Phylogenetic Conservation of Individual Differences in Impulsivity

Although impulsivity, in most of its forms, is typically envisioned as inefficient, unadaptive or suboptimal, this is clearly not universally the case. The remarkable conservation of “impulsive” traits across phylogeny suggests that manifestations of rapid action, quick decision making, and hasty reward seeking can represent a phenotypic advantage for many species. This is true across vertebrate species, at the least. The observation that the rapid, inflexible approach to potentially risky social stimuli in adolescent vervet monkeys predicts adult dominance to a significant, albeit limited, degree is a particularly powerful illustration of this point (Fairbanks et al., 2004a). Although the full impact of laboratory studies of impulsive temperament in rodents must be considered in light of the restricted phenotypic and genotypic variability that has been attained through selective breeding, as well as because of the very different manifestations of impulsivity that may stem from the different ecological niche occupied by rodents, they nevertheless represent efficient models for investigating the neural systems underlying aspects of impulse control. Many species of nonhuman primates, including those commonly used in laboratory investigations, retain a good deal of natural variation in these traits (Fairbanks, 2001; Fairbanks et al., 2004b; Higley and Linnoila, 1997; Manuck et al., 2003). Hence, these animals represent a unique opportunity to investigate the genetic, molecular, and neural basis of impulsivity and its neurocognitive correlates (James et al., 2007) in a manner that may directly inform our understanding of these processes in humans.

It appears that genetic and neurochemical determinants of impulsive temperament are conserved at least across Old World monkeys and humans (Bailey et al., 2007; Forbes et al., 2009; James et al., 2007; Keltikangas-Jarvinen et al., 2003). Remarkably, a 48-basepair variable number tandem repeat polymorphism in exon 3 of the dopamine D4 receptor (DRD4) gene is found across a broad number of mammals, and this similar allelic variant is responsible for phenotypic variation in aspects of impulsive temperament in several species. This result further endorses the notion that the forces acting upon impulsivity are phylogenetically very old, using common mechanisms in species having diverged millions of years ago. Although genetic factors mediating impulsivity may not be uniformly conserved across all mammals, there is considerable evidence suggesting that shared neurochemical mechanisms contribute to these processes. For example, relatively low levels of brain serotonin turnover (as indicated by cerebrospinal fluid levels of 5-hydroxyindoleacetic acid, the primary acidic catabolite of serotonin) are predictive of a variety of impulsive behaviors in both monkeys and humans (Brown and Linnoila, 1990; Fairbanks et al., 1999, 2001; Faustman et al., 1991; Spreux-Varoquaux et al., 2001; Westergaard et al., 2003). Clearly, this particular phenomenon dovetails nicely with the observation that some aspects of impulsivity in rodents are also affected by serotonergic manipulations (Soubrie, 1986; Winstanley et al., 2004b), showing how rodents can be used to understand mechanistic relationships discovered in more complex species in a directed and hypothesis-driven way.

Fractionating Impulsivity

In keeping with the diverse nature of impulsive behaviors, there are currently a variety of very different tests used in both human and nonhuman subjects to measure forms of impulsivity, leading some to question whether “impulsivity” is even a valid or useful construct. However, its nonunitary nature is actually more typical than atypical when comparing it with other psychological concepts. For example, there are many different classes of memory, short-term, long-term, iconic, episodic, semantic, procedural, verbal, working, reference, spatial, to name but a few (see Kesner and Rogers, 2004; Squire, 2004 for recent reviews). In this way, memory is an umbrella term that usefully covers these subclasses of mnemonic processes, and the fact that these mental faculties are distributed across different brain regions does not invalidate the term.

