The first line of empirical support comes from experiments that examined the decision-making capacity of individuals with substance dependence problems using laboratory instruments that measure decision-making, such as the Iowa Gambling Task (IGT). The neural circuitry that is critical for processing emotions (or somatic state activation) overlap considerably with that subserving decision-making, as measured by complex laboratory decision-making tasks, such as the IGT. In other words, performance on this task is impaired by damage to various key structures that make up somatic marker circuitry. Alternatively, performance of this task activates neural components of the somatic marker circuitry in functional neuroimaging studies. In a PET activation study, which examined patterns of brain activation during IGT performance in healthy participants (Ernst et al., 2002
), the authors observed that decision-making was associated with increased activation in the VMPC, anterior cingulate, parietal/insular cortices and the amygdala, predominantly on the right side. Later imaging studies have confirmed and extended these findings, and implicated additional neural regions, e.g., the striatum (Verney et al., 2003
), and the nucleus accumbens (Mathews et al., 2004
), in processes that are critical for decision-making.
In substance dependent individuals (SDI), Bolla et al. (2003)
used oxygen labelled PET to examine brain activation in 25-day abstinent cocaine dependent patients while performing the Iowa Gambling Task (IGT). Group analyses showed increased activation during IGT performance in the right orbitofrontal cortex, and less activation in the left dorsolateral cortex of cocaine patients, with regard to healthy subjects. Activation of the orbitofrontal cortex was directly correlated with better performance on both groups, and negatively correlated with amount of cocaine used in the patient group. In another study, Ersche et al. (2005)
tested current opiate and amphetamine users and ex-users in the Cambridge Gamble task. SDI and matched healthy comparison participants were subjected to oxygen labelled PET while performing this decision-making task. Results revealed that drug abusers performing the risk task showed increased activation of the left orbitofrontal cortex and decreased activation of the right dorsolateral cortex (identical localization but reversed lateralization with respect to the results of the Bolla et al. study). No significant differences were found between the current users consuming pharmacologically different components (amphetamine vs. opiates). A more recent study scanned three different groups including polydrug abusers with pathological gambling, polydrug abusers without pathological gambling and healthy controls while performing a fMRI version of the IGT (Tanabe et al., 2007
). Results were broadly consistent with the somatic marker framework; both groups of drug abusers showed reduced task related activation in the VMPC, the key structure to trigger somatic states associated with long-term prospects, when compared to controls. In addition, only non-gambling drug abusers had reduced superior frontal and frontal pole activation in comparison with gambling drug users and controls. This finding was supported by behavioral results, where both gamblers and controls outperformed drug abusers for the IGT learning curve.
Other studies have examined patterns of brain regional activation in abstinent methamphetamine abusers and young stimulant recreational users using a two-choice prediction task (Paulus et al., 2002
). This task also taps into decision-making function under conditions of uncertainty, by requiring the prediction of an uncertain outcome, which can be predicted correctly (success) or incorrectly (failure). However, unlike the IGT or Cambridge Gamble tasks, this task does not involve incentive evaluation of rewards and punishments. Overall, the behavioural results of these studies demonstrated a more rigid stimulus driven decision-making pattern in the methamphetamine group, as opposed to a more outcome driven pattern in the healthy participants. Imaging patterns in methamphetamine individuals while performing the two-choice prediction task showed decreased activation of the orbitofrontal, dorsolateral, insular and inferior parietal cortices; orbitofrontal activation was inversely correlated with duration of methamphetamine use (Paulus et al., 2003
). This decreased activation was particularly observed in low error-rate phases of the task (i.e., when the subjects successfully predicted the correct outcomes). Interestingly, the patterns of brain activation during this task were strong predictors of amphetamine relapse in a 1-year follow-up study (Paulus et al., 2005
). In a later study that tested young stimulant recreational users employing the same task, results showed that these users were less able to flexibly modify their choice pattern in response to the frequency of errors, which was related to the degree of insular activation (Paulus et al., 2007
3.1. Behavioural and Physiological Studies
If decision-making is a process guided by emotions (Bechara, 2004
; Damasio, 1994
), then there should be a link between abnormalities in expressing emotions and experiencing feelings in substance dependent individuals (SDI) on the one hand, and impairments in decision-making on the other hand.
