An Addiction Science Network Resource

Reprinted from N.K. Mello and J.H. Mendelson (1987), Operant analysis of human drug self-adminstration: Marihuana, alcohol, heroin and polydrug use. In M.A. Bozarth (Ed.), Methods of assessing the reinforcing properties of abused drugs (pp. 525-558). New York: Springer-Verlag.
Brain Reward System
Back to book TOC
(Search book contents) (Search entire ASNet)


Chapter 25

Operant Analysis of Human Drug Self-Administration:
Marihuana, Alcohol, Heroin and Polydrug Use

Nancy K. Mello and Jack H. Mendelson

Harvard Medical School - McLean Hospital
115 Mill Street
Belmont, Massachusetts 02178

The evolution of operant procedures to study human drug self-administration in clinical research ward settings is described. The advantages and limitations of several techniques developed between 1965 and 1984 are discussed. Illustrative data from studies of alcohol, marihuana, heroin, and polydrug self-administration are summarized. The advantages of objective operant behavioral measures of drug acquisition, in contrast to anecdotal or retrospective reports, for evaluating new pharmacotherapies for the treatment of drug abuse are described. The importance of direct observation of drug use patterns and drug effects on behavioral and biological variables under controlled clinical research ward conditions is emphasized.



Although the examination of drug self-administration patterns in human subjects is a relatively recent development, the value of this approach for the study of substance abuse has now been amply demonstrated in many laboratories. Advances in techniques for the analysis of behavior have been paralleled by an expansion of and refinement in the type of questions asked. This review will discuss some dependent variables that can be studied in a human drug self-administration paradigm. These issues will be considered in the context of a quasi-historical review of some operant procedures developed in our laboratory and their advantages and limitations. Finally, selected data obtained on patterns of alcohol, heroin, marihuana, and multiple-drug self-administration will be described.

Dependent Variables in Human Drug Self-Administration

Since 1965, we have studied the self-administration of a number of abused drugs in human volunteers with a history of drug use or addiction (Mello & Mendelson, 1965). Multidisciplinary studies have been conducted on an inpatient clinical research ward, and each subject has been observed before, during, and after a period of self-regulated drug intoxication. Drug self-administration was observed over several days or weeks. This approach permits study of an encapsulated sequence of the basic behavioral disorder, drug abuse, under clinical research ward conditions. Our objective has been to study natural or relatively unconstrained drug use patterns and to correlate these with a number of biological and behavioral variables. We have attempted to devise situations where drug self-administration would simulate drug use patterns in the natural environment as closely as possible. Two general categories of dependent variables can be examined in such studies: drug-effect variables and patterns of drug self-administration.

Drug-Effect Variables

The pharmacological actions of drugs are defined by effects on many aspects of behavioral and biological function. Studies of physiological responses to drugs may include examination of many systems such as neuroendocrine hormones, plasma lipid levels, and sleep EEG (Mendelson, 1964, 1970; Mendelson, Kuehnle, Ellingboe, & Babor, 1974a; Mendelson & Mello, 1976, 1984a; Mendelson, Mello, & Ellingboe, 1978; Mendelson, Rossi, & Meyer, 1974b). Behavioral studies may explore drug-related changes in subjective states, perceptual and cognitive function, social interaction, and objective measures of performance (Mello & Mendelson, 1978). Another drug-effect variable is the fine-grain analysis of operant responding, including interresponse time analysis. Many other examples of drug-effect variables could be listed. However, our main point is to distinguish between drug-effect variables and patterns of drug self-administration.

Patterns of Drug Self-Administration

The pattern of drug self-administration is itself an important dependent variable which may be central to our understanding of human substance abuse. Basic and human behavioral pharmacology have repeatedly shown that the schedule of drug reinforcement—the dose and frequency of drug availability—influences the effects of drugs on behavior (Kelleher, Goldberg, & Krasnegor, 1976). By observing an individual’s drug self-administration pattern, we can study the self-imposed schedule of reinforcement. However, this is only possible if the individual determines his or her own pattern of drug use.

Examination of the self-imposed pattern of drug use may assist in the identification of factors which contribute to the maintenance of drug abuse. The diversity of individuals with drug abuse problems and the fact that no single psychological, social, or biological factor appears to predict drug abuse suggest that it may be more productive to try to determine how drug abuse is maintained than to focus exclusively on etiological factors (for review, see Mello, 1983). Moreover, as we study the self-administration of different drugs under similar conditions, we hope to be able to compare use patterns across drugs. Eventually, this type of analysis should help to identify some commonalities and differences in the determinants of patterns of heroin abuse, alcohol abuse, marihuana and tobacco use, stimulant and hallucinogen abuse, et cetera. If an analysis of drug use patterns can reveal commonalities which transcend particular drug effects, such information might generalize to future forms of drug abuse as well. Although we cannot now predict the type of tomorrow’s drug abuse problems, some form of substance abuse seems almost inevitable. Clarification of factors that maintain drug self-administration should facilitate the development of more effective intervention procedures.

The pattern of drug self-administration can be operationally defined by the following measures:

(1) Number of drug self-administration occasions (per hour; per day; per week),
(2) Drug dose per occasion, and
(3) Interval between drug self-administration occasions—the distribution of drug doses over a 24-hour period.
Although it could be argued that drug use also affects the subsequent pattern of drug self-administration and that this also should be classed as a drug-effect variable, we believe a distinction between drug use patterns and drug effects has both conceptual and methodological advantages.

This approach to the study of human drug self-administration and the questions posed are different from studies which attempt to manipulate human drug use patterns by varying the conditions necessary for drug acquisition. It has been demonstrated that manipulation of conditions such as response-cost, the time of drug availability, or the dose of drug available can change both the amount and frequency of drug use (Babor, Mendelson, Greenberg, & Kuehnle, 1978; Bigelow, Griffiths, & Liebson, 1976; Griffiths, Bigelow, & Liebson, 1976a; Mello, McNamee, & Mendelson, 1968; Pickens, Cunningham, Heston, Eckert, & Gustafson, 1977). The generality of these relationships across drugs as well as between man and animal models has been repeatedly demonstrated (Griffiths, Bigelow, & Henningfield, 1980).

Since the use patterns of several drugs have been shown to be modified by manipulation of acquisition variables, these studies may have implications for social controls of drug use (e.g., hours of drug availability, taxation, price; Popham, Schmidt, & de Lint, 1975). However, the identification of common controlling variables does not suggest that the use patterns of different drugs are identical. Rather, use patterns appear to be quite different depending upon the pharmacological action and the rate of absorption and disposition of the particular drug.

Limitations of Drug Self-Administration Studies

One disadvantage of the study of spontaneous drug self-administration patterns is that precise time-dose-response relationships between the various drug-effect variables cannot be established since drug dose, frequency, and inter-dose intervals will vary on an unpredictable basis. Yet, this variability constitutes the drug self-administration pattern which is our primary dependent variable.

One alternative is to ignore drug self-administration patterns and to focus instead on drug-effect variables. The most efficient way to do this is to use a programmed drug administration regimen in which fixed drug doses are administered every four to six hours. This permits precise dose-time correlations with whatever drug-effect variable has been selected for study.

Programmed drug administration is the traditional method used to examine the basic pharmacological effects of drugs. The pioneering studies of drug effects conducted at the Lexington Addiction Research Center employed programmed drug administration with only one exception (Wikler, 1952). The first studies of the effects of chronic alcohol intoxication in alcoholics were conducted in a programmed administration paradigm (Mendelson, 1964).

It is evident that programmed dose paradigms are useful for asking different types of questions than drug self-administration paradigms. However, there are also other factors which may limit the utility of a programmed dosage paradigm, even for the study of drug-effect variables. When the consequences of programmed alcohol administration and spontaneous self-regulated alcohol consumption were compared in eight alcoholic subjects, it was found that biological drug-effect variables varied markedly between the two alcohol administration conditions (Mello & Mendelson, 1970). Each subject served as his own control during a 20 day spontaneous alcohol administration paradigm and 20 days of programmed alcohol administration. Programmed alcohol administration was associated with greater toxicity (e.g., gastritis, vomiting) during intoxication than spontaneous drinking. Subjects were able to tolerate higher doses of alcohol during the spontaneous self-administration paradigm and distributed alcohol consumption to achieve higher peak blood alcohol levels than were measured during programmed alcohol administration. The severity of alcohol withdrawal signs and symptoms after cessation of drinking was markedly greater after spontaneous drinking than after programmed alcohol administration, even in those subjects who drank equivalent quantities of alcohol in each condition.

We concluded that the pattern of drinking was more critical than the duration of drinking as a determinant of biological reactions to alcohol intoxication and withdrawal. These data seemed to justify our pursuit of self-determined patterns of drug self-administration both as a primary dependent variable and as the most sensitive and reality-concordant baseline against which to correlate biological as well as behavioral drug-effect variables.

Techniques to Study Human Drug Self-Administration

Operant techniques to study human drug self-administration have developed in parallel with those used in basic behavioral pharmacology and are derived from concepts and procedures for the experimental analysis of behavior (Skinner, 1938, 1953). Operant procedures have been shown to produce orderly sequences of responding which provide an objective index of the relative reinforcing consequences of various drugs (or competing reinforcers such as money) at any point in time. It is possible to directly observe the amount and frequency of drug self-administration and the behavioral consequences of drug intoxication without reliance upon retrospective self-reports.

Drug acquisition in real life involves engaging in a variety of behaviors, since drugs are not available without some expenditure of effort or money. Consequently, it seems realistic to require performance on an operant task to obtain drugs in a clinical research setting. The nature of the task can vary from performance on a relatively simple schedule of reinforcement to complex procedures which concurrently assess variables such as timing behavior or memory function. When we designed our first operant paradigm to study alcohol self-administration by alcoholics, we thought it necessary to use a very simple task so that subjects could perform and earn alcohol irrespective of their intoxication level (Mello & Mendelson, 1965). We anticipated that if alcohol reinforcement was made contingent upon successful performance of a complex discrimination task, the subject would not be able to sustain his initial performance as he became progressively more inebriated. Of course, this would preclude examination of behavior to acquire alcohol and would only yield data on the effects of alcohol on some aspect of perceptual or cognitive function, which was not the primary goal. Subsequently, we learned that behavioral tolerance for alcohol permits alcohol addicts to perform very complex tasks with accuracy, even when blood alcohol levels exceed 250 mg/dl (Mello, 1973).