It may be important to note that isolating the exact nature of the memory or attentional deficit associated with certain psychiatric or neurological condition has improved, rather than hindered, diagnostic tools (Brooks and Baddeley, 1976; Budson, 2009). In addition, it has been possible for research into the nonunitary nature of such processes to contribute to more fundamental hypotheses regarding how the brain works. For example, the realization that there is no one “memory centre” supports the view that our mental operations are conducted within a distributed network of “multi-tasking” brain regions rather than a discrete “one region—one function” model (e.g. Poldrack and Foerde, 2008). Rather than fearing the plethora of tasks claiming to measure impulsivity, researchers may be better served by embracing the diversity of impulsive behaviors and considering what can be learned from their comparison.

The majority of research into impulsivity at a clinical level involves the use of self-report questionnaires designed to measure impulsive tendencies, the most widely used of which is the Barratt Impulsivity Scale (BIS-11) (Patton et al., 1995). Factor analysis of the output from this 30-item scale indicates that 3 separate dimensions of impulsivity can be identified: attentional, motor, and nonplanning. The attentional/cognitive domain is reflective of the degree to which an individual can focus on the task at hand or tolerate cognitive complexity, the motor component reflects spontaneity or action without due consideration, whereas nonplanning impulsivity reflects a lack of regard for the future. Although useful indicators of cognitive well-being, self-report questionnaires are always limited by the subjects’ insight and view of the self (Moeller et al., 2001a). Behavioral tests of impulsivity have therefore been developed which are a better medium for neurobiological experiments, in terms of both clinical fMRI and pharmacological challenges and preclinical modeling studies. Such tests use as their base the ideas encapsulated within the self-report questionnaires. However, despite the best efforts of researchers in the field, it is not always possible to provide a 1:1 relationship with the dimensions of impulsivity extracted from the BIS-11 and the psychological processes measured in a behavioral paradigm.

Behavioral tasks that measure impulsivity have often been classified into two, rather than 3, camps: those measuring impulsive action (motor impulsivity) and those measuring impulsive choice or decision making (Dalley et al., 2008; Pattij and Vanderschuren, 2008; Winstanley et al., 2006). A diagram illustrating how distinct aspects of impulsivity are reflected in different behavioral paradigms is provided in Fig. 1. The concept of impulsive action, or behavioral disinhibition, maps well onto the BIS-11 factor of motor impulsivity, and there are a number of behavioral tasks that appear to provide valid measurements of this phenomenon in both human and nonhuman subjects. Nevertheless, it may be unwise to consider them all as measuring identical processes. In contrast, the concept of impulsive choice appears to span both the cognitive/attentive and nonplanning domains of the BIS-11 (Patton et al., 1995). However, the majority of tests used to assess this construct focus on delayed gratification and discounting functions and therefore probably reflect non-planning more directly. The inability to process complex tasks overlaps with a wide range of executive functions and is therefore difficult to isolate. The matching familiar figures test of “reflection” impulsivity is probably the task most directly aligned with this BIS-11 factor. However, this is fundamentally difficult to model in animals (see Evenden, 1999) which has hindered exploration of its biological basis from this perspective. Indeed, the degree to which the application of “cognitive effort” or the analysis of cost-benefit decisions overlaps with impulsivity is an open question that merits further discussion (see Floresco et al., 2008; Pattij and Vanderschuren, 2008 for reviews of both literatures).

It is worth noting, however, that dimensions of temperament or personality are likely to be secondary to individual differences in more basic psychological processes, including aspects of cognition, reward sensitivity, and approach/avoidance. In both humans and monkeys, it is now clear that individuals exhibiting relatively high naturally occurring impulsivity exhibit poor working memory capabilities (Cools et al., 2007; Dowson et al., 2007; James et al., 2007; Romer et al., 2009) and that this association may be mediated by impairments in the function of the frontal cortex (Cools et al., 2007). Perhaps the basic mechanism mediating this relationship is that poor working memory leads to deficiencies in developing and maintaining goal-directed plans related to behavior—the result being a more myopic behavioral phenotype. If correct, this would suggest that working memory and its related component processes should also be considered as a focus on endophenotype-oriented working for disorders of impulse control (Loo et al., 2008). Of particular relevance to the current review, recent data suggest that working memory performance is impaired following chronic, escalating administration of cocaine, and that such impairments are correlated with cell loss within the frontal cortex (George et al., 2008). Furthermore, the authors suggest that the working memory impairments observed may reflect aspects of increased perseveration, compulsivity, and incentive motivation induced by repeated drug intake, and these processes may likewise influence impulsive behaviors.

