Motivational Processes Underlying Substance Abuse Disorder

Abstract

Drug addiction is a syndrome of dysregulated motivation, evidenced by intense drug craving and compulsive drug-seeking behavior. In the search for common neurobiological substrates of addiction to different classes of drugs, behavioral neuroscientists have attempted to determine the neural basis for a number of motivational concepts and describe how they are changed by repeated drug use. Here, we describe these concepts and summarize previous work describing three major neural systems that play distinct roles in different conceptual aspects of motivation: (1) a nigrostriatal system that is involved in two forms of instrumental learning, (2) a ventral striatal system that is involved in Pavlovian incentive motivation and negative reinforcement, and (3) frontal cortical areas that regulate decision making and motivational processes. Within striatal systems, drug addiction can involve a transition from goal-oriented, incentive processes to automatic, habit-based responding. In the cortex, weak inhibitory control is a predisposing factor to, as well as a consequence of, repeated drug intake. However, these transitions are not absolute, and addiction can occur without a transition to habit-based responding, occurring as a result of the overvaluation of drug outcomes and hypersensitivity to incentive properties of drug-associated cues. Finally, we point out that addiction is not monolithic and can depend not only on individual differences between addicts, but also on the neurochemical action of specific drug classes.

1 Introduction

Drug addiction is characterized by intense drug craving and compulsive drug use, with an excessive amount of time and effort spent procuring drug¹ (Jaffe 1975; Koob and Le Moal 2008; Tiffany 1999; Skinner and Aubin 2010; Gass et al. 2014; American Psychiatric Association 2013). In many cases, the drug has no direct benefit to the subject’s survival, and drug-taking occurs despite adverse consequences to one’s health, livelihood, and family. Smoking cigarettes is a classic example of this, and becoming addicted to prescription opioids highlights the complex balance between the benefits and harms of many drugs. So what drives drug-taking in the first place, and what shifts motivational systems into “overdrive,” resulting in addiction? The obvious starting point is that the individual must first try the drug. This initiation depends on a number of factors, including individual personality traits, cultural norms, social context, and interactions with comorbid mental illness (Bucholz 1999; Compton et al. 2007; Homberg et al. 2014). Following this initiation, a drug is taken repeatedly to obtain some perceived positive effect, be it physiological or psychological (e.g., euphoria or social acceptance), or to relieve a perceived negative state (e.g., to reduce anxiety). The experienced outcome of drug use reinforces the drug-taking behavior; that is, the experience of a positive outcome or the removal of a negative outcome increases the likelihood that the drug will be used again. These forms of positive and negative reinforcement, respectively (Khantzian 1997; Wise and Bozarth 1987), help maintain drug-taking behaviors, particularly in early, often experimental stages.

The transition from initial controlled drug use to uncontrolled use and addiction is not well understood, in part because this transition is gradual, and is not defined by a precise behavioral or biological tipping point (Ahmed and Koob 1998). However, it is clear that most drug users begin by experimenting with a variety of drugs, that multiple drug experiences are needed, and that not all drug-taking experiences will result in addiction (even if use is prolonged; think of four years on an American college campus). Below, we describe several key psychological processes that underlie initial drug-taking, how they are altered in the addicted state, and the neurobiological substrates thought to underlie these processes. Because these ideas have been presented in many reviews (Robinson et al. 2014; Belin et al. 2009; Wise and Koob 2014; Hogarth et al. 2013; Wiers et al. 2007; Kalivas and Volkow 2005; Redish et al. 2008), we will focus on core motivational processes involved in addiction and their key neurobiological substrates. In addition, we highlight how different motivational processes may be involved in different aspects of addiction depending on the individual, the drug class, and the duration of drug-taking behavior.

1.1 Rodent Models of Addiction

Much of what is known about the psychological and neurobiological substrates of addictive behavior has been learned using animal models (Lesscher and Vanderschuren 2012; Deroche-Gamonet et al. 2004). Here, we review a few key approaches to orient the reader to studies referred to in subsequent sections. Early studies used repeated administration of potentially addictive drugs to an animal (usually rats or mice) and measured the consequences on behavior and brain function. This approach was, and still is, very useful for understanding how drug exposure changes the brain and influences behavior, but lacks an aspect of voluntary drug-taking (Wolf and Ferrario 2010; Vezina and Leyton 2009; Schmidt and Pierce 2010). Ultimately, self-administration approaches became the gold standard not only for studying drug addiction, but also for assessing how likely it is that a drug will be abused (Schuster and Thompson 1969; Weeks 1962). More recently, self-administration models that attempt to differentiate “casual” drug-taking from “addiction-like” overconsumption have been developed. These models typically use extended access procedures in which drug is freely available over prolonged periods of time. These procedures induce escalated drug self-administration and incorporate the idea of continued pursuit of the drug despite adverse consequences (e.g., crossing an electrified grid to obtain the drug). Indeed, a number of studies have found that extended access self-administration procedures produce “addiction-like” behaviors and neuroadaptations that are distinct from those associated with limited access “drug-taking” behaviors (Loweth et al. 2014; Wolf and Ferrario 2010; Deroche-Gamonet et al. 2004; Lesscher and Vanderschuren 2012; Ahmed 2010). In addition to modeling different aspects of addiction, self-administration models have been informed by concepts from learning theory and psychology in order to ask questions not only about drug-taking and drug-seeking behaviors, but also to determine how these behaviors can be influenced and modified by non-drug experiences (e.g., stressors) and stimuli in the environment that are associated with drug (i.e., Pavlovian-conditioned cues; referred to as drug cues throughout). For example, neuroadaptations induced by drug self-administration have long-term behavioral consequences that can be measured using rodent models of relapse or “reinstatement.” These relapse tests involve “extinguishing” drug self-administration by omitting drug delivery for several sessions and then presenting the rat with stressors, drug cues, or the drug itself. The degree to which these stimuli “reinstate” drug-seeking behavior is a model of relapse severity. Not surprisingly, there are different neurobiological systems underlying these different forms of reinstatement, which are themselves distinct from those involved in drug-taking (Torregrossa and Kalivas 2008; Shalev et al. 2002).