The most frequently used paradigm to assess decision-making is the Iowa Gambling Task (IGT) (Bechara et al., 2001
; Ernst et al., 2003
; Grant et al., 2000
; Monterosso et al., 2001
; Petry et al., 1998
), which was initially developed to investigate the decision-making defects of neurological patients with VMPC damage, and to provide empirical support for the somatic-marker hypothesis. This task factors a number of aspects: immediate rewards and delayed punishments, risk, and uncertainty of outcomes (Bechara, 2003
). The task has been described in detail elsewhere (Bechara et al., 1994
); briefly, in this task, the participants have to choose between decks of cards that yield high immediate gain but larger future loss (long-term disadvantageous decks), and decks that yield lower immediate gain but a smaller future loss (long-term advantageous decks). A number of studies that used this paradigm have shown impairments in decision-making performance among alcohol, cannabis, cocaine, opioids, MDMA and methamphetamines abusers (Bechara et al., 2001
; Fein et al., 2004
; Grant et al., 2000
; Hanson et al., 2008
; Quednow et al., 2007
; Verdejo-García et al., 2007a
; Whitlow et al., 2004
). Decision-making deficits have also been reported in populations who are at high risk for drug abuse, such as drug-naive individuals with higher familiar density of alcohol abuse (Fein and Chang, 2008
; Lovallo et al., 2006
), adolescents with externalizing behaviour disorders (Ernst et al., 2003
), or adolescent binge drinkers (Johnson et al., 2008
; Goudriaan et al., 2007
). Interestingly, impaired decision-making has been observed also in individuals with Antisocial Personality Disorder (APD) (Mazas et al., 2000
; van Honk et al., 2002
), a psychiatric disorder that is robustly associated with substance dependence, and which involves severe disturbances in emotion processing (Blair and Utah, 2000
); and pathological gambling, a form of behavioral addiction where the neurotoxic effects of drugs are minimized (Goudriaan et al., 2006
Additional evidence for impaired decision-making in SDI stems from studies that used different decision-making paradigms, including tasks of delay discounting (Kirby et al., 1999
; Monterosso et al., 2001
), tasks that involve betting, such as the Cambridge Gamble task (Rogers et al., 1999
), and probabilistic choice tasks (Heiman and Dunn, 2002
; Paulus et al., 2002
). Although the behavioural results from these studies reveal evidence for impaired decision-making among SDI, they do not necessarily link poor decision-making capacity to any abnormality in emotional processing. However, other psychophysiological studies have attributed these decision-making impairments to defective processing of affective/emotional signals (or somatic markers) (Bechara et al., 1997
; Crone et al., 2004
). Furthermore, cognitive modelling analyses (Stout et al., 2004
) strongly support the notion that affective/emotional factors associated with the processing of reward and punishment play a significant role in the decision-making performance of SDI on the IGT.
More specifically, in a series of studies using the IGT, the performance of SDI was compared to that of patients with damage to the VMPC (Bechara and Damasio, 2002
; Bechara et al., 2002
). These studies also included physiological measures of autonomic activity before and after making a choice in the IGT. The physiological responses triggered after making the choice and seeing the outcome (i.e., gain or loss of a certain amount of money) were called (i) reward/punishment responses; and those generated before making the choice were called (ii) anticipatory response, i.e., responses triggered during the time the participant was pondering from which deck to chose. Good performance in the IGT has been shown to be linked to the development of these anticipatory emotional responses, which in this case were changes in the skin conductance response (SCR), especially before selecting cards from the disadvantageous decks (A and B). It was suggested that these anticipatory emotional responses help guide decision-making away from disadvantageous choices (i.e., avoid decks A and B, and choose from decks C and D) (Bechara et al., 1997
). In a study that used the IGT to measure behavioural decisions, and the skin conductance response to measure anticipatory and reward/punishment responses in SDI, the behavioural and physiological results revealed that there were at least two different subgroups within this SDI population (Bechara and Damasio, 2002
). One subgroup of SDI (a minority of the sample) showed a behavioural profile similar to that of healthy participants, i.e., they selected more cards from the advantageous decks. They also showed a physiological profile similar to healthy participants, in that they began to trigger anticipatory SCRs before selecting cards from the bad decks. By contrast, another subgroup of SDI (a majority of the sample) exhibited behavioural and physiological profiles that were different from healthy participants, and more similar to patients with VMPC damage, i.e., they chose disadvantageously on the task, and they failed to acquire anticipatory SCRs.