IMPULSIVE ACTION VERSUS IMPULSIVE CHOICE

The level of motor impulsivity has largely been conceptualized as the inability to withhold from making a motor response. Three of the most widely used laboratory-based tests used to assess this process are the stop-signal reaction time task (SSRTT), the continuous performance test (CPT), and the go/no-go paradigms. In the SSRTT, subjects are required to make a rapid response upon the presentation of a go signal (Logan and Cowan, 1984). On a subset of trials, a stop signal is presented at varying delays after the presentation of the go signal, and the subject must then cancel their planned response. Go/no-go paradigms likewise use 2 signals, one indicating that a go response is required, and the other that this response should be inhibited. However, only one signal is presented on any trial, and the signal occurs at the beginning of each trial, before any action has been initiated. Both SSRTT and go/no-go tasks have been developed for use in rodents and nonhuman primates (SSRTT: Eagle and Robbins, 2003; Feola et al., 2000; Liu et al., 2009; go/no-go Iversen and Mishkin, 1970; Terman and Terman, 1973).

In contrast, the CPT is ostensibly a test of attentional function that incorporates an aspect of motor impulsivity (Rosvold et al., 1956; Wilkinson, 1963). In this task, subjects are required to scan a 5-digit sequence and respond when the numbers match a target stimulus. “False alarm” errors occur when the subject responds positively to a sequence which is identical to the target, with the exception of the final number. When making such errors, the subject therefore responds prematurely before processing the full sequence, hence providing an index of motor impulsivity. The five-choice serial reaction time task (5CSRT) was developed as a rodent analog of the CPT (Carli et al., 1983). Animals learn to respond to a stimulus light which is briefly presented in 1 of 5 spatially distinct apertures. However, subsequent to initiating a trial and prior to presentation of the light, there is a 5-s inter-trial interval during which the animal must withhold from responding at the aperture array. Any responses made during this time are described as premature responses and are punished by time-out periods. Such premature responses are thought to be analogous to the false alarms errors made in the CPT.

Despite the fact that all 3 of these tests purport to measure motor impulsivity and are often referred to interchangeably, there are subtle but perhaps critical differences in the inhibitory processes involved (see Eagle et al., 2008 for a recent review). In particular, it appears possible to dissociate action restraint (the inhibition of a motor response prior to its initiation, as in the go/no-go paradigm) and action cancelation (the inhibition of an already-initiated motor response, as in the SSRTT) in terms of their neurochemical modulation. As one example, the 5-HT system is strongly implicated in action restraint but not action cancelation, whereas noradrenaline may play a more fundamental role in action cancelation. Furthermore, although performance of action restraint and cancelation tasks was correlated in control subjects, there was no such relationship between task scores in a group of attention-deficit hyperactivity disorder (ADHD) patients, despite the fact that ADHD patients were collectively worse on both inhibitory measures (Schachar et al., 2007). Such findings suggest that patients with different inhibitory deficits may obtain therapeutic benefit from different pharmacological treatments and highlights the importance of considering individual differences in impulsive symptomatology. Within this framework, premature responding on the 5CSRT may have more in common with errors on go/no-go tasks rather than deficits in responding to stop signals on SSRTT. However, in the 5CSRT there is no “no-go” signal: the subjects regulate their own behavior in anticipation of the presentation of a “go” signal. Whether this turns out to be an important factor when considering translation between tasks and species remains to be determined, but the response to reward-related stimuli is clearly affected by addictive drugs.

