Internal Stimulus Control and Subjective Effects of Drugs


For many years psychotropic drugs have been characterized and classified using methods designed to measure their subjective effects in humans (). This research approach has two principal purposes: 1) to investigate the efficacy of a drug in attenuating unwanted subjective states in patients (e.g., pain, anxiety, depression), 2) to investigate the abuse potential of new drugs by comparing their subjective effects in experienced drug abusers to those produced by known drugs of abuse. In regard to the latter, such methods have been used to determine whether there are any common subjective states produced by all drugs of abuse (e.g., euphoria).

Systematic studies of subjective methods for drug classification have been conducted at the Addiction Research Center (ARC) in Lexington, Kentucky, now part of the National Institute on Drug Abuse. A major mission of the ARC has been to evaluate new analgesic compounds to determine whether they produced morphine-like effects. The subjective effects of morphine and related compounds were an important aspect of this evaluation. The research demonstrated that morphine and related narcotic analgesics produced a unique spectrum of subjective effects that can be reliably discriminated from subjective effects produced by other psychotropic drugs in experienced narcotic addicts (). Even within the analgesic class, mixed agonist-antagonists (e.g., cyclazocine) can be readily discriminated from morphine in terms of their subjective effects (). Other studies have also shown that the methods for measuring the subjective effects of drugs are useful for characterizing and differentiating other classes of abused drugs (e.g., psychomotor stimulants; sedative-hypnotics; hallucinogens). Thus, it is possible to determine whether an unknown drug belongs to the opiate, psychomotor stimulant, sedative-hypnotic, or hallucinogenic drug class on the basis of its subjective effects.

Until recently, measurement of drug-induced changes in subjective effects was possible only with humans, since only this species has the necessary verbal skills to describe how a drug makes them feel. However, behavioral methods have been developed over the past decade which allow animals to report on discriminations between psychotropic drugs (). There is a striking concordance between drug classes based on similarities in the subjective effects produced in humans and on similarities as discriminative stimuli in animals (). This has led many researchers in behavioral pharmacology to make the working assumption that the components of drug action responsible for the discrimination among various classes of psychotropic drugs by animals are the same as those responsible for the differences in the subjective effects of these drugs in humans. In part, the purpose of the present paper is to show that this concordance across species is not surprising, since the same learning processes are involved in both species. The fact that humans learn to apply a topographically unique response (verbal) to drug-induced discriminative stimuli should not mask the fact that the same fundamental processes are involved. A second major purpose of this paper is to examine the hypothesis that although each class produces certain distinctive subjective effects, all drugs of abuse produce certain common subjective effects (e.g., euphoria) and it is these effects which are responsible for their abuse.

Behavior analysis of drug-induced changes in self-reports

An analysis of the processes involved in measuring drug-induced changes in subjective states requires a review of precisely what subjects are asked to do in these experiments. The most common instruments used to measure subjective states are paper and pencil inventories. Some of these instruments are composed of a list of adjectives commonly used to describe mood (e.g., happy, angry) and the subject is asked to rate how he/she feels in relation to that mood (e.g., the Profile of Mood States: POMS). Other instruments consist of statements related to sensations and perceptions about which subjects are asked to indicate their agreement or disagreement (e.g., the Addiction Research Center Inventory developed at the ARC). In some procedures, greater quantification is obtained by having subjects indicate the strength of then mood or agreement with each adjective or statement on an ordinal scale. Instructions to the subjects indicate that they should respond in a manner which best reflects how they feel at that moment. The responses before and after drug or under drug and placebo conditions are compared to determine whether the drug has produced a significant change. It is usually assumed that the verbal behavior accurately represents a matching between the subject’s feelings and the statements or adjectives checked. Since the individual’s feelings are a private event, there is no way for the investigator to determine the degree of accuracy of the subject’s verbal report, i.e., how precisely it reflects a feeling state. For most purposes, such as drug classification, this problem can be ignored as long as the data produced are lawful (i.e., show comparable dose-related changes across individuals).

The self-reporting response is operant behavior controlled by its consequences and thus susceptible to change by a variety of influences besides the drug administered. It is well established for example, that verbal behavior can be markedly altered by social contingencies. The Greenspoon phenomenon () has amply demonstrated that powerful control can be exerted over the verbal behavior of subjects who were unaware that their behavior was being manipulated by subtle responses of the experimenter. In the Greenspoon study, human subjects were instructed to “say words” during a 50-minute experimental session. When the experimenter made a specific verbal response (“mmm-hmm”) following every plural noun uttered by the subject, there was a significant increase in that class of verbal responses (plural nouns). The existence of such influences make one less comfortable in assuming the accuracy of the reporting of private events.

