kaplan, Goodglass, & Weintraub, 1983), comprehension (Token Test; Benton, Hamsher, & Sivan, 1994), visuospatial perception (Judgment of Line Orientation; Benton, Sivan, Hamsher, Varney, & Spreen, 1983), verbal attention span [Weschler Memory Scale-Revised (WMS-R) Digit Span; Wechsler, 1997], working memory (Consonant Trigrams; Brown, 1958; Peterson & Peterson, 1959), verbal fluency (phonemic and semantic; Benton et al., 1994), associative learning (WMS-R Verbal Paired Associates I and II; Wechsler, 1997), and executive function (Trail Making Test, Army Individual Test Battery, 1944; Wisconsin Card Sorting Test, Heaton, 1981), as well as questionnaires of depression symptoms (Beck Depression Inventory, Beck & Beck, 1972), absentmindedness (Cognitive Failures Questionnaire, Broadbent, Cooper, Fitzgerald, & Parkes, 1982), and diurnal rhythms (Morningness Eveningness Questionnaire, Horne & Ostberg, 1976). Table 2 Lesion characteristics Subject no. Sex Hand Pol Orb IM AC SM DL VL NF Etiology Chronicity (months) Lesion size a Frontal Nonfrontal Frontal 500 b M R B R B R R R R 0 Ruptured aneurysm c,d M R L L 0 Sx, low grade glioma c,d M L L L L Sx, late onset seizures b,d M R R R R 0 0 Sx, late onset seizures c,d F R L L 0 Sx, benign meningioma b M Amb B B B 0 0 Sx, falx meningioma F R R R R R R R R R CVA c,d M R 0 L L 0 Trauma M R R 0 R 0 R Trauma c F R B B R L 0 Trauma c,d M R B B R R Trauma Nonfrontal 511 M R L Sx, late onset seizures b M R L Sx, early onset seizures M R L Sx, early onset seizures > b F R R Sx, late onset seizures b F R L CVA 10 NA 537 F R R Sx, late onset seizures Pol = polar [10], Orb = orbitofrontal [11, 13, 14, 47/12 (orbital)], IM = inferior medial [14 (medial), inferior 24, 25, 32], AC = dorsal anterior cingulate [superior 24, 32], SM = superior medial [4 (medial), 6A (medial), 8B, 9 (medial)], DL = dorsolateral [4 (lateral), 6A (lateral), 8Ad, 8Av, 9 (lateral), 46, 9/46D, 9/46V], VL = ventrolateral [4 (ventral), 6B, 44, 45A, 45B, 47/12 (lateral)], NF = nonfrontal, Sx = surgical resection, CVA = cerebrovascular accident. a Lesion volume expressed as percentage of whole brain. b Impulsive. c Risk-taking. d Reduced response to negative outcomes.

4 216 D. Floden et al. / Neuropsychologia 46 (2008) Barratt Impulsiveness Scale All subjects completed the Barratt Impulsiveness Scale-11 (BIS-11; Patton et al., 1995), a self-report measure of real-world impulsive behaviors. Others (Berlin, Rolls, & Kischka, 2004) have found that patients with orbital frontal lesions report more impulsivity on BIS-11. The questionnaire contains subscales for attentional, planning, and motor aspects of IMP. Our definition of impulsivity is most similar to the behaviors reflected in the motor or nonplanning subscales. To assess changes from premorbid status, patients also reported the incidence of the same behaviors before any neurological event (Pre- BIS) Gambling task Trial events Five cards (4 cm 6 cm) were presented on a touch screen, face-down in a horizontal array (see Fig. 1). There were two Order conditions (Add and Subtract). In the Add condition, the screen was initially blank and one card was added to the display every 2 s, in a left-to-right order. In the Subtract condition, all five cards were initially present and the right-most card disappeared every 2 s. Subjects were informed that one of the five cards displayed the word WIN on its face whereas the other four cards were blank. Subjects were instructed that they could touch the screen at any time to stop the adding or removing process and turn over the cards present. If the WIN card was among the cards present at the response, the subject earned points. If the WIN card was absent at the response, no points were awarded. Thus, outcome information was available immediately after each response. The position of the WIN card was random on each trial, meaning that the more cards on the screen, the higher the probability that the WIN card was present. However, the point value of the WIN card was inversely related to the number of cards on the screen (i.e., more cards/higher probability = fewer points). The random nature of the WIN card location on each trial, the independence of each trial from every other, and the probability/reward contingencies were explicitly described to each subject. The likelihood of finding the WIN card and its associated value were displayed on the screen at all times during the trial to minimize memory demands. These values are indicated in Fig. 1. Subjects were instructed to try to win as many points as possible. Given the complexity of the task, instructions were repeated/paraphrased as needed to ensure that all subjects understood the nature of the task. Subjects