In addition to these motor-based paradigms, impulsivity is also reflected in more cognitive tests of choice and preference. One of the most widely used tests of impulsive decision making in both animals and humans is the delay-discounting paradigm. Delay discounting describes the process by which the subjective value of a reward diminishes as the delay to its delivery increases, and rates of discounting are assessed by measuring choice between smaller rewards delivered relatively soon versus larger rewards delivered later in time (e.g. Rachlin et al., 1991). Selection of the smaller-sooner reward is indicative of a more impulsive choice, reflecting an intolerance to delayed gratification that results in a negative outcome i.e. less reward in the long run. The delay-discounting curves for both human and nonhuman subjects are best represented by the same form of hyperbolic equation (Ainslie, 1975; Mazur, 1987), and there is a general concordance between the effects of brain damage and drug administration on delay-discounting behavior between species (see Winstanley et al., 2006 for a review).

Just as delay to its delivery can devalue a larger reward, so can the probability that such a reward will actually be received. It has been suggested that perceived uncertainty regarding the stability of an environment can lead to enhanced delay discounting, as the delayed reward appears probabilistic rather than guaranteed (Mischel and Grusec, 1967). The question of whether decision making related to delayed versus probabilistic rewards is controlled by similar psychological and neural processes can be addressed empirically using simple probability-discounting schedules (Adriani and Laviola, 2006; Cardinal and Howes, 2005; Mobini et al., 2002) and also more complex gambling-like paradigms (e.g. Rivalan et al., 2009; Zeeb et al., 2009). The extent to which decision making under risk or uncertainty is analogous to delay-discounting tests more commonly used to assess impulsive decision making is an important question, particularly given both the association with pathological gambling, impulsivity, and substance abuse. Whether engaging in impulsive-like behaviors is itself rewarding and the contribution such processes make to the development of pathological gambling are likewise key issues in terms of understanding the relationship between behavioral and chemical addictions. However, in the interests of brevity and focus, we will restrict our discussion to the more conventional, delay-discounting models of impulsive choice, as more data is available regarding the relationship between impulsive choice on such models and the response to addictive drugs.

WHY MIGHT ADDICTIVE DRUGS INFLUENCE IMPULSIVITY AND VICE VERSA?

One of the reasons why addictive drugs exert such a powerful control over the behavior of people affected by substance use disorders is that the dopamine response to drug-paired cues is potentiated, such that these cues trigger craving for drug and drug seeking (Bossert et al., 2005; Everitt et al., 2003; Parkinson et al., 2001, 1999; Shaham et al., 2003; Taylor and Robbins, 1986; Wyvell and Berridge, 2000). Both acute and repeated administration of cocaine and other addictive drugs leads to the enhancement of conditioned reinforcement in rodents (Jentsch and Taylor, 2001; Mead et al., 2004; Olausson et al., 2004; Taylor and Horger, 1999), see also (Mead et al., 2004; Miles et al., 2004; Wyvell and Berridge, 2001). This is also true in nonhuman primates as monkeys that had previously been exposed to cocaine showed augmented responding to a reward-associated cue when tested on the acquisition of a second-order schedule of reinforcement (Olausson et al., 2007). Conditioned stimuli (lights, tones, etc.) are also thought to make a significant contribution to the addictive nature of gambling behavior (Kushner et al., 2008). Understanding how manipulation of the salience of reward-paired cues influences different forms of impulsive responding may prove fruitful when considering how addictive drugs influence impulsivity under different circumstances. For example, it has been suggested that animals that respond more to reward-paired cues in autoshaping (“sign-tracking”) paradigms, theoretically as a result of individual differences in the attribution of incentive salience to cues, are more sensitive to cocaine and may likewise be more impulsive (Flagel et al., 2009; Tomie, 1996).

