A contextualized reinforcer pathology approach to addiction - Nature.com

Abstract

Behavioural economic accounts of addiction conceptualize harmful drug use as an operant reinforcer pathology, emphasizing that a drug is consumed because of overvaluation of smaller immediate rewards relative to larger delayed rewards (delay discounting) and high drug reinforcing value (drug demand). These motivational processes are within-individual determinants of behaviour. A third element of learning theory posits that harmful drug use depends on the relative constraints on access to other available activities and commodities in the choice context (alternative reinforcers), reflecting the substantial influence of environmental factors. In this Perspective, we integrate alternative reinforcers into the contemporary behavioural economic account of harmful drug use — the contextualized reinforcer pathology model — and review empirical literature across the translational spectrum in support of this model. Furthermore, we consider how increases in drug-related mortality and health disparities in addiction can be understood and potentially ameliorated via a contextualized reinforcer pathology model in which lack of alternative reinforcement is a major risk factor for addiction.

Introduction

In 2020, an estimated 138.5 million Americans aged 12 or older reported alcohol use in the past month, and 61.6 million reported a binge drinking episode (four or five drinks for women or men, respectively) in the past month¹. Rates of alcohol consumption have remained relatively stable over the past decade, with some upward trends for women and older adults possibly amplified by the COVID-19 pandemic^2,3. Illicit drug use is also rising: nearly one in five Americans reported illicit drug use of some kind in the past year, and 43.5 million Americans reported cannabis use in the past year¹. These trends in the USA are generally mirrored around the world^1,4,5. Although many use alcohol or illicit drugs without suffering notable clinical concerns, a minority use at high levels, resulting in diagnoses of substance-use disorder and a range of acute to chronic substance-related problems. Serious harms, such as alcohol-related mortality, overdose and cirrhosis, are on the rise^6,7, with notable increases during the COVID-19 pandemic⁸, even as other causes of death are decreasing⁹. The public health impact of these harms are monumental and result in billions of dollars in financial costs each year¹⁰.

Myriad policies have been developed to reduce or eliminate the burden of substance use. Yet alcohol- and drug-related morbidity and mortality have increased over the past two decades^7,11,12, suggesting that existing strategies are far from adequate. Robust and valid theories are needed to guide treatment and policy development. Specifically, it is incumbent upon psychological theories of addiction to explain the following: why people consume drugs; why some people use alcohol and drugs in a manner that contributes to health and social problems, while others are able to use moderately with minimal consequences; why many people are able to reduce patterns of alcohol and drug use, often without participating in formal treatment, while others experience chronic, escalating patterns of use; and why substance-related morbidity and mortality have increased over the past two decades and in particular over the past few years^7,11,12.

In this Perspective, we provide an overview of a contemporary behavioural economic theoretical account of addiction, the reinforcer pathology model, which suggests that drug consumption is the result of overvaluation of smaller immediate rewards and drug-specific reinforcement. Next, we highlight limitations of the reinforcer pathology model and describe an extension, the contextualized reinforcer pathology model, that highlights the critical role of alternative reinforcers in addiction motivation. We then review empirical literature across the translational spectrum that supports this model. Finally, we review relevant literature on increases in addiction-related morbidity and mortality and addiction-related health disparities that might be understood and potentially ameliorated via contextualized reinforcer pathology.

Reinforcer pathology

A behavioural economic account of substance use refers to a set of empirical methods and models of decision-making that integrate microeconomic and operant learning theory principles to understand the decision-making processes and contextual features that influence substance consumption over time.

The most popular contemporary behavioural economic model — reinforcer pathology^13,14 — suggests that addiction is marked by within-individual differences in relative reinforcing value (high drug demand) and a more general decision-making bias that overvalues smaller immediate rewards relative to larger delayed rewards (high delay discounting) as central aetiological risk factors. These two key concepts of the reinforcer pathology model are described below.