2 Overview of Motivational Processes and Their Neurobiological Substrates

Shifts in behavioral and psychological responses to drugs and drug cues underlie a transition from controlled or causal drug use to addiction. Experimental evidence from human and animal studies shows that drugs produce alterations in mesocorti-colimbic circuits that underlie drug-seeking and drug-taking behaviors, particularly within basal ganglia circuits (Fig. 1; Nestler 2005; Koob and Nestler 1997; Belin et al. 2009; Fields et al. 2007; Sesack and Grace 2010; Vanderschuren and Kalivas 2000; Pierce and Kalivas 1997; Wolf 1998; Steketee 2003). The basal ganglia can be divided into a ventral subsystem in which the nucleus accumbens (NAcc) receives dopaminergic input from the ventral tegmental area (VTA) and glutamatergic input from the prefrontal cortex and amygdala, and a dorsal subsystem (comprised of dorsolateral and dorsomedial subregions) which receives input from the substantia nigra. The NAcc and dorsal striatum project to the ventral pallidum and globus pallidus, respectively, and both have reciprocal connections with frontal cortical areas, including the prefrontal, anterior cingulate, and orbitofrontal cortices. These reciprocal connections form parallel cortical–striatal–pallidal–cortical loops (Humphries and Prescott 2010; Alexander et al. 1990; Haber 2003) that are intimately involved in the initiation and maintenance of drug-taking and drug-seeking behaviors (as well as other motivated behaviors). This system, particularly the ventral subsystem, projects to and receives input from limbic areas including the amygdala and hippocampus (Schmidt et al. 2005; Berendse et al. 1992; Brog et al. 1993; Klitenick et al. 1992). Because of its connectivity with these emotion and learning centers, as well as the hypothalamus, the ventral subsystem of the basal ganglia is often referred to as a “functional interface” between motivation and action (Mogenson et al. 1980; Kelley et al. 2005). In comparison, the dorsal subsystem has been more tightly linked to the processing of goal-directed behavior and habit, which is discussed below (Belin et al. 2009; Yin et al. 2004; Faure et al. 2005). For further discussions of the role of these circuits in motivation, see chapters by Learning and Motivational Processes Contributing to Pavlovian–Instrumental Transfer and Their Neural Bases: Dopamine and Beyond, Multiple Systems for the Motivational Control of Behavior and Associated Neural Substrates in Humans, The Computational Complexity of Valuation and Motivational Forces in Decision-Making Processes, and Neurophysiology of Reward-Guided Behavior: Correlates Related to Predictions, Value, Motivation, Errors, Attention, and Action, in this volume.

Fig. 1.

An oversimplified diagram of key neural projections underlying motivational aspects of addiction. Note that this diagram does not reflect that these projections participate in multiple processes. For example, the amygdala is also involved in the hedonic response to drug rewards and in incentive–salience attribution to reward cues. Also, reciprocal connections between the cortex and midbrain areas are involved in all the concepts listed in the figure

Addiction is mediated by several different motivational processes that can be dissociated experimentally and neurobiologically. For example, the maintenance of drug-taking by negative reinforcement processes results from the removal of a negative drug withdrawal state by the drug (Koob and Le Moal 2001). This hedonic allostasis model also emphasizes that drugs produce a reward deficit in which previously pleasurable events now fail to produce the same hedonic effect. Thus, addiction is maintained by a need both to increase the amount of drug used in order to reach the new hedonic threshold and to overcome anhedonia (i.e., a generalized inability to feel pleasure) induced by repeated drug exposure. From a biological standpoint, this new set point is mediated by reduced activity within dopamine, opioid, and GABA neurotransmitter systems, while anhedonia is mediated by alterations in the function of the amygdala and corticotropin-releasing hormone (CRH) systems (Koob and Le Moal 2008; George et al. 2012) (Fig. 1). In comparison, the incentive–sensitization model of addiction (Robinson and Berridge 1993) proposes that in the addicted state, drugs and drug cues have a heightened ability to direct and influence behavior, ultimately focusing behavior toward drug-seeking and drug-taking. This concept also stresses that in the addicted state, drugs and drug cues are better able to direct behavior, not because they are pleasurable or “liked,” but because they are highly noticeable, attractive, and “wanted” (for more detailed information on the distinct roles of wanting and liking in motivating behavior, please see Robinson et al., in this volume). Neurobiologically, the enhanced influence of drug cues over behavior (i.e., incentive sensitization) is linked to increased responsivity of mesolimbic dopamine systems, including inputs into this system from cortical areas (Fig. 1). Thus, alterations in dopamine- and glutamate-mediated transmission, particularly in the NAcc, drive incentive sensitization that underlies cue-triggered drug-seeking behaviors.

These first two models of addiction propose bivalent changes in brain reward systems to increase motivation: one in which the function of mesolimbic systems and hedonic responses are reduced and drug is needed to return to a “normal” state and one in which heightened responsivity of mesocorticolimbic systems to drug cues drives drug-seeking and drug-taking. These ideas describe different aspects of motivation that may be involved in different phases of addiction.² For example, hedonic allostasis may maintain drug-taking behaviors during early withdrawal or attempts to reduce drug use, whereas incentive sensitization may promote relapse even after physiological and emotional withdrawal symptoms have subsided.

Additional concepts of addiction focus on the relationship between behavioral actions (e.g., lighting a cigarette) and an outcome (e.g., smoking a cigarette). Within this framework, initial drug-taking involves learning that an action (drug-seeking) results in a given outcome (euphoria, anxiety reduction, etc.), but the nature of the relationship between the action and outcome changes as a consequence of drug use to produce addiction. Goal-directed (i.e., action–outcome) models posit that the expected drug outcome is improperly overvalued such that behaviors directed toward obtaining drug override all others. Similarly, drug-induced anhedonia may also contribute to this improper valuation through the underestimating of the value of alternative goals. On the other hand, habit-based models of addiction (i.e., stimulus–response) posit that after initial learning and repeated drug exposure, the outcome no longer drives drug-seeking behavior, but instead, drug-seeking is an automated response triggered by drug cues, emotional states, stress, etc. Like hedonic allostasis and incentive sensitization, goal-directed and habit models of addiction also propose that changes in mesolimbic function underlie these behaviors. However, habit-based ideas emphasize a transition from behaviors mediated by the ventral subsystem of the basal ganglia to behaviors mediated by the dorsal subsystem (Fig. 1). In addition, goal-directed, habit-based, and incentive sensitization all incorporate the idea that the relationship between drug-seeking behavior and consumption of the drug itself changes as addiction develops, though they differ in the nature of the relationship that is thought to change. Below, we provide more specific details of the processes involved in each of these models described above as well as evidence in support of and contrary to them.