Therefore, one approach in dealing with self-report data is to treat the verbal response as devoid of a referent. Thus, if after being given a sedative drug, Subjects say “I feel sleepy,” one can record this as a change in verbal behavior induced by the drug, without any inference about changed in subjective state. Thus, a positive answer to that statement conceived of in this way would have no value in predicting other behavior of the individual, such as the likelihood of reclining on a bed, or of exhibiting a sleep-appropriate EEG. However this “black box” approach is inadequate to account for the subjective drug effects data. If drug-induced changes in verbal responses are treated as devoid of a referent the meaning of the verbal response should be irrelevant. Subjects could be asked to check off boxes labeled with color names or numbers rather than mood descriptors. Since we usually have no discriminative training for applying such color names or numbers to internal states, we would probably get little consistency in drug effects. On the other hand, as we will illustrate, when we allow subjects to respond using common adjective or simple descriptive statements of mood words, we see a fair degree of agreement in responses across individuals who are given certain psychotropic drugs. This agreement is based on a common conditioning history in which certain adjectives or mood descriptions have been associated with certain internal states. Whether or not one chooses to ignore the internal cues, we are taking advantage of a conditioning history based on these internal cues when self-report methods are used for investigating drug effects.

How do humans learn to apply verbal labels to private events? It is clear how children can be differentially reinforced for correctly labeling colors, sounds, and other publicly observable stimulus events. Internal stimuli represent a special problem for such differential conditioning since the mediator of reinforcement cannot observe the private event to determine the accuracy of the labeling. Under these conditions, the trainer uses a combination of observing the external environment for significant cues and collateral responses as an indication of the veracity of the verbal label. For example, we would agree with (i.e., reinforce) a child who says she is sad when found sitting hunch-shouldered in her bedroom with tears streaming down her face if we also sew that her favorite toy had been ruined by the family dog. Conversely, if she comes bounding in the door, whistling and swinging her lunch pail with a smile on her face to show us her good report card, we would reinforce her for saying she is happy. As a child matures, some of the more observable parts of these behavior patterns (e.g., crying, whistling) may diminish, but the label is still accurately associated with the internal stimulus events. It is this type of conditioning history which we utilize when subjects are asked to match their internal state with a list of adjectives or statements. It is most remarkable that such conditioning histories are consistent enough across individuals so that drugs induce a fairly close agreement in self-reports of internal states. Although some radical behaviorists may still choose to deal with such self-reports as simply verbal behavior and ignore the internal cues setting the occasion for the responses, it is obvious that the subjects do not ignore them.

Drug discrimination studies in animals

It is well established that animals can be trained to discriminate between the internal cues associated with food and water deprivation (). For example, we could arrange conditions so that following 24 hrs of food deprivation an animal would be reinforced with the termination of electric shock for turning right in a T-maze. Turning left would not be reinforced under these stimulus conditions but would be reinforced when the animal was not food deprived. After several trials, the animal’s behavior would become appropriate to the deprivation conditions. The animal is correctly identifying an internal state in the same way that a human subject might check the adjective “hungry” under similar food deprivation conditions. In the case of the animal, we have controlled the conditioning history, whereas in the human we usually assume that such discriminative training has already occurred.

In the same manner, animals can be taught to discriminate between various drugs (). Holtzman and his colleagues have developed a method using a discrete-trial avoidance-escape paradigm in which animals (rats and monkeys) can prevent the onset of or terminate an electric shock by pressing one of two choice levers (). The animals were trained to press one lever after drug injection (morphine or cyclazocine) and the other lever after placebo administration. Specifically, rats were injected 30 minutes prior to each daily 20-trial session. During the session, a light was illuminated for 5 seconds prior to the onset of electric shock. Depression of the correct choice lever terminated the light (an avoidance response) or both the light and shock (escape response). On days when a rat had received morphine, one of the choice levers was correct; on days when saline was administered, the other choice lever was correct. Rats were trained until they completed at least 18 trials on the appropriate choice lever (90% correct). After this criterion for discrimination between drug and placebo was reached, drug test sessions were periodically conducted with a new drug or new dose. During training trials only one of the two choice levers was operable and could terminate a trial. Thus, after the first trial was completed, the reinforcement, i.e., the shock and/or warning stimulus termination, might then serve as the cue to the correct choice lever for the remainder of the session. To deal with this problem, on test days depression of either lever satisfied the avoidance-escape contingency. Thus, the effectiveness of the response could not function as a cue to signal the animal which choice lever was correct. Amazingly, when drug pretreatment times were arranged so that the onset of drug effect occurred half way through the session, animals switched from responding on the saline-appropriate lever to the drug-appropriate lever. Clearly, the drug cues are a more effective discriminative stimulus than even the reinforcer (i.e., light and/or shock terminations).