In a subsequent study, a variant version of the IGT was used, in which they reversed the order of reward and punishment contingencies, so that the advantageous decks yielded high immediate punishment but even higher future reward, and the disadvantageous decks had lower immediate punishment, but even lower long-term reward (Bechara et al., 2002
). The combination of the behavioral results from the original and variant tasks, in conjunction with physiological responses recorded during the performance of both tasks, helped discriminate between two other sub-groups: those who were hypersensitive to immediate reward and those who were insensitive to long-term consequences. In this way, they identified three sub-populations of SDI: one small sub-population of SDI that was indistinguishable from healthy participants, a second small subpopulation that was indistinguishable from VMPC lesion patients, and a third larger sub-population of SDI that was different from the other two; these SDI exhibited signs of hypersensitivity to reward, as evidenced by impaired performance on the original IGT, normal performance on the variant version, and abnormally high reward SCR in both tasks. Interestingly, these subpopulations did not differ in terms of basic neuropsychological abilities or clinical characteristics such as severity of drug use. A subsequent study has replicated this pattern of behavioral and physiological response using the Cambridge Gamble Task paradigm (Fishbein et al., 2005
) that was designed to isolate different components of the IGT (see Clark and Robbins, 2002
). In this study, polydrug users selected more risky choices in the high-risk conditions of the task, and failed to generate increased SCR responses when making riskier decisions with regard to healthy participants. Similar results have been obtained in a sample of pathological gamblers with negligible exposure to alcohol and drugs (Goudriaan et al., 2006
It is important to note here that the problem in one of the subpopulations of SDI is not apparently linked to a primary dysfunction in the VMPC (or reflective) system itself, but rather an overactivity in the impulsive system, thereby exaggerating the incentive impact of reward stimuli. This deficit is still consistent with the proposed somatic marker model of addiction, except that the deficit is related to a different part of the neuronal circuitry. The one surprising finding is that one subpopulation of SDI seemed normal and did not exhibit any signs of “somatic marker” deficits. As a result, one argument can be made that the proposed somatic marker model does not explain all instances of addiction. We argue that the somatic marker model should explain all instances of addiction, except that this particular subgroup should not be considered as true addicts, although they meet the DSM-IV diagnostic criteria for drug dependence. We have suggested the use of the term “functional addicts” since this particular subgroup seem functional in their real-life (Bechara et al., 2002
). Nonetheless, further research should be conducted to investigate if the emotional/decision-making profile of this subgroups is associated with better prognosis of their dependence problems.
Importantly, decision-making deficits are strong predictors of clinical outcome and relapse in addiction. Recent studies in alcohol and opiate users have demonstrated that poor performance on the IGT and the Cambridge Gamble Task predicts treatment retention and the probability of relapse (Bowden-Jones et al., 2005
; Passetti et al., 2008
). In addition, imaging studies have revealed that patterns of brain activation within the neural substrates of the somatic marker model can prospectively predict drug use relapse as far as one year after cessation of drug use (Grüsser et al., 2004
; Paulus et al., 2005
There are relatively few studies that examined emotional perception and experience, and its relationship to decision-making, in SDI. Most studies on emotional perception have focused on analyzing possible alterations in the processing of emotional facial expressions in long-term substance abusers, which seem to vary depending on drug of choice (Hoshi et al., 2004
; Kemmis et al., 2007
; Kornreich et al., 2001
). Cocaine abusers have particular difficulties in recognizing emotional expressions of fear (Kemmis et al., 2007
; Verdejo-García et al., 2007
). In the first study, recently abstinent regular cocaine users had lower recognition hits for expressions of fear than occasional cocaine users and healthy controls. In the latter study, cocaine polysubstance users abstinent for at least four months had lower recognition hits for expressions of fear, surprise, and the global recognition score (including all emotions). Furthermore, the fear recognition and global scores were positively correlated with decision-making performance in the IGT, supporting the link between emotion processing and decision-making. In alcoholics, several studies have revealed significant alterations in the processing of facial expressions, although the range of emotions affected is still controversial. One study showed that alcohol dependent individuals showed specific impairments for recognizing facial expressions portraying happiness and anger (Kornreich et al., 2001
). These alterations were characterized as overestimation of the intensity of the emotion displayed. By contrast, other studies showed that overestimation of the intensity of emotion in facial expressions reported by alcoholics related mainly to the facial expression of fear (Townshend and Duka, 2003