initially completed five practice trials of each presentation condition (Add and Subtract) to ensure comprehension for the task. Subjects then completed 20 Add trials and 20 Subtract trials (block order counterbalanced across subjects) with a 2 s delay between response/reward delivery and the beginning of the next trial (Fast ITI). To evaluate the Common Difference Effect, Add and Subtract blocks were repeated (order reversed) with a 10 s delay between trials (Slow ITI). This procedure is somewhat idiosyncratic in that we use multiple probabilistic options rather than surety of two rewards to evaluate the Common Difference Effect. The probabilistic nature of the choices means that, in effect, choice of any gamble other than a sure thing is actually a selection of delay to reward. For example, choosing the first card means that, on average, one will have to make this choice 5 times (or minimum of 10 s given a 2 s intertrial interval) to obtain a reward, whereas choosing the third card means that, on average, one will only have to make this choice 3 times (or minimum of 18 s given a 2 s ITI plus 2 card presentations at a 2 s rate) to obtain a reward. Extending the intertrial adds a temporal constant to each choice, although it does not change the expected value of each choice. This is the crux of the Common Difference Effect manipulation. Note that the timing of card presentation (2 s interstimulus interval) did not change. Fig. 1. Gambling task displays and contingencies. A schematic diagram of successive screen displays during the gambling task. In the Add condition, presentation order moves from top to bottom. In the Subtract condition, presentation order moves from bottom to top. Each display is present for 2 s or until the subject makes a response and the trial is terminated. Each trial is then followed by a blank screen for the intertrial interval (2 s in the Fast ITI condition and 10 s in the Slow ITI condition). Reward contingencies are shown adjacent to each display, P win = the probability of finding the WIN card, WIN = point value for the WIN card. Note that, as the probability of finding the win card increases, the point value decreases.

5 D. Floden et al. / Neuropsychologia 46 (2008) Statistical analysis RTB and IMP were dissociated through comparison of Add and Subtract performance. A risk-taking subject will respond on the basis of few cards in an attempt to gain a large reward (despite the low probability of obtaining that reward). This means a risk-taker will respond quickly in the Add condition but wait for some cards to be removed in the Subtract condition. In contrast, an impulsive subject will show disinhibited, rapid responses, regardless of stimulus presentation order. This means they will respond with few cards present in the Add condition but many cards present in the Subtract condition. Therefore, the difference between the number of cards present at response for the two conditions (Subtract and Add) should be small for risk-taking subjects and large for impulsive subjects. Impulsivity was operationally defined as use of significantly fewer cards in the Add compared with the Subtract condition (to reflect large differences according to a Wilcoxon signed-rank test). Risk-taking was operationally defined as use of fewer than three cards on average in both presentation conditions (high rewards with an average probability of success less than.6). Performance patterns were evaluated for each subject and coded for presence or absence of impulsivity and risk-taking. As noted in Section 1, these patterns were mutually exclusive given that risk-taking reflected a performance strategy whereas impulsivity represented the absence of strategy. Chi-square analysis was employed to identify group differences in the proportion of subjects exhibiting each performance pattern. A trial-by-trial analysis examined performance change in response to negative and positive outcomes. If no reward is obtained on a trial (negative outcome), an adaptive strategy would be to respond on the basis of more cards on the next trial, thereby improving the odds of obtaining a reward. Conversely, if a reward is obtained on a trial (positive outcome), a subject might chose to respond on the basis of fewer cards on the next trial, in an attempt to obtain a larger reward. Subjects with behavior change less than 1.5S.D. from the control mean were considered to show reduced or low responses to negative or positive outcomes Lesion analysis We grouped patients on the basis of performance and used chi-square to examine the relationship between lesion location and performance pattern. This approach differs from the typical procedure of dividing patients into