Drug demand

Across behavioural economic models, the reinforcing value (or 'demand' in behavioural economic terms) of substance use is measured by the level of consumption or the amount of behavioural (or monetary) output emitted to engage in the activity¹⁵. Real or hypothetical purchase tasks are usually used to measure drug or alcohol value. In a typical alcohol purchase task, individuals report how many drinks they would purchase during a hypothetical drinking scenario at each price in a series of escalating prices¹⁶. Responses across each of a specific (monetary) cost are plotted to create a demand curve and produce indices that reflect an index of drug value (Fig. 1a). Human and laboratory animal research has consistently demonstrated that, within a closed economy (a choice economy with defined constraints on access to drugs and no access to any commodities outside the economy), response to a drug reinforcer decreases as the cost of acquiring the substance increases¹⁷ (Fig. 1a). Purchase tasks mirror (but are more cost- and time-effective than^18,19) laboratory progressive ratio tasks in which animals or human participants have the opportunity to self-administer a substance as the cost for doing so (for example, the number of button presses required or the monetary expenditure) progressively increases²⁰.

Fig. 1: Behavioural economic demand and delayed reward discounting.

a, Representation of a behavioural economic demand curve, which can be plotted using data from purchase tasks. As cost increases, consumption decreases. Demand indices that can be extracted from the data include intensity (consumption at zero cost), breakpoint (the price at which consumption is fully suppressed) and elasticity (the rate of change in consumption as a function of cost). b, Representation of the change in value of two rewards as a function of the delay to reward receipt. The reward from substance use is smaller, but the receipt is nearer in time, whereas the reward from an alternative is larger, but receipt is further delayed in time. When receipt of both rewards are distant in time, the value of the larger reward is greater. However, owing to the hyperbolic nature of delayed reward discounting, the value of the immediate reward increases at a greater rate as reward receipt becomes closer in time. Thus, an individual might experience a preference reversal, in which the value of the immediate reward surpasses the value of the delayed reward when receipt of the immediate reward is imminent.

Full size image

Individual differences in the degree to which costs lead to a decrease in responding index between-person valuation of a drug or drug demand (Fig. 1a). Reinforcer pathology suggests that these individual differences reflect strength of motivation for the drug and should be correlated with levels of alcohol use and alcohol-related problems^13,14. Indeed, indices of drug demand, in particular maximum consumption, expenditure level and sensitivity to changes in drug price, show robust correlations with consumption^{16,21,22,23,24}, substance-use problems^25,26, and substance-use disorder²⁷. Demand indices are also robust prospective predictors of drinking behaviour even after controlling for past alcohol consumption^28,29. In other words, measures that aggregate a series of hypothetical drinking decisions across escalating costs have predictive utility over and above measures of recent drinking practices.

Delay discounting

Costs and benefits across choice options are unevenly distributed across time, such that some reinforcers, like drug use, have relatively greater immediate benefits (for example, intoxication, euphoria, social facilitation, anxiety reduction and withdrawal relief) and health and social costs that are substantially delayed (and probabilistic). By contrast, other reinforcers, such as earning a college degree, have relatively immediate costs (attending class, studying and paying tuition) and delayed (and probabilistic) rewards (satisfaction of earning good grades and graduating; higher-quality employment and salary). This critical temporal element to reward valuation for drugs versus alternatives is captured by delayed reward discounting, which is the relative preference for smaller, sooner rewards compared to larger, later rewards³⁰. Delayed reward discounting describes how much the value of an activity or commodity decreases as a function of its temporal 'distance' from the current moment. Empirical research suggests that the subjective current value of delayed rewards decreases more steeply with initial delays, consistent with a hyperbolic decay function, rather than at a constant rate³¹ (Fig. 1b). One implication of hyperbolic discounting is that the preference for smaller immediate versus larger delayed rewards shifts dynamically as a function of time to reward availability, exhibiting steep devaluation at initial delays and shallower devaluation at further delays. Thus, humans and laboratory animals generally prefer larger, later rewards when reward receipt for both options is distal, but preference often reverses as the availability of the smaller, sooner reward becomes imminent³¹ (Fig. 1b).

Individuals vary in their time horizons for behavioural allocation, influencing the rate at which they devalue delayed rewards. Thus, utility maximization is relative to the temporal frame of reference. A local (shorter) time frame of reference typically compares discrete, independent choices (for example, should I drink alcohol tonight or study for my exam?) to maximize short-term utility (enjoyment from drinking and socializing). By contrast, a temporally extended global or molar frame of reference compares two choices on the basis of their anticipated value over the course of an extended pattern of behaviour that comprises many discrete choices that might accrue value exponentially over time^32,33. For example, consider a series of discrete choices between watching TV and drinking alcohol versus exercising each night over the course of a month. An evening spent watching TV and drinking alcohol might have high immediate value that does not necessarily aggregate over time (whereas costs might aggregate). By contrast, exercise might lead to benefits that are not immediately evident after one discrete event but instead emerge after consistently engaging in a pattern of behaviour. This intertemporal choice dynamic is foundational to behavioural economics, including applications to substance-related harms³³.