Study of drug acquisition using operant techniques permits examination of a wide range of behavioral variables (e.g., time, duration, and pattern of operant work for the drug); the rate of operant work (assessed by both analog—cumulative recorder—measures and quantitative interresponse time measures); the time and number of drug purchases; and the effects of each successive drug use occasion on both rate and duration of operant responding. The effects of drug use on operant response patterns and choices between alternative reinforcers (e.g., marihuana vs. money or alcohol vs. marihuana) can also be examined. Such data provide direct measures of performance capacity and permit inferences about drug effects or intervening variables such as "motivation," sometimes postulated to affect performance.

Alcohol Acquisition on Multiple Schedules

There are a variety of ways to examine human drug self-administration, and the technical aspects of the behavioral procedures define and limit the type of data acquired. It may be of some historical interest to review the types of operant procedures developed in our laboratory and comment on their advantages and limitations. The first machine designed for the study of alcohol self-administration in alcoholic subjects is shown in Figure 1 (Mello & Mendelson, 1965). The subject could select whether to work for alcohol or for money reinforcement and could work at the machine at any time.

Operant manipulandum used to study alcohol self-administration
Figure 1: Operant manipulandum used to study alcohol self-administration by alcoholic subjects. Reprinted from Mello, 1972. Copyright 1972 by Plenum Press.

The subjects’ task was to press the center response key which was trans-luminated with a number of colored stimulus lights associated with a series of simple schedules of reinforcement. These schedules occurred in an irregular sequence and included: fixed ratios of 60, 120, 240, 360; fixed intervals of 1, 2, and 3 minutes; extinction of one minute; and differential reinforcement of no responding. Subjects were told to press the translucent response key in order to make the key color change as often as possible, since reinforcement occurred only when the colors changed. Upon completion of the response requirement, 10 ml of bourbon or three nickels were directly dispensed, and the key color and schedule changed. The value of a single alcohol or money reinforcement was equated, and subjects could use money earned to buy alcohol. In an effort to make the task more interesting, a gambling contingency was added. When reinforcement became available, subjects could choose to take that reinforcement or to try for double-or-nothing by pressing the yes or no key at the right of the operant panel.

Subjects worked at the operant task for alcohol for 14 days. After seven days when subjects’ response behavior failed to come under control of the various schedules of reinforcement, explicit verbal descriptions of the schedule requirements associated with the various colored lights were provided. Subjects still failed to respond under schedule control, except for one component—differential reinforcement of no responding. When that particular stimulus light appeared on the response key, subjects usually left the room. This behavior was adventitiously reinforced since the light associated with another schedule came on during their absence.

Rates of operant responding and time spent working at the machine were unimpaired by alcohol intoxication. Subjects maintained relatively high blood alcohol levels (200 to 300 mg/dl) throughout the period of alcohol access. Subjects usually worked for 1.5 to 2 oz of alcohol before removing the glass. Each glass removal shut off the machine for a period of 10 minutes. An immediate alcohol reinforcement was consistently preferred to money reinforcement, even though money could be used to buy an equivalent amount of alcohol at any time.

Despite the ease of working at this simple operant task, subjects complained continually about the machine. They were bored, and they did not gamble (double-or-nothing) except on rare occasions. Distaste for the machine was illustrated by the fact that one subject incorporated distorted thoughts and perceptions about the operant instrument in his hallucinatory experiences and delusional ideations during alcohol withdrawal.

Negative reactions to the task did not prevent subjects from working for alcohol. However, since response behavior did not come under control of any of the operant schedules of reinforcement provided, these data suggested that analysis of drug effects on schedule control, in the usual sense, would be very difficult in alcoholic subjects. Subsequently, we have used a simple fixed ratio or fixed interval schedule rather than multiple schedules for drug acquisition studies.

Alcohol Acquisition With Observing Response Procedures

An additional 14 subjects were studied under comparable experimental conditions over a 7-day period of alcohol availability (Mello et al., 1968). The machine shown in Figure 1 was modified so that the subjects’ task was to press the response key whenever a light of 500 msec duration appeared (i.e., an observing response). The light onset occurred at irregular intervals which ranged between 2.5 and 10 seconds. Errors of omission or of commission resulted in the loss of all accumulated points. Alcohol and money acquisition were contingent upon completion of a fixed ratio of 16 (FR-16) or 32 (FR-32) consecutive responses to the signal light.

Although differences in response-cost did not change the amount of alcohol earned per session (as defined by removal of the receptacle glass), the average blood alcohol levels maintained by the FR-16 group were almost twice as high as in the FR-32 group. Subjects required to work twice as hard for alcohol tended to drink half as much. Individual blood alcohol levels were highly variable within and between days, and no subject earned all the alcohol potentially available.

Subjects complained vociferously about the demands of the operant task and pounded the machine in frustration after an omission or commission error. However, they were able to perform at all levels of intoxication, and the accuracy of performance was unrelated to blood alcohol levels. Most subjects did not work for money, and only five subjects gambled (double-or-nothing) with any consistency. The occurrence of gambling was also unrelated to blood alcohol levels. These data suggest that risk taking as defined by this task cannot be predicted on the basis of intoxication. The accuracy of performance at blood alcohol levels which averaged 200 mg/dl testifies to the behavioral tolerance for alcohol developed by alcohol addicts.

Alcohol Acquisition Using a Simulated Driving Machine

Figure 2 shows a driving machine that was developed to study group drinking behavior (Mendelson et al., 1968). The subjects’ task was to steer a model automobile and keep it on the moving road on a revolving drum. Each time the auto touched a metal contact on the road, a point was registered. After 120 points were earned, 10 ml of bourbon was automatically dispensed into a common reservoir. Since the subjects could earn about 60 points per minute or 3600 points per hour, the maximum amount of alcohol that could be earned each hour was about 300 ml or 10 oz. Each subject had an ignition key and could work at the machine at any time. Each subject could also withdraw as much alcohol from the reservoir as he wished at any time by activating the ignition switch. The total time each subject worked, the number of 10 ml alcohol reinforcements earned, and the amount of alcohol each subject withdrew were automatically recorded. Subjects were permitted to work at the driving machine to earn alcohol for a period of 30 days.

Subjects took turns working at the driving machine and their ability to perform was not discernibly impaired, even at blood alcohol levels over 200 mg/dl. No subject drank continuously throughout the 30-day period, and no subject drank as much alcohol as was available. Subjects spontaneously terminated drinking episodes on several occasions; four of these occasions were correlated with stressful situations on the ward. There were seldom clearly definable events which accompanied resumption of drinking by these subjects. All subjects showed discernible increases in anxiety and agitation during intoxication and appeared far more depressed and anxious when they terminated drinking than when they initiated a subsequent drinking episode. Two subjects drank more than one fifth of bourbon per day on an average and maintained blood alcohol levels that fluctuated between 50 and 250 mg/dl. A third subject drank during three separate episodes of 8, 6, and 7 days respectively. A fourth subject became agitated, depressed, and assaultive and left the study after three days of drinking.

Operant apparatus used to study alcohol self-administration
Figure 2: Schematic diagram of an operant apparatus used to study alcohol self-administration by alcoholic subjects. Each subject could activate the instrument by placing his key in the appropriate ignition switch located at the right of the revolving road. The modified driving machine had a steering wheel, brake, and accelerator which controlled the movement of a model automobile on the revolving road drum. Periodically, signals to stop, go, or turn appeared on the panel above the revolving road. Points were earned only for keeping the car on the road. Every 120 points earned resulted in 10 ml of alcohol being directly dispensed from the alcohol IN dispenser into the group reservoir. Each subject had a key to one of the four ignition switches located to the right of the reservoir. Any subject could withdraw alcohol from the group reservoir by placing his key in the ignition switch. Alcohol was dispensed into a glass underneath the OUT dispenser and would flow for as long as the ignition switch was activated. Reprinted from Mendelson, Mello, and Solomon, 1968. Copyright 1968 by Williams and Wilkins.

These subjects evolved a stable pattern of group interaction and maintained their mutually defined roles in relation to alcohol acquisition. One subject consistently removed more from the group alcohol reservoir than he contributed, and another consistently contributed more than he removed. The free-loader appeared to be the leader of the group. Another subject contributed an amount of alcohol to the group reservoir approximately equivalent to the amount that he withdrew. This machine was considerably more acceptable to the subjects than the machine shown in Figure 1. Subjects perceived the driving machine as more of a game than a performance task.

Alcohol Acquisition Using a Portable Operant Manipulandum (FR-1000)

Relocation of our laboratory to Washington prevented further studies with these instruments and necessitated the development of a non-automated, portable instrument shown in Figure 3 (Mello & Mendelson, 1972). While new automated instruments were being constructed, these simple hand-held manipulanda permitted study of operant work-contingent drinking patterns in alcoholic men. Subjects could work for alcohol or for cigarettes by depressing a button which activated a mechanical counter inside the box. Subjects could earn one ounce of alcohol or one cigarette within about 5 minutes of performance on a fixed ratio-1000 schedule of reinforcement. Points earned were exchanged for color-coded tokens each day.

Tokens could be used to buy alcohol or cigarettes directly dispensed from an apparatus shown in Figure 4. To activate the dispenser, the subject turned on an ignition switch with a coded ignition key which told the programming circuitry who was activating the dispenser and when. After the subject set a glass receptacle over the photocell, a signal light came on indicating that the dispenser would receive tokens. One ounce of alcohol was dispensed into the glass for each token deposited. The circuitry recorded the time of purchase, the number of purchases, and the subjects’ identification numbers. Subjects could also buy alternative reinforcers, such as 15 minutes of television time.

Hand-held operant manipulandum used to study self-administration
Figure 3: Operant manipulandum used to study alcohol self-administration by alcoholic subjects. Reprinted from Mello, 1972. Copyright 1972 by Plenum Press.

Alcohol and cigarette dispenser
Figure 4: Schematic diagram of an alcohol and cigarette dispenser. Subjects could use poker chips earned at a simple operant task to purchase one of three types of alcohol (vodka, bourbon, or gin) or cigarettes. To activate the dispenser, the subject first turned on the ignition switch which automatically informed the circuitry of his identity and the time. He then selected which type of reinforcer he wanted and deposited a colored poker chip in the token slot when the signal light came on. A photocell had to be interrupted—by the drinking glass—before the liquid was dispensed and the signal light illuminated.