mutually exclusive groups based on gross lesion location and then comparing performance patterns across groups. Such groupings are, to a greater or lesser degree, artificial in that lesions rarely respect these divisions, either invading or undercutting other cortical territories. At best, large samples can reveal group differences but do not allow identification of critical areas within the larger region. At worst, such groupings can obscure meaningful effects when they contain heterogeneous patients. Our approach is therefore to use a priori criteria for identifying particular behavior patterns in individual patients and only then investigate whether meaningful relationships exist with lesion location. Others have adopted similar behavior-based strategies to identify brain-behavior relationships in a range of cognitive domains, including decision-making (Bechara et al., 1994), amnesia (von Cramon, Hebel, & Schuri, 1985), alexia (Cohen et al., 2003), and unilateral neglect (Hillis et al., 2005). However, care must be taken in using this technique with small sample sizes and conclusions based on few cases must be viewed as preliminary. 3. Results 3.1. Baseline neuropsychological measures Neuropsychological test performance was highly similar across groups, demonstrating the excellent recovery in the present patient sample. Only the Paired Associates I WMS- R subtest showed a group difference (F(2, 25) = 8.4, p <.005), where the nonfrontal group performed significantly worse than either frontal patients or controls (all t(15) > 2.8, p <.05). This is consistent with associative memory deficits frequently noted in mesial temporal lobe dysfunction secondary to temporal lobe epilepsy and/or anterior temporal lobe resection. Group performance for all tests is shown in Table Barratt Impulsiveness Scale Self-reported impulsivity did not differ across groups (see Table 1) although there was a trend for frontal lobe patients to report more cognitive impulsivity after the injury (t(10) = 2, p =.07). The frontal group also showed a weak relationship between cognitive impulsivity and increased absentmindedness on the Cognitive Failures Questionnaire (r(10) =.58, p =.06). Patients with larger frontal lobe lesions also reported more postinjury motor impulsivity (r(10) =.76, p <.01) Gambling task Fig. 2 displays the number of cards used during the Add and Subtract conditions for each a priori defined performance pattern (RTB, IMP) as well as for subjects whose performance was neither risky nor impulsive Impulsivity Three frontal subjects (500, 507, 517), three nonfrontal subjects (513, 521, 532), and seven control subjects showed impulsive responding in at least one of the ITI conditions (Fast ITI or Slow ITI). There was no group difference in the proportion of impulsive subjects (χ 2 = 1.64, p =.4) Risk-taking RTB was specific to frontal lobe damage; six frontal lobe patients (504, 505, 509, 525, 529, 531) demonstrated risky performance in at least one ITI condition while NO nonfrontal or control subjects were classified as risk-taking (χ 2 = 8.3, p <.005). All six risk-taking patients had either left ventrolateral or left orbitofrontal damage whereas patients who were not risk-taking did not have damage to this region (χ 2 = 11.0, p <.001; see Fig. 3). Within the frontal lobe group, RTB correlated negatively with lesion size (point bi-serial r(10) =.64, p <.05), indicating that risky performance was not an artifact of large lesions. RTB was not related to any other lesion characteristic (chronicity r(10) =.16, p =.64; etiology, χ 2 =.45, p =.52) Response to negative and positive outcomes On average, all subject groups responded on the basis of more cards on trials following a negative outcome (non-win trials); frontal subjects used an average of.47 additional cards (S.D. =.32), nonfrontal subjects used an average of.81 additional cards (S.D. =.49), and controls used an additional.79 cards (S.D. =.40). Six frontal subjects (504, 505, 507, 509, 525, 531), two control subjects, and no nonfrontal subjects were classified as low responders (less than 1.5S.D. from the control mean or an average increase of less than.19 cards). Five of the six frontal lobe patients identified as low responders had also been identified as risk-taking, suggesting that risk-taking may be related to reduced behavioral correction after negative outcomes. Just as with risk-taking, then, there