With these considerations in mind, it may be particularly important to understand what drives impulsive behavior within different tests of impulse control, particularly if impulsive responding is generated by the reaction to an external cue. For example, it could be argued that inhibitory failure in the SSRTT and go/no-go task arise through the failure to process or attend to the stop/no-go signal, whereas premature responding in the 5CSRT is “self-generated”: there is no inhibitory signal that is ignored. Alternatively, elevated impulsive responding in all 3 tests could arise through a hyper-response to the perceived or anticipated “go” signal. Drug-induced potentiation of the power of reward-related cues to elicit a behavioral response may therefore directly contribute to high impulsivity on these motor tasks.

It is also worth considering how such a heightened response to reward-paired cues might influence delay-discounting processes. When an animal selects the small immediate reward over a larger delayed one, such a decision is regarded as an impulsive choice, because the desire for immediacy outweighs the larger reward on offer following a delay. However, it could also be argued that elevated choice of the larger reward is indicative of an increase in incentive motivation for the large reward, such that the delay fails to sufficiently devalue its subjective value, and it is selected despite the negative consequences (Uslaner and Robinson, 2006; Winstanley et al., 2004a). This view shares similarities with the idea that persistent drug-taking results in a short-sighted view of the future such that decisions are made with little regard for their future consequences. If choice of the large reward is signaled by a cue, immediate feedback is provided that could be amplified by exposure to psychostimulant drugs, or alcohol (Grant and Macdonald, 2005; Olmstead et al., 2006). It has been observed that using a cue light to bridge the gap between responding for the large reward and its delivery can facilitate the acquisition of delay-discounting performance and alter the response to amphetamine (Cardinal et al., 2000). Recent unpublished findings also suggest that use of such a cue leads to virtually exclusive selection of the larger delayed reward in a subset of individuals, and that this may reflect enhanced dopamine signaling within the OFC (Zeeb et al., 2010; see Floresco et al., 2008 for further discussion of cues in delay-discounting tasks). Whether these “cue-sensitive” rats are like-wise more vulnerable to addictive behaviors remains to be determined.

The inability to withhold from responding to a reward-paired stimulus may also impact the performance of reversal learning tasks, where subjects are required to inhibit responding to a previously rewarded option, and instead select a new or previously unrewarded stimulus (see Birrell and Brown, 2000; Boulougouris et al., 2007; Chudasama and Robbins, 2003; Cools et al., 2002; Dias et al., 1997; Seu et al., 2009 for recent studies using such methodology). Although deficits in shifting away from the previously rewarded stimulus are generally regarded as evidence for cognitive inflexibility or perseverative/habitual responding, such behavior could also reflect a difficulty in inhibiting a prepotent reward-directed response (Groman et al., 2009; Tait and Brown, 2007). Such comparisons may help explain the pattern of behavioral impairments reported following chronic intake of addictive drugs.

HIGH IMPULSIVITY AS A PREDICTOR OF DRUG ABUSE

Given the diversity of impulsive behavioral tests just described, an important question to consider is whether it matters what kind of test is used to probe the relationship between addiction and impulsivity: do all kinds of impulsivity show similar relationships with substance abuse, or is there a particular impulsive subtype that stands out? Furthermore, is there a particular phase in the development of addiction where impulsivity is more prominent, and where interventions aimed at reducing impulsivity may be maximally effective? There is also the similarly critical question of whether the specific drug used or the sex or age of the subjects influences the importance of different impulsive behaviors in the development of addiction. However, for brevity, we will largely restrict our discussion to alcohol and cocaine, as the links between impulsivity and addiction appear particularly strong for these drugs, and their use has been heavily studied.