The major disadvantage of the portable manipulandum was that it was not possible to analyze time, duration or rate of operant work since responses could not be recorded by the programming circuitry. However, the manipulandum proved to be tamper proof and yielded reliable data on drinking patterns, as described in a later section of this review. The manipulandum was also used to study marihuana self-administration patterns in young men (Mendelson et al., 1974b) and heroin self-administration by heroin addicts (Meyer & Mirin, 1979).

Alcohol Acquisition Using a Titrated Delayed Matching-To-Sample Task

The behavioral tolerance for alcohol shown by alcoholics in several studies persuaded us that it was possible to combine work-contingent alcohol acquisition with the assessment of drug-effect variables in addition to operant performance. Consequently, we designed and constructed an operant system that could be used to evaluate various aspects of perceptual or cognitive function. Each response panel was located in a separate operant booth, and six subjects could work simultaneously at their individual machines in relative privacy.

There has been considerable debate about whether alcohol directly affects short-term memory function. One way to evaluate this question was to use a titrated delayed matching-to-sample procedure. The operant response panel is shown in Figure 5 (Mello, 1973). The subjects’ task was to select the comparison key which contained a picture identical to that which had previously appeared on the sample key. Short-term memory was defined as the interval between the offset of a picture on the sample key and the onset of pictures on four comparisons keys. Attention to the sample stimulus was insured by requiring the subject to make an observing response (FR-10) to turn on the sample stimulus. The subject then pressed the sample key until the picture projected on it went off. After a delay interval elapsed, four pictures were projected on the comparison keys. Selection of one of the comparison keys ended the trial. The length of the delay interval increased with each correct match trial and decreased with each incorrect trial in 4 second increments. The possible delay intervals ranged from 0 to 6 minutes. The matching-to-sample stimuli included pictures of the ward staff, movie stars and political figures, household objects, liquor and cigarette labels, nude figures, trigrams and abstract geometric designs—a total of 120 stimulus sets in several different sequences.

Apparatus to study alcohol self-administration and memory effects
Figure 5: Apparatus used to study alcohol self-administration and the effects of alcohol on short-term memory function in alcoholic subjects. Reprinted from Mello, 1973. Copyright 1973 by Plenum Press.

In order to encourage the subject to perform as well as possible, completion of four correct trials was required to earn a single token. Correct trials were indicated on the "correct" counter at the upper left of the panel. However, if four error trials occurred before completion of four correct trials, each counter reset to zero and the subject was required to begin accumulating correct trials again in order to earn tokens to buy alcohol. After completion of four correct trials, a single token was directly dispensed into the token bank shown on the right of the operant panel. The bank was clear plastic, and the subject could see the number of tokens he had earned at any time. Each subject had the only key to his token bank and could remove tokens to purchase alcohol whenever he wished from the dispenser shown in Figure 4.

All subjects maintained high blood alcohol levels, and performance for alcohol was sustained throughout a 12-day period of alcohol access. Alcohol had no discernible effect on "short-term memory" defined operationally by the interval between sample stimulus offset and comparison stimulus onset. Subjects were able to match correctly at the longest delay intervals even when blood alcohol levels exceeded 300 mg/dl. Subjects with a history of alcoholic "blackouts" performed as well as subjects who had never experienced blackouts during intoxication. The conclusion that alcohol does not impair "short-term memory" and that the alcoholic "blackout" probably cannot be accounted for by an alcohol-specific disruption of memory (Mello, 1973) has been confirmed in other laboratories (for review, see Mello & Mendelson, 1978).

This machine was probably the most powerful tool we have developed for assessing both alcohol acquisition behavior and the effects of alcohol on cognitive function. Subjects found the task challenging and interesting, perhaps because a continuously changing array of visual stimuli was provided. Since subjects were highly motivated to acquire alcohol, they appeared to perform at the limit of their capacity. Unfortunately, a de-emphasis on intramural research by the National Institute on Alcohol Abuse and Alcoholism during the early 1970s prevented further studies with this instrument. However, this study demonstrated the feasibility of asking questions related to drug effects in combination with questions about drug self-administration patterns.

Drug Acquisition on a Second-Order Schedule of Reinforcement

After return of the laboratory to Boston in 1974, we again developed a simple hand-held operant manipulandum for the study of drug acquisition patterns. The manipulandum was about the size of a pack of cigarettes and weighed about 198 grams. The prototype manipulandum was attached to a movable cable which could be connected to coded terminals permitting subjects to work at the operant task in their individual bedrooms or in a central day room. Subsequently, the manipulandum was modified to be completely portable.

A schematic drawing of the portable manipulandum is shown in Figure 6. Each response transmits a radio frequency signal on a discrete band which activates the programming and recording circuitry in an adjacent room. Points earned are registered on a central panel, and subjects always have a record of their earnings. Unlike the portable manipulandum shown in Figure 3, operant response patterns are automatically recorded by the programming circuitry, and both rate of response and interresponse times can be measured. Each manipulandum is color coded and labeled with the subject’s number to permit easy identification by the ward staff and to discourage subjects from exchanging manipulanda.

Figure 6: Portable operant manipulandum used to study heroin, marihuana, and alcohol self-administration. Each response transmits a radio frequency signal to the programming circuitry.
Portable operant manipulandum

Subjects were required to press the button on the manipulandum on a fixed interval one-second schedule of reinforcement (FI-1 sec). Only the first response after one second elapsed was recorded as a point by the programming circuitry. A signal light flashed each time an effective response was made. Subjects could earn one purchase point for 300 effective responses on an FI-1 second schedule or for 5 minutes of sustained operant work. This schedule is designated as a second-order FR-300 (FI-1 sec:S).

The prices of different drugs or money reinforcers are assigned a purchase-point cost which can be adjusted to reflect the current price prevailing in the Boston area. One advantage of a time-based schedule is that price can easily be translated into time required at the operant task. Whenever a subject elects to purchase a drug, the points spent are immediately deducted from the accumulated reinforcement points. Subjects are allowed to work on the operant task at any time, and a record of their point accumulation is continuously available.

We have used this simple procedure to study alcohol, marihuana, and heroin self-administration (Mello, Mendelson, & Kuehnle, 1978, 1982; Mello, Mendelson, Kuehnle, & Sellers, 1981; Mendelson, Babor, Kuehnle, Rossi, Bernstein, Mello, & Greenberg, 1976a; Mendelson, Kuehnle, Greenberg, & Mello, 1976b, 1976c). The manipulandum and the task are well tolerated by subjects, and they have not been able to tamper with or destroy the device. Subjects are able to perform the operant task while talking, reading, watching television, and eating. Although the manipulandum can be used with any schedule of reinforcement, we have continued to employ an FI-1 second schedule to permit comparisons across successive studies with different samples of drug users.

In summary, our experience with these several operant drug-acquisition procedures suggests the following general conclusions. Subjects will accept complicated and challenging procedures which maintain their interest (e.g., Figure 5). If acquisition of a meaningful reinforcer (drugs or money) is contingent upon accurate performance, subjects will usually perform to the limit of their capacity. Severe intoxication produces surprisingly little performance impairment in alcohol addicts because of behavioral tolerance. The extent to which comparable behavioral tolerance occurs with other classes of abused drugs remains to be determined. However, it does appear feasible to study alcohol and drug self-administration patterns with a task that simultaneously assesses some drug-effect variable.

A task which requires minimal attention or effort to perform, such as the simple second-order schedule procedure we currently use, is also accepted by subjects. This task yields reliable data on drug self-administration patterns as well as rates of operant responding. The relative ease of construction and maintenance of a portable manipulandum must be balanced against the complexities of construction and maintenance of a device which involves coded filmstrips, special recording procedures, and continual adjustment as did the machine shown in Figure 5. The most realistic compromise is probably to use a simple procedure to study drug self-administration patterns and to assess cognitive and perceptual variables separately in a situation where relatively larger amounts of drug are provided as a reinforcer for accurate performance. Unfortunately, there have been very few studies in which accurate performance is reinforced with a consequence that is significant for the subject. This failure to establish contingent relationships between accuracy and reinforcers has contributed to the numerous inconsistencies in data on the behavioral effects of alcohol (for review, see Mello & Mendelson, 1978).

The importance of using a mechanical dispenser for most drug reinforcers cannot be over-emphasized. The machine is consistently neutral and cannot encourage or discourage drug purchase. Although staff can be trained to dispense drugs without comment, it is impossible to insure that some attitude about further drug use by an intoxicated individual is not conveyed. There is no way to evaluate the extent to which the attitude of a human drug-dispenser may have influenced the basic datum—drug self-administration patterns. Alcohol, marihuana and tobacco cigarettes, and any drug that can be given in capsule form can be automatically dispensed. Intravenously administered drugs, such as heroin, are an obvious exception. Staff-supervised heroin administration under limited drug availability conditions is necessary for patient safety.

Patterns of Alcohol, Marihuana, Polydrug,
and Heroin Self-Administration

This section will summarize selected data on alcohol, marihuana, polydrug and heroin self-administration and illustrate the application of several of the techniques previously described. In each of these studies, volunteer subjects lived on a clinical research ward for several weeks. Behavioral studies were conducted simultaneously with physiological, biochemical, and neuroendocrine studies designed to examine the biological effects of chronic drug use. Subjects were observed during a drug-free baseline, a period of drug self-administration, and a post-drug baseline period. An own-control design is essential for human drug self-administration studies, since the use of "normal" control groups in the conventional sense is precluded by medical and ethical considerations. A control group consisting of "occasional drug users" is usually ill advised since such subjects are not sufficiently drug tolerant to permit meaningful comparisons with heavy users or drug addicts.

Alcohol Self-Administration

A number of stereotypic beliefs about alcoholics often appear in the clinical literature despite accumulating evidence to the contrary. One persistent belief is that the alcohol addict has a predictable and invariant drinking pattern—to drink as much as possible. This stereotype is linked to the concept of "craving," usually defined as a loss of control over drinking, with the implication that each time an alcoholic starts to drink he is compelled to continue until he reaches a state of severe intoxication. The circularity inherent in this reasoning is evident (i.e., craving is defined by the behavior it is invoked to explain). Empirical observations of alcoholics allowed to self-administer alcohol have not supported this view (for review, see Mello, 1975).