6 218 D. Floden et al. / Neuropsychologia 46 (2008) Table 3 Neuropsychological performance Frontal Nonfrontal Control N Mean S.D. N Mean S.D. N Mean S.D. Boston Naming Test a Token Test a,b Judgment of Line Orientation a Digit Span WMS-R Forward a Backward a Consonant Trigrams 0s a s a,b s a,b s a,b N-back Task Lag 0 a Lag 1 a Lag 2 a,b Paired Associates WMS-R Immed-easy a,b Immed-hard a Delay-easy a,b Delay-hard a Trail Making Test A Time B Time Wisconsin Card Sorting (128 cards) Categories Perseverative Errors Phonemic Fluency a Semantic Fluency a Beck Depression Inventory Cognitive Failures Questionnaire b Morningness Eveningness a Value reflects total number correct. b Required non-parametric comparison. was a significant relationship between left orbital or left ventrolateral lesions and reduced response to nonreward (χ 2 = 4.4, p <.05), and a negative correlation between low response to nonreward and lesion size (r(10) =.65, p <.05), indicating that this also was not an effect of large lesions. There was no relation of low response to nonreward to other lesion characteristics (chronicity, r(10) =.06, p =.86; etiology, χ 2 =.62, p =.23). However, response to nonreward did correlate with education (r(27) =.58, p <.001) such that low responders were also less educated. Responses to positive outcomes, or reward on the previous trial, were likewise in the expected direction. Frontal subjects responded with an average of.29 fewer cards (S.D. =.21), nonfrontal subjects used.37 fewer cards (S.D. =.23), and controls used.48 fewer cards (S.D. =.24) following a successful trial. However, there was no observed relationship between response to reward (positive outcomes) and lesion location or RTB, indicating that RTB was more associated with processing of negative outcomes Delayed reward opportunities There was a significant effect of ITI condition (F(1, 25) = 5.9, p <.05, MSE =.594) which did not interact with presentation condition (Add, Subtract), or group (all F < 1). Fig. 4 shows the general trend to use fewer cards (i.e., prefer larger reward values) when reinforcement was delayed (Slow ITI), consistent with the Common Difference Effect. 4. Discussion This study used a novel gambling task to dissociate IMP and RTB. In our sample, IMP was NOT exacerbated in patients with frontal lobe damage. RTB, on the other hand, was exaggerated following frontal lobe damage in some patients. Several studies using the Cambridge Gambling Task have also reported increased risk-taking (rather than impulsivity) in patients with frontal lobe damage (Clark et al., 2003; Manes et al., 2002; Mavaddat et al., 2000). Both the current procedure and the Cambridge Gambling Task emphasize probabilistic decision-making

7 D. Floden et al. / Neuropsychologia 46 (2008) Fig. 2. Performance-based groups. Average cards employed for the Add and Subtract conditions for risk-taking and impulsive subjects, as well as for subjects whose performance did not meet a priori criteria for either category. Consistent with those criteria, risk-taking subjects used fewer than 3 cards on both conditions whereas impulsive subjects showed large differences between Add and Subtract performance. Sample size is denoted for each group. Error bars represent the standard error of the mean. with explicit reward contingencies. In contrast, Miller (1992) used a skilled problem-solving task without explicit reward contingencies and found increased impulsivity rather than risktaking. Together, this suggests that probabilistic decision tasks encourage RTB, not impulsivity, in patients with frontal lobe damage. Risk-taking was not a general consequence of frontal lobe damage; it occurred only in patients with ventrolateral and orbitofrontal lesions. Others have reported poor probabilistic decision-making after lesions to the ventral surface (Mavaddat et al., 2000; Sanfey, Hastie, Colvin, & Grafman, 2003). Many studies using the Iowa Gambling Task (i.e., Bechara et al., 1994; Fellows & Farah, 2005a) have also reported poor decisionmaking in patients with ventromedial lesions. Discrepancies in ventral lesion location across studies may be largely terminological rather than anatomical. Surface depictions of the Iowa patients lesion appear almost discretely orbital and medial, but the projections on coronal slices reveal extensive undercutting of the ventrolateral region we have identified here. However, not all studies in patients with well-defined frontal lesion locations have identified RTB in patients with orbital and ventral lesions. Manes et al. (2002) found RTB was greater only in patients with large Fig. 3. Lesion location comparison for frontal patients. Left hemisphere is to the right, according to radiological convention. Lesion density is indicated by colour bar on the right. (A) Patients showing risky performance show maximal overlap in the inferior regions of the left frontal lobe (green arrows). (B) Patients who do not use a risk-taking strategy do not have lesions in these areas.