Importantly, discounting applies to all delayed rewards, and rates of discounting vary considerably across commodities^34,35. Furthermore, steep discounting (a greater preference for smaller immediate rewards over larger delayed rewards) can be adaptive when it comes to securing reinforcers in dangerous or deprived environments in which delayed rewards are also highly uncertain (as in the idiom, 'a bird in the hand is worth two in the bush'). Nevertheless, steep delay discounting might be an especially relevant decision-making bias that contributes to frequent drug use because positive drug effects (euphoria, enhanced focus, or reduced pain and anxiety) tend to occur immediately, whereas their costs or negative effects are generally delayed (ranging from hours for acute illness or hangover to years for health and social impacts). Additionally, in modern society, legal and illegal drug reward is often easy to obtain and imposes little upfront cost.

A central tenet of the contemporary reinforcer pathology model is that the temporal window of value allocation substantially determines the relative value of the reinforcers operating in that window¹³. In other words, higher substance value, and therefore persistent substance consumption even at high costs, can be attributed to a preference for immediate rewards because the value of more distal rewards diminishes very quickly as they fall outside a person's time horizon. In turn, the reinforcer pathology model has led to intervention approaches focused on expanding the temporal horizon of decision making (that is, reducing delay discounting)^36,37,38,39.

Alternative reinforcement

The contemporary reinforcer pathology model is subject to several underemphasized considerations that are critical for understanding choice behaviour. The reinforcer pathology model emphasizes that steep discounting contributes to preference for drug rewards relative to alternative rewards. However, it does not emphasize the environment, including the availability of drugs and the relative reinforcing efficacy of substance-free alternative reinforcers, as contributing factors to elevated demand or discounting, or as direct contributors to risk for harmful alcohol and drug use⁴⁰. The contemporary reinforcer pathology approach emphasizes, and typically measures, drug reinforcing value and delay discounting with the assumption of all other things being equal, but the influence of alterative reinforcers frequently violates this assumption. Thus, studies in the addiction literature over the past two decades have focused on reductionistic accounts of absolute responding, closed economies, and the differences in between-person choice preferences for delayed and substance rewards. Consequently, these critical behavioural economic variables are now often considered static individual difference variables, which deviates from decades of research showing that the economy and choice context influence the relative value of a drug^15,41,42,43.

Addiction might be better understood by simultaneously considering temporal discounting and drug-specific reinforcing value, alongside immediate and delayed costs and benefits of both the substance and alternatives over extended patterns of behaviour. Indeed, real-world decision making occurs in an open economy in which an individual can typically choose between two or more options in a choice context. When considering this broader choice context^41,42, distal causal influences exerted by the characteristics of the choice economy emerge that cannot be described by models of proximal causation⁴⁴ and emerge only through a molar analysis of behaviour^42,45. Although the contemporary reinforcer pathology model acknowledges the importance of relative value and often compares the immediate value of drugs to the delayed value of some alternative, the central tenets explicitly ignore the distal causal influence of the choice environment^30,41,42,45.

Indeed, the trenchancy of behavioural economics — and a distinguishing factor from other theories of addiction — is its explicit scaffolding to reconcile person-level and environmental factors (Fig. 2a). Other prominent theories of addiction emphasize factors within the person, be it through neurobiological^46,47 or psychological mechanisms. By contrast, sociological⁴⁸ and anthropological⁴⁹ models emphasize environmental conditions over person-level factors. Although proponents of other person-level theories have begun to integrate environmental factors⁴⁶, behavioural economics provides a robust conceptualization that quantitatively and intuitively accounts for both within-individual and environmental factors, making this theory ideally positioned to enhance addiction research, intervention and prevention. In a discrete choice context, a person's intertemporal orientation, the constraints on the drug itself, and constraints on alternatives are all mutable environmental factors implicated as determinants of the likelihood of drug consumption. Over time, each of these form distinct, predictable, aggregate patterns of behaviour that can be measured and used as individual difference variables (Fig. 2b).

Fig. 2: Situating behavioural economic theories of addiction.

a, Although most theories of addiction recognize diverse influences, disciplinary foci tend to be oriented toward person-level factors or environmental factors. b, Behavioural economics bridges the connection between environmental and within-individual determinants by framing behaviour within a discrete choice context that is heavily influenced by environmental factors; these discrete choice contexts are building blocks for patterns of behaviour over time, which aggregate into measurable individual difference variables.

