The overarching hypothesis examined in this review is that chronic drug abuse produces neuronal plasticity in brain circuits underlying stimulus-response habit learning. These changes result in maladaptive, drug-directed forms of automatic behavior that, in close interaction with certain goal-directed aspects of drug use, result in the addicted phenotype (Everitt and Robbins, 2005; Robbins and Everitt, 1999; Tiffany, 1990). This is not to say that stimulus-response habits in themselves equate to compulsive, addictive behaviors, that drug-associated habitual behaviors and addiction are synonymous or that addictive behavior no longer has goal-directed properties. Rather, it is suggested that automatic aspects of drug seeking and taking play a critical role in translating the powerful influence of drugs and drug-associated cues into action. That is, environmental stimuli and other factors that precipitate drug craving also trigger ingrained behaviors that facilitate drug seeking and taking. We argue that cocaine-induced changes in neuronal circuits including the ventral (i.e. nucleus accumbens) and dorsal striatum underlie the motivation/craving and habitual components of drug seeking behavior, respectively.

One common way of testing the extent to which responding for a positive reinforcer occurs according to a stimulus-response (i.e. habitual) or action-outcome (i.e. goal-directed) associative structure is by evaluating instrumental responding for a devalued reinforcer (Adams, 1982; Dickinson, 1985; Killcross and Coutureau, 2003). Ingestive reinforcers, for example, can be devalued either by pre-feeding the subject with the food (or liquid) reinforcer, by degrading the taste of the reinforcer using bitter substances like quinine, or by associating the food with illness (induced by post-ingestion injection with lithium chloride, for example). If responding for the reinforcer is subsequently decreased, it is interpreted as being mediated by an action-outcome process, because behavior is driven by a representation of the value of the outcome, which is reduced by devaluing the reinforcer. On the other hand, if after devaluation the subject shows no difference in responding for valued and devalued reinforcers, responding is interpreted as being under the control of a stimulus-response process, independent of outcome value. Research in this area has shown that drug seeking can indeed become dependent on a habitual stimulus-response associative structure, and that this may occur more quickly for drugs than for food. Thus, it has been shown that lever pressing for food or sucrose was markedly suppressed after pairing with lithium chloride, whereas responding for cocaine and ethanol was impervious to this kind of devaluation (Dickinson et al., 2002; Miles et al., 2003). In these studies, both the drug (ethanol, cocaine) and natural (food, sucrose) reinforcers were ingestive rewards (i.e. ethanol, cocaine and sucrose solutions, food pellets, with clear and distinguishable taste characteristics). It is therefore unlikely that differences in stimulus associability (contiguity) explain the differential sensitivity to lithium devaluation between ethanol and cocaine on the one hand, and food pellets and sucrose on the other. Remarkably, both during taste aversion conditioning with lithium (which took place in four daily sessions) and during re-acquisition of responding for drug after the extinction test, responding for and intake of the drug solution associated with lithium-induced malaise was markedly decreased (Dickinson et al., 2002; Miles et al., 2003). This demonstrates that intake of the reinforcer itself can remain sensitive to devaluation even when responding for a drug reinforcer in extinction, using just the internal representation of the reinforcer to guide behavior, is impervious to devaluation. Thus, after relatively modest operant training under a random-interval schedule, responding for drugs relies on a stimulus-response associative structure of behavior, whereas intake of the drug is still a goal-directed action (Dickinson et al., 2002; Miles et al., 2003). It is possible that the pharmacological effects of the drug itself cause drug-directed behavior to change from a habitual to a goal-directed structure. This is not to say that drug intake always remains goal-directed. Insensitivity to devaluation of intake of a drug reinforcer has been observed in studies investigating the oral intake of ethanol and amphetamine, but this took prolonged drug experience to develop (Galli and Wolffgramm, 2004; Wolffgramm, 1991). In these experiments, rats became insensitive to the aversive properties of quinine since they did not reduce their intake of the drug solution after it had been rendered bitter. In early phases of the experiments, intake of the drug solution was reduced by quinine. In fact, the insensitivity to devaluation required no less than nine months of drug experience followed by an abstinence period to develop (Galli and Wolffgramm, 2004; Wolffgramm, 1991). On the other hand, as described above, the insensitivity to devaluation of operant responding for drug already developed after a relatively moderate amount of operant training (Dickinson et al., 2002; Miles et al., 2003). It should be born in mind that there are methodological differences between the studies by Dickinson et al. (2002) and Miles et al. (2003) on the one hand, and Galli and Wolffgramm (2004) and Wolffgramm (1991) on the other, such as differences in the devaluation procedure (lithium vs. quinine) and output parameters (lever pressing vs. drug intake). Nevertheless, these studies, taken together, suggest that operant responding for drugs (an act relatively distal from the actual subjective drug effects) more readily gains habitual properties than intake of the drug itself. This points to the possibility that the development of inflexible, habitual drug use occurs in several distinct phases, where distal drug cues or acts gain habitual properties before intake of the drug itself does, perhaps representing a gradual worsening of the addiction syndrome with increasing drug experience. These results also clearly indicate that drug seeking progresses more readily from a goal-directed action-outcome to a habitual stimulus-response associative structure than the seeking of natural rewards.