8 220 D. Floden et al. / Neuropsychologia 46 (2008) Fig. 4. Delayed reward opportunities. Average number of cards used by each group in the Add and Subtract conditions in the Fast and Slow ITI conditions. Consistent with the Common Difference Effect, there was a significant effect of ITI such that subjects used fewer cards during the Slow ITI condition when reward opportunities were delayed relative to the Fast ITI. Note, however, that the small nonfrontal group did not show this tendency in the Add condition. Error bars represent standard error. lesions whereas Rogers et al. (1999) found that patients with ventromedial lesions were risk-averse rather than risk-taking. The reason for these conflicting findings is unclear although several lesion and demographic variables have been proposed to contribute to decision-making patterns. For example, lesion laterality and gender may play a role in decision-making performance. The effect of laterality is a difficult question to address in patient studies given the relative rarity of unilateral lesions in ventral areas. Left-sided lesions were most relevant in the present study, although the small sample size prevents strong conclusions regarding lateralization. Fellows and Farah (2005a) also found that left-sided lesions were most associated with RTB. Others have argued that right lesions are critical for observing impairment on gambling tasks (Clark et al., 2003; Tranel, Bechara, & Denburg, 2002). Tranel, Damasio, Denburg, and Bechara (2005) have recently suggested that laterality alone does not determine performance but rather, that laterality interacts with gender. In a series of directed comparisons, they demonstrated that decision-making deficits seem to occur more frequently after left-sided lesions in men and rightsided lesions in women. They propose that this is consistent with sex differences in approach to problem-solving. We looked for similar effects in our small sample of risk-taking patients but did not find a consistent interaction of gender and lesion location. Nonetheless, this is a very intriguing idea that deserves further study in an effort to understand the possible contribution of lesion laterality. Risk-taking was also associated with a reduced behavioral reaction to negative outcomes. That is, orbital and ventrolateral patients who showed RTB did not alter their performance to the same degree as non-risk-taking subjects on trials following nonreward. Similar observations have been made in reversal learning studies where patients (Fellows & Farah, 2003) or nonhuman primates (Rolls, 2000) with lesions to ventral frontal regions fail to alter established stimulus-reward associations in the face of repeated negative feedback. Convergent functional neuroimaging studies (Elliott, Frith, & Dolan, 1997; Paulus, Feinstein, Tapert, & Liu, 2004) have repeatedly demonstrated that orbitofrontal cortex is active during flexible modification of stimulus-reward associations. It may be that RTB in our patients may be one manifestation of impaired reversal learning. Alternatively, RTB may be a consequence of a general impairment in setting stimulus-response criteria and flexibly modifying that criteria based on experience or feedback. Our prior work has demonstrated that left ventrolateral lesions (Petrides & Pandya areas 44/45) impair contingent criterion setting and subsequent response bias on multi-dimensional choice reaction time tests (Stuss, Binns, Murphy, & Alexander, 2002), sustained attention tests (Alexander, Stuss, Shallice, Picton, & Gillingham, 2005), and recognition memory tests (Alexander, Stuss, & Fansabedian, 2003). Similarly, an event-related fmri and EEG study reported left ventral frontal activation only on error trials when reaction time on the subsequent trial slowed (Garavan, Ross, Murphy, Roche, & Stein, 2002). This is consistent with our suggestion that this region calibrates response selection as behavior unfolds. The reversal learning and criterion setting accounts are not necessarily mutually exclusive and could account for complementary aspects of impaired adaptive decision-making or problem-solving. Our findings may represent separable frontal systems defined by unique patterns of cortico-cortical (Petrides & Pandya, 2002) and cortico-striatal (Alexander, DeLong, & Strick, 1986) connectivity. One system involves the ventromedial and medial orbital surfaces that have major connections with limbic system: hypothalamus, amygdala, and ventral (limbic) striatum. Lesions in this region may disrupt the drive states or emotional responses to reward value that motivate behavior and underlie reversal learning (Rolls, 2000). Another system involves the ventrolateral and lateral orbital surfaces that have bidirectional connections with parietal and temporal association cortex and discrete, topographically distinct cortical-striatal networks. Based on the connectivity of this region, lesions here could disrupt the attention-dependent capacity to establish and modify response criteria based on