Following establishment of such stimulus control, it is of interest to determine to what degree such control will generalize to other drug stimuli. There are two ways in which discriminative stimuli may be varied for generalization testing-quantitatively and qualitatively. When a drug is used as the discriminative stimulus, quantitative generalization tests are accomplished by varying the dose. When this is done in animals trained to discriminate 3.0 mg/kg of morphine from saline, lowering the morphine dose results in dose-related decrements in responding on the morphine-appropriate choice lever with a concomitant increase in responding on the saline-appropriate choice lever. Doses higher than that used in training produce similar or even greater discriminative stimulus control (i.e., responding on the morphine-appropriate lever) until behaviorally toxic doses are reached. This relation between morphine dose and response choice is similar to that observed when exteroceptive discriminative stimuli (e.g., light) are varied along a quantitative dimension (e.g., intensity). It is also the same relationship as that shown between dose and the intensity of the subjective effects produced by a drug in humans ().

When conducting generalization studies in which the discriminative stimulus is varied qualitatively, the situation is more complex. With an auditory discriminative stimulus, the unidimensional continuum of frequency can be manipulated. For a visual stimulus the continuum is wavelength. When using drug states as discriminative stimuli, however, we do not know the relevant continua along which changes might show a lawful relatlonship to behavior. This deficit is not unique to drugs, however, as the same problem exists with olfactory stimuli. Nevertheless, it is possible to do generalization tests from training drugs to other drugs with different structures and pharmacologic properties. For example, after approximately 8 to 10 weeks of training in the Holtzman experiments, most animals responded almost exclusively on the appropriate lever when given either morphine or saline. Subsequently, a variety of psychotropic drugs were investigated to determine which produced “morphine-like” discriminative effects (i.e., animals responding on the morphine-appropriate choice lever 18 out of 20 trials). For the following reasons, the results of these generalization tests indicate that the discriminative control exerted by morphine is a specific narcotic effect:

  1. (1) all narcotic drugs tested showed morphine-like discriminative control in a dose-related manner;
  2. (2) these narcotics showed a ranking in potencies highly correlated with their potencies in producing morphine-like subjective effects in humans;
  3. (3) the stimulus control exerted was stereospecific with only analgesically active isomers exerting morphine-like effects;
  4. (4) naloxone administration produced a pronounced shift in the dose response curve relating dose of morphine to its discriminative control;
  5. (5) tolerance to the discriminative effects of morphine developed after repeated administration and there was cross tolerance to methadone; and finally
  6. (6) d-amphetamine, chlorpromazine, ketamine, mescaline, pentobarbital, physostigmine, and scopolamine failed to exert morphine-like discriminative stimulus control.

The results of studies in which monkeys were trained to discriminate between cyclazocme and saline were comparable to those described for the rat with morphine. Naloxone diminished the stimulus control exerted by cyclazocine, i.e., on days when the animals were given both drugs, most of their responses were made on the choice lever associated with saline administration. Studies in humans have shown that cyclazocine produces subjective effects distinctly different from morphine. Accordingly, in the monkey experiments, morphine did not substitute for cyclazocine as a discriminative stimulus. These results indicate that cyclazocine and morphine produce distinctive stimulus effects in animals and humans. In contrast, in monkeys trained to discriminate cyclazocine, there was generalization to drugs such as nalorphine, levallorphan, and ketocyclazocine, all of which produce a common set of dysphoric subjective reactions in humans ().

This series of experiments conducted by Holtzman and his colleagues has convincingly demonstrated that generalization tests in animals can be used to classify drugs in the opiate class as well as those with mixed opiate agonist-antagonist properties. Further, the classification derived from animal experiments is in concordance with that based upon the subjective effects of these drugs in humans (). Similarly, animals can be trained to discriminate prototypic agents from other classes of abused drugs (e.g., cocaine as a prototypic stimulant) and then generalization tests can be conducted by testing other psychotropic drugs (e.g., amphetamines, barbiturates, etc.). Again the drug classes based upon discriminative effects in animals and upon subjective effects in humans are in striking concordance.

