While a great deal of current thought about drug addiction is dominated by the concept of "the hijacked brain" altered by its exposure to drugs, such theories have little to offer to the clinician. On the other hand, the self-medication hypothesis, which proposes that addicts are using their drug of choice to relieve symptoms of an underlying disorder or condition (such as stress), provides the clinician with a useful conceptual model to guide treatment. The often mystifying problem of relapse is also clearly understandable when viewed from the perspective of the self-medication hypothesis.

The greatest inadequacy of the "hijacked brain" model of addiction is its inability to explain why most users of drugs do not become addicted (Anthony & Helzer, 1991; Anthony, Warner, & Kessler, 1994). Population studies show that only about one out of every five users of cocaine becomes addicted or develops the sort of problems that would justify a DSM diagnosis of substance abuse or dependence. The rate of alcoholism or alcohol abuse in alcohol users is about the same. For users of heroin and marijuana the rates of addiction is more like one out of ten.

Edward J. Khantzian and David F. Duncan are usually credited with being the originators of the self-medication hypothesis but while theirs are the most fully developed versions there are earlier precedents for the idea. Both Fenichel (1945) and Rado (1957) pointed to an underlying depression as the motive for drug abuse. Glover (1956) went further in suggesting that drugs were used to cope with overwhelming and psychotogenic aggresssion and rage. Rosenfeld (1965) described drug addiction as a defense against psychotic suffering. None of these contributions, however, have had the degree of influence that Duncan and Khantzian have.

Khantzians Model of Drug Abuse as Self-Medication

One of the two major versions of the self-medication hypothesis is the psychoanalytic perspective developed by Edward J. Khantzian and his colleagues at Harvard Medical School. This model arose from Khantzians clinical experience evaluating and treating heroin addicts. He noted that his addict patients had histories of difficulties with aggression and derivative problems of rage and depression that long preceded their use of any illegal drugs. He also found that many of them reported that use of heroin gave them relief from dysphoric feelings of restlessnees, anger, and rage.

Khantzian concluded that the predisposition to become a heroin addict resulted from these problems with aggression -- specifically from inadequate ego mechanisms for controlling and directing aggression. The repeated use of heroin or other opiates as a means of coping with the addicts poorly controlled aggressive drive result in the development of physical dependence. Methadones effectiveness in treating heroin addiction, he argued, is due not only to its prevention of withdrawal illness but also due to its relief of those same dysphoric feelings. This limited formulation of what would become the Self-Medication Hypothesis was published by Khantzian, Mack, and Schatzberg in 1974 in the American Journal of Psychiatry.

Eleven years later the original hypothesis about heroin addiction (Khantzian, Mack, & Schatzberg, 1974) was named the Self-Medication Hypothesis and was elaborated to include cocaine addiction as well (Khantzian, 1985). He now speculated that cocaine has its appeal because of its ability to relieve the distress associated with depression, hypomania, or hyperactivity. The hypothesis was subsequently expanded to include alcoholism, speculating that the addicts use of alcohol, permits the experience of affection, aggression, and closeness in an individuals who is otherwise cut off from their feelings and relationships (Khantzian, 1990). And finally developed into a theory of all drug addictions (Khantzian, 1997 & 1999).

In its fully developed version, Khantzians version of the Self-Medication Hypothesis holds that addiction occurs in a context of self-regulation vulnerabilities -- primarily difficulties in regulating affects, self-esteem, relationships, and self-care. Potential addicts suffer severely from their feelings, either being overwhelmed with painful emotions or seeming not to feel any emotions at all. Drugs of abuse help such individuals to relieve painful emotions or to experience emotions that are confusing or threatening.

Regardless of specific symptoms or personality styles, Khantzian believes that certain character traits are typical of all drug addicts. These include problems in affect management, self esteem, object relations, judgment, and self-care. He argued that these developmentally and structurally determined problems predispose individuals to drug dependence because they are the basis of the distresses that are relieved by drug-taking.

Vulnerability to drug dependence varies greatly both between individuals and for the same individual at different times. In many cases addiction develops in a previous abstainer (or perhaps even in a non-dependent drug user) following some precipitating event that involves a severe crisis in which the individuals adaptive capacities are diminished and narcissistic vulnerability is intense.