3 Hedonic Allostasis

The hedonic allostasis model of addiction (Koob and Le Moal 2001) is based on Solomon’s opponent process theory, but has been modified to account for the chronic, pathological nature of drug-taking and relapse (Solomon and Corbit 1974). In Solomon’s original work, he proposed that drug-taking consists of two dissociable processes: a short-lived a-process, which might include elevated heart rate, a state of well-being or euphoria (i.e., a hedonic affective state), or motoric effects (e.g., punding or chewing that are common with psychostimulants). At the same time, the drug also produces a b-process that has two key features: (1) It is in opposition to the a-process (e.g., reduced heart rate) and (2) it has a slower onset and offset relative to the a-process. The b-process can thus counteract aspects of the a-process. Solomon argued that the b-process, because it is longer lasting than the a-process, could explain the phenomenon of drug withdrawal. Further, he suggested that the b-process becomes larger after repeated drug administration. This change in the b-process has two effects: first, to further counteract the a-process (resulting in drug tolerance and a need to increase drug intake to compensate) and second, to worsen the withdrawal syndrome.

The hedonic allostasis model adds to this idea by positing that a portion of the b-process involves a reduction in the hedonic affective state (to counter euphoria induced by drug) that is maintained after drug use is discontinued. This maintained anhedonic b-process changes the “set point” for activation of the a-process such that the person remains in a relatively anhedonic state in the absence of drug. Koob and colleagues call this “allostatic state … a new equilibrium, a state of chronic deviation [from the] normal (homeostatic) operating level to a pathological (allo-static) operating level” (George et al. 2012).³ This anhedonic deficit creates a motivational drive to take drugs in order to reverse anhedonia (through negative reinforcement). However, because the set point for the desired a-process has changed, people must now take larger doses of the drug to activate the a-process once more and overcome the enhanced b-process. In this manner, a shift from positive to negative reinforcement underlies drug addiction. In addition, the imbalance between a- and b-processes produces an unending “downward anhe-donic spiral” where the b-process (anhedonia) continually grows stronger due to drug use and thus does not produce the desired a-process effect (euphoria to remove anhedonia). The motivational drive to relieve this hedonic deficit becomes ever stronger and leads to more drug-taking (Ahmed and Koob 1998; Koob and Le Moal 2001). Thus, drug withdrawal induces both physiological withdrawal (e.g., shakes in alcoholism or temperature dysregulation in the case of opioids) and “hedonic” withdrawal (i.e., anhedonia and malaise).

Drug-induced alterations in a- and b-processes are mediated by within-systems adaptations, in which changes occur in systems responsible for initial drug effects, and between-systems adaptations, which involve the recruitment of additional systems not initially engaged by the drug (Koob and Le Moal 2008). Neurobiologically, within-systems changes mediating a reduction in the set point a-process are due to drug-induced reductions in mesolimbic dopamine. For example, acute administration of alcohol stimulates VTA dopamine input into the NAcc (Brodie et al. 1999; Meyer et al. 2009; Nimitvilai et al. 2012; Gonzales et al. 2004). However, repeated alcohol exposure reduces the activity of VTA neurons (Diana et al. 1993; Shen 2003; Shen et al. 2007), including during withdrawal from chronic alcohol drinking (Bailey et al. 1998). This is consistent with the idea that increased alcohol intake is needed to restore reduced function of the mesolimbic system to prealcohol exposure levels. In another study, Barak et al. (2014) used a chronic intermittent alcohol procedure to induce escalated intake; this escalation was associated with a decrease in the firing rate of VTA dopamine neurons. When administered glial-derived neurotrophic factor (GDNF), the escalation of alcohol intake and the decrease in neural firing rates were blocked. These data suggest that normalizing an alcohol-induced deficit in VTA neuron firing (and by extension NAcc dopamine levels) resulted in a reversal of escalated alcohol intake. Additional evidence for a-process set point alterations comes from animal studies of intracranial self-stimulation (ICSS), in which lever pressing is reinforced by electrical stimulation of the medial forebrain bundle (Edmonds and Gallistel 1974; Miliaressis et al. 1986). ICSS reward thresholds are defined as the intensity of stimulation needed to support self-stimulation; increases in ICSS thresholds are interpreted as a need for stronger stimulation to achieve the same level or reinforcement (Diana 2011). Indeed, chronic exposure to many drugs increases ICSS thresholds (Kenny et al. 2006; Epping-Jordan et al. 1998; Kenny et al. 2003; but see Kenny and Markou 2006). Importantly, increased ICSS thresholds are long-lasting and persist even during protracted abstinence (Ahmed et al. 2002; Kenny et al. 2003). ICSS activates dopamine neurons (Yeomans 1989), and maintenance of ICSS requires activation of the hypothalamus (Kempadoo et al. 2013; Olds 1962). Thus, these experiments match electrophysiological data from VTA neurons supporting the hypothesis that a reduction in the activity of dopamine neurons drives the escalation of drug intake, but also engage additional hypothalamic circuits and transmitter systems.

Between-systems b-process adaptations that lead to enhanced anhedonia after repeated drug use may involve CRH within the extended amygdala as well as the dynorphin–κ opioid system (George et al. 2012). CRH receptor antagonists block withdrawal-induced anxiety (Basso et al. 1999; Logrip et al. 2011), withdrawal-induced increases in nicotine and alcohol self-administration (George et al. 2007; Finn et al. 2007) and escalated cocaine and heroin self-administration during extended access tests (Specio et al. 2008; Greenwell et al. 2009). Further, blockade of κ receptors reduces negative hedonic aspects of drug withdrawal and decreases drug-seeking (Koob 2013; Koob et al. 2014; Chartoff et al. 2012). After extended drug intake, input from CRH and dynorphin systems into the striatum is strengthened (e.g., see Turchan et al. 1997), leading to a further increase in the anhedonic b-process. Ultimately, these changes in b-processes are mediated by between-systems adaptations in CRH that lead to a further masking of the a-processes mediated by within-systems adaptations in mesolimbic dopamine systems and the persistence of drug-taking behavior (Aston-Jones et al. 1999; Nestler 2001; Koob 2003; George et al. 2012).