The progression from goal-directed to habitual forms of drug seeking and taking appears to be mediated by the gradual recruitment of ventromedial-to-dorsolateral striatal regions with prolonged drug self-administration. In an elegant series of studies, it was shown that initial exposure to cocaine self-administration was correlated with changes in glucose utilization in the ventral striatum, whereas prolonged exposure resulted in a metabolic response that expanded to encompass the entire dorsal striatum, including its most caudal and lateral portions (Porrino et al., 2004a; Porrino et al., 2007). Thus, when adult rhesus monkeys were allowed only five days of experience with cocaine self-administration, significant alterations in functional activity were restricted to limbic nuclei including the nucleus accumbens core and shell as well as the ventromedial prefrontal cortex, which sends major glutamatergic projections to the nucleus accumbens (Porrino et al., 2002). After one hundred days of cocaine self-administration, the intensity and topography of changes in metabolic activity expanded markedly to encompass additional limbic structures such as the amygdala and hippocampus as well as the dorsolateral striatum (Beveridge et al., 2006; Porrino et al., 2004b). Parallel changes were found when the effects of cocaine self-administration on expression of dopamine transporters and D2 dopamine receptors in the striatum were assessed. Initially, cocaine taking induced down-regulation of dopamine transporter expression, but up-regulation was found after prolonged cocaine self-administration. This up-regulation was most pronounced in ventral portions of the striatum, but intensified and spread caudally and dorsally as cocaine experience progressed (Letchworth et al., 2001). Cocaine self-administration reduced D2 dopamine receptor densities, initially in ventral portions of striatum only, but with extended cocaine experience these changes also became larger and were found in caudal and dorsal striatal regions as well (Moore et al., 1998; Nader et al., 2002; Porrino et al., 2004b). Similar effects have been observed in humans, where decreases in D2 dopamine receptor binding were found throughout the striatum (including dorsolateral portions) in cocaine addicts (Martinez et al., 2004). It should be noted that many rodent studies examining the effect of cocaine exposure on dopamine receptor densities in the striatal complex have produced inconsistent findings (see Anderson and Pierce, 2005). These divergent results are likely due to multiple methodological differences including dose, route, extent of exposure to cocaine as well as the length of forced abstinence prior to tissue collection.

In sum, there is evidence to support three main points: i) Drug seeking and taking becomes inflexible or habitual after prolonged use; ii) After lengthy exposure to self-administered cocaine, drug-induced neuronal activation patterns and adaptive changes spread from ventromedial into dorsolateral striatal regions; iii) The dorsolateral striatum plays an important role in drug seeking in drug-experienced animals, especially in well-established drug seeking driven by drug-associated environmental stimuli. These converging lines of evidence provide substantial experimental support for the hypothesis that when patterns of drug intake become ingrained, they come to rely on dorsal striatal mechanisms. There are, however, several remaining questions that need to be addressed in future studies, which are discussed below.

The lion’s share of the evidence discussed here was obtained from cocaine-using subjects. It is important to determine if the dorsolateral striatal regions engaged by prolonged cocaine use also underlie the development of inflexible behavioral patterns associated with other drugs of abuse. It is clear that drugs of abuse such as heroin, alcohol, amphetamine and nicotine also evoke addictive behavior characterized by inflexible and habitual patterns of intake (American Psychiatric Association, 2000; Nutt et al., 2007; O'Brien, 2001). Indeed, results from animal studies indicate that prolonged alcohol and amphetamine intake in rats becomes inflexible and habitual (Dickinson et al., 2002; Galli and Wolffgramm, 2004; Miles et al., 2003; Wolffgramm, 1991). There is also some evidence to support the involvement of dorsolateral striatal regions in addictive behavior maintained by other drugs of abuse. Thus, stimulation of dorsolateral striatal dopamine D1 receptors is necessary for relapse to heroin seeking evoked by a drug-associated context (Bossert et al., 2009). Furthermore, studies in humans have demonstrated activation of the dorsolateral striatum by alcohol- and tobacco-associated cues (Grusser et al., 2004; McClernon et al., 2009).