IMPULSIVITY AND COGNITIVE DEFICITS AS A CONSEQUENCE OF COCAINE ABUSE: FOCUS ON THE OFC

Although high levels of impulsivity appear to predispose individuals to cocaine abuse, data from both rats and monkeys indicate that drug exposure can likewise affect a range of cognitive behaviors, including impulse control, and such behavioral change may be integral to the development and maintenance of addiction. Short-term cocaine exposure is sufficient to increase subsequent cocaine self-administration in rodents (Vezina, 2004), indicating that the cocaine-induced neuroadaptations that occur early during drug experimentation can promote further cocaine consumption. One hypothesis is therefore that drug exposure can dysregulate the inhibitory control functions dependent on regions of the prefrontal cortex (PFC) to facilitate cue-motivated drug seeking and the development of an “addictive” state in vulnerable individuals (Everitt et al., 2007; Jentsch and Taylor, 1999; Robbins and Everitt, 1999; Robinson and Berridge, 2003). A significant challenge is therefore to determine the neurobiological substrates associated with cocaine-induced alterations in behavior and motivation.

In addition to the nucleus accumbens and striatum (e.g., Belin and Everitt, 2008; Calu and Schoenbaum, 2008; Cornish and Kalivas, 2001; Everitt et al., 2008; Nestler, 2004; Robinson and Berridge, 1993), one major focus of such research has been the OFC. Hypoactivity within this region has been observed repeatedly in human cocaine addicts (Volkow et al., 1993, 1991). Cocaine self-administration reduces glucose utilization in the primate OFC and PFC (Beveridge et al., 2006; Lyons et al., 1996), and these alterations progressively impact cortical-striatal regions with increased exposure (Porrino et al., 2007). It has been suggested that this reduced function contributes to impulse-control deficits in addiction (Rogers et al., 1999), as OFC-damaged patients score more highly on the BIS-11 and show high levels of impulsive responding on gambling tasks (Bechara et al., 1999; Berlin et al., 2005) but see (Fellows and Farah, 2005). In rats, damage to this region slows SSRTT and transiently increases premature responses, indicating an increase in motor impulsivity (Chudasama et al., 2003; Eagle and Baunez, 2009). Interestingly, both increases and decreases in impulsive choice have been observed following OFC lesions, which may reflect differences in the design of the delay-discounting tasks used (Rudebeck et al., 2006; Winstanley et al., 2004a; see Floresco et al., 2008 for discussion). In addition, OFC damage across species leads to cognitive inflexibility and perseveration, as evidenced by impairments on reversal-learning tasks (Boulougouris et al., 2007; Chudasama and Robbins, 2003; McAlonan and Brown, 2003; Schoenbaum et al., 2003). Such deficits have also been observed in rats, monkeys, and humans following repeated cocaine intake (Calu et al., 2007; Fillmore and Rush, 2002; Jentsch et al., 2002; Schoenbaum et al., 2004), again indicating that compromised OFC function may be central to the cognitive dysfunction associated with cocaine addiction.

OFC lesions reduce the ability of reward-associated cues to elicit or maintain reward-motivated behavior (Pears et al., 2003) and play an important role in the ability of reward-associated stimuli to influence behavioral output (O’Doherty et al., 2003; Roesch and Olson, 2004; Schoenbaum et al., 1998). The OFC may therefore play a critical role in mediating the motivational effects of drug-associated stimuli. This hypothesis is supported by observations that OFC lesions disrupt cocaine self-administration under a second-order schedule of reinforcement (Everitt et al., 2007; Hutcheson and Everitt, 2003), and reversible inactivation of the lateral OFC reduced the motivational effects of cocaine-associated cues in a relapse model (Fuchs et al., 2004). As outlined above, such increased sensitivity to reward-paired cues could promote drug seeking, but also influence impulsive responding.

