Decision-making involves the outcome of cognitive processes leading to a choice between alternative courses of action. Poor decision-making has been described as “deciding against one’s best interests and inability to learn from previous mistakes, with repeated decisions leading to negative consequences” (Bechara, 2005).

A commonly used measure of decision-making is the Iowa Gambling Task (IGT) (Bechara et al., 1994). The IGT was originally developed to measure decision-making in patients with focal brain injury. Patients with prefrontal brain damage were noted to make poor choices, especially in social interactions, leading to negative consequences in spite of having otherwise intact intellectual function. Using the IGT, Bechara et al., have shown that the real life problems with decision-making in patients with ventromedial prefrontal cortical injury are paralleled by their performance on the IGT. Subjects with brain lesions in these areas repeatedly respond disadvantageously, making few changes in their behavior in spite of the losses that accumulate (Bechara et al., 2000). Thus, disadvantageous decision-making as measured by the IGT overlaps conceptually with lack of planning, which is a critical component of impulsivity. The IGT has also been used to study other clinical populations that are known to have problems with decision-making and impulsivity, including subjects with pathological gambling (Goudriaan et al., 2005) and substance abuse (Bechara and Damasio, 2002; Stout et al., 2004; Verdejo-Garcia et al., 2007a, b). Most of these studies find that performance on the IGT parallels problems with real life decision-making.

Since its development, the IGT has looked at decision-making in terms of overall net gain from choices, though its authors acknowledge the role of individual components of the decision-making process such as sensitivity to gains and losses and learning from past mistakes. Like impulsivity, there are many aspects that shape decision-making including expected outcomes, memory, and learning. In order to thoroughly assess the relationship between impulsivity and decision-making, it is necessary to examine the different processes involved in decision-making. One method that has been used to examine these processes in the IGT is the expectancy-valence learning model or EVM (Busemeyer and Stout, 2002). Using this model, each trial block’s gains and losses are assigned a valence, with expectancies regarding each valence learned by an adaptive learning mechanism. Probability of choice on trials can then be selected by looking at expectancies based on previous decisions. The EVM has three parameters that represent stages in the decision-making process: valence of gains/losses, expectancies about outcomes, and consistency of choices with regard to valence and expectancy. Each of these components allows for a more specific measure of different aspects of choice by isolating attention and learning rate. These parameters may correspond in different ways to components of impulsivity.

Several previous studies have examined IGT performance and response inhibition in cocaine-dependent subjects. Outcomes in these studies have not conclusively linked IGT performance to response inhibition tasks, though many have reported poor decision-making and increased impulsivity in cocaine dependent subjects, and in substance-dependent individuals more broadly. Stout et al., found that cocaine-dependent subjects showed a preference for the disadvantageous decks relative to control subjects, and that these choices appeared to be related to motivational and choice consistency factors (Stout et al., 2004). Hanson et al., (2008) compared MDMA (3,4 methylenedioxymethamphetamine, also know as ecstasy) users with other drug users and controls on outcomes from BIS-11 and IGT, finding that both study groups tended to have increased impulsivity and fewer advantageous choices on IGT (Hanson et al., 2008). Both groups tended to show dose-related effects, with heavier use correlating with more impulsivity and more decision-making deficits (Hanson et al., 2008). Another study compared outcomes on IGT and various response-inhibition tasks in cocaine and heroin polysubstance abusers. Findings showed that cocaine polysubstance abusers had significant deficits in response inhibition as measured by go no-go and Stroop tasks compared to heroin polysubstance abusers, additionally both groups made less advantageous choices on the IGT than controls (Verdejo-Garcia et al., 2007b) In a recent report, cocaine-, methamphetamine- and alcohol-dependent men and women were compared to controls by performance on the IGT. This study showed fewer advantageous choices on IGT with methamphetamine- and cocaine-dependent individuals, as well as gender differences in performance, with methamphetamine- and cocaine-dependent women making more disadvantageous choices than men (van der Plas et al., 2008). Additionally, this study assessed response inhibition in these groups via stop-signal task, and did not find a correlation between categorical drug dependence and response inhibition, or a correlation between decision-making and response inhibition (van der Plas et al., 2008).

In summary, previous studies in substance abusing individuals have shown higher scores on questionnaire measures of impulsivity, higher rates of impulsive errors on behavioral inhibition tasks, and poorer decision making as measured by the IGT, however few studies have examined the relationship among all these constructs.