The negative reinforcement process of the hedonic allostasis model can account for drug-taking during early and late phases of addiction (especially in the case of alcoholics who can experience shakes and life-threatening seizures if alcohol use is stopped abruptly) and addresses why drug users may continually increase the amount of drug used. While the physiological, flu-like symptoms of withdrawal for most commonly abused drugs (shakes, nausea, malaise) subside within two to three weeks after drug use is discontinued, the hedonic effects, including mood alterations that manifest as depression and anxiety, may last longer (Soto et al. 1985; Ashton 1991; Rothwell et al. 2012; Hughes 2007; Satel et al. 1993). This may explain why many recovering addicts relapse weeks to months after drug use has stopped and when physical withdrawal symptoms are absent. However, it is important to note that drug-taking behavior can occur even in the absence of subjective, hedonic effects (Fischman and Foltin 1992; Fischman 1989; Lamb et al. 1991; Robinson and Berridge 2001), suggesting that the reinforcing properties of drugs are not mediated by their hedonic impact alone (Robinson and Berridge 2001). In addition, drug-seeking can often be context specific (e.g., only occurring in particular places) and triggered by drug cues (like hearing the click of a lighter when trying to quit smoking). An increase in hedonic set point is intrinsic and thus should not be modified by drug cues and drug contexts (Crombag et al. 2000, 2001; Badiani 2013). Thus, the hedonic allostasis model does not explain cue-induced reinstatement, context-dependent self-administration behaviors, or craving reported by recovering addicts when faced with a drug cue (Fox et al. 2005; Witteman et al. 2015; Celentano et al. 2009; Bossert et al. 2013). Further, a decrease in the a-process would be expected to generalize to other types of rewards, because there is a partial overlap in the neural circuits underlying the responses to “natural” rewards, such as food and sex, and drugs. However, it is not clear whether addicts are impaired in their ability to experience pleasure from non-drug sources (DiLeone et al. 2012; Hone-Blanchet and Fecteau 2014; Ziauddeen and Fletcher 2013; Volkow et al. 2012). Thus, the idea that drug history produces a generalized, long-lasting anhedonia cannot explain all instances of addiction, and is unlikely to drive relapse after long-term abstinence.

4 Incentive Sensitization

The incentive–sensitization concept of addiction proposes that addiction is due to adaptations in brain regions within the mesocorticolimbic systems that mediate a specific aspect of “incentive motivation”: the process of “incentive–salience attribution” (Robinson and Berridge 1993, 2001; also see Robinson et al., this volume). Drug-induced neuroadaptations result in the sensitization of incentive–motivational processes such that drugs and drug-related stimuli acquire enhanced (sensitized) salience. Ultimately, it is sensitization of the incentive–motivational properties of drugs and drug-related stimuli that govern compulsive pursuit of drug.

Incentive motivation refers to a multi-component processes by which behavior is directed toward “incentives” such as food, a potential mate, drugs, and stimuli associated with these incentives through Pavlovian learning. During early drug use, the activation of neural systems that mediate pleasure triggers incentive–motivational processes that become progressively stronger across repeated drug use. An association between the act, event, or cues in the environment in which the pleasure occurs is then formed through Pavlovian learning processes. These Pavlovian cues can themselves gain motivational properties through the separate process of incentive–salience attribution. In this process, after the initial association has been formed, salience is then attributed to the mental representation of the reward/drug-associated cue. Incentive sensitization posits that the process of incentive–salience attribution occurs in a neural system that is separable from the system responsible for the pleasurable effects of the initial incentive (liking) and separable from the neural systems responsible for the initial Pavlovian associative learning processes. Thus, cues that have incentive salience are able to direct behavior, not because they are pleasurable or “liked,” but because they are highly noticeable, attractive, and “wanted.” Within this conceptual framework, addiction develops and persists due to drug-induced neuroadaptations in the brain systems that mediate incentive salience (wanting), making these systems hypersensitive (sensitized) to drugs and drug cues. Further, sensitization of brain circuits that mediate incentive salience, a component of incentive motivation, results in drug-seeking behavior. The sensitization within these neural circuits persists even after long periods without drug use, making addicts susceptible to relapse long after drug use has ceased. Furthermore, progressive alterations in incentive salience through associative learning mechanisms engaged by repeated drug-taking experiences enhance the overall incentive–motivational properties of drug cues leading to increasingly compulsive patterns of drug-seeking and drug-taking. Rodent studies have shown that “liking” and “wanting” processes can be dissociated neurobiologically (Berridge and Kringelbach 2015). This is discussed further by Robinson et al., in this volume.

Neurobiologically, incentive sensitization has traditionally been attributed to progressive enhancement (i.e., sensitization) of mesolimbic responses to drugs and drug cues (Berridge and Robinson 2003; Di Chiara et al. 1999; Wise 1987, 1988). Many studies have focused on the role of dopamine projections from the VTA to the NAcc in this process (Vanderschuren and Kalivas 2000; Flagel et al. 2008; Vezina 2004; Vezina and Leyton 2009), as well as the contribution of enhanced glutamatergic transmission particularly within the NAcc (Wolf and Ferrario 2010; Loweth et al. 2014; Kalivas and Nakamura 1999; Vanderschuren and Kalivas 2000; Cardinal et al. 2002). The focus on dopamine systems arose because many commonly abused drugs share the ability to enhance dopaminergic neurotransmission within the mesocorticolimbic system (Vanderschuren and Kalivas 2000; Vezina 2004; Wolf 2002) and to produce psychomotor sensitization (Babbini and Davis 1972; Joyce and Iversen 1979; Kita et al. 1992; Benwell and Balfour 1992; Littleton 1998; Nestby et al. 1997), an enduring hypersensitivity to the psy-chomotor activating effects of drugs that includes locomotion and stereotyped sniffing and rearing behaviors (Robinson and Becker 1986; Segal 1975).

The development of psychomotor sensitization is a progressive, incremental process, much like the development of addiction itself. Furthermore, psychomotor sensitization can persist for months to years after drug use has stopped (Paulson et al. 1991; Robinson and Becker 1986). Thus, incentive sensitization underlying addiction is thought to be mediated by sensitization-related adaptations in brain systems that mediate the incentive–salience and incentive–motivational properties of drug and drug cues.⁴ Consistent with this, drug pretreatments that produce psychomotor sensitization facilitate the acquisition of drug self-administration in the monkey, rat, and mouse (Woolverton et al. 1984; Horger et al. 1990; Lessov et al. 2001), increase motivation for drug assessed by breakpoint on a progressive ratio schedule (Mendrek et al. 1998; Lorrain et al. 2000; Deroche et al. 1999), and enhance the escalation of drug intake in extended access models (Ferrario and Robinson 2007). Thus, psychomotor sensitization is linked to enhanced motivation for drugs and with addiction-like behaviors (Ferrario et al. 2005; Morgan and Roberts 2004 though see also Ahmed and Koob 1998). The incentive–sensitization model additionally proposes that drug-induced increases in incentive salience contribute to aberrant motivation underlying addiction. In support of this, drug cues themselves can support conditioned approach behavior (Meyer et al. 2012b; Uslaner et al. 2006; Yager and Robinson 2013), and pretreatment with a sensitizing regimen of amphetamine enhances the subsequent acquisition of Pavlovian-conditioned approach behavior (Harmer and Phillips 1998).