); the degree of this overestimation correlated with the number of prior detoxifications. In this study, alcoholics also presented with difficulties in distinguishing between the facial expressions of anger and disgust. The significant alteration in anger and fear processing among SDI was supported by a recent fMRI study, where alcoholics showed lower task-related activation in several regions of the somatic marker neural system, including the subgenual anterior cingulate cortex (which is included in what we define here as the VMPC region), the insular cortex, the amygdala, and the striatum, when exposed to angry and fearful facial expressions (Salloum et al., 2007
). Alcohol abusers also have problems in identifying prosody in sentences with incongruent semantic content, and in matching affective prosody with facial expression (Uekermann et al., 2005
). In opiate users, results have shown a generally slower reaction time and poorer emotion recognition accuracy (Kornreich et al., 2003
; Martin et al., 2006
). Other studies have analyzed the effects of controlled acute doses of different drugs on the perception of emotions. These studies have shown that acute low doses of alcohol and MDMA can improve the recognition of emotional facial expressions in current users, although recognition accuracy significantly decreased during the following days (Hoshi et al., 2004
; Kano et al., 2002
). Detrimental effects of acute drug doses on the recognition of emotions in facial expressions have also been reported using ketamine (Abel et al., 2003
). These results indicate that SDI are impaired in the recognition of facial expressions portraying different emotions, including fear, anger, disgust, and happiness. The poorer recognition of facial emotional expressions can affect SDI’s interpretation of social cues, so that they can be less able to manage and regulate emotions, and to make decisions and solve problems of an interpersonal or social nature. In this sense, their poor ability to recognize facial emotional expressions has been attributed to several aspects of their addictive behaviors, such as diminished empathy, increased levels of aggression (Hoshi et al., 2004
), and a higher frequency of relapse and ensuing alcohol detoxification (Townshend and Duka, 2003
). In particular, poor recognition of fear expressions, which is thought to depend on the amygdala, can be associated with impaired conditioning of fear responses to drug related environments, increasing the probability of relapses.
Finally, very few studies have examined the emotional experience of SDI. The most frequently used paradigm in the study of the experience of emotions in SDI is the presentation of affective images that induce emotional states, such as the International Affective Picture System (IAPS). The IAPS consists of a large set of images classified according to their normative values in three relevant dimensions: valence (indicating if the emotional response induced is pleasant or unpleasant), arousal (if the emotional response induced is arousing or relaxing), and control (if the emotional response induced can/can not be controlled by the subject). Gerra et al. (2003)
used this paradigm to analyze the neuroendocrine response of SDI and healthy participants to experimentally induced pleasant and unpleasant emotions. Their results showed that in response to unpleasant images, SDI showed decreased activity in several neuroendocrine markers, including norepinephrine, cortisol, and adrenocorticotropic hormone levels. Similar results have been obtained using a different response modality, the subjective affective response to IAPS images (Aguilar de Arcos et al., 2005
). SDI showed a more flattened response pattern to both pleasant and unpleasant images. SDI scored as less positive the images considered by normal participants to be very pleasant and arousing. SDI also scored as less negative the images considered by normal participants to be highly unpleasant and arousing. A subsquent study used the same IAPS images to examine the emotional experience of current heroin users enrolled in a clinical trial and using heroin as a prescribed treatment. It was observed that these current heroin users had even lower arousal response to pleasant affective images when compared to both abstinent opioid users and healthy individuals (Aguilar de Arcos et al., 2008
). The fact that SDI showed such a flattened emotional response to pleasant affective images may suggest that they also have a diminished emotional response to natural reinforcers in general, except for drugs, which begin to exert exaggerated rewarding effects. This notion is strongly supported by imaging studies on craving in drug addiction, which show that drug related stimuli are able to strongly activate brain regions involved in emotional evaluation and reward processing (Garavan et al., 2000
; George et al., 2001
; Grant et al., 1996
; Kilts et al., 2001
; Wang et al., 1999
; Wexler et al., 2001). In contrast, the same brain regions show blunted activation to other natural reinforcing stimuli such as food or sex (Garavan et al., 2000
). Although these studies are consistent with Somatic Marker Model in that neural systems supporting affective and emotional processing become altered among SDI, the proposed model does not really explain the specificity of these alterations. In other words, the proposed model does not really explain why, for example, the trigger structures for somatic states begin to exaggerate the reward signals elicited by drug related cues, at the expense of blunting the reward signals from non-drug related rewards. However, the implications of these differences are such that somatic states associated with natural reinforcers may not be strong enough to bias decisions in SDI, while strong somatic states associated with the prospect of abusing drugs become so powerful to drive decisions towards drug use.