The relationship of subjective and reinforcing effects of psychotropic drugs

Although the classes of abused drugs can be differentiated on the basis of their spectrum of subjective effects, certain effects in common are produced by all such drugs. When hospitalized exaddicts are tested with the Addiction Research Center Inventory (ARCI), scores on the Morphine-Benzedrine Group Scale (MBG) show dose-related increases when subjects are administered narcotic analgesics () barbiturates () or amphetamine-like drugs ().

William Martin, Director of the ARC for several years, believes that LSD-like hallucinogens, alcohol, and marijuana would also produce similar results on the MBG Scale if tested in an appropriate subject population (). The items in the MBG Scale are related to feelings of popularity, efficiency, social effectiveness, pleasant feelings, absence of worry, good self-image, and feelings of insight and satisfaction. This scale is designed to measure a drug’s ability to produce a subjective state of “euphoria” (). It is the opinion of many researchers that drugs are abused by humans because they produce this state of “euphoria” ().

Another approach used to study factors contributing to drug abuse in humans has been to develop an animal drug self-administration model (see Griffiths et al., this volume). In these studies animals are given an opportunity to emit a response which is followed by the drug delivery. If responding is maintained by a drug it is said to possess positive reinforcing properties, i.e., the drug is a positive reinforcer.

Previously we discussed the similarity in drug classifications formed on the basis of subjective effects in humans and discriminative stimulus effects in animals. If animals and humans have similar subjective” responses to drugs one might predict that drugs which serve as reinforcing stimuli in animals should produce “:euphoria” in humans. If we operationally define “euphoria” as the state measured by the MBG Scale of the ARCI, this relationship can easily be determined. Table “Changes in “Euphoria” Rating on the ARC1 (MBG Scale) Correlated with Animal Drug Self-Administration’” shows that there is a good correlation between these two procedures. Drugs in the opiate agonist, psychomotor stimulant, and barbiturate classes generally serve as reinforcers in animal studies and, as well, produce dose-related increases in MBG Scale scores (i.e., “euphoria” Further, both opiate agonist/antagonists, such as nalorphine and cyclazocine, and neuroleptics produce “dysphoria” and are generally not self-administered by animals. These results are also in accord with actual street abuse of these various drugs. That is, commonly abused drugs serve as reinforcers in animals and produce “euphoria,” whereas drugs producing “dysphoria” are neither abused nor do they generally serve as reinforcers in animals. It remains to be determined whether this pattern generalizes to alcohol, marijuana, and LSD-like hallucinogens. The data available suggest that both drug self-administration experiments in animals and investigations of the subjective effects of drugs in humans can be used to predict whether a new drug has significant abuse potential. This has led, in our opinion, to the incorrect use of the term “reinforcing” as synonymous with “euphorigenic,” and has produced both theoretical and practical problems.

Since it would be difficult to measure “euphoria” in animals, animal self-administration studies cannot shed light on whether the reinforcing effects of drugs are based upon their ability to produce “euphoria.” In order to answer this question, human experimentation is required in which measures of both a drug’s reinforcing properties and its subjective effects are obtained. Only if the concordance between the two measures is invariant can a causal hypothesis be tenable.

The administration of single (or a limited number of) doses of drugs in the opiate, psychomotor stimulant, and sedative-hypnotic class produces “euphoria” in “appropriate” subjects. To demonstrate that a drug serves as a reinforcer, however, it is necessary to show that the drug increases the response rate on which its administration is contingent. When this is done, a dissociation between the reinforcing effects and mood effects appears. Clinical data indicate that the continued use of alcohol and opiates is associated with progressive dysphoria, anxiety, irritability, and aggressiveness (). Unfortunately, most studies have not used the MBG Scale from the ARCI as their measure of euphoria. Nonetheless the measures of “euphoria” show a progressive decline although the drug continues to maintain self-administration and is by definition a reinforcer. Furthermore, Johanson and Uhlenhuth () have shown that when normal human subjects are allowed to choose between self-administering d-amphetamme (5 mg orally) or a placebo they initially prefer the drug. This preference is associated with a spectrum of subjective changes on the Profile of Mood States () indicative of “euphoria.” With repeated opportunities to choose, however, the number of drug choices declines despite the fact that when the drug is taken (during forced “sampling” administrations), it still produces the same changes on the POMS. Thus, a drug can continue to serve as a reinforcer despite the development of progressive “dysphoria,” and, moreover, a drug can continue to produce “euphoria” but not continue to serve as a reinforcer. Thus, mood changes do not necessarily covary with changes in the reinforcing efficacy of drugs. This weakens the hypothesis that drugs serve as reinforcers because of their ability to produce “euphoria.”