Full size image

In this section, we present an extension to the reinforcer pathology model, the contextualized reinforcer pathology model, which addresses the limitations described above and highlights the critical role of alternative reinforcers in addiction motivation. We then describe the matching law, which serves as a primary theoretical premise of alternative reinforcement.

Contextualized reinforcer pathology

Contextualized reinforcer pathology posits that drug value, and consequently the likelihood of drug consumption, is critically determined not only by temporal windows of value allocation, but also by the characteristics of environmental choice contexts (Fig. 3). The contextualized reinforcer pathology model is a molar theory of behaviour: behaviour is measured over extended temporal windows and diverse sets of constraints to characterize the most likely behavioural output over time⁴⁵. Constraints can be anything that influence the value of the commodities in the choice context⁵⁰. From this perspective, behaviour can be broadly explained through utility (value) maximization, in which choice outcomes maximize benefits and minimize costs over a specified and varying temporal window (that is, there is no assumption that human or non-human laboratory animals maximize utility in an ultimate sense).

Fig. 3: A contextualized reinforcer pathology approach.

Two theoretical examples depicting the effects of environmental constraints on the value of alternatives and alcohol at three time points. Blue represents reinforcement from alternative activities and red represents reinforcement from alcohol-related activities. The left panel shows a scenario most likely to result in increasing levels of substance use. Initially, substance value is low, and the individual engages in many alternative activities. However, over time the environmental context places increasingly high constraints on alternatives (the local park shuts down, the individual cannot afford to go to college, the roads are bad for biking) and low constraints on alcohol (easily available from local store, cheap, social reinforcement from drinking). Consequently, alcohol value increases over time while engagement and availability of alternatives decrease. The right panel shows a scenario that would result in stable or decreasing levels of alcohol use over time. Initially, substance value is low, and the individual engages in many alternative activities. Over time, constraints on alternatives remain low. As the individual enters emerging adulthood, they connect with friends through drinking, and therefore the value of alcohol rises slightly. However, the individual maximizes more global utility and continues to engage in available alternatives that effectively compete with the immediate rewarding effects of alcohol. Consequently, when the individual leaves college and drinking among friends declines, the individual's drinking declines as well.

Full size image

A key assumption of contextualized reinforcer pathology is that a drug's reinforcing value is not an innate quality of a drug but is instead critically determined by characteristics of the choice environment. Although delayed reward discounting and behavioural economic demand have been operationalized as stable, individual difference variables, this is a feature of measurement; the stability in these constructs is due in part to the stability of the environmental choice context and (lack of) availability of alternatives in the instructional sets. Value is influenced by factors across varying temporal and environmental (spatial) frames in a way that requires explanations of distal causation (the level of public health in the environmental context influences individual drug value). In other words, a narrow spatial analysis might ignore the environment altogether and focus on within-individual or between-individual variables that predict alcohol use. Expanding the spatial analysis might reveal county-level differences in the availability of liquor stores and alternatives, such as parks and recreational opportunities, that might explain additional variance across populations in various counties. Expanding the spatial analysis further might reveal country-level differences in the acceptability of consumption or state-level differences (for example, in legal status of cannabis), or cultural differences across nations in the acceptability of public alcohol consumption.

In the contemporary reinforcer pathology model, the pathology of overuse of a specific reinforcer resides in the internal decision-making processes of the individual, and the influence of the broader context is unaccounted for, whereas pathology in the contextualized reinforcer pathology model resides in the interaction between the person and the context. Studies have demonstrated that substance demand is malleable to numerous experimental manipulations, such as cue exposure (controlled exposure to substance-related environmental stimuli)^51,52, opportunity cost (choosing the substance reinforcer at the expense of an alternative that also carries value)^53,54, the social context^55,56, and both pharmacological and psychosocial treatments^57,58,59. Delay discounting is also influenced by context^36,60, including through exposure to natural (as compared to man-made) environments⁶¹, shifts in the time to receipt of alternatives³¹, and manipulations targeting the temporal frame, such as episodic future thinking^62,63,64. Indeed, although the effects of alternative reinforcers, demand, and delayed reward discounting are often studied in isolation, these factors may interact to influence behaviour during a discrete choice (Fig. 4).

