Drug intake escalation can be defined as a progressive increase in individual drug consumption over time that becomes excessive, overwhelming, and difficult to control. Everything happens as if drug use begets further drug use, a phenomenon dubbed “adjacent complementarity” by prominent economists interested in addictive behaviors (13). There is a large and long-standing consensus across many scientific disciplines for considering escalation of drug consumption as a core feature of addiction or even as constitutive to this behavioral disorder. There are several reasons for this consensus. First, it is simply difficult to conceptualize the transition to addiction without postulating a progression toward heavy drug use. For instance, it would be odd to speak of tobacco addiction in so-called tobacco “chippers” who continue to smoke a few cigarettes a week as during their first encounters with cigarettes (14). Indeed, without escalation to exaggerated levels of drug use, there would be no harm (except those accidentally caused by acute intoxication) and no motivated attempts to stop or to reduce drug use. Second, escalation of drug use is one of the few diagnostic criteria common to all known forms of addiction, including addictions that involve licit or illicit drugs and so-called behavioral addictions that do not involve a substance (e.g., pathological gambling). In contrast, other symptoms of addiction are specific to some drug categories and/or legal status (i.e., licit or illicit). For instance, the criterion “a great deal of time spent in activities necessary to obtain the substance, use the substance or recover from its effects” is not currently applicable to tobacco addiction, though this will probably change with the application of more strict regulation (15). Finally, among the different behavioral criteria being used to define or diagnose addiction, as opposed to recreational drug use or drug abuse, escalation of drug intake has stood the passage of time despite several successive major revisions in diagnostic classifications. This formidable resistance to change suggests that this criterion is unlikely to change in future diagnostic revisions, except perhaps to further increase its operationality (4).

To prevent confusion, it is important from the outset to establish a clear conceptual distinction between escalation of drug use and drug tolerance. For historical reasons, drug intake escalation has always been associated with drug tolerance in diagnostic systems. In fact, escalating drug use is often presented as one of the definitions of, or a surrogate for, drug tolerance. In the current edition of the Diagnostic and Statistical Manual for Mental Disorders (DSM), tolerance is defined as either a “markedly diminished effect with continued use of the same amount of the substance” or a “need for markedly increased amounts of a substance to achieve intoxication or desired effect” (3). However, as defined above, escalating drug use is an individual behavioral process that can and should be conceptually separated from pharmacodynamic tolerance (16). Indeed, depending on the theoretical framework considered, one can conceive of a wide array of different underlying mechanisms – not necessarily mutually exclusive. Certainly, under some conditions, escalation of drug use can directly arise from tolerance to some of the effects of the drug (e.g., tolerance to the drug’s rewarding or aversive effects) brought about by molecular adaptations within drug receptor signaling pathways and other cellular mechanisms. However, many alternative mechanisms operating at different, higher levels of neurobiological organization can also be envisioned. Here is a short, certainly not exhaustive, list of potential mechanisms that could account for escalating patterns of drug use. First, within an affective self-­medication or self-regulatory framework, drug use is maintained by relief of a preexisting affective condition. Escalation of drug use then arises, not because the drug progressively loses its efficacy in ameliorating the affective condition, but because this condition worsens with extended drug use, either as a direct or indirect consequence of drug use (e.g., through “reward allostasis” or “disuse atrophy”) (16, 17). Second, in a multisystems learning framework, escalation of drug use could result from a shift from goal-directed to habitual drug-taking behavior (18). Specifically, goal-directed drug use is intrinsically limited by deliberation over potential alternative courses of actions. With extended experience, however, drug use becomes habitual and thus less limited by time-consuming deliberative decision-making processes. One can predict, however, that such a mechanism will only account for drug intake escalation in stationary and highly predictable environments (e.g., drug self-administration chambers for laboratory animals). In more variable, uncertain environments that require strategic decision-making, such a mechanism is unlikely to be recruited. Third, in a behavioral economics framework, drug consumption is influenced by microeconomic factors, including drug availability, prices and accessibility to substitutes or alternative commodities (19). Thus, escalating drug use can result from increased drug availability, decreased drug price and/or reduced accessibility to nondrug substitutes. Finally, from a psychosocial perspective, drug use is limited in a specific society or culture by cultural norms, taboos, and/or overarching value systems (20). Here escalation of drug consumption can result from a progressive loss of influence of prevailing social or cultural norms on drug taking due, for instance, to social exclusion or marginalization processes that are directly or indirectly related to drug use. Within the same framework, escalating drug use can also arise from a progressive depletion in the individual’s social capital brought about by extended drug use. Thus, to sum up, the historical link established between drug intake escalation and tolerance has no apparent conceptual necessity. It is an empirical question to determine whether escalation of drug use is more often than not driven by drug tolerance and/or other mechanisms. In the remaining text, I will focus on drug intake escalation as a behavioral process without theoretical commitment to a specific underlying mechanism.