As shown in Table 1, cocaine dependent subjects had significantly higher BIS total scores as well as higher scores on each of the BIS subscales compared to controls. In addition, cocaine-dependent subjects had significantly higher commission errors on the IMT. The repeated measures ANCOVA for the IGT total score showed a significant between subjects-group effect (F =6.21, df = 1, 83, p = 0.015), but no significant effect of Block (F = 0.528, df = 4, 332, p = 0.715) and no block by group interaction (F = 0.561, df = 4, 332, p = 0.691). The overall pattern of responding showed that control subjects showed a greater preference for the advantageous decks compared to cocaine-dependent subjects (See Figure 1). On EVM analysis, the consistency parameter was significantly lower on cocaine subjects than controls on Mann-Whitney U test (z = −2.31, p = 0.021). Parameters of valence and update were not significantly different in the cocaine-dependent population(z = −0.540, p = 0.589, and z = −1.128, p = 0.259, respectively). For complete results of the EVM parameter values please see Table 3 in the online appendix.

On the IGT, cocaine-dependent subjects shifted to advantageous decks more slowly and to a lesser degree than controls. This is compatible with previous findings for the IGT in cocaine users (Verdejo-Garcia et al., 2007a). In the EVM analysis, the consistency parameter was lower in cocaine-dependent subjects. Valence (motivation) and updating (learning) were not significantly different in cocaine-dependent subjects. These findings are partially consistent with those from Stout et al. 2005, which found consistency (choice) and valence (motivation) to be significantly different between cocaine users and controls. There were no significant correlations between EVM model parameters and BIS or IMT/DMT.

As seen in previous studies examining outcomes of behavioral paradigms assessing attention, impulsivity, and response inhibition in cocaine dependent subjects, cocaine dependent individuals showed deficits related to increased impulsivity, decreased response inhibition and decreased attention (Moeller et al., 2005, Garavan et al., 2008, Colzato et al., 2007, Lane et al., 2007). Previous studies have also shown that decision-making as assessed through various paradigms is influenced by cocaine use and dependence (van der Plas et al., 2008, Green et al., 2009, Verdejo-Garcia et al., 2007a & b, Strickland et al., 1993). However, previous studies have not assessed correlations between these various measures of impulsivity, response inhibition and decision-making in this specific population. Previous studies have looked at “substance dependent populations” (multiple substance types) or cocaine users as a subset of other populations, with fewer looking at cocaine dependent populations specifically. Previous studies by Stout et al. employing EVM analysis of IGT performance in cocaine dependent subjects have shown similar outcomes on EVM parameters, but had smaller samples of cocaine dependent subjects (Stout et al., 2004a).

Analyzing our findings in light of previous EVM analysis in cocaine dependent subjects (Stout et al., 2004a), our findings partially replicate the previous findings, which showed consistency and valence to be lower in cocaine dependent individuals compared to controls. Our findings are also similar to findings in patients with bilateral ventromedial prefrontal cortex (vmPFC) lesions, who had decreased degree of consistency and update rate on EVM analysis (Yechiam et al., 2005). This may represent prefrontal gray matter volume loss observed in substance dependent people (Tanabe et al., 2009).

The results of examination of correlations between questionnaire measures of attentional, motor, and nonplanning impulsivity and response inhibition as well as decision-making joins a group of previous studies examining the relationship between different components of impulsivity and decision making in different subject groups. Previous studies have also described a non-correlating or partially correlating relationship between measures of impulsivity and decision-making. Examining the relationship between similar decision-making tasks (Delay Discounting Procedure or DDP, Gambling Task or GT, and Rogers Decision-Making Task or RDMT) in substance abusers, Monterosso et al. found GT performance correlated with DDP performance, response time on RDMT correlated with GT and DDP performance, and no task was observed to correlate with measures of impulsivity (Monterosso et al., 2001). They attribute these findings to the similarity of the constructs they were measuring; risk, delay, and impulsivity, which overlapped with one another, but may require more refinement in their assessment. A correlation between measures of trait impulsivity and commission errors on a behavioral measure of motor impulsivity (go/no-go task) and lack of correlation between trait impulsivity and decision-making, delayed discounting, reflection impulsivity and self-regulation was also observed in a sample of high and low trait impulsivity women (Perales et al., 2009). There have been other studies that have described a relationship between measures of impulsivity and decision-making. In a study comparing 50 adults with ADHD with 51 controls, there was a negative correlation with BIS-11 nonplanning and net scores on IGT second, third, fourth and fifth blocks, a negative correlation between IGT total net score and BIS-11 nonplanning subscale, and BIS-11 total score negatively correlated with the difference between advantageous and disadvantageous choices in the second and third block of the IGT (Malloy-Diniz et al., 2007). The difference in correlation between IGT and BIS-11 in this population compared with our cocaine-dependent sample may result from differences in the presentation of impulsivity, nonplanning and decision-making in these two populations, as well as inclusion of combined- type ADHD in that study. To compare the findings of the current paper to those of Malloy-Diniz, exploratory correlation analyses were performed on all 5 blocks of the IGT with the other measures (BIS-11, IMT). After correction for multiple comparisons, this analysis revealed no statistically significant correlations. The relatively small size of the control group (N=20) may have contributed to the null findings. Future studies would benefit from larger control sample sizes.