Once drug taking has been initiated in a susceptible state and the user has experienced adaptive benefit from the use of the drug, a variety of other processes are set in motion that often lead to addiction. These processes include both regressive effects that can exacerbate the original vulnerability and progressive effects that promote stable functioning. In both cases, however, Khantzian argued that this may interfere with further emotional maturation, particularly when the onset of addiction occurs in adolescence.

The addict's choice of drug, according to Khantzian, was a result of the interaction between the psychopharmacologic properties of the drug and the "primary feeling states" the addict was seeking relief from. The drugs effects substitute for defective or non-existent ego mechanisms of defense. The addicts "drug of choice," therefore, is neither random nor simply the result of current fashion or fad, but rather, results from a process of "self selection" that has been referred to as "preferential drug use" (Milkman & Frosch, 1973) Thus, narcotic addicts prefer opiates because of the relief they provide from the disorganizing and threatening affects of rage and aggression. Cocaine, in turn, has its appeal due to its ability to relieve the distress associated with depression, hypomania, or hyperactivity.

According to Treece and Khantzian (1986) the development of drug dependence involves the gradual incorporation of the drug effects and the need for them into the defensive structure-building activity of the ego itself. Overcoming an addiction, therefore, involves dealing with the unconscious and conscious components of this outcome. The user must be able to relinquish behaviors and drug effects that have come to be experienced as a valued (even if also hated) part of the self-capacity to function, cope, and be comforted in distress.

The second major version of the self-medication hypothesis was developed by David F. Duncan of the Texas Research Institute of Mental Sciences. Duncan also based his hypothesis on extensive clinical experience with addicts. Whereas Khantzians formulation was rooted in psychoanalytic theory, Duncan took a behavioristic perspective.

Duncans version of the self-medication hypothesis was initially published in two papers in 1974. The first (Duncan, 1974a) was a discussion of reinforcement of drug abuse that appeared in the Clinical Toxicology Bulletin. The second (Duncan, 1974b) was a response in the American Journal of Psychiatry to Khantzian, Mack, and Schatzbergs (1974) paper on heroin use as a coping mechanism. It was in this second publication that Duncan compared drug abuse to a bandaid. A further elaboration of the hypothesis appeared the following year in an invited paper for the Journal of Psychedelic Drugs (Duncan, 1975).

The distinction between drug use and drug abuse is critical to Duncans model. He asserts that most of the people who take illegal drugs do not meet criteria for substance abuse, let alone for dependence. The great majority exercise control and restraint over their drug use and suffer no harm from using. His model is concerned with describing why a minority (10-20%) of those who take drugs non-medically do lose control over their use and expose themselves to serious social, interpersonal and medical riskss. He and Dr. Robert S. Gold have also explored the implications of his model for the primary prevention of drug abuse (Duncan & Gold, 1982 & 1983) and the tertiary prevention of the destructive consequences of drug abuse (Duncan,et al, 1994).

Duncan was by no means the first to develop a behavioral model of addiction. Other formulations of drug abuse and addiction as an operant behavior have dealt with the euphorogenic effects of drugs as positive reinforcement and avoidance of withdrawal sickness as negative reinforcement (Valdman & Zvartau, 1982; Schulteis & Koob, 1996; Bigelow, Brooner, & Silverman, 1998). Duncan, on the other hand, asserted that drug use is maintained by positive reinforcement (principally from the pleasurable effects of the drug) but that drug dependence is not. It is drug dependence that Duncans version of the self-medication hypothesis attempts to explain.

Drug dependence is the term which has formally replaced addiction in medical terminology. In 1964 the World Health Organizations Expert Committee on Drug Abuse proposed that the terms addiction and habituation be replaced with the term dependence and distinguished between two types psychological dependence and physical dependence. Psychological dependence "refers to the experience of impaired control over drug use" while "physical dependence involves the development of tolerance and withdrawal symptoms upon cessation of use of the drug, as a consequence of the body's adaptation to the continued presence of a drug" (UNIDCP, 1998). In Duncans view, physical dependence is a lesser problem which occurs in addiction to some drugs, such as alcohol and opiates, but not in addiction to cocaine, heroin, etc., nor in non-drug addictions such as compulsive gambling.

Duncan essentially argues that drug dependence is just another name for avoidance or escape behavior when the operant behavior being reinforced is drug taking. Drug addicts, in his view, have found a drug which provided them with temporary escape from an ongoing state of emotional distress which might be due to a mental disorder, to stress, or to an aversive environment. Non-drug addictions, in his opinion, represent similar negatively reinforced behavior.