Fig. 4: Interactions between substance cost and alternative reward.

The likelihood of using a substance is based on the cost of the substance, the delay to the receipt of the alternative reward, and the value of the alternative reward. Across both plots, substance use is most likely when the substance cost is low and when the delay to the alternative is high. As the cost of the substance increases, the likelihood of use decreases. Further, as the alternative reward receipt becomes closer in time, the likelihood of use decreases. However, when there is a high-value alternative available (an activity that generates a positive affective state or sense of accomplishment or alleviates an aversive state; bottom panel) the likelihood of use across all delays and substance costs are attenuated relative to when a low-value alternative (a non-stimulating or aversive activity; top panel) is available.

Full size image

Moreover, in the contextualized reinforcer pathology model, pathology can be determined only within an individual's functional context. Reinforcement learning is an adaptive process that occurs because it results in reward or alleviation of distress; the reinforced behaviour serves a function and drug behaviour is only 'pathological' when the behaviour leads to functional impairment in the short term (for example, accidents, hangovers or missing work) or long term (for example, declining health or social functioning).

The matching law

The importance of alternative reinforcement in decision making broadly, and drug use specifically, is grounded in the behavioural matching law⁶⁵, a behavioural principle which states that the relative rate of responding approximates the relative rate of reinforcement at each alternative⁶⁶ (Box 1). In an exemplar experiment⁶⁶, pigeons were concurrently reinforced to peck two keys in an experimental chamber under independent variable-interval schedules of reinforcement. In other words, each reinforcer was delivered following a specified amount of time after the first key peck response, and the time between reinforcers varied throughout the task. Across five experimental sessions, rates of responding corresponded almost perfectly with frequency of reinforcement. These findings illustrate that the observed response rate for each choice option is approximately equivalent to reinforcement from that option and that behaviour is reflected proportionally based on the frequency of responding relative to other reinforcers^65,67. That is, consummatory behaviour matches the local reinforcement contingencies. Behavioural allocation consistent with the matching law is generally adaptive and does not imply a pathological pattern of responding but is instead a quantitative codification of the assumptions made about choice under different conditions⁶⁸.

The matching law has two critical implications for theories of substance use. First, the response rate for various reinforcers might serve as a measure of reinforcement value. In line with this theoretical premise, applied human researchers have developed indices of reinforcement for humans based on time allocation and discretionary spending for alcohol and non-alcohol activities as well as activity enjoyment, with activities that are engaged in frequently and rated as subjectively enjoyable classified as highly reinforcing⁶⁹.

Second, operant behaviour is a zero-sum outcome set dependent upon the available reinforcers and an individual's response rate to each reinforcer. The introduction of any alternative eliciting a non-zero response rate will shift response rates, and therefore the reinforcement, for all other choices in the choice context⁶⁵. This is consistent with other research showing that preference for drug reinforcers varies as a function of the availability of non-drug reinforcers^{70,71,72,73,74,75}. To illustrate the zero-sum nature of reinforcement and the importance of alternatives, consider patterns of reinforcement for two individuals. Both individuals drink the same amount of alcohol per week, but person 1 allocates less time to other activities, such as work, family, and hobbies compared to person 2. Consequently, reinforcement from substance use is a larger percentage of the overall 'reinforcement pie' for person 1, and they are therefore more likely to have problems related to drinking and are more vulnerable to chronic alcohol-related harms.

These lines of research provide support for the premise that preference for a substance depends on the relative constraints on access to other available reinforcers in the choice context. Thus, drug consumption is in part an inverse function of access to alternative rewards. This conclusion contrasts with early work in laboratory animals and humans that consistently found that alcohol and other drugs often continue to be self-administered at high rates under schedules of reinforcement in which a reward is only provided after a specified number of responses (fixed ratio or progressive ratio schedules) and when no alternative reinforcers are available^76,77. These results are consistent with the law of absolute responding, which suggests that as the amount of reinforcement for a given commodity increases, the amount of behaviour allocated toward that commodity increases hyperbolically⁶⁶. However, these experiments lack validity: people are rarely presented with only one choice in daily life. Indeed, substance consumption decreases when alternatives are concurrently available alongside substance self-administration^{78,79,80,81,82,83}.