feedback. Thus, differences in placement within limbic or cognitive networks yield discrete expectations about performance deficits following lesions to ventrolateral versus ventromedial portions of the orbital surface. Functional imaging work has suggested similar conclusions with anatomical specificity for these processes consistent with the two lesion locations identified in the present study (Elliott, Dolan, & Frith, 2000; O Doherty, Critchley, Deichmann, & Dolan, 2003). All subjects, regardless of strategy and despite knowledge that each trial was independent, showed a tendency to reduce risk following a nonreward trial and increase risk following a rewarded trial. From the viewpoint of formal decision-making theory, this could be interpreted as a demonstration of susceptibility to the Gambler s Fallacy, which is the mistaken belief that outcomes across random events are actually related. The patients who altered their performance less could therefore be seen as more rational decision makers. Similar observations have been made in a recent study of investment decisions in patients with frontal lobe damage (Shiv, Loewenstein, Bechara,

9 D. Floden et al. / Neuropsychologia 46 (2008) Damasio, & Damasio, 2005). These authors gave participants the option to invest or pass on coin tosses and found that control subjects were significantly more risk-averse following a loss compared to target patients with damage to orbital frontal lobe or other regions involved in emotional processing (i.e., amygdala, somatosensory cortex). These studies suggest that blunted reaction to negative outcomes or failure to use negative feedback to guide behavior is a beneficial factor when decisions are unrelated. In contrast, in contexts where trials are related and one must use feedback to learn to avoid options that were once beneficial, frontal lobe patients appear irrational. There is some evidence that patients with focal frontal lobe damage show normal susceptibility to other irrational decision-making tendencies that reflect the subjective value of rewards. In the present study, we manipulated the delay between reward opportunities and found that frontal lobe lesions neither increased nor decreased susceptibility to the Common Difference Effect (Loewenstein & Prelec, 1992) such that when response options are equally delayed, the size of the reward has more influence over decision-making. This effect was apparent despite the atypical structure of our task relative to traditional Common Difference Effect procedures. Likewise, Fellows and Farah (2005b) have demonstrated that frontal lobe patients and control subjects show a similar tendency to undervalue delayed rewards relative to immediate rewards, an effect known as temporal discounting. Steeper discounting slopes have been identified in some studies of Attention Deficit Hyperactivity Disorder (Barkley, Edwards, Laneri, Fletcher, & Metevia, 2001) and drug abuse (Bickel & Marsch, 2001; Coffey, Gudleski, Saladin, & Brady, 2003; Dom, D haene, Hulstijn, & Sabbe, 2006). It may be that more widespread neurochemical abnormalities are necessary to disrupt these types of decision-making processes Clinical relevance for impulsivity Our task evoked performance consistent with our definition of IMP but, contrary to our hypothesis, it was equally prevalent in all subject groups. This result differs from Miller s (1992) finding of increased IMP following frontal lobe lesions. In Miller s task, decisions were based on problem-solving skill on visual or verbal puzzles. Patients with frontal injury show deficient strategy use on a variety of tasks (Burgess & Shallice, 1996; Stuss et al., 1994). With failure to adopt a successful strategy, frontal subjects may default to a stimulus-bound tactic, thereby appearing impulsive in comparison with control subjects. In the current task (and the Cambridge Gambling Task), strategy gains nothing because success is probabilistic; there is no winning strategy. Performance in control subjects and frontal patients would not be differentially influenced by use of behavioral strategies. Our work and others (Clark et al., 2003; Manes et al., 2002; Mavaddat et al., 2000) reinforce the fact that some degree of impulsive behavior is normal. Probabilistic decisions (as in the current study and the Cambridge Gambling Task; Rogers et al., 1999) may evoke IMP in normal subjects and, as such, the selective problems in frontal lobe patients are only seen in situations where normal controls use behavioral strategies that would prevent disinhibited responses. The chronicity of the lesions may also be a factor. IMP is often noted during the acute phase of recovery from frontal lobe damage. In Miller s study (Miller, 1992), approximately a third of the sample was tested in the subacute (2 3 weeks post-surgery) stage of recovery. In contrast, the patients included in the work with the Cambridge Gambling Task (Clark et al., 2003; Manes et al., 2002) and the current sample were tested during the chronic (>4 months) stage of recovery. IMP