Duncan asserts that the characteristics of dependence are all typical of operant behaviors maintained by negative reinforcement. Negative reinforcement maintains high rates of behavior. Animals that have been negatively reinforced for performing a task such as pressing a bar will often do so to the exclusion of eating, sleeping, sexual activity, etc.. Avoidance behaviors are highly resistant to extinction and even when they appear to have been eliminated they tend to recur spontaneously. Thus the intensity, compulsiveness and proneness to relapse that characterize drug dependence all result, in Duncans opinion, from the fact that the behavior is maintained by negative reinforcement.

Dackis and Gold (1984, 1985) assert that depression in cocaine addicts, rather than being the cause of the addiction, is a direct result of abstinence symptomology. Withdrawal illness encourages increased cocaine use, which in turn results in alterations in brain chemistry (dopamine depletion). They conclude that the addiction itself is the cause of painful emotional states rather than a response to them.

Cocores et al., (1987) also advance a dopamine deficiency hypothesis as an alternative to the self-medication hypothesis, in this case as an explanation of the apparent correlation between attention deficit hyperactivity disorder (ADHD) and chronic cocaine abuse. They assert that cocaine further depletes dopamine in already dopamine-compromised individuals. They argue that the resultant dopamine deficiency may then induce a temporary and reversible ADHD even in those with no prior history of the disorder. Thus they also argue that the Self-Medication Hypothesis has causal directionality backwards and that the comorbid psychiatric condition follows rather than preceding addiction.

In the opinion of Goldsmith (1993) the Self-Medication Hypothesis ignores important biological research which has explored the mechanisms of reward, motivation to use drugs, and the impact on mood of chronic, excessive drug use. He argues that a new psychology of addiction is needed which includes this biological research as well as the psychological observations contained in the self-medication hypothesis. He believes that Self Psychology will provide the basis for such an integrated psychology for the addictions.

Frances (1997) urges that additional longitudinal research is needed on the Self-Medication Hypothesis. He warns that there is a danger that clinicians could use the hypothesis in overly reductionistic ways and that they must be alert to potential rationalization and retrospective distortion when listening to their patients' accounts of the causes of their substance abuse.

In particular, Frances warns that this is a "cart/horse problem" in which it is often unclear which came first the drug abuse or the disorder that the Self-Medication Hypothesis suggests underlies the abuse. He suggests that it can be clinically useful to tease apart the temporal sequence of onset of each disorder, the ways in which the two disorders interact, and the ways in which each may amplify the suffering caused by the other. In his view, it is "the wrath of grapes that leads to the greater part of suffering in substance use disorders."

Schiffer (1988) reported on a series of nine cocaine abusers successfully treated with long-term, in-depth, dynamic psychotherapy begun on an inpatient drug abuse unit and continued after hospitalization. He found his patients to have been victims of unrecognized psychological trauma in childhood. He argued that their cocaine abuse, in addition to functioning as a form of self-medication, was functioning as a component of a repetition compulsion in which old psychological traumas were symbolically recreated in the post-drug dysphoria. This perspective guided his clinical approach which involved: 1) identifying traumatic or abusive experiences in the patients history, 2) established emotional contact, 3) helping the patient to appreciate how they had been affected by the trauma. and 4) helping them to master the traumatic experiences.

Dixon et al. (1990) reported that substance abuse among schizophrenic patients is an increasingly recognized clinical phenomenon. They reviewed experimental and observed clinical effects of drug abuse and patients' subjective experiences of acute intoxication. Though drug abuse may exacerbate psychotic symptoms, abused drugs may also lead to transient symptom reduction in subgroups of schizophrenic patients. Some patients report feeling less dysphoric, less anxious, and more energetic while intoxicated. This relief from unpleasant symptoms motivates these patients to become chronic drug takers.

Silver and Abboud (1994) examined the critical issue of the relationship between the onset of drug abuse and onset of illness (defined as first hospitalization) and its correlates in 42 hospitalized schizophrenics identified as drug abusers. Sixty per cent of the patients began drug use before their first hospitalization. No differences on socio-demographic or clinical variables were found between patients who began drug use before their first hospitalization and those who began afterward. The findings are consistent with self-medication models of comorbidity of drug abuse and schizophrenia.