Although VTA-NAcc circuitry is heavily involved in psychomotor and incentive sensitization, alterations in glutamatergic transmission within the NAcc mediated by both cortical and amygdalar inputs play important roles (Ma et al. 2014; MacAskill et al. 2014; Everitt and Wolf 2002). For example, cocaine self-administration increases glutamatergic transmission in the NAcc, drug exposure increases the number of excitatory inputs from the basolateral amygdala to the NAcc shell, and remodeling of glutamatergic synapses from the PFC to NAcc contributes to cue-triggered drug-seeking behavior (MacAskill et al. 2014; Wolf 2003; Cornish et al. 1999; Wolf and Ferrario 2010; Robinson and Kolb 2004; Ma et al. 2014). Furthermore, the central nucleus of the amygdala (CeN) is necessary for Pavlovian-conditioned approach to reward predictive incentive stimuli (Everitt et al. 2003), and activation of dopaminergic input into the CeN is sufficient to elicit approach to drug cues (Cardinal et al. 2002).

The induction and expression of psychomotor sensitization is modulated by learning and the circumstances surrounding drug administration (Robinson et al. 1998). This concept is particularly important to the incentive–sensitization model of addiction because it posits that associative learning processes focus “wanting” onto drug cues enabling the expression of sensitization-related neuroadaptations to be influenced by factors that contribute to associative learning (Robinson and Berridge 1993). For example, even when neural sensitization is induced by repeated drug exposure, the expression of sensitization can be context specific. That is, if the drug is given in an environment that is different from where initial drug treatments were received, the sensitized response is not evoked (Post et al. 1981; Anagnostaras and Robinson 1996; Terelli and Terry 1999). However, these same animals will express a sensitized response when given drug in the initial treatment environment (Anagnostaras and Robinson 1996; Post et al. 1981). Thus, although neural sensitization occurred, the behavioral expression of sensitization is context dependent, i.e., it is influenced by conditioned stimuli paired with drug administration (Anagnostaras and Robinson 1996). Conversely, animals placed in a drug-paired environment will show conditioned locomotor hyperactivity in the absence of any drug (see Stewart 1992 for review). Enhanced incentive salience of drug cues and increased incentive–motivational power of these cues in rodent models are consistent with behavioral studies in human drug addicts, as well as self-reports of enhanced cue-triggered craving during abstinence (Childress et al. 1993; Gawin and Kleber 1986; for review, see Veilleux and Skinner 2015). In addition, incentive sensitization accounts for the context dependence of drug craving. More recent studies of incentive sensitization have focused on individual differences in the attribution of incentive salience in order to address why some individuals are able to maintain controlled drug use, while others develop addiction (Robinson et al. 2014; Meyer et al. 2012a).

In summary, at the heart of incentive sensitization is the concept that drugs produce neuroadaptations in systems that mediate the attribution of incentive salience to drug cues. Through these sensitization-related processes, Pavlovian cues associated with drug-taking become increasingly able to direct and focus behaviors, inducing cue-triggered motivational states that drive pathological drug “wanting” and addiction. Several other concepts of addiction presented below incorporate the idea that drug cues influence drug-seeking behaviors. However, they differ from incentive sensitization in that not all concepts include the induction of a motivational state by Pavlovian cues.

5 Addiction as a Disorder of Learning and Decision Making

In the above sections, we discussed ideas about addiction that focused on hedonic responses to drug- and emotion-driven states, versus conditioned incentive–motivational states elicited by drug cues. Addiction can also be described in the context of aberrant decision-making processes in which the drug becomes irresistible and leads to persistent drug-taking even in the face of negative consequences. Decision-making conceptualizations of motivated behavior are based on the idea that there is some value, either real or expected, to the outcome of a particular action. These actions have potential costs and rewards, and individuals vary in their ability to estimate these costs and rewards, just as they differ in how much they value these costs and rewards in the first place. Thus, making a decision to engage in some action depends on the individual’s ability to weigh out the costs, benefits, and consequences of their actions. There are at least two strategies in forming this appreciation. In the forward-search strategy, the outcome of each possible action is independently considered every time the person makes a decision. In caching, the person learns essentially through trial and error that a particular action in a given situation is likely to produce a certain outcome; then, the appropriate action is chosen when that particular situation is encountered. Thus, the forward-search strategy involves planning, is flexible, and is sensitive to changes in the action–outcome relationship, while the cache strategy is inflexible, is fast, and leads to habitual responding (Redish et al. 2008; also see Redish et al., in this volume). In many instances, these strategies occur sequentially; the initial phase of motivated decision making involves forward-search planning and evolves later into a caching strategy. These forward-search and caching strategies are analogous to reflective (goal-directed) control and reflexive (habitual) control of decision-making processes (Balleine and Dickinson 1998; Dolan and Dayan 2013; Doll et al. 2012).

Many studies of drug addiction consider the transition from forward-search goal-directed behavior to caching or habitual behaviors a hallmark of addiction (Corbit et al. 2012; Zapata et al. 2010; Everitt and Robbins 2015; Redish et al. 2008). Below, we describe general processes involved in goal-directed and habit-based responding, and how alterations in these processes may contribute to addiction. The main distinction between these two ideas is that for goal-directed behavior, the outcome (i.e., the subjective or physiological drug experience) maintains addiction, whereas for habit-based responding, the outcome itself is no longer relevant; drug-taking behavior is instead maintained by automated processes elicited by drug cues, stress, etc.