3.3. Somatic Markers and Neural Correlates of Craving
People still disagree on how craving should be defined (Franken, 2003
; Tiffany, 1999
). We will define craving as the accompanied emotional state that is produced by emotionally competent stimuli that are associated with the reinforcing effects of drugs of abuse (Bechara, 2003
). To study craving in the laboratory, a number of studies have used different strategies to induce craving. These strategies include the visual presentation of the drug, or drug related paraphernalia, through images (pictures), or videos; the auditory presentation of audio-taped scripts containing autobiographical experiences related to drug use; or the direct infusion of the drug itself. The aim of this section is to review the evidence showing an overlap between the neural structures that represent critical components of the somatic marker neural circuitry, and neural structures activated during states of craving in substance abusers.
Grant et al. (1996)
used PET to study craving in recently abstinent cocaine abusers. In the scanner, participants were shown videos of drug related paraphernalia, and cocaine self-administration. A pattern of significantly increased activation associated with craving was detected primarily in frontal (VMPC and dorsolateral –DLPC), parietal, temporal, and striatal regions. The correlation analyses showed that the subjective response of craving was associated with changes in the activation of the amygdala in cocaine abusers. A follow up study by the same group (Bonson et al., 2002
) used auditory presentation of drug cues, i.e., a script describing sensations associated with being “high” on cocaine. The results revealed increased activation in the DLPC, orbitofrontal cortex, amygdala and adjacent entorhinal cortices. The correlation analyses showed that the subjective response of craving correlated with the degree of activity within these neural regions.
Numerous subsequent PET studies in cocaine abusers have also revealed a significant overlap between the neural regions activated during craving and those considered as components of the somatic marker neural circuitry. Wang et al. (1999)
detected higher metabolism in the orbitofrontal and insular cortices of abstinent cocaine abusers after an interview in which participants were allowed to manipulate drug paraphernalia. Furthermore, activation of the left insular cortex was significantly correlated with self-reported craving. Similarly, Childress et al. (1999)
found amygdalar and anterior cingulate activation in cocaine abusers exposed to a drug-related video.
In two subsequent PET studies, Kilts et al. (2001
examined the craving response of cocaine dependent men and women to auditory scripts describing events from their own previous drug experience. These studies reveal that there are important gender differences in regional patterns of brain activation in relation to drug cue induced craving. However, irrespective of these differences, the general neural circuitry engaged during craving, in both males and females tend to reveal activation within neural regions that are considered critical components of the neural circuitry underlying somatic state activation and decision-making.
Using a different neuroimaging technique (functional magnetic resonance imaging, fMRI), Garavan et al. (2000)
contrasted the responses of crack-cocaine dependent individuals to films containing natural outdoors, sexual explicit, and drug use related scenes. The drug-associated film induced increased activation of an extensive neural circuitry that included the prefrontal and parietal regions, temporal, insular, anterior and posterior cingulate cortices, and the striatum. A similar pattern of activation was observed in healthy participants when exposed to the film containing explicit sexual scenes (an emotionally competent natural reinforcing stimulus), which contrasted with the fainter activation of these regions in the cocaine abusers. These findings suggest that cocaine dependent individuals present with a reduced sensitivity to the rewarding properties of natural reinforcers, while at the same time, they present with an increased sensitivity to drug related stimuli. Although the somatic-marker model does not explain to the differential response to drug related stimuli versus non-drug stimuli or natural reinforcers, this abnormal capacity to process the emotional value of a stimulus has a significant impact on decision-making, in that it can shift the decision-making process towards short-term horizons, i.e., the seeking of drugs. In support, several fMRI studies have shown an exaggerated brain response to drug cues (George et al., 2001
; Maas et al., 1999; Tapert et al., 2004
; Wang et al., 1999
; Wexler et al., 2004), even when they are presented at a pre-attentive level (Childress et al., 2008
The brain activation studies on craving were not restricted to individuals who abused stimulant drugs, but similar findings were obtained from individuals addicted to opioids (Daglish et al., 2002, 2003
; Langleben et al., 2008
). A relevant finding from these studies was the correlation between length of abstinence and anterior cingulate activation, i.e. longer abstinence duration predicted larger cerebral blood flow changes in the anterior cingulate. This finding seems to suggest that the affective evaluation of drug related stimuli do not decrease. Rather it increases after prolonged abstinence, thus suggesting a persistent sensitization effect.
To mimic the acute dopamine response induced by cocaine, Volkow et al. (1999)
injected two sequential doses of the drug methylphenidate to a sample of cocaine dependent individuals. The correlation analyses showed that changes in the subjective report of craving were significantly correlated with changes in the right striatum and orbitofrontal cortex. Further studies using raclopride traced PET in cocaine dependents have shown that craving was specifically associated with the rate of D2 receptor occupancy in the dorsal striatum (Volkow et al., 2006; Wong et al., 2007). Another pharmacological study showed that cocaine patients, compared to controls, had increased response to a procaine challenge in the right orbitofrontal, midfrontal, midtemporal, and parietal cortices, and in the brainstem (Adinoff et al., 2001
). In contrast, saline administration was associated with deactivation of the orbitofrontal region in the cocaine patients. There was also a trend of a significant relationship between duration of cocaine abuse and right orbitofrontal cortex change rate in response to craving.
Another set of studies has directly employed the administration of the abuser’s drug of choice to evoke craving states. Breiter et al. (1997)
administered repeated cocaine infusions to examine the “rush” and “craving” patterns of brain activation using fMRI in cocaine dependent patients. Their results showed a wide array of regional activations on regions involved in the ongoing evaluation of the emotional significance of emotionally competent stimuli, and the modification of behaviour through somatic signals (Fresnois et al., 2005
; Kilts, 2001
). Similarly, Kufahl et al. (2005)
showed that cocaine users had significantly increased activation in the anterior prefrontal cortex, striatum, amygdala and ventral tegmental area after administration of a 20 mg/70 kg cocaine infusion. Using a cocaine “self-administration” paradigm during fMRI (i.e., participants were asked to press a buttom each time they wanted to receive a small cocaine dose, although the doses were indeed constantly administered at 5 minutes intervals) Risinger et al. (2005)
also showed prominent activation of the orbitofrontal cortex, anterior cingulate, insula, striatum, thalamus and cerebellum which were correlated with self-reports of “high” and “craving”. In another study (Sell et al., 2000
), a combination of heroin injection and visual presentation of films containing drug related scenes was used to induce craving in a sample of heroin dependent individuals undergoing methadone treatment. The correlation analyses showed that the subjective response of “urge to use heroin” was significantly associated with increased activation in the inferior frontal, right orbitofrontal, and insular cortices of the patients. The insula activation was significantly correlated with physiological measures of pulse rate, outlining its role in autonomic regulation. Significant activations were also observed in the brainstem in response to both the heroin injection and the drug-related film.
Together, these studies reveal a great overlap between the neural structures known to be critical components of the somatic-marker neural circuit, and the neural systems engaged in emotional states such as craving in substance abusers. Activities within the orbitofrontal cortex (included in what we define here as VMPC) and amygdala, two key areas for triggering somatic states, were shown to correlate with subjective self-reports of craving across several different studies (Adinoff et al., 2001
; Bonson et al., 2002
; Grant et al., 1996
; Tapert et al., 2004
). Areas involved in body mapping and emotional representation, as well as the generation of feelings, such as the insula and anterior cingulate, were also engaged during the experience of craving (Breiter et al., 1997
; Daglish et al., 2001
; Wang et al., 1999
Most of the regions activated during the experience of craving induced by drug cues in SDI were also engaged in healthy participants when they were exposed to natural reinforcers, such as sex cues (Garavan et al., 2000
). This suggests that the neural substrates that mediate craving in drug addicts have actually evolved to subserve natural emotional functions, such as those related to food and sex. Nonetheless, it is important to note that drug addicts present with some degree of dissociation between the emotional responses to drug versus non-drug related stimuli. In other words, drug addicts tend to trigger strong emotional responses (or somatic states) in response to drug related cues, but they trigger relatively weak somatic states in response to non-drug related, natural reinforcers.