As suggested above, several different factors can precipitate – at least in principle – escalation to higher levels of drug consumption. Despite such potentially diverse etiology, however, research on laboratory animals (i.e., rats) has so far been limited, mainly, to investigating the role of increased drug accessibility in promoting the progression toward excessive drug use (see Sect. 9, for additional information regarding this limitation). Specifically, escalation of drug consumption is induced reliably and rapidly by increasing the length of daily sessions of drug self-administration, as compared to much shorter daily sessions. The aim of this section is to summarize this research, but before proceeding, some comments on several relevant methodological issues are in order.

A first, obvious issue concerns the method for distinguishing the escalation process per se from the increase in behavior related to the acquisition of the instrumental response (i.e., positive reinforcement). At least two different approaches can be used to achieve this goal. In a within-subjects design, subjects can be pre-trained before having long access to drug self-administration. For instance, in some published studies, animals were first trained to acquire the drug self-administration response before increased drug availability. In this approach, escalation of drug self-administration manifests as a gradual increase in drug consumption over time above the pre-escalation level of consumption. In a between-subjects design, at least two groups of untrained animals are tested: one group has a short, daily access (ShA) to drug self-administration while the other group has a much longer daily access (LgA). In this approach, drug intake escalation appears in the LgA group as a gradual increase in the rate of drug consumption above the control level of the ShA group. This between-subjects design is less time consuming than the first approach because it does not require pretraining. Of course, these two approaches can be combined to better assess the development and maintenance of drug intake escalation following increased access time to the drug.

Another important methodological issue concerns the minimization of potential experimental restrictions on drug consumption. Obviously, to assess changes in drug consumption as a function of drug availability, individuals should be relatively free to regulate their drug intake (21). Thus, the response ratio requirement to obtain a fixed drug dose should be reduced to the minimum as in the continuous reinforcement (CRF) schedule. In addition, the postinjection time-out period should be selected to allow for the highest possible number of injections while protecting animals from the risk of overdose (which increases with the unit dose available). Finally, limiting the maximum number of injections per session should be avoided as this may interfere with the escalation process (22–24).

A final issue concerns the choice of the dependent variable for assessing the effects of drug access time on drug self-­administration in the between-subjects design (i.e., ShA versus LgA). Specifically, the problem is to select a dependent variable with the highest degree of between-subjects comparability (obviously, this choice is not a problem in the within-subjects design because the assessment of drug intake escalation only involves within-subject comparisons). The most comparable dependent variable is the number of drug injections during the time period common to all access durations (e.g., during the first hour when comparing 1-h versus 6-h sessions of drug self-administration). Inversely, the least comparable variable is the total number of injections simply because it is directly influenced by the independent variable. Another potential dependent variable could be the average hourly number of injections obtained by dividing the total number of injections by the length of the self-administration session. However, this variable may be a source of inaccurate comparisons when there are pronounced within-session variations in drug self-administration – which is often the case for intravenous cocaine or heroin self-administration (e.g., initial drug loading). Thus, whenever possible, it will be preferable to compare drug use during the time period common to all access durations.