5.1 Goal- Versus Habit-Based Responding

During initial drug-taking experiences, an individual learns that an action (drug-seeking) produces some outcome (drug experience). This action–outcome relationship, or A-O relationship, can be dependent on the presence of environmental stimuli (or, as the notation goes, S:A-O). In other words, the individual learns that drug-seeking actions produce drug-taking outcomes only in the presence of certain stimuli such as drug cues. As addiction develops, the drug-taking outcome, either actual or expected, becomes overvalued. This overvaluation can be a result of drug withdrawal (Hutcheson et al. 2001), or an inflated expectation of the euphoric effects of the drug, even when the euphoric event is actually decreased in magnitude (Kennett et al. 2013). Overvaluation of the drug-taking outcome (O) can lead to pathological drug-seeking (A), especially when triggered by drug cues (S). In this manner, goal-directed responding can lead to addiction when the expected drug outcome is improperly overvalued such that behaviors directed toward obtaining drug override all others.

In contrast, habit-based ideas of addiction suggest that a drug-seeking action is an automated response (R) elicited by a drug cue (S). This stimulus–response, or S-R, learning process leads to habitual drug-seeking and drug-taking, despite changes in the outcome of the drug-seeking action. Further, habitual, S-R drug-taking behaviors are automatically elicited in the absence of craving or urges, and occur independently of antecedent motivational states or the actual outcome of drug use itself (Hogarth et al. 2013; Belin et al. 2013; Ostlund and Balleine 2008; Everitt et al. 2001; Tiffany 1990). As frequency of use increases, more automated and routine actions are prompted that can further perpetuate drug-taking. These automatic drug-taking behaviors are initiated by drug cues and have been conceptualized as “incentive habits,” or habit processes prompted by incentive stimuli (Pavlovian cues) that have undergone incentive sensitization (Belin et al. 2013). Importantly, incentive habits differ from incentive motivation because cue-driven S-R responses occur in the absence of an explicit motivational state, whereas Pavlovian stimuli with incentive motivational properties induce motivational states that produce drug-taking.

S-R habits are further characterized by the blunting of executive function such that conscious decision making (i.e., ability to weigh costs, benefits, and potential consequences) becomes increasingly compromised. A decline in function of executive control processes further enhances habit-based behaviors and reduces outcome-based decision-making processes (Jentsch and Taylor 1999). These features of habitual use develop over prolonged drug use, further drive drug-taking behavior, and likely occur in parallel with the other features of addiction. Therefore, through learning routine drug-taking patterns, automatic drug-use behaviors crystallize, becoming highly efficient and repetitive behaviors that are resilient to intervention.

In humans, many aspects of drug-taking behavior can be categorized as habitual, such as the tapping and packing of a pack of cigarettes, or the preparation of a needle, which can occur in an automated fashion analogous to driving the same route to work each day. The automation of complex behaviors can be adaptive, but these processes can also be usurped by drugs. Thus, habitual behaviors in the addicted state are aberrant not because these behaviors have become automated, but because these behaviors are fixed in a habitual state and do not move back into a goal-directed system (Everitt and Robbins 2005; Dickinson et al. 2002; Miles et al. 2003; Robbins and Everitt 1999). However, some aspects of human addiction do not fit as well with habit-based ideas of drug-taking. For example, addicts often engage in novel and even creative behaviors in order to maintain their drug use (e.g., a prescription opioid pill abuser that switches from oral administration to intravenous street drug use). This transition requires an entirely different set of actions that were never previously used to obtain drug and thus do not fit well with an automated, habitual behavioral response.

Another layer can be added to these concepts of goal-directed and habitual behaviors; that is the ability of Pavlovian-conditioned motivational states⁵ (as described above in the Incentive Sensitization section) to invoke the retrieval of specific outcomes, such as drug-specific effects. Thus, while addiction can transition from A-O to S-R, addiction may also be marked by a transition from the ability of Pavlovian stimuli to elicit the representations of specific drug outcomes to elicit general motivational states (Hogarth 2012; Hogarth et al. 2013). As discussed next, these instrumental and Pavlovian processes are subserved by different neurobio-logical systems and thus may interact to produce craving and drug addiction.

5.2 Neural Circuits Governing Goal- and Habit-Based Decision Making

Studies of food reinforcement show that prefrontal cortical inputs to the dorso-medial section of the striatum are critical for the formation of associations between actions and outcomes (Shiflett and Balleine 2011; Balleine 2005). The neuro-chemical basis for goal-directed responding may be related to dopamine signals, where drug-induced dopamine release maladaptively increases the value of the outcome (i.e., drug intake; Redish 2004; Naude et al. 2014; Belin et al. 2009). Because the likelihood of engaging in a drug-seeking action is a function of the perceived value of the outcome (O), this “hijacking” of the dopamine system would increase drug-seeking progressively. A change in the valuation of an outcome as a result of drug exposure involves alterations in the amygdala and hippocampus (Wassum et al. 2011; Mahler and Berridge 2012), and the retrieval of this now altered outcome likely involves a circuit connecting areas of the prefrontal cortex (including the orbitofrontal cortex) and the NAcc (Shiflett and Balleine 2010; Jentsch and Taylor 1999; Puumala and Sirvio 1998; Simon et al. 2013; Winstanley et al. 2004; Johnson et al. 2007; Schoenbaum et al. 2006; Kalivas and Volkow 2005).

As described above, initial drug-seeking responses are mediated by dopamine transmission in the ventral striatum. However, as goal-directed drug-taking transitions to habitual drug-taking, prolonged activation of dopamine neurotransmission triggers a transition to behavior mediated by the dorsolateral striatum (Balleine et al. 2014; Yin et al. 2006). This transition from ventral to dorsal striatal activation reduces goal-directed behaviors and supports habit learning such that behavior is driven by drug cues, but no longer requires a representation of or motivational state driven by the outcome (Corbit et al. 2012; Vanderschuren et al. 2005; Willuhn et al. 2012). Dorsal and ventral regions of the striatum are connected through a complex “spiraling” or “looping” of reciprocal connections to the VTA and substantia nigra that allow one striatal subregion to regulate dopaminergic tone in adjacent regions. For example, inputs from the NAcc shell to the VTA can inhibit dopamine release within the NAcc core (Belin and Everitt 2008). Similarly, the NAcc core regulates dopaminergic tone within the dorsal striatum through the substantia nigra. The dorsal striatum is divided into multiple domains as well, and dopaminergic input is regulated through reciprocal nigrostriatal projections that are organized anatomically from ventral to dorsal. Each domain regulates the dopaminergic tone of the adjacent domain and, through their spatial location, produces a “spiral” of reciprocal connections that can regulate dopaminergic tone throughout the striatum. This anatomical arrangement allows for the transition from goal-directed behaviors mediated by the NAcc to habit-based behaviors mediated by the dorsal striatum. Specifically, initial exposure to a drug would induce dopaminergic responses and neuroadaptations in the ventral striatum, but because of the spiraling reciprocal connections, these responses and neuroadaptations would engage and alter dorsal striatal projections during extended drug-taking (e.g., Porrino et al. 2004; Letchworth et al. 2001; Ito et al. 2000, 2002; Belin and Everitt 2008). Consistent with the idea of a transition between goal-directed and habit-based processes, an intact connection between NAcc core and the dorsal striatum is required to elicit habit behaviors (Belin and Everitt 2008). The involvement of the dorsal striatum in habit learning is also dynamic throughout prolonged drug use, with activity shifting from the dorsal–medial striatum (DMS) to the dorsal–lateral striatum (DLS) in driving habit-based, stimulus–response behaviors (Murray et al. 2012). Consistent with this, inactivation of the DLS after prolonged cocaine exposure renders rats sensitive to changes in goal value and hence restores goal-directed behavior while attenuating automated stimulus–response processes (Zapata et al. 2010). Thus, reciprocal connections between these circuits allow for transitions between habitual and goal-directed behaviors as needed, whereas drug-induced alterations within these circuits perturb this flexibility. This is further compounded by drug-induced cortical dysregulation that compromises executive function and planning.