Two groups of four subjects were allowed to work for alcohol and cigarettes for periods of 30 and 62 consecutive days, respectively (Mello & Mendelson, 1972). The portable operant manipulandum shown in Figure 3 was used, and a fixed ratio of 1,000 responses was required to earn a single token which could be used to buy one cigarette or one ounce of bourbon from the automated dispenser shown in Figure 4. Subjects were able to earn one token in about 5 minutes of rapid performance or the equivalent of about 12 ounces of alcohol in an hour. The task could easily be performed while watching television, eating, drinking, or talking. Each subject had a color-coded manipulandum and an identical colored token to prevent exchanges between subjects. The volume of alcohol and the number of cigarettes purchased by each subject were recorded by the programming circuitry and checked against the number of colored tokens in the dispenser. Blood alcohol levels were measured at 8 a.m., 4 p.m., and 12 midnight each day.

The earning and spending patterns of a typical subject during the pre-alcohol baseline, alcohol availability, and the alcohol withdrawal periods are shown in Figure 7. The pattern of earning and spending for cigarettes is shown in the top row. The pattern of earning and spending for alcohol is shown in the middle row. Daily mean blood alcohol levels are shown in the bottom row. Partial withdrawal signs (e.g., tremor and gastritis) are indicated as asterisks. The type and duration of alcohol withdrawal signs and symptoms are shown at the right of the middle row.

Each subject could work for points for alcohol during the last day of the pre-alcohol baseline period. The number of points earned that day usually was sufficient to sustain a period of drinking of 3 to 6 days. Throughout the remainder of the drinking period, there was a clear dissociation between periods of earning and spending. Subjects alternated between working to earn points for alcohol and spending points earned for a work-free drinking spree. This pattern persisted throughout the 30- and 62-day alcohol-available periods and was strikingly similar in all subjects. Fluctuations in the average blood alcohol levels correlated roughly with the pattern of spending for alcohol. Subjects sustained relatively high blood alcohol levels averaging between 130 and 200 mg/dl for periods up to 62 days. No subject drank all the alcohol available, and all subjects tolerated the discomfort of withdrawal symptoms during the intermittent periods of self-imposed abstinence. These abstinent periods were unexpected in view of the relatively trivial performance requirement involved. Intoxication did not impair the subjects’ ability to work at this simple task.

Although all subjects showed a dissociation between working and drinking, subjects worked and drank at different times. Some member of the group was always working, while others were drinking. It is unlikely that the observed behavior represented satiation for alcohol, since other subjects given alcohol with no operant work requirement sustained blood alcohol levels which averaged above 200 mg/dl for periods of 14 to 20 days (Mello & Mendelson, 1972), Similarly, the decrease in operant work for cigarettes did not reflect decreased interest in smoking since subjects attempted to acquire free cigarettes from staff throughout the study.

Earning and spending patterns of a single subject
Figure 7: Earning and spending pattern of a single subject working for cigarettes during a 10-day baseline period, for both cigarettes and alcohol during a 62-day alcohol-available period, and for cigarettes during a 10-day withdrawal period. Subjects worked to earn tokens to buy alcohol and cigarettes by pressing a button on a portable operant box (fixed ratio = 1000). Tokens earned for alcohol were not interchangeable with tokens earned for cigarettes. Patterns of earning (filled circles, shaded area) and spending (open circles) for cigarettes are shown in top row. Patterns of earning (filled circles, shaded area) and spending (open circles) for alcohol are shown in the middle row. Subjects were allowed to work for alcohol tokens during the last 24 hours of the baseline period, and these tokens could be spent after 8 a.m. on the first day of the drinking period. Tokens earned during this period are shown at arrow as First Days Earnings. Reprinted from Mello and Mendelson, 1972. Copyright 1972 by the American Psychosomatic Society.

These men described themselves as periodic spree drinkers. The observed pattern of discordant working and drinking was probably more comparable to their real world experience than a stable alcohol intake permitted by an unlimited supply. Since the subjects determined their pattern of alcohol self-administration, it is reasonable to assume that this was their preferred or accustomed pattern. This technique appeared to result in an adequate simulation of normal drinking patterns by chronic alcoholic individuals in a clinical research ward context. A similar alternation between working and drinking has also been reported by Nathan and co-workers in alcoholics who worked at a comparably simple task (photo cell interruption) for points that could be converted into alcohol (Nathan, O’Brien, & Lowenstein, 1971; Nathan, Titler, Lowenstein, Solomon, & Rossi, 1970 ).

Marihuana Self-Administration

A number of multidisciplinary studies of marihuana self-administration have been conducted in our laboratory (Mendelson et al., 1974a, 1974b, 1976a). Biological studies of marihuana effects examined in an operant drug self-administration paradigm have included studies of the effects of marihuana on CNS structure (Kuehnle, Mendelson, Davis, & New, 1977), on cardiac and pulmonary function (Bernstein, Kuehnle, & Mendelson, 1976; Goldenheim, Mendelson, Mello, Tilles, & Bavli, 1985), and on reproductive function in men (Mendelson et al., 1974a, 1974b, 1978) and women (Mendelson & Mello, 1984a, 1984b; Mendelson, Mello, Ellingboe, Skupny, Lex, Griffin, & Bavli, 1984). Behavioral studies have examined the effects of marihuana on mood, memory, and social interactions (Mendelson et al., 1976a). In addition to studies of the pattern of marihuana self-administration, the hypothesis that marihuana induces an "amotivational" syndrome was also examined (Mendelson et al., 1976b; Mendelson & Mello, 1984b). Among the effects often ascribed to marihuana are apathy, lethargy and indolence, diminished "drive" and ambition, and decreased productivity and goal directedness.

Marihuana self-administration patterns were examined in 12 casual and 15 heavy users allowed to work for marihuana for 21 consecutive days. "Motivation" was inferred from time spent at the operant task working for marihuana and for money. Subjects worked on a second-order FR-300 (FR-1 sec:S) schedule at the portable operant manipulandum shown in Figure 6. One marihuana cigarette cost 6 purchase points or 30 minutes of sustained operant work. Each marihuana cigarette contained approximately 1 gm of marihuana (1.8 to 2.3% THC). Subjects could also work for money at the cost of 50 cents per 6 points or 30 minutes of sustained operant work. Points earned could be exchanged either for marihuana or for money.

All subjects smoked some marihuana every day. Casual marihuana smokers smoked an average of 2.6 cigarettes per day, and heavy marihuana users smoked an average of 5.7 cigarettes per day. However, both groups showed a linear increase in marihuana smoking over the 21 day period of marihuana availability. On the final day of marihuana availability, the casual users smoked an average of 5.8 cigarettes and the heavy users smoked an average of 14.3 cigarettes (Mendelson et al., 1976b).

All subjects worked longer at the operant task and earned far more points than were required to buy the quantity of marihuana they actually smoked. The heavy user group worked between 6.7 and 14.4 hours per day even though 2.2 to 3 hours of work were required for the 4.3 to 6 cigarettes usually smoked per day. Throughout the period of marihuana use, heavy users worked up to an average of 10 hours each day. The casual marihuana users worked between 5 and 11 hours each day even though the number of cigarettes smoked (2 to 3 per day) required only 1 to 1.5 hours of operant work. No subject stopped operant work even when he smoked 10 or more marihuana cigarettes per day. Moreover, periods of maximal operant work coincided with periods of maximal marihuana smoking (i.e., between 4 p.m. and 12 midnight each day).

Subjects worked more for money than for marihuana, and the dollars saved far exceeded dollars spent on marihuana by both the casual and the heavy marihuana users. At the conclusion of the study, the heavy user group had saved an average of $242.38 (± $19.22 S.E.M.) and the casual users had saved an average of $233.17 (± $26.31 S.E.M.). These earnings reflected sustained operant work and savings during the period of marihuana availability. Since both casual and heavy marihuana users worked for both money and marihuana reinforcement during a period of unrestricted marihuana smoking, these data appear to argue strongly against simplistic descriptions of marihuana effects on motivation (Mendelson et al., 1976b).

Subjects also worked at the operant task at far higher rates of responding than were required. Only the first response after an interval of one second had elapsed was counted as an effective response by the programming circuitry. The response requirements were carefully explained to the subjects but most reported they preferred to respond at a comfortable rate. In most instances this resulted in the emission of approximately 600 responses for each purchase point earned, when 300 responses distributed over 5 minutes would have sufficed. A rate of 120 responses per minute (two responses per second) was typical. Sustained high rates of operant responding by casual and heavy marihuana users during the period of active marihuana smoking are also inconsistent with the notion that marihuana induces an "amotivational" syndrome.

Polydrug Use: Marihuana, Alcohol, and Tobacco

Operant procedures for the examination of single drug self-administration patterns can also be extended to study the concurrent self-administration of two or more drugs. Polydrug use appears to be an increasingly frequent drug use pattern, according to clinical and epidemiological studies (Benvenuto, Lau, & Cohen, 1975; Bourne, 1975). The possible combinations of abused drugs appear almost infinite and defy any effort at simple categorization. However, survey data suggest that marihuana is often used in combination with alcohol (Carlin & Post, 1971; Goode, 1969; Grupp, 1972; Tec, 1973), and alcohol is perhaps the most commonly used and abused recreational drug available today. Tobacco use frequently accompanies alcohol use, and it has been shown that alcohol availability increases cigarette smoking in alcoholics (Griffiths, Bigelow, & Liebson, 1976b) and social drinkers (Mello, Mendelson, Sellers, & Kuehnle, 1980a).

The way in which alcohol and marihuana interact and influence concurrent use patterns has long been a subject of speculation. However, there is a prevailing impression that the combined use of marihuana and alcohol leads to a subjective enhancement of the positive or euphorigenic properties of marihuana (Hollister, 1976; Manno, Manno, Kiplinger, & Forney, 1974). Since the combined effects of alcohol and marihuana are thought to be facilitory, we were interested in exploring the effects of concurrent access to marihuana and alcohol. We were interested in learning whether concurrent access to marihuana and alcohol had led to an increase, a decrease, or no change in use patterns of these drugs. On the basis of data demonstrating that alcohol induced an enhancement of tobacco use, we postulated that marihuana and alcohol use would be increased under concurrent access conditions (Mello et al., 1978).

