It has been clear for some time that a variety of animal species will voluntarily self-administer a opiates and mu-opioid agonists via various routes (1). When the delivery of a drug is the consequence of a particular behavior, and the behavior changes according to the schedule of drug delivery, then the drug is considered to serve as a reinforcing stimulus (3). That opiates could reinforce arbitrarily learned responses was presaged by Shirley Spragg, who showed that physically dependent chimpanzees would gradually learn to select a box concealing a syringe filled with morphine (4). Perhaps, the first demonstration of operant self-administration was by Headlee, Coppock, and Nichols, who reported that morphine/cocaine-pretreated rats would learn to make head movements in order to received IP injections of morphine and codeine (5).

An initial issue of contention was whether animals required some type of preexposure to opiates in order to subsequently acquire self-administration behavior. It soon became clear, however, that this was a methodological rather than a pharmacological issue (see below), and a distinction between “inducing” versus “non-inducing” procedures was proposed (3). Inducing procedures are employed in situations where drug-naïve animals have to surmount some aversive effect of the drug such as, for example, the bitter taste of morphine in oral self-administration studies. Thus, naïve rats do not normally drink solutions of morphine, but they will learn to prefer such solutions over water if they are repeatedly “forced” to drink them in order to relieve thirst (6), or if they are chronically pretreated with morphine prior to oral self-administration training (7). In contrast, non-inducing procedures simply make the drug available; acquisition of self-administration is accomplished with no further behavioral or pharmacological manipulation (3). These procedures are typically employed when opiates are available intravenously to freely moving animals, a method pioneered in the 1960s by Schuster and Thompson and by Daneau and Yanagita in monkeys, and by Weeks in rats (8). Pre-training animals to respond for food, or other nondrug reinforcers, is generally not necessary for the acquisition of intravenous self-administration of opiates. In addition, although this type of induction procedure is likely to accelerate the rate of learning, and can be useful in identifying “nonresponders,” food pre-training involves food restriction, which is a powerful stressor that can alter subsequent responses to opiates (9, 10). Food pre-training is also not appropriate for studies designed to determine whether animals acquire self-administration of new compounds with untested reinforcing efficacy.

The most convincing evidence suggesting that opiate self-administration is goal-directed is the observation that animals regulate drug intake to achieve and maintain some level of satiety. James Weeks (11) may have been the first to report that intravenous opiate self-administration in rats is characterized by prolonged periods of not responding alternating with brief periods of high rates of responding terminated by an injection. From this, he concluded that morphine was a reinforcer that produced almost immediate satiation. It is now generally accepted that animals dynamically regulate drug intake by adjusting operant responding to the available unit dose. That is, at low ratio schedules, the rate of opiate-reinforced behavior is inversely related to drug dosage (12), and the number of infusions increases or decreases with decreases or increases in the unit dose, respectively (13). This is achieved by altering post-infusion pauses in responding: inter-dose intervals increase following increases in doses, and decrease following decreases in doses (14–16). In addition, mu-antagonist pretreatment usually increases opiate intake, while agonist pretreatment decreases it (1). It should be noted, however, that the relationship between dose and responding is also modulated by the schedule of reinforcement. For example, Roberts et al. (17) explored the progressive ratio (PR) schedule, which restricts drug intake by requiring progressively more operant behavior for each successive infusion. At heroin doses ranging between 12.5 and 100 μg/kg/infusion, an inverted U-curve was observed. Although the exact shape of the curve is regulated by the rate of ratio progression escalation (18) and by the pharmacological properties of the opiate used in self-administration (19), it is clear that unit dose and intensity of responding for the drug are not always linearly related. Therefore, studies using PR schedules have indicated that it may be misleading to estimate the level of motivation to self-administer opiates from an analysis of self-administration behavior at low ratio schedules.