Evaluation of substance abuse and pathological gambling via different measures of impulsivity, decision making, sensation seeking and time orientation revealed that substance abusing subjects did poorly on the IGT, but scores did not correlate with performance on impulsivity measures such as BIS-11 (Petry, 2001). Comparisons of performance across the IGT, the Wisconsin Card Sorting Task (a measure of set-shifting), a delayed discounting procedure, and the BIS-11 revealed that higher impulsivity was associated with poorer performance on the IGT, but was not related to card sorting performance in nicotine dependent individuals. (Sweitzer et al., 2008). In a college student sample, self-reported impulsivity and decision making on the IGT were examined. Only lack of premeditation was correlated with performance on the IGT (Zermatten et al., 2005).

Limitations of this study include differences in education level in our cocaine dependent and control groups, noting that education was a covariate in the analyses. In regards to the difference in education between the groups, ANCOVA scoring was based on SCID-I education rating (1 = grade 6 or less, 2 = grade 7–12, 3 = high school graduate, 4 = education beyond high school). On further analysis, the level of education with most frequency in the cocaine dependent group was 3, or high school graduate (50%). For controls, it was 4, or partial college (35%). This signifies that even though there was some difference in level of education, education levels of completion were not highly discrepant (e.g., only a minority of cocaine-users or controls were college graduates). Importantly, this study used a control group closely matched in age and recruited from identical zip codes, thereby approximating a match to the cocaine dependent group in sociodemographics. Notably, IGT performance in the control group may appear inconsistent with performances of control subjects in prior studies (specifically, inconsistent performance across IGT blocks and a large standard deviation even in blocks 4 and 5). This may be due to our control group being recruited from the same zip codes as the cocaine-dependent group, and therefore more comparable in regards to socioeconomic status to our cocaine group than to control groups in previous studies comprised of middle-class adults or college students. Using multinomial logistic model, Fuentes et al. (2006) distinguished pathological gamblers from controls with regression models that included both Go No-Go and BIS-11 scores. With larger datasets, similar approaches may be useful in future studies to examine combinations of factors (e.g., IGT EVM parameters, BIS-11 parameters, IMT outcomes) that may be revealing with regards to differences in cocaine dependent and control populations.

We found no significant correlation between EVM parameters of the IGT (meant to distinguish the underlying processes of decisions made on the IGT) with either measures of response inhibition and attention (IMT), or self-reports of impulsivity (BIS-11). One possible reason is that while similar, the IMT and IGT are measuring behavioral processes with unique features and sources of variance. Additionally, to more completely examine the relationship between decision-making, impulsivity, and behavioral inhibition, future studies might benefit from measures such as reflection impulsivity (the ability to gather information prior to making decisions) to further elucidate the processes of nonplanning and decision-making (Clark et al., 2006). The concept of reversal learning, or the inability to terminate repetitive responses in the presence of previous learning or environmental cues may also play a role in decision-making (Fillmore and Rush, 2006). One study found that in a population of polydrug users (cocaine and heavy alcohol use), subjects displayed acquisition of learning, but impaired discrimination-reversal learning of inhibitory and activational responses compared to controls (Fillmore and Rush, 2006). It is possible that reinforcement contingencies may also play a role in the inability to reverse responses to stimuli in cocaine dependent subjects (Ersche et al., 2008). Further examination of measures of decision-making with reversal learning in cocaine-dependent populations may be worthwhile to better elucidate cognitive deficits in this population.

01