For the most part, the available evidence shows that the dorsolateral striatum is involved in addictive behavior maintained or evoked by drug-associated cues. In principle, this is consistent with the notion of a dorsolateral striatum-mediated stimulus-response associative structure of behavior where exposure to reward-associated cues results in the execution of an ingrained response, irrespective of the current value of the outcome or cost of the response. In the case of addictive behaviors, it is easy to appreciate that responding for drugs under a second-order schedule of reinforcement has habitual properties. Establishment of this behavior takes extensive training during which the animal receives hundreds of drug-cue pairings. Furthermore, responding for drug-associated cues themselves is persistent in the absence of further presentations of the primary drug reinforcer and it is relatively resistant to extinction (Di Ciano and Everitt, 2004b; Weiss et al., 2001; Panlilio et al., 2005). However, responding for drugs under a second-order schedule of reinforcement also has goal-directed properties, mediated by drugs and conditioned reinforcers (Arroyo et al., 1998). In fact, responding under a second-order schedule of reinforcement having both goal-directed and habitual properties is consistent with the findings that responding for cocaine under a second-order schedule of reinforcement depends on the nucleus accumbens core (Di Ciano and Everitt, 2001) and the dorsolateral striatum (Vanderschuren et al., 2005), likely working in concert (Belin and Everitt, 2008). Indeed, responding for cocaine-associated cues in the absence of cocaine reinforcement also depends on both the nucleus accumbens core and the dorsolateral striatum (Di Ciano et al., 2008). However, involvement of the dorsolateral striatum in a certain behavior does not necessarily mean that this behavior is habitual, as there is recent evidence to suggest that the dorsolateral striatum has functions that extend beyond habitual behaviors (Corbit and Janak, 2007; Luo et al., 2009).

On the other hand, the demonstration that dorsolateral striatal mechanisms also underlie forms of cue-evoked drug-seeking that require only limited self-administration training followed by extinction of operant behavior (Bossert et al., 2009; Fuchs et al., 2006; See et al., 2007) suggests that either striatally driven habits can develop after quite limited drug experience or that the dorsolateral striatum is also involved in certain non-habitual aspects of addictive behavior, such as the general appetitive arousal evoked by reward-associated cues (Corbit and Janak, 2007). In some cases responding was not reinforced by presentations of drug-associated cues but evoked by placement in a self-administration environment (Fuchs et al., 2006; See et al., 2007) and this drug-context-evoked responding extinguished upon repeated testing (See et al., 2007) as did responding for heroin-associated cues in a self-administration environment using a within-session paradigm (Bossert et al., 2009). Thus, it remains unclear if all dorsolateral striatum-dependent operant behavior is indeed habitual (see also Kantak et al., 2002). Therefore, an essential line of future research would be to investigate whether drug seeking that is inflexible in the sense that it is insensitive to punishment (Deroche-Gamonet et al., 2004; Vanderschuren and Everitt, 2004) also depends on dorsolateral striatal mechanisms.

Another issue is whether habitual addictive behavior is specific to drug-associated cues or whether it also extends to the drug itself. There is evidence from animal studies that cocaine self-administration under a simple fixed-ratio 1 schedule depends on stimulation of dorsal striatal dopamine receptors (McGregor and Roberts, 1995; Vanderschuren et al., 2005 but see also Caine et al., 1995). In addition, self-administration of cocaine by experienced cocaine users was found to decrease neural activity and increase dopamine transmission in both the ventral and dorsolateral striatum (Cox et al., 2009; Risinger et al., 2005). In his original proposal, Tiffany (1990) stated that automatized drug use is stimulus-bound (i.e. evoked by drug-associated stimuli), but the automaticity applies to both the procurement and consumption of drugs. Thus, habitual forms of drug use not only involve drug seeking (the dependent measure used in most experiments focused on the dorsolateral striatum reviewed above) but also drug taking, as in tobacco chain smoking as well as binge cocaine or alcohol use. Thus far, however, there is relatively little evidence for the involvement of the dorsolateral striatum in the habitual aspects of drug taking.