Additional evidence for regulation of opiate intake by satiety levels comes from studies whereby animals are given long periods of access to drugs. In fact, when given unlimited (i.e., 24 h a day) access to opiates, self-administration typically shows an escalation followed by a stabilization. For example, Gracin et al. (20) found that morphine intake in rats increased over 14 days of unlimited access and then stabilized for the rest of the self-administration period (over 30 days) at a relatively constant amount of daily morphine intake (about 240 mg/kg/day). Bozarth and Wise (21) reported that rats given unlimited access to heroin self-administration (0.1 mg/kg/infusion) showed a gradual increase in 24 h drug intake during the first 2 weeks of testing, but after this time, daily intake remained constant. Interestingly, Ahmed et al. (22) demonstrated that escalation induced by extending the duration of access to opiates can largely be attributed to increases in opiate intake during the initial period of each daily session, and a recent study by Chen et al. (23) suggested that these changes are associated with the development of dependence. More specifically, rats given unlimited (23 h a day) access to heroin for 35 days stabilized their intake after approximately 14 days of self-administration, and escalation was marked by an increase of heroin intake in the first hour of each session. Importantly, somatic signs of naloxone (1 mg/kg)-precipitated withdrawal were maximal after 14 days of self-administration, suggesting that stabilization of heroin intake occurred when a new “dependence level” of satiety was achieved.

It should be noted, however, that dependence does not necessarily require escalation of intake. In fact, when heroin access per hour is limited and animals still have access to the drug 24 h a day (i.e., discrete-trials procedure), escalation is not observed, but dependence still develops as indicated by loss of body-weight within 24 h following the termination of self-administration (24). Clearly, escalation and dependence are not all-or-none phenomena, and their relationship may be more or less clear depending on the experimental conditions employed.

One possible explanation for why animals self-impose limits on drug intake within- and between-sessions of self-administration may be impaired motor functions rather than attainment of satiety. After all, behavior on the ascending limb of the opiate dose–effect curve is thought to be modulated by reinforcing value, whereas the descending limb is thought to be controlled both by the reinforcing and response-suppressing effects of the self-administered drug (see (25) for review). However, the results of intracranial self-administration studies argue against this possibility. In fact, drug-naïve and untrained rats will readily learn to self-administer morphine and other mu-agonists directly into the ventral tegmental area (26, 27). Importantly, as observed in intravenous studies, rates of operant responding during the first hour of intra-VTA self-administration are significantly higher than that of each subsequent hour of testing. Given that activation of mu-receptors in the VTA cause increases in motor activity (28), not decreases, it is quite likely that changes in within-session response rates are indeed related to the differential requirements for establishing and maintaining satiating drug concentrations in the brain (26).

During the development of inducing and non-inducing protocols, the role of physiological dependence and withdrawal in opiate self-administration became a subject of critical interest. In turn, the analysis of when and how animals self-administer opiates has been instrumental to the understanding of motivational drives for opiate-seeking and opiate-taking behaviors. The central question of this research has been: is learning to seek opiates dependent on drive reduction (i.e., reduction of a need state) or on some pleasurable euphoric effect? (29). The studies of Spragg in chimpanzees strongly supported the drive-reduction hypothesis (4). In fact, he showed that chimpanzees gradually learned to voluntarily accept morphine injections applied to different parts of their body. Over the period of regular drug injections (twice per day, about 160 injections), animals developed tolerance as well as physiological dependence characterized by a variety of emotional and physiological signs of withdrawal. In rats, chronic injections of morphine (100 mg/kg/day × 48 days) lead to the development of dependence, which manifests itself as the occurrence of signs of abstinence within 6 h from the last injection. These signs, which include decreases in eating, drinking, body weight, locomotion, and increases in defecation, are dose-dependently alleviated by administration of additional morphine (6), and can be precipitated by administration of opioid antagonists such as naloxone or naltrexone (30).