In a recent experiment, 5CSRT testing and cocaine self-administration were combined in a concurrent design such that animals self-administered cocaine in the afternoons and were tested on the 5CSRT in the mornings (Winstanley et al., 2009). Thus, it was possible to track the development of any changes in impulsive behavior as drug use developed. During the initial phase of acquisition of cocaine self-administration, premature responding increased, but this was no longer evident after 4 weeks of drug exposure. In contrast, when animals were withdrawn from drug, premature responding sharply increased. Furthermore, these deficits could be exacer-bated by overexpression of the transcription factor ΔFosB, a member of the Fos family of immediate early genes, within the OFC. ΔFosB has previously been associated with the generation and maintenance of the addicted state through its drug-dependent induction within the striatum (Nestler, 2004), but self-administration of cocaine also increases ΔFosB within the OFC (Winstanley et al., 2007). Furthermore, microarray data suggests that this increase leads to upregulation of local inhibitory networks within this region, potentially providing a mechanism by which OFC function is reduced in human addicts (Volkow and Fowler, 2000; Winstanley et al., 2007).

It is worth noting that animals with a history of cocaine self-administration showed a tolerance to the ability of an acute cocaine challenge to increase premature responding, and this tolerance could be induced by overexpression of ΔFosB within the OFC (Winstanley et al., 2009, 2007). Such data somewhat parallels the observation that HI-5CSRT rats become less impulsive following binge access to cocaine (Dalley et al., 2007). One possibility is that the level of impulsivity exhibited by individuals changes as addiction develops, with levels of impulsivity decreasing after initial exposure to drug, potentially, because of adaptive neuroplastic changes that allow the addict to function while taking large quantities of stimulant drugs. However, when drug is withdrawn, impulse control deficits are unmasked as the brain can no longer function optimally without drug onboard (see Winstanley, 2007 for discussion).

Understanding the biochemical changes that occur within the OFC, and other regions of the frontal cortex, after cocaine exposure could potentially provide valuable information into the mechanisms responsible for deficits in impulse control and behavioral flexibility associated with addiction. Using genomics and proteomics, it is possible to identify both the effects of chronic drug exposure and, by simultaneously sampling hundreds or thousands of gene products and proteins, also the global patterns of changes that are more relevant to the neurobiology of addiction than any individual alteration alone. An increasing number of studies have, however, demonstrated limited correlations between changes in mRNA and protein expression using these techniques, indicating that mechanisms that control protein translation are important in determining the biochemical effects of drug exposure. As proteins represent functional components of the cellular machinery, methods for unbiased proteomic analysis therefore offer substantial advantages. Initial proteomics studies in cocaine-exposed monkeys, which show established deficits in reversal learning deficits (Jentsch et al., 2002), have found alterations in metabolic proteins, protein folding and turnover, and cyto-skeletal rearrangements in striatum and OFC when measured 4 weeks after the final cocaine injection (Krueger et al., 2005; Olausson et al., 2007). The prominence of alterations in proteins related to energy metabolism at this time point was striking. It is possible that these changes reflect an alteration in synapse number or function as neurotransmission is a highly energy-intensive process, possibly consistent with the structural changes observed in the OFC in abstinent cocaine users (Franklin et al., 2002; Matochik et al., 2003).

In cocaine-dependent individuals, decreased metabolic activity within the OFC was associated with reduced striatal D2 receptor availability (Volkow et al., 1993); studies in monkeys demonstrate that this effect is most likely rapidly induced by cocaine intake itself (Nader et al., 2006). Similarly, quantitative real-time PCR data show that chronic cocaine exposure reduces D2, but not D1, receptor mRNA in the caudate nucleus in monkeys (P. Olausson, N.C. Tronson, D. Krueger, A.T. Kedves, A.C. Nairn, J.R. Taylor, unpublished observations). This observation may be directly comparable with the reduction in D2/3 receptor binding observed in the ventral striatum of highly impulsive rats (Dalley et al., 2007). In rats, extended cocaine self-administration was reported to reduce the levels of D2 receptors in OFC (Briand et al., 2008) and to disrupt D2-dependent modulation of prefrontal cortical neuron excitability (Nogueira et al., 2006). Whether cocaine exposure also reduces D2 receptor levels in the primate OFC and whether such alterations contribute to the reversal learning deficits observed in cocaine-exposed animals remains to be determined, although this link is generally supported by behavioral pharmacological studies (Lee et al., 2007).