may resolve over time, at least in well-recovered individuals. This has been demonstrated in Utilization Behavior, a disinhibition syndrome that may reflect IMP (Lhermitte, 1983). To our knowledge, no work has focused on the evolution or recovery of IMP. The lack of significant group differences in self-reported IMP on a real-world behavior questionnaire suggests that our sample of patients is not excessively impulsive in other contexts. One might argue that failure to report IMP in daily life simply reflects reduced insight. While this is always a possibility, particularly in patients with frontal lobe damage, we are hesitant to assume that this is overwhelmingly the case in the present sample. Disturbed awareness correlates with the severity of cognitive dysfunction (Wagner & Cushman, 1994) and the present sample showed no significant differences from controls on baseline neuropsychological testing. Case studies (i.e., Eslinger & Damasio, 1985) of patients with intact cognitive function who nonetheless lack awareness into severe behavior disturbances typically have had large bilateral frontal lesions and have been too behaviorally impaired not cognitively impaired to return to work. In our patient group, six subjects with frontal lobe damage had returned to their previous employment and two others had resumed working, albeit in a different or modified capacity. Others (Berlin et al., 2004) have found that patients with orbital frontal lesions report more impulsivity on the same questionnaire used here. Subtle differences in lesion size, precise frontal location, or chronicity may underlie different results, however the presence of IMP appears to predict a poor functional outcome despite good cognitive recovery. In addition, the only significant trend in the patient responses was for increased motor impulsivity in patients with larger lesions, whereas lack of insight might be expected to be greater in patients with larger lesions. Regardless, we consider IMP to be a multifaceted construct and it is possible that our operational definition of IMP would be unrelated to fully insightful responses on the BIS Summary and implications This study explored the neural bases of impulsivity and risk-taking behavior. Contrary to clinical descriptions, we did not find evidence of impulsivity in a group of well-recovered patients with frontal lobe damage. Rather, poor decision-making in some patients appeared to be a function of impaired rewardrelated processing as patients with left orbital and ventrolateral lesions engaged in risk-taking behavior and showed blunted reaction to negative outcomes. This may be due to poor calibration of stimulus-response criteria, failure to reverse learned reward associations, or some combination of these mechanisms depending upon precise lesion location. Our data also suggest

10 222 D. Floden et al. / Neuropsychologia 46 (2008) that behavior tends to become more risky when reinforcement opportunities are delayed. These preliminary findings have important implications for understanding and managing behavior problems. First, they highlight the fact that impaired decision-making may arise from dissociable sources and suggest that the location of damage or dysfunction may provide clues to the underlying mechanisms. For example, in the presence of ventral lesions, behavioral interventions might target knowledge of behavioral contingencies and training of response-reward associations. Our findings also suggest that delaying response opportunities may exacerbate, rather than prevent, poor decision-making. As a clearer picture emerges of the neural and cognitive mechanisms underlying these behaviors, behavioral and pharmacological treatment should translate into increased independence and quality of life for patients with frontal injuries. Acknowledgements This work was supported by a Rotman Group Grant (# MGC 14974), Operating Grant (DTS # MT 12853), and Doctoral Research Award (DF) from the Canadian Institutes of Health Research. D.T. Stuss was funded by the Reva James Leeds Chair in Neuroscience and Research Leadership. M.P. Alexander was supported by the NICHHD, R01-HD We are very grateful to anonymous reviewers for helpful comments and suggestions and to D. Derksen, S. Gillingham, and C. Gojmerac for testing participants. D.F. is now at the Cleveland Clinic, Center for Neurological Restoration and Department of Psychiatry and Psychology, Cleveland, OH. References Alexander, G. E., DeLong, M. R., & Strick, P. L. (1986). Parallel organization of functionally segregated circuits linking basal ganglia and cortex. Annual Review of Neuroscience, 9, Alexander, M. P., Stuss, D. T., & Fansabedian, N. (2003). California Verbal Learning Test: Performance by patients with focal frontal and non-frontal lesions. Brain, 126, Alexander, M. P., Stuss, D. T., Shallice, T., Picton, T. W., & Gillingham, S. (2005). Impaired concentration due to frontal lobe damage from two distinct lesion sites. Neurology, 65, Army Individual Test Battery. (1944). Manual of directions