5.3 Pavlovian Cues: Reward Prediction, Incentive Salience, and Transfer to Instrumental Responding

Pavlovian cues play an important role in these processes by triggering sequences of actions that can be goal-directed and habitual, or by inducing a conditioned motivational state (incentive motivation). The ability of cues to guide goal-directed behavior likely involves the dopamine system as well, in which changes in dopa-mine neuron activity signal prediction error—outcomes that have a perceived value that was either greater or less than expected based on the presence of reward-predicting cues (Schultz et al. 1997). Therefore, increases in dopamine neuron activity, and the resultant increases in dopamine (e.g., Hart et al. 2014), would lead to more rapid associations between drugs and their cues. However, dopamine is also involved in the attribution of incentive salience to drug cues (Flagel et al. 2011; Saddoris et al. 2015). In other words, drug-associated cues acquire incentive salience as a consequence of drug-induced increases in dopamine. The incentive motivational state induced by these cues can then transfer to instrumental drug-taking, thereby increasing the drug-seeking and drug-taking behavior. These Pavlovian transfer effects during food self-administration are dependent on the orbitofrontal cortex (Ostlund and Balleine 2007a, b), while the devaluation of an instrumental outcome during A-O learning is dependent on the prelimbic cortex (Corbit and Balleine 2003). However, whether Pavlovian transfer effects during drug-taking also involve these brain areas is not known.

6 Impulsive Action and Impaired Executive Function

Layered on top of the concept of enhanced goal-directed behavior described above is the contribution of impulsivity and impulsive action. One consequence of the overvaluation of a drug is impulsive responding in which individuals (1) value immediate drug delivery over more adaptive, but long-term, outcomes and/or (2) value less probable drug availability over more certain, alternative outcomes. These patterns of responding could be due to the overvaluation of the drug itself and/or to altered discounting of the temporal delay between the action and receipt of the outcome, the probability of drug availability, or the negative consequences of drug-taking. This aberrant outcome valuation, in the form of impulsivity, may explain the correlation between impulsive behavior and drug-taking (Moeller and Dougherty 2002; de Wit and Richards 2004; Perry and Carroll 2008). However, impulsivity may also reflect a loss of inhibitory control over drug urges, including those elicited by drug cues (Jentsch and Taylor 1999; Belin et al. 2013). In this case, in contrast to overvaluing drug outcomes, individuals are simply unable to make choices based on long-term outcomes (Bechara 2001, 2005).

The crystallization of habit appears to be further exacerbated by cortical dys-regulation, particularly in the prefrontal areas involved in planning (Volkow et al. 2003). Under normal circumstances, habitual behaviors can be inhibited by cortical systems. However, repeated drug use also produces neural alterations in prefrontal cortical systems that exacerbate habitual behaviors. These adaptations in human addicts occur within the orbitofrontal cortex and anterior cingulate cortex (Lubman et al. 2004; van Huijstee and Mansvelder 2014). In addition, dysregulations in these regions are implicated in other behavioral control disorders involving impulsivity, evaluation and planning, and inhibitory control (Verdejo-Garcia et al. 2006, 2007; Sakagami et al. 2006). Dampened cortical activation and reduced executive function after repeated drug exposure may explain the addict’s failure to control habitual behaviors as well as the inflexibility in response to changing outcomes (Jentsch and Taylor 1999). Together, the loss of prefrontal inhibitory control and consolidation of dopaminergic signaling in the dorsal striatum are thought to perpetuate responses to drug cues by bypassing goal-directed behaviors entirely, and prompting drug use in a stimulus-dependent manner. These changes may also explain the ineffectiveness of cognitive control approaches to reducing drug-taking behaviors.

7 Models that Integrate Motivational Concepts

7.1 Dual-Process Models

Dual-process models of drug addiction describe drug-taking behavior as the output of two competing forces: an automatic, impulse-generating system that is approach-oriented and generates drug-taking behavior, and a controlled, reflective system that uses cognitive control to regulate the impulse-generating system. For example, in the case of alcoholism, Wiers et al. (2007) argues that alcohol-use disorders occur from an imbalance between automatic and controlled processes. In adolescence, where much of the evidence for this model is derived, this controlled process is weak—as any parent can attest. Upon repeated alcohol exposure, the appetitive/approach system is sensitized, and recreational users are attracted to and influenced by drug-associated cues (Duka and Townshend 2004; Field et al. 2005). In contrast, the controlled process degrades as a result of chronic alcohol use (White et al. 2000). Thus, adolescents with already weak control systems are particularly prone to the further sensitization and weakening of approach and control systems, respectively.

While these ideas may seem analogous to the habit-based and goal-directed response systems described above, they should not be conflated. Instead, incentive motivation, action–outcome systems, and habit-based systems together produce drug-seeking behaviors. However, because each process can be localized to some extent to different neural circuits, they can be independently modified by drug exposure. Specifically, cue-triggered behaviors rely heavily on the ventral and dorsal striatum, as well as prelimbic and orbitofrontal cortices (Daw et al. 2011; de Wit et al. 2009, 2012). These systems are in turn under negative control of a regulatory gating system mediated by cortical regions. This cortical regulatory system can then influence whether forward-search or cached response patterns (discussed earlier in this section) are employed. While such an arbiter of response strategy selection has not been identified, it may involve the infralimbic and dor-solateral components of the prefrontal cortex, as manipulations of these regions can induce switches from one strategy to the other (McClure and Bickel 2014; Jentsch and Taylor 1999; Coutureau and Killcross 2003; Killcross and Coutureau 2003; Daw et al. 2005).