Collectively, these data converge to support an important role for D2-like receptor function in aspects of impulsivity in the context of drug addiction. Furthermore, considering impulsivity as part of a spectrum of neurocognitive functioning, highly impulsive volunteers showed improved working memory performance following administration of a D2/3 agonist (bromocriptine) which was accompanied by the modulation of frontostriatal activity, yet low impulsive people did not (Cools et al., 2007). Hence, genetically or environmentally determined differences in dopaminergic transmission, particularly through the D2 receptor, could alter neurocognitive mechanisms, and this may increase engagement in hasty, risky, and/or impulsive behaviors.

IMPULSIVITY AND COMPULSIVITY

The precise nature of the relationship between impulsive and compulsive behaviors remains an area of active debate. One conceptual distinction between these forms of behavior may lie in the goal-directed nature of incentively motivated impulsive action and choice, as opposed to the habitual nature of relatively outcome-insensitive compulsions (Balleine and O’Doherty, 2009). In other words, impulsive behavior, as we have characterized it here, often depends very much on motivational systems that mediate voluntary reward seeking, while compulsive behaviors are usually thought of as almost reflexive actions triggered by contextual stimuli paired with outcome availability. These stimulus-response actions are also thought to be relatively independent of input from the prefrontal cortices and are largely mediated by areas within the striatum, thalamus, and basal ganglia.

Prospectively, it appears that impulsive tendencies are the predictors of risk for substance abuse and that impulsive behavior is fueled in the early stages of drug use. More recently, however, increased attention has been paid to the idea that—as drug use behavior continues—a “drug habit” may form (Belin et al., 2009; Everitt et al., 2008; Torregrossa et al., 2008). For example, recent preclinical studies have suggested that drug-taking behavior in rats becomes increasingly inflexible with longer experience (Deroche-Gamonet et al., 2004; Vanderschuren and Everitt, 2004). As such, the transition from impulsive to compulsive behaviors may reflect a progressive decline in the influence the prefrontal cortices exert over behavioral output and a shift in emphasis from the ventral to the dorsal striatum (Kalivas, 2008; Robbins and Everitt, 1999). Moreover, impulsive animals that are at risk for the rapid development of drug-taking behavior are subsequently more likely to exhibit evidence of the transition to “compulsive” drug taking (Belin et al., 2008). Conceptually, this type of research has a lot of appeal for our understanding of addiction processes, but a limitation of these models is that it is assumed that inflexible drug taking is synonymous with compulsive actions. This relationship is simply not clear. For that reason, more work in this area is required.

SUMMARY

Data across different species is generally convergent in that high levels of a number of forms of impulsivity are not only predictive of addiction, but can also be affected by drug intake, and that the increases in impulsivity and other cognitive deficits associated with drug dependency may reflect compromised function within the frontal cortex (see Fig. 2 for a summary of this position). Based on current findings looking at cocaine abuse, it would appear that motor impulsivity in particular is associated with the development of drug dependence, whereas intolerance to delayed gratification may contribute to relapse to drug use. Such high levels of delay discounting also appear to be predictive of increased alcohol intake. Such a hypothesis is open to verification by future studies in which different forms of impulsivity are systematically compared, and it would be useful to determine whether pharmacological interventions to reduce different forms of impulsivity would likewise decrease drug taking or seeking. However, as indicated at various points throughout this review, impulsive behavior does not exist in a vacuum, but is part of a wider set of cognitive vulnerabilities for addictive and risky behavior. Key nodes within the frontostriatal circuitry regulating impulsivity are likewise implicated in other cognitive processes that relate to addiction. The studies discussed here largely focused on the OFC, and it is clear that this region not only plays an important role in impulse control deficits, but also the response to reward-paired stimuli and cognitive flexibility. By considering how these behaviors relate to one another, we may be better positioned to identify biomarkers for susceptibility to addiction.