and scoring. Washington: War Department, Adjutant General s Office. Barkley, R. A., Edwards, G., Laneri, M., Fletcher, K., & Metevia, L. (2001). Executive functioning, temporal discounting, and sense of time in adolescents with attention deficit hyperactivity disorder (ADHD) and oppositional defiant disorder (ODD). Journal of Abnormal Child Psychology, 29, Bechara, A., Damasio, A. R., Damasio, H., & Anderson, S. W. (1994). Insensitivity to future consequences following damage to human prefrontal cortex. Cognition, 50, Beck, A. T., & Beck, R. W. (1972). Screening depressed patients in family practice. Postgraduate Medicine, 52, Benton, A. L., Hamsher, K. S., & Sivan, A. B. (1994). Multilingual aphasia examination (3rd ed.). Iowa City: University of Iowa. Benton, A. L., Sivan, A. B., Hamsher, K. S., Varney, N. R., & Spreen, O. (1983). Judgment of line orientations. Lutz: Psychological Assessment Resources, Inc.. Berlin, H. A., Rolls, E. T., & Kischka, U. (2004). Impulsivity, time perception, emotion and reinforcement sensitivity in patients with orbitofrontal cortex lesions. Brain, 127, Bickel, W. K., & Marsch, L. A. (2001). Toward a behavioral economic understanding of drug dependence: Delay discounting processes. Addiction, 96, Broadbent, D. E., Cooper, P. F., Fitzgerald, P., & Parkes, K. R. (1982). The Cognitive Failures Questionnaire (CFQ) and its correlates. British Journal of Clinical Psychology, 21, Brown, J. (1958). Some tests of the decay theory of immediate memory. Quarterly Journal of Experimental Psychology, 10, Burgess, P. W., & Shallice, T. (1996). Response suppression, initiation and strategy use following frontal lobe lesions. Neuropsychologia, 34, Busemeyer, J. R., & Stout, J. C. (2002). A contribution of cognitive decision models to clinical assessment: Decomposing performance on the Bechara gambling task. Psychological Assessment, 14, Clark, L., Manes, F., Antoun, N., Sahakian, B. J., & Robbins, T. W. (2003). The contributions of lesion laterality and lesion volume to decision making impairment following frontal lobe damage. Neuropsychologia, 41, Coffey, S. F., Gudleski, G. D., Saladin, M. E., & Brady, K. T. (2003). Impulsivity and rapid discounting of delayed hypothetical rewards in cocaine-dependent individuals. Experimental and Clinical Psychopharmacology, 11, Cohen, L., Martinaud, O., Lemer, C., Lehericy, S., Samson, Y., Obadia, M., et al. (2003). Visual word recognition in the left and right hemispheres: Anatomical and functional correlates of peripheral alexias. Cerebral Cortex, 13, Dom, G., D haene, P., Hulstijn, W., & Sabbe, B. (2006). Impulsivity in abstinent early- and late-onset alcoholics: Differences in self-report measures and a discounting task. Addiction, 101, Elliott, R., Dolan, R. J., & Frith, C. D. (2000). Dissociable neural responses in human reward systems. Journal of Neuroscience, 20, Elliott, R., Frith, C. D., & Dolan, R. J. (1997). Differential neural response to positive and negative feedback in planning and guessing tasks. Neuropsychologia, 35, Eslinger, P. J., & Damasio, A. R. (1985). Severe disturbance of higher cognition after bilateral frontal lobe ablation: Patient EVR. Neurology, 35, Fellows, L. K., & Farah, M. J. (2003). Ventromedial frontal cortex mediates affective shifting in humans: Evidence from a reversal learning paradigm. Brain, 126, Fellows, L. K., & Farah, M. J. (2005a). Different underlying impairments in decision making following ventromedial and dorsolateral frontal lobe damage in humans. Cerebral Cortex, 15, Fellows, L. K., & Farah, M. J. (2005b). Dissociable elements of human foresight: a role for the ventromedial frontal lobes in framing the future, but not in discounting future rewards. Neuropsychologia, 43, Floden, D., & Stuss, D. T. (2006). Inhibitory control is slowed in patients with right superior medial frontal damage. Journal of Cognitive Neuroscience, 18, Garavan, H., Ross, T. J., Murphy, K., Roche, R. A. P., & Stein, E. A. (2002). Dissociable executive functions in the dynamic control of behavior: Inhibition, error detection, and correction. Neuroimage, 17, Heaton, R. (1981). A manual for the Wisconsin Card Sorting Test. Odessa: Psychological Assessment Resources. Hillis, A. E., Newhart, M., Heidler, J., Barker, P. B., Herskovits, E. H., & Degaonkar, M. (2005). Anatomy of spatial attention: Insights from perfusion imaging and hemispatial neglect in acute stroke. Journal of Neuroscience, 25, Horne, J. A., & Ostberg, O. (1976). A self-assessment questionnaire to determine morningness eveningness in human circadian rhythms. International Journal of Chronobiology, 4, Kahneman, D., & Tversky, A. (1984). Choices, values, and frames. Americal Psychologist, 39, Kaplan, E., Goodglass, H., & Weintraub, S. (1983). Boston Naming Test. Philadelphia: Lea & Febiger. Lhermitte, F. (1983). Utilization behavior and its relation to lesions of the frontal lobes. Brain, 106, Loewenstein, G., & Prelec, D. (1992). Anomalies in intertemporal choice: Evidence and an interpretation. Quarterly Journal of Economics, 107,