Sixteen adult male volunteers with a history of concurrent alcohol and marihuana use were studied in groups of four on a clinical research ward. Patterns of drug use during 10 days of concurrent access to marihuana and alcohol were compared with successive 5-day periods when only alcohol or only marihuana was available. Two groups were studied in the alcohol first sequence, and two groups were studied in the marihuana first sequence. A drug free control period preceded and followed the 20-day period of spontaneous drug self-administration.

Drug use patterns were assessed by performance on the simple operant task used in studies of marihuana self-administration described previously. Subjects could earn money (50 cents) or marihuana (a 1-gm cigarette containing 1.8 to 2.3% THC) by working at the operant task on an FR-300 (FI-1 sec:S) schedule of reinforcement for 30 minutes. Alcohol (1 ounce) was available as wine, beer, or distilled spirits for 15 minutes of operant work. Subjects could work for only one type of reinforcer at a time, and points were not interchangeable between categories.

The major finding of this study was that concurrent access to alcohol and marihuana resulted in a significant decrease in alcohol consumption in comparison to a 5-day period when only alcohol was available. Fourteen of the 16 subjects studied decreased alcohol use when marihuana was also available (p < 0.01), and the magnitude of the decrease in drinking was significant for seven subjects (p < 0.05).

During the 10-day period of alcohol and marihuana access, subjects gradually increased marihuana smoking and this increase was significant (p < 0.001) as evaluated by a trend analysis. However, this increase cannot be attributed to the concurrent availability of alcohol, since a similar trend was seen in our previous study of casual and heavy marihuana use under comparable experimental conditions (Mendelson et al., 1976b, 1976c). Although 12 subjects smoked more marihuana when alcohol was also available (p < 0.05), the magnitude of this increase was statistically significant in only two instances.

Figure 8 illustrates the most common drug use pattern observed: a slight increase in marihuana use and a decrease in alcohol use during the period of concurrent marihuana and alcohol availability. This subject smoked an average of 5 cigarettes per day during the baseline period of marihuana availability. He was also a heavy drinker and consumed an average of 20 drinks per day during the baseline period of alcohol availability. Peak blood alcohol levels ranged between 50 and 140 mg/dl during the hours of maximum drinking. When both alcohol and marihuana were concurrently available, marihuana smoking increased slightly to an average of 6 cigarettes per day. Alcohol consumption decreased to a mean of 5 drinks per day. Peak blood alcohol levels never exceeded 110 mg/dl during the period of concurrent marihuana and alcohol access.

Subjects usually used alcohol and marihuana together during the period of concurrent availability. Despite the temporal concordance of marihuana and alcohol use, there were no instances of adverse reactions or other evidence of toxic drug interactions as has been reported by others following low acute doses of alcohol and marihuana (Sulkowski & Vachon , 1977).

Only 6 of the 16 subjects studied were consistent tobacco users who smoked an average of 15.9 cigarettes per day. Tobacco use was significantly correlated (p < 0.05) with both alcohol and marihuana use. Tobacco use also accompanied alcohol use and marihuana use during the single drug availability period. These data are consistent with previous reports of alcohol-induced increases in tobacco use (Griffiths et al., 1976b) and with survey reports of a high correlation between marihuana and tobacco use (O’Donnell, 1976).

Drug use and purchase points earned over 34 consecutive days
Figure 8: Marihuana and alcohol use and purchase points earned for money and drugs over 34 consecutive days. The successive conditions of drug availability are shown at the top of the figure. The first row shows the number of marihuana cigarettes smoked each day (filled squares). The average number of marihuana cigarettes smoked during each marihuana availability period is indicated by the dotted line and cross hatched area (± S.E.M.). The second row shows the number of alcoholic drinks consumed each day (open circles) and the 24 hour mean and range of blood alcohol levels observed each day (filled circles). The average number of drinks consumed during the 5- or 10-day period is shown as a dotted line and cross hatched area (mean ± S.E.M.). The third row shows the number of purchase points earned for money (open triangles), marihuana (filled squares), and alcohol (open circles). Consecutive days of the study are shown on the abscissa. On the single day (D) immediately following the pre-drug baseline and preceding the post-drug baseline, an acute combined dose of alcohol and marihuana was given. No drugs were available on the single day (D/) following the 5-day period of marihuana availability and of alcohol availability. Adapted from Mello, Mendelson, and Kuehnle, 1978.

Data obtained are not consistent with the hypothesis that the simultaneous availability of marihuana and alcohol will lead to a significant increase in the use of both drugs. Only two of the 16 subjects increased consumption of both alcohol and marihuana during the simultaneous access conditions, even though alcohol and marihuana were usually used together. These data suggest the importance of defining conditions under which multiple drug access will result in a depression of the use of one or more drugs, an increase in the use of one or both drugs, or no change in drug use as a function of single or multiple drug access. In future studies it will be important to try to identify the interacting pharmacological and behavioral variables which control patterns of multiple drug use. It will also be necessary to determine the generality of these findings with other groups of heavy drinkers and alcoholic individuals.

Heroin Self-Administration

Although opiate self-administration has been studied extensively with animal models (see Johanson & Schuster, 1981; Schuster & Johanson, 1974), there have been relatively few clinical studies of opiate self-administration under controlled research ward conditions. Most of our information about heroin abuse has come from retrospective reports by heroin addicts, usually during a period of drug abstinence. This quasi-anecdotal view has been balanced by the meticulous, empirical studies of the physiological and subjective effects of acute and chronic morphine administration in incarcerated drug addicts who volunteered for research at the Addiction Research Center (ARC) in Lexington, Kentucky (Martin & Isbell, 1978). The long term effects of chronic opiate administration as well as the nature and persistence of the opiate withdrawal syndrome have been examined (for reviews, see Mansky, 1978; Martin & Jasinski, 1969). New therapeutic drugs have been compared with morphine in an effort to predict abuse liability (Jasinski, 1977), and the effects of potential pharmacotherapies for opiate addiction on subjective and physiological reactions to opiates have been evaluated (Martin, 1977; Martin, Jasinski, Haertzen, Kay, Jones, Mansky, & Carpenter, 1973a; Martin, Jasinski, & Mansky, 1973b).

With few exceptions the ARC studies have used a fixed dose opiate administration paradigm. In 1952, Wikler reported the first systematic study in which the dose and frequency of morphine administration was determined by an opiate addict rather than by the investigator. Over a two month period, the subject progressively increased his morphine dose from 30 to 1,000 mg on the last day of morphine availability. During this first simulation of naturalistic opiate use patterns, dreams, fantasies, and subjective drug reactions were studied in clinical interviews (Wikler, 1952). This important study was the beginning of a new approach to the study of drug abuse which has many adherents today—the direct observation of drug self-administration behavior and its consequences (for review, see Krasnegor, 1978).

Methadone Effects on Heroin Self-Administration

During the 1970s, as new pharmacotherapies became available for the treatment of opiate abuse, it became evident that evaluation would be greatly facilitated by the availability of objective and quantifiable measures of drug-seeking behavior. Martin and co-workers (1973a) and Jones and Prada (1975) evaluated the effects of methadone maintenance on operant work for Dilaudid (hydromorphone) in heroin addicts. They reported that some heroin addicts maintained on 50 to 100 mg/day of methadone continued to work for Dilaudid (4 mg/injection, i.v.) for approximately 2.5 months by riding an exercycle 10 miles within one hour. Continued heroin use during methadone maintenance has also been observed clinically.

Naltrexone Effects on Heroin Self-Administration

In 1979, Meyer, Mirin and co-workers reported a series of studies designed to evaluate the efficacy of the long-acting narcotic antagonist, naltrexone, on heroin self-administration by heroin addicts (Meyer & Mirin, 1979). This study demonstrated that it was possible to conduct long term in-patient studies with heroin addict volunteers who were not incarcerated and who were not under any legal or other constraints.

Subjects were allowed to work for heroin on a simple mechanical operant manipulandum similar to that shown in Figure 3. Heroin cost either 300 points or 2100 points per 0.5 mg. Points could be accumulated at about 10,000 per hour. Subjects were allowed to determine the interval between heroin doses and could increase a dose (e.g., from 2.5 to 7.5 mg) by waiting 6 hours rather than taking a smaller dose (2.5 mg) every 2 hours. Although the operant task was relatively simple, the heroin addict subjects rapidly learned how to tamper with the operant manipulanda and advance the counters. They were unrelentingly ingenious in thwarting all efforts to reliably relate operant work for heroin to actual heroin use (Meyer & Mirin, 1979).

Narcotic addicts maintained on naltrexone reported no subjective effects of heroin. In most subjects there was no objective evidence (e.g., myosis, changes in vital signs) that heroin produces physiological effects during naltrexone blockade. These data confirmed the observation of Martin and co-workers that 50 mg of naltrexone produced a total blockade of narcotic effects for a period of 24 hours (Martin, Jasinski, & Mansky, 1973b).

All subjects self-administered significantly more heroin under placebo conditions than under naltrexone blockade. Meyer & Mirin (1979) reported that when naltrexone was administered to informed subjects, there was little experimentation with heroin: Seven of nine subjects sampled heroin an average of 13 times (range 2 to 46) over a 10-day period of heroin availability while maintained on 75 mg/day of naltrexone (p.o.). However, when naltrexone was administered under double blind conditions, each of 22 subjects sampled heroin occasionally. Over a 10-day period of heroin access, 11 subjects took heroin on an average of 15.9 occasions, whereas the other 11 took heroin on an average of 4.3 occasions. Frequency of heroin self-administration among the other members of the group appeared to be the best predictor of heroin-use frequency in any individual (Meyer & Mirin, 1979). It is also possible that the escalating dose schedule used encouraged heroin sampling, since naltrexone-maintained subjects continued to test the antagonist blockade at progressively higher heroin doses.

Despite the effectiveness of the naltrexone blockade, some subjects who sampled heroin frequently during naltrexone maintenance showed respiratory depression and pupillary constriction after the first several heroin doses. Meyer and Mirin (1979) suggested that these autonomic effects were not due to inadequate antagonist blockade but rather were classically conditioned responses which extinguished after repeated blocked heroin injections. The importance of conditioning effects associated with the ritual of heroin self-injection has been clearly demonstrated (Grabowski & O’Brien, 1981; O’Brien, 1976).