Another line of evidence bearing on the issue of concordance between a drug’s “euphorigenic” and reinforcing actions is based on individual differences in response to drugs. In the preceding section we have qualified the description of drug-induced mood changes by saying that these occurred in “appropriate” subjects. This implies that mood changes should differentiate those individuals for whom drugs serve as reinforcers and those for whom they do not. There is very little acceptable evidence on this point. Drug-induced mood changes often differ between humans who abuse drugs and those that do not. Beecher () demonstrated that morphine generally produced “dysphoria” and aversion in normal subjects whereas it produced “euphoria” and a desire to repeat the drug experience in ex-heroin addicts. In contrast, amphetamines produced “euphoria” in normals but not in the ex-heroin addicts. Unfortunately, since the reinforcing actions of these drugs were not determined with the same subjects, we cannot state they would have covaried with the mood measures.

It is commonly assumed by many clinicians that patients who experience a “euphoric” response to medically prescribed drugs are at greater risk for iatrogenic addiction. In one recent study (), the mood changes induced by &amphetamine or placebo were compared in normal human subjects. POMS scores revealed that d-amphetamine produced an increase in Arousal, Positive Mood, and Elation. After this experience subjects were given repeated opportunities to choose between ingesting c-amphetamine or placebo. Though most preferred d-amphetamine, some did not self-administer it at every opportunity despite the fact that the drug produced comparable mood changes in all subjects. Thus one could not predict on the basis of similarities in drug-induced mood changes whether the drug would serve as a reinforcer. On the other hand, it was demonstrated that there was a subset of individuals who chose d-amphetamine on every opportunity and these individuals did not differ in their subjective response to the drug. They did, however, show significant differences in mood prior to ingestion of the drug. Subjects who showed this decided preference for d-amphetamine were more anxious and depressed as measured by the FOMS (). It is important to stress that despite these differences in pre-drug mood, their subjective responses to drug were no different from those of the other subjects.

It is clear that we need a great deal more information on how individuals differ in their subjective responses to drugs and the importance of these differences as a determinant of whether the drug serves as a reinforcer. Ideally these studies should be done in drug-naive subjects, but there are limitations on the types of drugs, range of doses, and duration of exposure which must be imposed for ethical and practical reasons. With therapeutic drugs such studies are, however, of extreme importance in order to define populations which may be at greater risk for dependence when exposed to the drug during treatment.

The evidence reviewed suggests that drug-induced changes in the subjective state called “euphoria” are produced by many drugs which readily serve as reinforcers in both animals and humans. There are circumstances, however, in which these two measures of drug effect do not covary. Thus one effect cannot be caused by the other; rather both are produced by the interaction of the drug with the organism (with a unique genetic behavioral and pharmacologic history) under a particular set of environmental conditions. Both organismic and environmental variables may modify the reinforcing and subjective drug effects differentially. It would not be unexpected then, using appropriate subjects under appropriate environmental circumstances (e.g., exaddicts in a controlled hospital setting), that one could predict the reinforcing effects of all drugs from their subjective effects. Clearly, the ARC has isolated effective procedures for selecting subjects and an environmental situation for predicting a drug’s reinforcing actions from measures of its subjective effects. This has had predictive utility for preventing the unwitting introduction of drugs with high reinforcing efficacy into medical practice. However, the close correlation between these two drug effects in these carefully selected subjects under highly controlled environmental conditions should not lead to the conclusion that the drug’s “euphorigenic” actions produce its reinforcing actions. These two aspects of drug action dissociate under a wide variety of conditions. Such dissociation can only lead to the conclusion that subjective and reinforcing effects are correlated but that neither is causal of the other.


Selections from the book: “Behavioral Pharmacology of Human Drug Dependence”. Travis Thompson, Ph.D., and Chris E. Johanson, Ph.D., eds. Presents a growing body of data, systematically derived, on the behavioral mechanisms involved in use and abuse of drugs. National Institute on Drug Abuse Research Monograph 37, July 1981.