The matching law provides a framework for understanding how addiction might develop in certain contexts. For example, the 'primrose path' model, which developed from early experiments with the matching law^84,85,86, suggests that addictive drugs have greater local utility than most competing alternatives, and therefore the addictive drug will almost always be selected when local utility is maximized (that is, when the individual uses a proximal frame of reference). However, characteristics of addictive drugs (such as tolerance and adverse physical and social effects associated with patterns of heavy use) reduce the value of both the drug itself and alternatives. Thus, as choices accumulate, the value of both options reduce over time, but the value of the addictive drug remains higher when maximizing local utility. When an organism maximizes global utility each successive choice is considered in the calculation of the potential value of the next choice option (choices and associated rewards are bundled together into an aggregated outcome). Most salutary alternative choices to drug consumption are distributed choices with immediate effort costs (for example, work or exercise) and delayed rewards (for example, affluence or health). Thus, maximization of global utility would result in a pattern of choices favouring the alternative reward. According to proponents of the primrose path model an analysis of the available commodities is needed to understand their effect on future choice^84,85,86, and addiction might be driven in part by stable, between-individual differences in the choice strategies dictated by either local or global frames of reference⁸⁵. However, what is lacking in these analyses, as in the reinforcer pathology model, is that drug use will be influenced by the substantial between- and within-individual differences in constraints on access to drugs versus alternative reinforcers across environmental contexts (Fig. 3).

Contemporary accounts of reinforcer pathology have underemphasized the matching law and overemphasized individual differences in reinforcing value and delay discounting^40,87. However, there is historical and theoretical precedent to assume a conceptual and nested interconnectedness between these three primary behavioural economic variables (Box 1), which each explain behaviour under increasingly specific conditions^30,41,42,88.

Box 1 Historical review

There is historical and theoretical precedent to assume a conceptual, hierarchical and nested interconnectedness between behavioural economic demand, the matching law and delay discounting, which each explain behaviour under increasingly specific conditions. Here we illustrate the relationships between these three variables.

The matching law and demand

The majority of decision-making occurs in open contexts with two or more reinforcers. The matching law was discovered through observations of allocated choice across available commodities and suggests that reinforcement from each choice option 'matches' behavioural allocation:

$$\frac{{B}_{1}}{{B}_{1}+{B}_{2}}=\frac{{R}_{1}}{{R}_{1}+{R}_{2}}$$

where B represents the rate of response (behaviour) at each available option (denoted by subscripts 1 and 2) and R represents the rate of reinforcement at each available option (denoted by subscripts 1 and 2). In an open economy choice context with two concurrently available reinforcers, reinforcement from commodity X might be independent or dependent upon the schedule of reinforcement of commodity Y (see figure). Two commodities are independent when changes in the price of commodity A has no influence on the consumption of commodity Y. Two commodities are complementary when consumption of commodity Y decreases as the price for commodity X increases, and vice versa. Two commodities are substitutes when consumption of commodity Y increases as the price for commodity X increases, and vice versa. These relationships are useful for understanding decisions between using substances and other, non-substance alternatives. For example, there might be activities that have complementary associations with alcohol use (spending time with friends or attending football games), activities that serve as substitutes for alcohol use (preparing for an exam, exercising or attending religious services), and activities that might have an independent association with alcohol use (dining or watching TV).

The generalized matching law was created in response to criticisms that the strict form of matching law represented only a special case of behavioural economic demand (when two commodities operate as perfect substitutes) and represents only a subset of all possible choice interactions posited by economic theories of consumer demand^41,42,196. The generalized matching law is:

$$\log \left(\frac{{B}_{1}}{{B}_{2}}\right)=s\,\log \left(\frac{{R}_{1}}{{R}_{2}}\right)+\log \,b$$

where B and R are as defined above, s represents the slope of the best-fitting line, and b represents the y-intercept. These two adjusting mathematical functions operate as free parameters and can account for variations in the substitutability between two reinforcers, consistent with economic utility theory^42,88,197. Unfortunately, a great deal of research using behavioural economic demand in the field of substance use has focused on the effect of price on behaviour under single-commodity conditions. These applications are more consistent with the matching law of absolute single responding⁶⁷ and do not account for the complexity of decision-making under conditions of varying costs across commodities.