The subjective consequences of heroin self-administration observed on the clinical research ward were completely at variance with the euphorigenic, pleasurable effects commonly ascribed to opiate intoxication in retrospective accounts by addicts. Meyer and Mirin (1979) confirmed and extended the previous observations of Wikler (1952) and Haertzen and Hooks (1969) that chronic opiate use is often accompanied by an increase in dysphoria, hypochondriasis and irritability, as well as increased psychopathology, belligerence, negativism, motor retardation, and social isolation. Although it appeared that each heroin injection was associated with a brief elevation in mood, even this transient mood change diminished as a function of chronic drug use (Meyer & Mirin, 1979; Mirin, McNamee, & Meyer, 1976). These data attesting to the dysphoric consequences of chronic heroin use are concordant with data on chronic alcohol intoxication and challenge the notion that drugs are used solely for their rewarding or euphorigenic properties (see Mello, 1977, 1978, 1983). These mood changes associated with chronic heroin use did not occur in subjects maintained on naltrexone for comparable periods. Consequently, the increased psychopathology and dysphoria were probably related to heroin use rather than residence on the research ward.

In the early 1980s, we evaluated the effects of naltrexone on heroin self-administration in comparison to placebo under double blind conditions in a clinical research ward setting (Mello et al., 1981). In order to achieve the primary goals of these studies, it was essential to devise a reliable operant procedure that would yield objective and quantifiable data about drug acquisition. The advantage of direct observation over reliance on retrospective self-reports or predictions of probable behavior by addict subjects is obvious. This design permitted examination of the efficacy and the limitations of this new pharmacotherapy operationally defined by the effect on the amount and frequency of heroin self-administration as well as the behavioral consequences of drug use.

Operant techniques were used to provide an objective and quantitative measure of performance for two alternative reinforcers. Since drugs are not available to most users without some expenditure of effort or money, we compared operant work for heroin with acquisition of an alternative reinforcer, money. The effects of naltrexone on heroin self-administration were measured in terms of duration, rate, and pattern of operant performance for heroin rather than inferred from verbal behavior. The effect of heroin intoxication on operant performance for money was examined in comparison to performance during drug-free conditions. The operant manipulandum shown in Figure 6 proved to be tamper proof and effective for this purpose.

Twelve male heroin addict volunteers lived on a clinical research ward for 34 days. After a 9-day drug-free period, naltrexone or placebo was given and heroin (40 mg/day) was available for 10 days. Subjects could earn money ($1.50) or heroin (10 mg/injection, i.v.) by responding on a second-order schedule of reinforcement (FR-300 [FI-1 sec:S]) for approximately 90 minutes. Subjects were limited to four 10-mg doses of heroin each day for 10 consecutive days for both medical and ethical considerations. Subjects could refuse to take any heroin dose earned but were not allowed to take fractional doses (e.g., 5 instead of 10 mg). Points earned for heroin could not be exchanged for money. However, naltrexone-maintained subjects who worked for heroin points during baseline and then elected not to purchase heroin were allowed to exchange heroin points for money at the end of the study.

Naltrexone effectively suppressed heroin self-administration. The three naltrexone-maintained subjects took only 2 to 7.5% of the total heroin available, two subjects stopped heroin self-administration after the first or second injection, and the third subject took a total of three injections over the 10-day period of availability. There were no discernible adverse side effects during 25 consecutive days on naltrexone maintenance. In contrast, the nine placebo naltrexone subjects used 57.5 to 100% of the total heroin available.

Cumulative records for money and for heroin responding
Figure 9: Cumulative records of responses for money and for heroin by a placebo naltrexone subject. Records of responding on an FR-300 (FI-1 sec:S) schedule of reinforcement are shown for the first day of the drug-free condition and for Days 1 and 5 of heroin availability. Cumulative records of responding are a direct read-out of response rates; each response advances the step pen 0.25 mm, and the paper advances at a constant speed of 28 mm/hour. The response pen resets after a fixed number of responses or when the subject changed from working for heroin to money or the converse. Each deflection of the bottom event pen indicates completion of 300 effective responses on the FI-1 second schedule of reinforcement. The top pen codes whether the subject was working for heroin (upward deflections) or money (no deflections). Only money was available on Days 1 to 7 of the drug-free conditions so no code is shown. Reprinted from Mello, Mendelson, Kuehnle, and Sellers, 1981. Copyright 1981 by the Society for Pharmacology and Experimental Therapeutics.

Cigarette smoking patterns of operant earning and spending for money and heroin
Figure 10: Cigarette smoking patterns of operant earning and spending for money and heroin by a subject assigned to the placebo naltrexone condition. The successive conditions—drug free baseline (Pre), heroin availability, methadone detoxification (M), drug-free period (Post), and naltrexone maintenance (N)—are shown across the top of the figure. The number of 10-mg doses of heroin taken each day are shown underneath the heroin heading. A maximum of four injections (40 mg/injection) was available each day. The number of tobacco cigarettes smoked each day is shown in the top row. The second row shows the number of hours worked for money (open bars) and heroin (shaded bars) each day throughout the study. The third row shows the number of purchase points earned for money (open circles) and for heroin (filled circles) each day. Acquisition of each purchase point required 300 effective responses on an FI-1 sec schedule of reinforcement (FR-300 [FI-1 sec:S]). On Days 6 and 8 of the pre-heroin base line, blood samples were collected for neuroendocrine studies over 10 consecutive hours. Adapted from Mello, Mendelson, Kuehnle, and Sellers, 1981, and Mello, Mendelson, Sellers, and Kuehnle, 1980b.

Heroin intoxication did not impair operant performance. Heroin users worked longer hours and earned significantly more purchase points during heroin self-administration and subsequent methadone detoxification than during the drug-free period. On the first day of heroin use, subjects worked 10 or more hours at the operant task. Ten days of chronic heroin intoxication were not associated with progressive decreases in operant work. Moreover, heroin intoxication did not compromise the ability of the subjects to titrate their operant work to acquire precisely the amount of heroin they wished to buy. Subjects continually monitored their accumulated points shown on the operant panel and did not earn more points for heroin than it was possible to spend.

A typical cumulative record illustrating that 10 mg of heroin had minimal effects on operant performance is shown in Figure 9. The lack of effect of heroin use on operant performance reflected in total points earned as illustrated in data from individual subjects is shown in Figure 10. These data also show that cigarette smoking increased during the period of heroin use (Mello et al., 1980b), an effect consistently observed with alcohol (Griffiths et al., 1976b; Mello et al., 1980a).

Buprenorphine Effects on Human Heroin Self-Administration

A similar experimental design was used to compare the effects of buprenorphine, an opioid mixed agonist/antagonist, with placebo on operant acquisition of heroin and money studied under double blind conditions (Mello et al., 1982). Buprenorphine combines the salient characteristics of two pharmacotherapies for heroin addiction, because it is both a partial opioid agonist similar to methadone and a potent opioid antagonist similar to naltrexone (Lewis, Rance, & Sanger, 1983). Ten male volunteers with a history of heroin abuse lived on a clinical research ward for 40 days. After a 5-day drug-free period, buprenorphine or placebo was given in gradually ascending doses (0.5 to 8 mg/day, s.c.) over 14 days. Subjects were maintained on 8 mg/day of buprenorphine for 10 days during which they could earn money ($1.50) or heroin (7 or 13.5 mg/injection, i.v.) by responding on a second-order schedule of reinforcement (FR-300 [FI-1 sec:S]) for approximately 90 minutes.

Buprenorphine-maintained subjects took significantly less heroin than subjects maintained on placebo (p < 0.001). Buprenorphine-maintained subjects took only between 2 and 31% of the total amount of heroin available, whereas placebo-maintained subjects took between 93 and 100% of the available heroin.

There were a number of similarities between the operant performance for heroin and money of the subjects studied under placebo buprenorphine and placebo naltrexone maintenance. In both studies subjects showed the capacity to precisely titrate operant work to acquire the desired amount of heroin and then resume working for money. There was no evidence of impairment of operant performance for heroin or for money during heroin intoxication. In fact, there were no significant differences in total operant points earned or in total hours worked during any phase of the study between the placebo group who used heroin and the buprenorphine-maintenance group who did not (Mello et al., 1982).

These points are clearly indicated by data shown in Figure 11. This subject was studied under both buprenorphine-maintenance conditions and placebo-maintenance conditions. It is obvious that maintenance of buprenorphine did not, in and of itself, significantly suppress operant performance for money or for time spent working at the operant task. These data indicate that heroin addicts can perform a simple operant task effectively for long periods of time (6 to 18 hours per day) during heroin intoxication at doses of 21 to 40.5 mg/day (Mello et al., 1982).

Summary and Conclusions

Data have been presented to illustrate some ways that operant procedures can be used to study self-administration patterns of a variety of substances. Operant procedures for drug self-administration studies have also been shown to be useful for concurrent examination of a variety of drug-effect variables, especially the effects of drugs on biological systems. Behavioral data were selected to illustrate some clinical research findings which are contrary to conventional wisdom and common expectation. We have seen that alcohol addicts do not maintain a constant pattern of alcohol intake and do not drink all the alcohol available when given unrestricted access to alcohol in a self-administration paradigm. Rather, alcohol addicts tend to alternate periods of alcohol intoxication and operant work, even though periods of abstinence working are accompanied by partial withdrawal symptoms. These data are inconsistent with the notion that alcohol abuse is maintained by either the avoidance of withdrawal signs and symptoms or an uncontrollable "craving" for alcohol (Mello, 1975; Mello & Mendelson, 1972).

Studies of marihuana self-administration are not consistent with the notion that marihuana intoxication produces an "amotivational" syndrome (Mendelson et al., 1976b, 1976c). Rather, subjects worked an average of 10 hours each day while smoking four to six marihuana cigarettes. Most subjects worked while they were smoking marihuana and earned far more money than they spent on marihuana cigarettes.

Similarly, heroin intoxication did not result in sedation and inactivity at doses up to 40 mg/day. Subjects often became more active after a heroin injection. During heroin intoxication subjects worked longer and earned more points than during drug-free or methadone detoxification phases of the study (Mello et al., 1981).