Interestingly, the chimpanzees of Spragg also displayed intense behavioral manifestations of “desire” for morphine when experiencing withdrawal. These included vigorous tugging of the experimenter toward the compound where injections were usually administered, and preference for the hypodermic syringe over food when both hungry and in withdrawal from morphine. However, when hungry and recently injected with morphine, dependent chimpanzees chose food. Furthermore, no interest for the syringe or injection room was noted 2 weeks following abrupt withdrawal from morphine. From this, Spragg concluded that alleviation of withdrawal was the critical drive for opiate-seeking behavior.

In support of the drive-reduction hypothesis, Beach and Hebb (29) reported that rats would learn a preference for a compartment associated with morphine injections. Although induction of morphine dependence prior to training did not seem to alter the development or expression of this preference, after 3 weeks of withdrawal, only animals trained when dependent still showed a preference. Thus, only rats that had presumably experienced drive reduction during training showed a persistent morphine-seeking habit.

The drive-reduction hypothesis of opiate self-administration was further developed and expanded in scope by Abraham Wikler (31–33), who argued that opiates reduce a variety of drives even in opiate-naïve organisms. These drives include pain, hunger, anxiety, and sex. When the drug effect wears off, these drives rebound with increased vigor as a result of homeostatic counter-adaptations. Their return is experienced as tensions, which eventually increase to the point of distress. From this analysis, it was concluded that opiates have greater behavior-enhancing effects if taken when the subject is dependent, since they would relieve the punishment of impeding withdrawal reactions (8, 34). Hence, “escape training” was considered the primary drive of opioid seeking and taking, and this form of learning was considered to be critically dependent on the development of a withdrawal state induced by homeostatic counter-adaptations to the direct effects of the drug.

Working within this framework, John Nichols and colleagues demonstrated sustained opiate-directed behavior in rats that drank morphine solutions while experiencing morphine withdrawal (7). Opiate dependence was considered essential for self-injection learning even in the absence of overt signs of abstinence. James Weeks (11, 35), for example, reported that rats whose physical dependence was significantly reduced by decreasing the dosage of morphine pre-exposure from 40 to 20 mg/kg/h, self-administered less morphine on a continuous schedule of reinforcement. Further, progressive decreases in infusion dose led to an increase in number of infusions taken. But compensation was found to be incomplete; that is, total daily morphine intake ended up decreasing. This was taken as additional evidence supporting the escape-learning hypothesis because self-administration of lower doses may have produced a loss of dependence and thus reduction of morphine “need.” Within similar lines, Wikler (36) reported that rats in acute withdrawal from morphine chose a solution containing 5–10 mg/mL of etonitazene (a potent mu-opioid agonist that is orally consumed by rats without severe water deprivation or pre-training) over water, and drank enough of it to the point of showing no physiological signs of withdrawal. This was in contrast to morphine-tolerant, but not abstinent, rats that drank a solution of etonitazene at volumes not significantly different from water.

Since these early studies, additional evidence accumulated suggesting that, indeed, physiological dependence modulates opiate self-administration. For example, opioid dependence facilitates self-administration of otherwise weak reinforcing drugs such as mixed opioid agonist–antagonists (see (37) for review), and reliably increases intravenous self-administration of fast acting mu-agonists such as remifentanil (38). But, the nature of dependence-induced enhancement of opiate reinforcement has been reconsidered. More specifically, the view that drug taking in dependent animals is maintained by reduction of withdrawal has been challenged by the incentive-motivational view initially proposed by Dalbir Bindra (39), which posits that the withdrawal state enhances the incentive value of the drug and of drug-predictive stimuli. In support of this interpretation, using an experimental design whereby responding on a “seeking” lever gave rats access to a “taking” lever, Hutcheson et al. (40) found that rats responded more vigorously on the seeking lever only after having self-administered heroin in a state of withdrawal, and only when tested in a state of withdrawal.

But, the evidence reviewed above does not imply that acquisition of opiate self-administration will occur only in the presence of dependence and withdrawal. In fact, Kumar (41) demonstrated that it was possible to eliminate morphine pretreatment and still convert an initial rejection of morphine into a marked preference, as measured by the proportion of morphine solution drunk by rats when both morphine and water were made available. In addition, pretreatment and withdrawal were found to have no effect on the development and expression of the preference (42, 43), and rats made passively dependent did not begin self-administration of morphine unless previously trained to drink morphine solutions when thirsty (44). From this, Kumar concluded that a morphine-seeking habit can be induced in animals to satisfy a drive different from relief of withdrawal.

A similar conclusion was drawn from a study of Weeks (45) who demonstrated that, within a certain dose range, morphine is self-administered in the absence of dependence, as revealed by sensitivity to precipitation of withdrawal by opioid antagonists. This observation was replicated by Dai et al. (46) studying intravenous heroin self-administration, and by Amit et al. (47) and De Vry et al. (48) studying intra-cerebroventricular (ICV) morphine and heroin self-administration, respectively. Amit et al. demonstrated that rats, which were neither shaped, food-deprived, nor pre-dependent on morphine, learned to perform an operant response for ICV infusions of morphine. At the termination of self-administration, some rats were injected with naloxone (1 mg/kg, IP). During the following 30 min, rats did not display signs of withdrawal (wet-dog shakes, tremors, or hyperactivity). In the study of De Vry et al., rats were tested for sensitivity to pain (hotplate) and withdrawal precipitated by naloxone (10 mg/kg), immediately after self-administration of different heroin doses. No differences were found in hotplate latencies between dose groups, and signs of withdrawal were small in all groups. The authors concluded that the effects of heroin on self-administration could be dissociated from its analgesic and physical-dependence-inducing actions.

Hence, opiates and other mu-opioid agonists can serve as reinforcing stimuli in drug-naïve animals because, at appropriate doses and administration routes, they will promote acquisition and maintenance of responses leading to their delivery. But, is this effect attributable to the induction of an positive affective state (i.e., reward) or to a reduction of unobservable drives? It may be too simplistic to attribute acquisition of self-administration solely to opiate-induced pleasure/reward as not all effects of opiates are pleasurable in humans (49), and rodents can develop both preferences and aversions to stimuli predictive of opiate effects (50, 51). Furthermore, the “onset” of dependence may be quite difficult to identify. That is, even when physical signs of withdrawal are not manifested, animals may be in a state of affective anhedonia (52) characterized by reduced sensitivity of brain reward systems (as measured by increases in intracranial self-stimulation (ICSS) thresholds (53)). In support of this possibility, Kenny et al. (54) demonstrated that increases in heroin self-administration occur in parallel to increases in ICSS thresholds in the absence of somatic signs (tremor, teeth chattering, wet-dog shakes) of withdrawal. Coupled with the observation that physiological tolerance can be observed even after a single administration of opiates (33), it may very well be that opiate self-administration is motivated by reduction of withdrawal, immediately after the first self-administered dose.

Beside dependence and withdrawal, what other drive could influence opiate self-administration? One obvious answer is pain, not only because most animal species seek to avoid it and will learn behaviors to escape it (55), but also because mu-opioid agonists are effective pain-relieving agents (49).

There is evidence that pain can influence opiate self-administration in animals, but whether this drive facilitates or impedes it remains unclear. From a drive-reduction point of view, pain should enhance opiate self-administration because it creates one additional drive to be reduced by the drug. The results of Beck and O’Brien (56) are consistent with this prediction. In fact, drug-naïve and previously untrained rats were allowed to press a lever to receive intravenous infusions of morphine (1.0–2.4 mg/kg, IV) 24 h a day, until morphine intake stabilized (i.e., 100 mg/kg a day for 3 consecutive days). Then, a period of classical conditioning was initiated whereby animals received electrical shocks to one foreleg (conditioned stimulus) just prior to the delivery of the drug. But, conditioning was unsuccessful because most animals developed an abnormal pattern of morphine intake characterized by clusters of several infusions followed by inactivity and eventually death by overdose (56). This result suggests that the acute pain caused by the shock may have elevated the dose required to achieve satiety beyond levels that caused respiratory depression and death. Similarly, Dib et al. (57) reported that nociceptive stimulation (electrical stimulation of the tail) enhanced ICV self-administration of morphine and produced overdose in some animals. Finally, Francis Colpaert (58, 59) reported that relative oral intake of fentanyl was higher in rats injected with Freund’s adjuvant to induce rheumatoid arthritis.

Others, however, have suggested that pain decreases opiate self-administration. For example, Lyness et al. (60) reported that arthritic rats self-administered less morphine (5 mg/kg, 24 h a day, FR1–FR15) over 35 self-administration sessions than controls. The authors argued that the nature of the reinforcement sought by the two groups was different and that these effects (analgesia vs reward) showed tolerance at different rates. But, there was no attempt to explain why morphine analgesia would substitute, rather than add to, morphine reward. Furthermore, if these drive sources are indeed additive rather than substitutive, then post-infusion drug effects should be larger in animals in pain, and this should decrease overall morphine intake. That is, at low response requirements, decreases in drug intake could reflect increases, not decreases, in drug potency. A similar interpretation can be applied to the findings of Martin et al. (61) who induced allodynia in rats by ligature of spinal nerves L5 and L6. These animals were found to self-administer (FR10) less heroin (9–15 μg/kg/inf), methadone (38–75 μg/kg/inf), morphine (180 μg/kg/inf), fentanyl (0.8–5 μg/kg/inf), and hydromorphone (5–20 μg/kg/inf). Arguably, the effect of pain states on the self-administration of opiates should be reexamined using more demanding schedules of reinforcement such as the PR schedule.

The question of what drives opiate self-administration becomes particularly interesting and clinically relevant when considering the problem of relapse. That is, what can motivate an individual to seek and use a drug after the body has had enough time to adapt to its absence? Wikler (36) argued for a complex interaction between pharmacological factors and learned responses. According to him, vulnerability to relapse in previously dependent individuals results from: (1) long-term persistence of low-grade physiological deviations from normal; (2) classical conditioning of abstinence phenomena to environmental situations frequently associated with acute withdrawal distress; and (3) operant conditioning of opioid-seeking behavior through reduction of acute withdrawal distress by self-administration of such drugs.

One of the first animal models of relapse was described by Thompson and Ostlund (62). Animals were forced to drink a morphine solution (acquisition), then water (extinction), and then morphine again. The interest was in studying factors that would modulate drug intake when animals had renewed access to the drug following extinction: a period they called “reacquisition.” A very similar approach was adopted by Garcin et al. (20) who used the term “primary dependence” to refer to dependence induced in drug-naïve animals by acquisition of opiate self-administration. This was followed by abstinence, which was characterized by three phases. During the early phase (4–5 days after termination of self-administration) there was a decrease in body weight as well as food and water intake. During the intermediary phase (5 days to 4 weeks), there was a gradual recovery of weight and food/water intake. In the late (or protracted) phase, all parameters were normal. Other researchers described similar phases of morphine abstinence and withdrawal in rats (although sometimes they used a biphasic model; see (20) for review). After all phases of abstinence, animals were allowed to reacquire morphine self-administration, and the term “secondary dependence” was used to refer to “relapse.”

During tests of secondary dependence to morphine in ex-addicted animals, Garcin et al. (20) found that morphine intake was significantly higher than that observed during primary dependence, and that the level of intake was only slightly modulated by the duration of the period of forced withdrawal. These results were then replicated by other investigators who found that ex-addicted rats self-administered higher amounts of opiates either intravenously or orally (46). Increases in opioid intake observed following protracted abstinence were taken to represent a long-lasting persistence of a “need” state. More specifically, long-lasting metabolic alterations (63) were considered to induce a hyper-sensitivity to stressful stimuli and therefore higher intake of substances that would reduce stress such as opiates (64).