Comorbidity Studies in Clinical Samples

Weiss, Griffin, and Mirin (1992) examined drug effects and motivation for drug use in 494 hospitalized drug abusers. Most patients reported that they used drugs in response to depressive symptoms and that they experienced mood elevation, regardless of their drug of choice. Drug use to relieve depressive symptoms was far more likely in men if they had major depression, but was equally common in women with and without major depression.

Greene, et al. (1993) tested Khantzians assertion that the drug of choice in patients with substance dependence disorders reflects the nature of the underlying disorder or deficiency being self-medicated. Cocaine and marijuana dependent inpatients were compared using the MMPI to test this hypothesis. These two groups of patients did not differ on the standard validity and clinical scales of the MMPI, and their scores were basically similar to a group of alcohol dependent inpatients and a group of similar age psychiatric patients with non-drug disorders. There was no single MMPI code type that was characteristic of either group of substance-dependent patients. They concluded that drug of choice bore little relationship to the MMPI scores of these subgroups of substance dependent patients. Their results are not consistent with Khantzians drug of choice hypothesis or with Milkman and Froschs (1973) preferential drug use theory.

Schinka, Curtiss and Mulloy (1994) conducted a similar study utilizing the Personality Assessment Inventory instead of the MMPI. Administering the inventory to four groups of drug dependent patients they, in contrast to Greene, et al (1993), found group differences in symptomatology and personality traits. Results suggest that there are traits or symptoms that separate various groups of drug-dependent patients, but not in accordance with Khantzians predictions.

Aharonovich, Nguyen, and Nunes (2001) tested the hypothesis that opiate abusers experience difficulty managing aggression and that cocaine abusers suffer from distress associated mostly with depression. They used the State-Trait Anger Expression Inventory and the Beck Depression Inventory to examine levels of anger and depression among three groups of substance abusers -- opiate, cocaine, and cannabis abusers. Anger and depression scores were elevated, but contrary to Khantzian's hypothesis, there were few differences between groups, and if anything, opiate addicts were more depressed and the cocaine abusers were angrier on several subscales.

Abraham and Fava (1999) examined the order of onset of substance abuse and depression in a sample of depressed drug abusers. They used the Structured Clinical Interview for DSM-III-R (SCID) to assess the drug use and depression histories of 375 outpatients with major depressive disorder. They found that, on average, alcohol dependence followed the onset of first life depression by 4.7 years. Among polydrug dependent patients, each drug abused followed the onset of depression, except for LSD, which coincided with the onset of depression. Among polydrug users, cocaine dependence occurred an average of 6.8 years after the first major depressive episode and alcohol dependence 4.5 years after the onset of depression. They concluded that alcohol and cocaine use in this sample of depressed outpatients conformed to a pattern of self-medication.

Voruganti, Heslegrave, and Awad (1997) tested the hypothesis "that schizophrenic patients take to the use of illicit drugs as a way of relieving or modulating the unpleasant dysphoric feelings experienced while on neuroleptics." They studied 223 patients receiving outpatient antipsychotic drug therapy for schizophrenia. Dysphoric reactions to the medication were prevalent among 38.7% of the sample and later development of substance abuse was noted in 30% of the patients. There was a statistically significant association between the 2 conditions (odds ratio = 4.08; p < 0.001).

Anhalt and Klein (1976) surveyed illegal drug use in a population of 3,807 students at a suburban junior high school. They found that illegal drug use was strongly correlated with family instability, personal problems, and poor academic performance. Non-prescribed use of tranquilizers, amphetamines, and sedatives was often motivated by attempts at self-medicationto reduce painful feelings.

Deykin, Levy, and Wells (1987) utilized the Diagnostic Interview Schedule (DIS) to ascertain the prevalences of major depressive disorder, alcohol abuse, and substance abuse (by DSM-III criteria) in a sample of 424 college students aged 16 to 19. The prevalence of major depression was 6.8%, while that of alcohol abuse was 8.2% and that of substance abuse was 0.4%. Alcohol abuse was associated with major depression, but not with other psychiatric diagnoses. Abuse of other drugs was associated both with major depression and with other psychiatric diagnoses as well. The onset of major depression almost always preceded alcohol or substance abuse, consistent with the self-medication hypotheses.