Studies of polydrug use involving alcohol and marihuana indicate that concurrent access to these drugs is not necessarily associated with increased use of alcohol and marihuana as would be predicted from anecdotal accounts and objective data on alcohol and tobacco (Griffiths et al., 1976b). Rather, the simultaneous availability of marihuana and alcohol was associated with a significant decrease in alcohol consumption in comparison to a baseline period when only alcohol was available (Mello et al., 1978).

We conclude that there is no substitute for direct clinical observation of drug use patterns and of the effects of drugs on behavioral and biological variables under controlled conditions. The direct observation of patterns of drug use and assessments of the subjective and objective consequences of drug intoxication have challenged many prevalent assumptions derived from retrospective reports by drug abusers during sobriety (Mello & Mendelson, 1978). It is our contention that direct clinical observation of drug use patterns and associated drug effects in the same individual over time is essential to an improved understanding of the behavioral bases of drug abuse. The experimental analysis of drug use patterns is one way to examine factors which maintain and perpetuate drug abuse. The periodicity of drug use and the rate and duration of operant behavior involved in drug acquisition are data which can be measured directly without reliance on alleged intervening variables such as "drug hunger."

Patterns of heroin self-administration
Figure 11: Patterns of heroin self-administration by a subject studied under both buprenorphine maintenance (8 mg/day; filled bars) and placebo maintenance (open bars) conditions. The daily dose of heroin or methadone (mg/day) is shown on the left ordinate. The daily dose of buprenorphine or equal volume placebo control maintenance is shown on the right ordinate. The sequence of conditions (pre-drug baseline, buprenorphine or placebo induction, heroin availability, detoxification with buprenorphine or methadone, and the post-drug baseline) is shown at the top of the figure. Consecutive days of the study are shown on the abscissa. Row 2 shows the hours worked for money under buprenorphine (filled bars) and placebo (open bars) conditions and hours worked for heroin (shaded bars) under each condition. The number of purchase points earned for money and for heroin under buprenorphine and placebo maintenance conditions are shown in the third row. The consecutive study days are shown on the abscissa. Adapted from Mello, Mendelson, and Kuehnle, 1982.

On the basis of such information, certain inferences can be made concerning broader questions about the phenomenology of drug abuse. It is of interest to determine if there are consistencies in drug use patterns within the behavior of a single individual or between individuals with comparable drug use histories. Given the diversity of individuals with substance abuse problems, an examination of factors which initiate and maintain periodic substance abuse episodes may prove more productive than a search for origins. A better understanding of how drug self-administration is maintained should permit more effective manipulation of critical maintenance variables and lead to the development of more effective forms of intervention. Both the conceptual and technical aspects of an operant analysis of drug self-administration behavior appear to be optimally designed to facilitate our understanding of how substance abuse is maintained. Comparisons of drug use patterns across addictive disorders may eventually permit identification of some reliable commonalities and differences that in turn will help to clarify the nature of drug-related reinforcers.


Preparation of this review was supported in part by Grants No. K05 DA00101, K05 DA0064, DA02519, and DA02905 from the National Institute on Drug Abuse and AA06252 from the National Institute on Alcohol Abuse and Alcoholism, ADAMHA. Portions of this paper were revised and reprinted with permission from N. Krasnegor (Ed.) (1978), Self-administration of abused substances: Methods for study (National Institute on Drug Abuse Research Monograph 20, pp. 93-127). Washington, DC: U.S. Government Printing Office. We thank Loretta Carvelli for her excellent assistance in preparing this manuscript.


Babor, T. F., Mendelson, J. H., Greenberg, I., & Kuehnle, J. C. (1978). Experimental analysis of the "happy hour"—effects of purchase price on alcohol consumption. Psychopharmacology, 58, 35-41.

Benvenuto, J. A., Lau, J., & Cohen, R. (1975). Patterns of nonopiate/polydrug abuse: Findings of a national collaborative research project. In Problems of drug dependence. Proceedings of the 37th Annual Scientific Meeting, Committee on Problems of Drug Dependence, (pp. 234-254). Washington, DC: National Academy of Sciences—National Research Council.

Bernstein, J. G., Kuehnle, J. C., & Mendelson, J. H. (1976). Medical implications of marihuana use. American Journal of Drug and Alcohol Abuse, 3, 347-361.

Bigelow, G. E., Griffiths, R. R., & Liebson, I. (1976). Effects of response requirement upon human sedative self-administration and drug-seeking behavior. Pharmacology Biochemistry & Behavior, 5, 681-685.

Bourne, P. G. (1975). Polydrug abuse—status report on the federal effort. In E. Senay, V. Shorty, & H. Alksen (Eds.), Developments in the field of drug abuse: National drug abuse conference (pp. 197-207). Cambridge, MA: Schenkman Publishing.

Carlin, A. S., & Post, R. D. (1971). Patterns of drug use among marihuana smokers. Journal of the American Medical Association, 218, 867-868.

Goldenheim, P., Mendelson, J. H., Mello, N. K., Tilles, D. , & Bavli, S. (1984). Marihuana smoking causes a reduction in single breath carbon monoxide diffusing capacity. American Journal of Medicine, 1985 (in press).

Goode, E. (1969). Marihuana and the politics of reality. Journal of Health and Social Behavior, 10, 83-94.

Grabowski, J., & O’Brien, C. P. (1981). Conditioning factors in opiate use. In N. K. Mello (Ed.), Advances in substance abuse, behavioral and biological research (Vol. 2, pp. 69-121). Greenwich: JAI Press.

Griffiths, R. R., Bigelow, G. E., & Henningfield, J. E. (1980). Similarities in animal and human drug taking behavior. In N. K. Mello (Ed.), Advances in substance abuse, behavioral and biological research (Vol. 1, pp. 1-90). Greenwich: JAI Press.

Griffiths, R. R., Bigelow, G. E., & Liebson, I. (1976a). Human sedative self-administration. Effects of interingestion interval and dose. Journal of Pharmacology and Experimental Therapeutics, 197, 488-494.

Griffiths, R. R., Bigelow, G. E., & Liebson, I. (1976b). Facilitation of human tobacco self-administration by ethanol: A behavioral analysis. Journal of Experimental Analysis and Behavior, 15, 279-292.

Grupp, S. E. (1972). Multiple drug use in a sample of experienced marihuana smokers. International Journal of Addictions, 7, 481-491.

Haertzen, C. A., & Hooks, N. T. (1969). Changes in personality and subjective experience associated with the chronic administration and withdrawal of opiates. Journal of Nervous and Mental Disease, 148, 606-614.

Hollister, L. E. (1976). Interactions of delta-9-tetrahydrocannabinol with other drugs. In E. S. Vesell & M. C. Braude (Eds.), Interactions of drugs of abuse (pp. 212-218). New York: New York Academy of Sciences.

Jasinski, D. R. (1977). Assessment of the abuse potentiality of morphine-like drugs (methods used in man). In W. R. Martin (Ed.), Drug addiction I: Morphine, sedative/hypnotic and alcohol dependence (Handbook of Experimental Pharmacology, Vol. 45, pp. 197-258). Berlin: Springer-Verlag.

Johanson, C. E., & Schuster, C. R. (1981). Animal models of drug self-administration. In N. K. Mello (Ed.), Advances in substance abuse, behavioral and biological research (Vol. 2, pp. 219-297). Greenwich: JAI Press.

Jones, B. E., & Prada, J. A. (1975). Drug seeking behavior during methadone maintenance. Psychopharmacologia, 41, 7-10.

Kelleher, R. T., Goldberg, S. R., & Krasnegor, N. (Eds.). (1976). Control of drug taking behavior by schedules of reinforcement. Baltimore: Williams & Wilkins.

Krasnegor, N. A. (Ed.). (1978). Self-administration of abused substances: Methods for study (National Institute of Drug Abuse Research Monograph 20). Washington, DC: U.S. Government Printing Office.

Kuehnle, J. C., Mendelson, J. H., Davis, K. R., & New, P. F. J. (1977). Computed tomographic examination of heavy marihuana smokers. Journal of the American Medical Association, 237, 1231-1232.

Lewis, J., Rance, M. J., & Sanger, D. J. (1983). The pharmacology and abuse potential of buprenorphine, a new antagonist analgesic. In N. K. Mello (Ed.), Advances in substance abuse, behavioral and biological research (Vol. 3, pp. 103-154). Greenwich: JAI Press.

Manno, J. E., Manno, B. R., Kiplinger, G. F., & Forney, R. B. (1974). Motor and mental performance with marihuana: Relationship to administered dose of delta-9-tetrahydrocannabinol and its interaction with alcohol. In L. L. Miller (Ed.), Marihuana, effects on human behavior (pp. 45-72). New York: Academic Press.

Mansky, P. A. (1978). Opiates: Human psychopharmacology. In L. L. Iversen, S. D. Iversen, & S. H. Snyder (Eds.), Handbook of psychopharmacology (Vol. 12, pp. 95-185). New York: Plenum Press.

Martin, W. R. (Ed.). (1977). Chemotherapy of narcotic addiction. In W. R. Martin (Ed.), Drug addiction I: Morphine, sedative/hypnotic and alcohol dependence (Handbook of Experimental Pharmacology, Vol. 45, pp. 279-318). Berlin: Springer-Verlag.

Martin, W. R., & Isbell, H. (1978). Drug addiction and the U.S. Public Health Service (Department of Health, Education, and Welfare Publication 77-434, pp. 325). Washington, DC: U.S. Government Printing Office.

Martin, W. R., & Jasinski, D. R. (1969). Physiological parameters of morphine dependence in man—tolerance, early abstinence, protracted abstinence. Journal of Psychiatric Research, 7, 9-17.

Martin, W. R., Jasinski, D. R., Haertzen, C. A., Kay, D. C., Jones, B. E., Mansky, P. A., & Carpenter, R. W. (1973a). Methadone—a reevaluation. Archives of General Psychiatry, 28, 286-295.

Martin, W. R., Jasinski, D. R., & Mansky, P. A. (1973b). Naltrexone, an antagonist for the treatment of heroin dependence. Archives of General Psychiatry, 28, 784-791.

Mello, N. K. (1972). Behavioral studies of alcoholism. In B. Kissin & H. Begleiter (Eds.), The biology of alcoholism: Vol. 2, Physiology and behavior (pp. 219-291). New York: Plenum Press.