The matching law and delayed reward discounting

Early research using the matching law explored only immediate reinforcement and therefore did not incorporate delay to reward receipt as a factor⁶⁵. Later iterations de-emphasized the strict behavioural operationalization of reinforcement and instead emphasized value, a derived function that is a product of obtained reinforcement and other factors that might influence preference (such as price or delay). The hyperbolic delayed reward discounting equation, which is considered to match actual behaviour better than the exponential equation, was developed as an extension of the matching law³⁰. The equation acknowledges that the temporal receipt of reinforcement plays a part in determining obtained reinforcement for a single reinforcer. A great deal of research suggests that responding for a single reinforcer fits the matching law of absolute single responding⁶⁷:

$${B}_{1}=\frac{k{R}_{1}}{{R}_{1}+{R}_{e}}$$

where B₁ represents the response rate, R₁ represents the obtained reinforcement, R_e represents 'distractions' from R₁ (or error) and k represents the total range of behaviour. The hyperbolic discounting equation explicates R₁ as a mathematical derivative³⁰:

$${R}_{1}=\frac{{A}_{1}}{1+K{D}_{1}}$$

where R₁ represents the obtained reinforcement, A₁ equals the amount of the delayed reinforcer, K is an impulsivity constant representing individual variation in the degree to which delay will affect obtained reinforcement and D₁ represents delay to receipt. Mathematically, delayed reward discounting is a special case of the matching law, which explicates a mechanism through which reinforcement, and therefore response rate, might systematically differ.

The nested nature of these three behavioural economic variables mathematically does not necessarily reflect their importance in understanding behaviour. Instead, the most frequently observed symptoms of substance-use disorder can at least partially explain the emphasis that has been placed on each of these variables in the scientific literature, in addition to the relative dearth of research on other aspects of the model (such as behavioural complements).

Translational evidence

Next, we review translational evidence that supports and extends the fundamental matching law⁶⁵ and shows that, across multiple levels of analysis, enhancing access to alternative rewards meaningfully affects engagement with substances over and above other necessary theoretical mechanisms of addiction. We begin with a discussion of basic non-human animal and human research and then discuss applied clinical translations in humans that demonstrate how increasing alternatives can be used as a treatment mechanism and intervention.

Experimental non-human animal laboratory research

One influential set of studies (known as 'rat park') provides a potent demonstration of the effect of the environment on drug administration behaviour. Specifically, experimenters tested the influence of social exposure as an alternative reinforcer competing with morphine^89,90. Rats randomized to either isolation or an enriched social environment (with running wheels and other activities) were given access to morphine for 57 days. In the experimental sessions, they were allowed to make concurrent choices between morphine and water. Rats in the enriched social environment consumed less of the morphine solution compared to isolated rats^89,90. This general finding that alternative reinforcers reduce drug self-administration in concurrent choice tasks among non-human laboratory animals has been replicated across substances^{78,91,92,93,94,95,96} and alternative reinforcers (such as food, sucrose and running wheels)^78,91,93,97.

The effects of alternatives do not seem to be limited to experimental paradigms in which rats choose between two rewards simultaneously. In another study⁹⁸, rats were trained to self-administer alcohol after only one lever press until stable responding was achieved, after which rats lever-pressed for alcohol or sucrose in alternating sessions. Rats reduced their responding to alcohol (that is, pressed the lever fewer times) after being introduced to sucrose, even in the sessions when sucrose was not available, suggesting that the effects of non-alcohol alternative reinforcers extend beyond the immediate choice context.

Other laboratory animal research suggests that the order in which the reinforcers become available might influence the impact of alternative reinforcers on drug self-administration. In one study⁹⁷, rats had access to d-methamphetamine self-administration for 21 experimental sessions. Access to a running wheel (an alternative reinforcer) was also available during sessions 1–14 for a first group, during sessions 8–21 for a second group, and during sessions 15–21 for a third group. Rats in the first group self-administered less methamphetamine across the first fourteen sessions compared to the rats in the second and third groups. Self-administration in the second and third groups decreased when the running wheel was introduced in sessions 8 and 15, respectively. When rats in the first group lost access to the running wheel, self-administration increased, but to similar levels as self-administration in the second and third groups when the running wheel was available. These findings suggest that early life access to alternative reinforcers might be protective against later substance use, even in the context of alternative reinforcement scarcity. However, this finding has not yet been extended to humans.

Finally, animal work has integrated other behavioural economic variables such as delayed reward discounting into models of alternative reinforcement⁹⁹. Experimenters trained rats on self-administration for an alternative reinforcer (60 seconds of social interaction with another rat) and for cocaine. Next, the rats were given choices between these two reinforcers over ten sessions. Across all sessions, rats showed a robust preference for social i...