As described above, drug intake escalation is observed across a broad spectrum of drugs of abuse, including stimulants, opiates, and ethanol. Whether there is cross-escalation between different drugs of abuse or within the same drug class (e.g., the opiates) has not yet been tested systematically. Drug cross-escalation is defined here as an increase in the consumption of a newly encountered drug after intake escalation of a different drug. The systematic study of the phenomenon of drug cross-escalation may provide critical information, not only about neuropsychopharmacological commonalities in the escalation process, but also for better understanding polydrug abuse and addiction. To begin to address this issue, it was determined whether there is a cross-escalation between cocaine and heroin consumption – two highly addictive drugs with both common and different neuropsychopharmacological effects. Two separate groups of rats were first allowed to self-administer either cocaine or heroin for 1 h/day and then, after stabilization of drug intake, escalation of cocaine or heroin self-administration was precipitated by increasing drug access time to 6 h/day (Lenoir and Ahmed, submitted). For each group, a dose-injection function for drug self-administration was generated both before and after increased drug access time. As shown in Fig. 4, there was virtually no cross-escalation between cocaine and heroin. Specifically, as shown previously, after prolonged access to one drug (cocaine or heroin), rats escalated their consumption of this drug, which resulted in a characteristic ­vertical shift of the corresponding dose-injection function. However, the same rats did not increase their intake of the alternative drug, regardless of the available dose. The lack of cross-escalation to heroin self-administration in cocaine-escalated rats was confirmed in another study. In this study, rats (N = 8) were initially trained to self-administer both cocaine and heroin during 1 h on alternate days until stabilization of cocaine and heroin intake. Then, access to cocaine was increased to 8 h/day during several days. As expected, rats escalated their cocaine intake during extended access to cocaine but again, regardless of the dose available, there was no cross-escalation between cocaine and heroin. In fact, there was even a nonsignificant trend for decreased heroin consumption compared to prior cocaine intake escalation (Guillem and Ahmed, unpublished data). The lack of cross-­escalation to cocaine self-administration in heroin-escalated rats is consistent with previous research showing no change in cocaine intake in rats with frequent heroin use under a discrete-trials procedure of drug self-administration (81, 82). Future research is clearly required to better assess the generality of these findings across a wider range of drugs of abuse, both between and within drug classes.

As illustrated above, escalation of drug consumption is clearly a transient and negatively accelerating process, which eventually levels off at a new, higher level of consumption. This process can thus be characterized by different dynamic and state parameters, such as, its shape, speed, and final, post-escalation steady-state. So far, comparatively little research has been conducted to study how these different features systematically vary as a function of drug access condition. In one systematic study, Wee and colleagues (57) showed that the final, post-escalation level of cocaine intake increases with the length of daily sessions of cocaine self-administration: the longer the self-administration session, the greater the steady-state level. Whether the final, post-escalation level of drug consumption can also be influenced by other factors is less clear. In two studies from the same laboratory, the level of escalated cocaine or methamphetamine intake decreased with the training dose (57, 62) while in another study, the level of escalated cocaine self-administration remained unchanged (61). Clearly, additional parametric research is needed to better quantify the dynamics of the escalation process. This research will eventually prove useful for future pharmacological and neurocomputational modeling approaches to drug addiction (83).

Knowledge about the maintenance of escalated levels of drug consumption is even more limited. In one study, the maintenance of escalated levels of cocaine self-administration was shown to depend on the session duration (21). In this experiment, after inducing cocaine intake escalation, the daily session duration was reduced from 6 to 1 h. As a result, escalated levels of cocaine intake significantly decreased over time and progressively, though very slowly, returned toward pre-escalation levels of consumption. Indeed, even after 2 months of reduced drug access, the de-escalation of cocaine self-administration was still incomplete. In contrast, in another, related experiment, escalated levels of cocaine self-administration completely de-escalated to pre-escalation levels after only 1 month of forced abstinence (7). This differential outcome shows that compared to total abstinence, intermittent, brief exposure to cocaine self-administration can considerably retard, without preventing, drug recovery. This finding may have important clinical implications.

There is now considerable evidence showing that increased access time to a variety of drugs of abuse can precipitate a rapid escalation to higher levels of drug consumption – a hallmark of addiction. More recent research now indicates that animals with escalating drug use also present other major behavioral signs of addiction, compared to animals with a more stable pattern and moderate level of drug use. These behavioral signs of addiction include: (1) an increased motivation to self-administer the drug, (2) greater difficulty in abstaining from drug use, and (3) a relative indifference to negative consequences. Finally, converging evidence from different laboratories shows that animals with increased drug use also become more sensitive to drug-induced and stress-induced reinstatement of drug seeking after extinction. Reinstatement of extinguished drug seeking is a well-validated animal model of craving and/or relapse (84). Though drug-induced craving is not a current diagnostic criterion of addiction, it nevertheless represents a selective marker of addiction since it is not seen in nondependent individuals (85, 86). Thus, as a whole, the differences between stable/moderate and excessive/escalating patterns of drug self-administration induced by differential drug access time in rats recapitulate the major behavioral differences between controlled and compulsive drug use in humans.