Mello, N. K. (1973). Short-term memory function in alcohol addicts during intoxication. In M. M. Gross (Ed.), Alcohol intoxication and withdrawal: Experimental studies (pp. 333-334). New York: Plenum Press.

Mello, N. K. (1975). A semantic aspect of alcoholism. In H. D. Cappell & A. E. LeBlanc (Eds.), Biological and behavioral approaches to drug dependence (pp. 73-87). Toronto: Addiction Research Foundation.

Mello, N. K. (1977). Stimulus self-administration: Some implications for the prediction of drug abuse liability. In T. Thompson & K. R. Unna (Eds.), Predicting dependence liability of stimulant and depressant drugs (pp. 243-260). Baltimore: University Park Press.

Mello, N. K. (1978). Control of drug self-administration: The role of aversive consequences. In R. C. Peterson & R. C. Stillman (Eds.), Phencyclidine abuse: An appraisal (National Institute on Drug Abuse Research Monograph 21, pp. 289-308).

Mello, N. K. (1983). A behavioral analysis of the reinforcing properties of alcohol and other drugs in man. In B. Kissin & H. Begleiter (Eds.), The pathogenesis of alcoholism, biological factors (Vol. 7, pp. 133-198). New York: Plenum Press.

Mello, N. K., McNamee, H. B., & Mendelson, J. H. (1968). Drinking patterns of chronic alcoholics: Gambling and motivation for alcohol. In J. O. Cole (Ed.), Clinical research in alcoholism (Psychiatric Research Report 24, pp. 83-118). Washington, DC: American Psychiatric Association.

Mello, N. K., & Mendelson, J. H. (1965). Operant analysis of drinking patterns of chronic alcoholics. Nature, 206, 43-46.

Mello, N. K., & Mendelson, J. H. (1970). Experimentally induced intoxication in alcoholics—A comparison between programmed and spontaneous drinking. Journal of Pharmacology and Experimental Therapeutics, 173, 101-116.

Mello, N. K., & Mendelson, J. H. (1972). Drinking patterns during work-contingent and non-contingent alcohol acquisition. Psychosomatic Medicine, 34, 139-264.

Mello, N. K., & Mendelson, J. H. (1978). Alcohol and human behavior. In L. L. Iversen, S. D. Iversen, & S. H. Snyder (Eds.), Handbook of psychopharmacology, Vol. 12: Drugs of abuse (pp. 235-317). New York: Plenum Press.

Mello, N. K., Mendelson, J. H., & Kuehnle, J. C. (1978). Human polydrug use: Marihuana, alcohol and tobacco. Journal of Pharmacology and Experimental Therapeutics, 207, 922-935.

Mello, N. K., Mendelson, J. H., & Kuehnle, J. C. (1982). Buprenorphine effects on human heroin self-administration: An operant analysis. Journal of Pharmacology and Experimental Therapeutics, 223, 30-39.

Mello, N. K., Mendelson, J. H., Kuehnle, J. C., & Sellers, M. (1981). Operant analysis of human heroin self-administration and the effects of naltrexone. Journal of Pharmacology and Experimental Therapeutics, 216, 45-54.

Mello, N. K., Mendelson, J. H., Sellers, M., & Kuehnle, J. C. (1980a). Effect of alcohol and marihuana on tobacco smoking. Clinical Pharmacology and Therapeutics, 27, 202-209.

Mello, N. K., Mendelson, J. H., Sellers, M. , & Kuehnle, J. C. (1980b). Effects of heroin self-administration on cigarette smoking. Psychopharmacology, 67, 45-52.

Mendelson, J. H. (Ed.). (1964). Experimentally induced chronic intoxication and withdrawal in alcoholics. Quarterly Journal of Studies on Alcoholism, Suppl. 2.

Mendelson, J. H. (1970). Biological concomitants of alcoholism. New England Journal of Medicine, 283, 24-32 & 71-81.

Mendelson, J. H., Babor, T. F., Kuehnle, J. C., Rossi, A. M., Bernstein, J. G., Mello, N. K., & Greenberg, I. (1976a). Behavioral and biological aspects of marihuana use. Annals of the New York Academy of Sciences, 282, 186-210.

Mendelson, J. H., Ellingboe, J., Kuehnle, J. C., & Mello, N. K. (1978). Effects of chronic marihuana use on integrated plasma testosterone and luteinizing hormone levels. Journal of Pharmacology and Experimental Therapeutics, 207, 611-617.

Mendelson, J. H., Kuehnle, J. C., Ellingboe, J., & Babor, T. F. (1974a). Plasma testosterone levels before, during, and after chronic marihuana smoking. New England Journal of Medicine, 291, 1051-1055.

Mendelson, J. H., Kuehnle, J. C., Greenberg, I., & Mello, N. K. (1976b). Operant acquisition of marihuana use in man. Journal of Pharmacology and Experimental Therapeutics, 198, 42-53.

Mendelson, J. H., Kuehnle, J. C., Greenberg, I., & Mello, N. K. (1976c). The effects of marihuana on human operant behavior: Individual data. In M. C. Braude & S. Szara (Eds.) Pharmacology of Marihuana (pp. 643-653). New York: Raven Press.

Mendelson, J. H., & Mello, N. K. (1976). Behavioral and biochemical interrelations in alcoholism. Annual Review of Medicine, 27, 321-333.

Mendelson, J. H., & Mello, N. K. (1984a). Effects of marihuana on endocrine hormones in human males and females. In M. C. Braude & U. P. Ludford (Eds.), Marihuana effects on endocrine and reproductive systems (National Institute on Drug Abuse Research Monograph 44, pp. 97-114). Washington, DC: U.S. Government Printing Office.

Mendelson, J. H., & Mello, N. K. (1984b). Behavioral pharmacology of marihuana use by men and women. In R. Balster & L. Seiden (Eds.), Behavioral pharmacology: The current status (pp. 451-463). New York: Alan R. Liss.

Mendelson, J. H., Mello, N. K., Cristofaro, P., Ellingboe, V., & Benedikt, R. (1985). Acute effects of marijuana on pituitary gonadal hormones during the periovulatory phase of the menstrual cycle. In L. S. Harris (Ed.), Problems of drug dependence, 1984. (National Institute on Drug Abuse Research Monograph 55, pp. 24-31). Washington, DC: U.S. Government Printing Office.

Mendelson, J. H., Mello, N. K., & Ellingboe, J. (1978). Effects of alcohol on pituitary-gonadal hormones, sexual function and aggression in human males. In M. A. Lipton, A. DiMascio, & K. F. Killam (Eds.), Psychopharmacology, A generation of progress (pp. 1677-1691). New York: Raven Press.

Mendelson, J. H., Mello, N. K., & Solomon, P. (1968). Small group drinking behavior: An experimental study of chronic alcoholics. In A. Wikler (Ed.), The addictive states (pp. 399-430). Baltimore: Williams & Wilkins.

Mendelson, J. H., Rossi, A. N., & Meyer, R. E. (Eds.) (1974b). The use of marihuana: A psychological and physiological inquiry. New York: Plenum Press.

Meyer, R. E., & Mirin, S. M. (1979). The heroin stimulus. New York: Plenum Press.

Mirin, S. M., McNamee, H. B., & Meyer, R. E. (1976). Psychopathology, craving and mood during heroin acquisition: An experimental study. International Journal of Addictions, 11, 525-543.

Nathan, P. E., O’Brien, J. S., & Lowenstein, L. M. (1971). Operant studies of chronic alcoholism: Interaction of alcohol and alcoholics. In M. K. Roach, W. M. McIsaac, & P. J. Creaven (Eds.), Biological aspects of alcohol (pp. 341-370). Austin: University of Texas.

Nathan, P. E., Titler, N. A., Lowenstein, L. M., Solomon, P., & Rossi, A. M. (1970). Behavioral analysis of chronic alcoholism: Interaction of alcohol and human contact. Archives of General Psychiatry, 22, 419-430.

O’Brien, C. P. (1976). Experimental analysis of conditioning factors in human narcotic addiction. In R. T. Kelleher, S. R. Goldberg, & N. A. Krasnegor (Eds.), Control of drug taking behavior by schedules of reinforcement (pp. 533-543). Baltimore: Williams & Wilkins.

O’Donnell, J. T. (Ed.). (1976). Young men and drugs: A nationwide survey. (National Institute on Drug Abuse Research Monograph 5). Washington, DC: U.S. Government Printing Office.

Pickens, R., Cunningham, M. R., Heston, L. L., Eckert, E., & Gustafson, L. K. (1977). Dose preference during pentobarbital self-administration by humans. Journal of Pharmacology and Experimental Therapeutics, 203, 310-318.

Popham, R. E., Schmidt, W., & de Lint, J. (1975). The prevention of alcoholism: Epidemiological studies of the effects of government control measures. British Journal of Addiction, 70, 124-144.

Schuster, C. R., & Johanson, C. E. (1974). The use of animal models for the study of drug abuse. In R. J. Gibbins, Y. Israel, H. Kalant, R. E. Popham, W. Schmidt, & R. G. Smart (Eds.), Research advances in alcohol and drug problems (Vol. 1, pp. 1-31). New York: Wiley.

Skinner, B. F. (1938). The behavior of organisms: An experimental analysis. New York: Appleton-Century-Crofts.

Skinner, B. F. (1953). Science and human behavior. New York: MacMillan.

Sulkowski, A., & Vachon, L. (1977). Side effects of simultaneous alcohol and marihuana use. American Journal of Psychiatry, 134, 691-692.

Tec, N. A. (1973). A clarification of the relationship between alcohol and marihuana. British Journal of Addiction, 68, 191-195.

Wikler, A. (1952). A psychodynamic study of a patient during experimental self-regulated re-addiction to morphine. Psychiatric Quarterly, 26, 279-293.

©1987 Springer-Verlag (printed version)
©2000-2009 Addiction Science Network (web-enhanced version)

ASNet Home

ASNet Profile

Illicit Drug Index

Virtual Lab Tour

Research Reports

Drug Classification

A Primer on Drug Addiction

Experimental Methods

Treatment Resources

Biological Basis of Addiction

Before Prohibition: Early Psychoactive Medicines

Links to Other Websites

Click here to enter the Addiction Science Network Discussion Forum

Brain Reward System©1999-2009 Addiction Science Network
This page was last revised 08 April 2009 18:27 EDT.
Send comments to:
Report technical problems to: