The present study used a recently developed IST to measure reflection impulsivity in current and former ecstasy users, cannabis users and drug-naïve controls. In the Fixed Reward condition (where there was no penalty for sampling further information), the drug-naïve control group sampled information to a point of 85% certainty (95% confidence intervals: 80–89%), similar to healthy performance in our previous study (Clark, et al., 2006). Moreover, the number of boxes opened and the level of certainty tolerated during decision-making were both inversely correlated with incorrect judgments in the overall sample (n = 94). This demonstrates the central feature of a test of reflection impulsivity: the extent of information sampling is predictive of eventual decision accuracy (Evenden, 1999a).

Regular cannabis users sampled significantly fewer boxes on the IST and tolerated more uncertainty in making the correct decision, compared with the other groups. This difference was statistically significant in a planned comparison against the drug-naïve control group, and the cannabis users sampled significantly less information than the Ex-ecstasy group in the more conservative Tukey’s post hoc group comparisons. The cannabis users altered their information sampling behaviour to a similar degree between the Fixed Reward and Reward Decrement conditions, compared with the other groups (i.e. the nonsignificant Group × Condition interaction term). This indicates comparable sensitivity to the change in reward contingencies, and suggests that the reduced information sampling behaviour was not simply attributable to a lack of motivation in the cannabis group. These data are also consistent with a report of risky decision-making (on the Iowa Gambling Task) in regular marijuana users (Whitlow, et al., 2004). Decision-making impairments on complex tests like the Iowa Gambling Task may putatively arise from a failure of pre-decisional information sampling or evaluation. Our findings extend our earlier observation of reduced information sampling in current and former users of amphetamines or opiates, who met DSM-IV criteria for dependence (Clark, et al., 2006). Given the presence of this effect across multiple substances of abuse with distinct pharmacological targets (amphetamines, opiates, cannabis), we suggest that impaired reflection impulsivity may represent a cognitive style associated with the pre-existing vulnerability to recreational drug use and later dependence, consistent with data from high-risk prospective studies (Nigg, et al., 2006; Tarter, et al., 2004).

We were unable to detect any significant group differences on the IST between the ecstasy-using groups and the drug-naïve controls. Several other studies have failed to substantiate the link between ecstasy use and other aspects of impulsivity, including the Stroop and Go/No Go tests (Dafters, 2006; Fox, et al., 2002; Gouzoulis-Mayfrank, et al., 2003). However, our findings fail to replicate several studies that have demonstrated impulsivity in regular ecstasy users on another widely used test of reflection, the MFFT (Morgan, 1998; Morgan, et al., 2002; Morgan, et al., 2006; Quednow, et al., 2007). The ecstasy users in the present study reported moderate use of other illicit substances, including similar cannabis usage to the cannabis group. The two ecstasy groups were also more likely than the cannabis group to have used a range of other substances, including amphetamine, cocaine and opiates. Consequently, if reduced reflection is a pre-existing cognitive style associated with general recreational drug use, we would expect this effect to have also been present in the two ecstasy groups. Lack of statistical power seems unlikely to explain the negative result, as the ecstasy groups actually sampled more information (in terms of boxes opened), on average, than the drug-naïve controls. In addition, the group size of 46 current ecstasy users is reasonably large for studies of this kind, and the level of ecstasy consumption was considerable (e.g. lifetime usage means of 609 and 1001 in the Current and Ex-ecstasy groups respectively), compared with the wider neuropsychological literature [e.g. 458 tablets in Quednow, et al. (2007)].

There are several possible explanations for the discrepancy with the studies by Morgan, et al. (1998, 2002, 2006), and Quednow, et al. (2007). One consideration is the duration of abstinence from ecstasy, which was relatively long in the present study (> 3 weeks) but much shorter in the positive studies, ranging from 3 days (mean 17 days; Quednow, et al., 2007) to 5 days (Morgan, et al., 2005). Studies in experimental animals reveal recoverable reductions in serotonin function 1–2 weeks after dosing that do not indicate neurotoxicity (Gouzoulis-Mayfrank & Daumann 2006). In addition, studies of other drugs indicate that short-term withdrawal may exacerbate behavioural impulsivity (see below). Lyvers and Hasking (2004) recommended a 1-month abstinence window for neuropsychological studies. Hence, the positive MFFT results by Morgan, et al. and Quednow, et al. could be caused by semi-acute effects of serotonin depletion upon task performance.