7.2 Transitions from Drug-Taking to Addiction

Given the neural overlap and interactions between the different motivational concepts described above, it seems likely that these concepts are explaining different aspects or stages of addiction (Piazza and Deroche-Gamonet 2013). For example, several have suggested that transition into addiction is characterized by a slow yet categorical change in the motivational processes underlying drug-taking. Koob and colleagues have argued that addiction transitions from positive reinforcement, to incentive sensitization, to allostasis (Koob et al. 2014). Goldstein and Volkow (2002) propose a model, based on neuroimaging data, where the intoxicating effects of drugs initially involve the frontal cortex, particularly the anterior cingulate and prefrontal cortices. As addiction develops, the degree of incentive motivation attributed to drugs (as measured by drug craving) is paralleled by activity in the orbitofrontal and anterior cingulate cortices.

7.3 Drug Addiction: A Compendium of Vulnerabilities

Redish (2004) and Redish et al. (2008) have utilized the goal-directed and habit distinction to describe how “vulnerabilities” in these decision-making processes can lead to addiction (Redish et al. 2008). Redish describes 10 such vulnerabilities or “failure points” in which the decision-making process can become maladaptive. These failure points can be grouped into general categories, including negative reinforcement, positive reinforcement, learning processes (including reward prediction and action–outcome learning), response inhibition and discounting, and habit. However, it should be mentioned here that even though we have grouped these vulnerabilities together for discussion purposes, they are to an extent neurobiologically dissociable and thus represent separate psychological processes.

Each of these failure points can lead to addiction. For example, a person may form incorrect action–outcome relationships, or improperly categorize situations in which action–outcome relationships occur. They would thus develop an “illusion of control” in which they view the success in achieving a positive outcome as result of their own actions, or because of changing situations, even if this is not the case. In this case, they would be less likely to select alternatives to drug-taking. For habit-based system vulnerabilities, behavior becomes inflexible, and insensitive to changes in the consequences of behavioral actions. While a non-addict may be able to switch from habitual to goal-directed responding without difficulty, an addict that suffers from improper response inhibition vulnerability will have difficulty doing so.

These vulnerabilities can be applied to positive and negative reinforcement as well. For negative reinforcement, the increase in “need” that happens as a result of withdrawal increases the probability that a person will take drugs to satisfy the need, and the allostatic load that accumulates (strengthened opponent process) is a result of the neuroadaptive response to repeated drug exposure. For positive reinforcement, the euphoric effect of drugs creates an intense reward signal that is stored in memory and becomes a goal to be sought after. Stimuli that induce recall of this euphoric event induce craving. Action–outcome vulnerabilities are also related to the reward signal. For example, incentive sensitization leads to a sought-after event, but the expected reward is much greater than the actual reward, and thus the distinction between “wanting” and “liking” arises, as discussed earlier. Impulsivity, then, is a similar improper valuation of future events in favor of choices that result in immediate outcomes. In this sense, Redish’s conceptualization is unique in that there is no single “culprit,” such as the habit system, that is responsible for the transition into addiction (Ahmed 2008; see chapters by Redish et al., and Cornwell et al., in this volume for further discussions of multiple vulnerabilities in motivational systems). Each of the predominant models of addiction explains one or more, but not all, of vulnerabilities to addiction. The advantage of describing multiple vulnerabilities is that it acknowledges that individuals likely become addicts for different reasons and that the biological underpinnings of drug-taking may depend on which vulnerability is being measured. Further, different drugs may be influenced by these vulnerabilities to different degrees. For example, opioid drugs may exploit the opponent process and positive reinforcement (reward-mimicking) vulnerabilities more than psychostimulant drugs, which may act more on the incentive salience and habit vulnerabilities. Nicotine, which facilitates learning by enhancing the reinforcing properties of cues (Caggiula et al. 2009), may drive drug-taking and relapse by enhancing initial learning about these cues.

8 Conclusion

There has been a great deal of focus on how personality traits or other pre-existing factors can influence drug-taking behavior, including sensation-seeking, impulsivity, and incentive–salience attribution (Flagel et al. 2009; Bickel et al. 2012; Bardo et al. 1996; Everitt 2014). These studies demonstrate that the predisposition to take drugs is not due to a single factor. Similarly, in this chapter, we have summarized some basic motivational concepts underlying drug-taking and how these processes are dysregulated as addiction develops. Behavioral neuroscience views largely ignore, or even avoid, the heterogeneity of motivational processes in addiction and are limited by studies in which only a single concept is tested. This point is illustrated in part by Fig. 2, in which we graphed the collaboration network of the references cited in this chapter. Each researcher cited is a node, and two researchers are connected if they have coauthored a paper together. A few prominent groups emerge that correspond to the main concepts discussed in the chapter. What is apparent is that there are very few connections between these groups and only a few researchers bridge these concepts. Of course, this is not a complete representation of the entire addiction field. However, the absence of cross-concept collaborations and integration is a major impediment to progress. Human studies strongly support the idea that different addicts take different drugs for different reasons and that these reasons may change throughout the course of the addict’s lifetime. Thus, ideas including hedonic allostasis, incentive sensitization, decision making and learning, impulsivity, and dual-process theory are applicable in some but not all situations. For example, much of what is known about negative reinforcement and allostasis comes from studies on opiates and alcohol, while incentive sensitization stems primarily from studies of psychostimulants (though see Robinson and Berridge 2000). Therefore, it is perhaps not surprising that none of these theories explain all aspects of addiction. Furthermore, it is likely that the transition into addiction is not driven by a single process, and people may transition to compulsive drug-taking for different reasons, both behavioral and neural. We believe that a unified framework of addiction is not a red herring. Instead, a framework that is flexible and adaptable to different drug classes, personality types, and environmental situations will be enormously useful in the development of specialized treatment for the wide variety of addicted individuals in the population.

Fig. 2.

Collaboration network of the references cited in this chapter. Each researcher cited is a node, and two researchers are connected if they have coauthored a paper together