Helzer and Pryzbeck (1988)used data from the Epidemiologic Catchment Area survey to examine the comorbidity between alcohol abuse and dependence, other substances abuse disorders and non-drug psychiatric disorders in a sample of approximately 20,000 persons drawn from the general population. Every one of the psychiatric diagnoses examined was more likely to occur in alcoholics than in non-alcoholics. Associations were particularly strong with antisocial personality disorder, other substance abuse and mania. The association between alcoholism and depressive disorders was positive but not very strong. They concluded that the impression widely reported by clinicians that depression and alcoholism are highly related is probably due to the fact that presence of depression increased the likelihood of treatment seeking by alcoholics - a fallacy known as Berksons bias.

Data derived from the National Longitudinal Alcohol Epidemiology Survey (NLAES), a national probability sample of the adult U.S. population, were examined by Grant (1995) for evidence of an association between drug use disorders and major depression. Comorbidity rates and odds ratios for associations between major depression and past-year, prior-to-past-year, and lifetime diagnoses of DSM-IV drug use disorders (i.e., prescription drugs, sedatives, tranquilizers, amphetamines, cannabis, cocaine, and hallucinogens) were calculated by gender, ethnicity. The results showed that virtually all of the odds ratios were significant, demonstrating that comorbidity of a variety of drug use disorders with major depression is pervasive in the general population. As predicted by Duncans version of the self-medication hypothesis, the association between drug dependence and major depression was greater than the association between abuse and major depression.

In further analyses of the NLAES data, Grant and Pickering (1998) examined the risk of cannabis abuse and dependence at different levels of cannabis use and in association with comorbidity with other psychiatric disorders. Two separate logistic regression analyses were conducted to determine the association between cannabis use, and abuse and dependence. The risk of cannabis abuse and dependence was found to increase with the frequency of smoking occasions and slightly decreased with age. More severe comorbidity was associated with dependence compared to abuse, suggesting that cannabis dependent persons were using cannabis to self-medicate major depression.

Gillman and Abraham (2001) used data from the first two waves of the Epidemiologic Catchment Area Study to estimate the odds of either major depression or alcohol dependence being followed by the other disorder after 1 year of follow-up. The odds of developing major depression associated with low, medium, and high levels of alcoholic symptoms at baseline were 1.66, 3.98, and 4.32 for females (P<0.001), and 1.19, 2.49, and 2.12 for males (P=0.026). Conversely, odds ratios indicating the 1-year follow-up risk of incident alcohol dependence within low, medium, and high categories of baseline depressive symptomatology were 2.75, 3.52, and 7.88 for females (P<0.001) and 1.50, 1.41, and 1.05 for males (P=0.091). Individuals with alcohol dependence appeared more likely to meet lifetime diagnostic criteria for both disorders after 1 year than individuals with depression. These results suggest that both alcohol dependence and major depression pose a significant risk for the development of the other disorder at 1 year.

As has already been noted, a number of critics of the self-medication hypotheses have raised what Frances (1997) calls the "cart/horse problem" of which came first, the substance abuse disorder or the comorbid disorder. Dackis and Gold (1984, 1985), specifically argue that the clinically observed relationship between depression and cocaine dependence is due to chronic cocaine use causing dopamine depletion which, in turn, causes the depression. A similar alternative hypothesis is proposed by Cocores et al., (1987) for the reported association between cocaine dependence and ADHD. Empirical evidence, from both clinical (Abraham & Fava, 1999) and community samples (Deykin, Levy, & Wells, 1987), shows that depression generally precedes substance abuse rather than following it. These findings support of the self-medication hypothesis.

A number of studies have attempted to test Khantzians suppositions about which underlying problems motivate abuse of which drugs problems with aggression and anger motivating opiate abuse, for instance, or depression, hypomania and hyperactivity motivating cocaine abuse. Psychometric studies of clinical samples fail to support Khantzians predictions (Greene, et al., 1993; Schinka, Curtis, & Mulloy, 1994). Aharonovich, Nguyen, and Nunes (2001) found that cocaine abusers actually showed greater problems with anger and opiate abusers with depression than the opposite as Khantzians model predicted.

The Epidemiologic Catchment Area Study found that alcoholism was more strongly associated with antisocial personality disorder, abuse of other drugs, or mania than with depression as predicted by Khantzians model (Helzer & Pryzbeck, 1988). The ECA data suggest that the clinically observed association between depression and alcoholism is actually due to depressed alcoholics being more likely to seek treatment than non-depressed alcoholics.

While there are some contrary finding, much of the relevant research supports the self-medication hypothesis. Khantzians drug of choice predictions do not stand up to an empirical test. Thus, it is Duncans model of the self-medication hypothesis that appears to have continuing value to the field of addiction studies.