We have shown previously that IST performance is related to MFFT performance in healthy volunteers: fast, inaccurate responders on the MFFT opened significantly fewer boxes on the IST than slow, accurate responders (Clark, et al., 2003). However, the MFFT involves a number of extraneous additional processes, including visual search, visual working memory and strategy implementation, which may be independently impaired in regular ecstasy users (Fox, et al., 2002; Halpern, et al., 2004; Wareing, et al., 2005). The design of the IST explicitly aimed to minimize these extraneous demands. In the MFFT study by Morgan (1998), there was a group difference in MFFT accuracy but not latency, which may be plausibly explained as a more general impairment. Other studies, however, reported significant differences in both speed and accuracy (Morgan, et al., 2006; Morgan, et al., 2002), which is likely to indicate impulsivity.

Additional factors may mediate the deficits in laboratory impulsivity in ecstasy users, and contribute to variability across studies. Gender may be one such variable: in the present study, the reduced information sampling in the cannabis group was mainly attributable to the male subjects, and other studies also described greater neuropsychological impairments in male drug users than in female drug users (Ersche, et al., 2006; Stout, et al., 2005), including ecstasy users (von Geusau, et al., 2004). In addition to gender, studies of other groups of drug-users with the delay discounting paradigm have indicated greater impulsivity in current users compared with ex-users (Bickel, et al., 1999; Petry, 2001). Two distinct mechanisms may contribute to this effect: withdrawal and/or craving may exacerbate impulsivity in current users (Field, et al., 2006; Giordano, et al., 2002), but also, less impulsive drug users may be more capable of achieving successful abstinence (Bickel, et al., 1999). In the present data, there was no evidence of the latter effect, as the Current and Ex-ecstasy users scored similarly on the IVE and IST measures. It is possible that the periods of abstinence from ecstasy in the Current (>3 weeks) and Ex (>1 year) ecstasy groups had attenuated impulsivity compared with the cannabis group, although against this explanation, the ecstasy groups did report moderate recent usage of other substances, including similar cannabis usage to the cannabis group in the past month.

Whilst we found no evidence of laboratory impulsivity on the IST in the ecstasy group, self-reported impulsivity on the Eysenck IVE questionnaire was significantly elevated in the current and former ecstasy users. Questionnaire impulsivity should indicate trait dispositions that are present prior to the initiation of drug use; for example, de Win, et al. (2006) showed no change on the Barratt Impulsiveness Scale before and after initiation of ecstasy use in a prospective cohort. Previous studies suggest large variability in trait impulsivity in ecstasy users, with a number of studies reporting elevations (Butler and Montgomery, 2004; Morgan, 1998; Parrott, et al., 2000), but other studies finding no differences (Travers and Lyvers, 2005) and one study even finding a significant reduction (McCann, et al., 1994). In our data, there was no association between the IVE score and performance on the IST. These data highlight the multi-factorial nature of impulsivity, and are in keeping with a number of other reports showing limited associations between state (laboratory) and trait (questionnaire) measures of impulsivity (Dom, et al., 2007; Lijffijt, et al., 2004; Reynolds, et al., 2006). Questionnaire ratings indicate general behavioural tendencies across a variety of situations, and rely on a subjective perception of one’s behaviour. In contrast, laboratory tasks provide an objective measure of a specific facet of impulsivity at a single point in time. Weak correlations between these two sets of variables may be a realistic expectation. Similarly, our findings do not refute the possibility that other domains of laboratory impulsivity (e.g. delay discounting, response inhibition) may be impaired in ecstasy groups. As discussed above, there are inconsistent findings using tasks of response inhibition in ecstasy users (Dafters, 2006; Fox, et al., 2002; Gouzoulis-Mayfrank, et al., 2003), and to our knowledge, no studies have yet explored delay-discounting in regular ecstasy users.

Some further limitations of the present study should be noted. Whilst the number of current ecstasy users was large, the group sizes for the former ecstasy users and the cannabis users were considerably smaller. In particular, the analyses split by gender should be treated as preliminary due to the reduced power, and need to be confirmed in a larger sample. In addition, the two groups of ecstasy users showed a high degree of polydrug use, although this arguably renders their intact IST performance even more surprising.

In conclusion, these data support the position of reflection impulsivity as a relevant cognitive dimension in regular drug users, by demonstrating reduced information sampling in a group of regular cannabis users. Reduced reflection is likely to have a detrimental impact on wider-scale decision-making capabilities, with potential relevance for treatment engagement and the ability to maintain long-term abstinence. Unexpectedly, the present study found no differences in reflection in current or former ecstasy users, despite evidence of trait impulsivity in these subjects. These data appear to challenge a simplistic pathway from ecstasy consumption to elevated impulsivity via serotonin neurotoxicity.