Human Drug Discrimination: A Primer and Methodological Review

Abstract

Drug-discrimination procedures empirically evaluate the control that internal drug states exert over behavior. They provide a highly selective method to investigate the neuropharmacological underpinnings of the interoceptive effects of drugs. Historically, drug discrimination has been one of the most widely used assays in the field of behavioral pharmacology. Drug-discrimination procedures have also been adapted for use with humans and are conceptually similar to preclinical drug-discrimination techniques in that a behavior is differentially reinforced contingent on the presence or absence of a specific interoceptive drug stimulus. This review gives some general history and background concerning the major theoretical concepts and principles of drug-discrimination research as well as its relevance to substance-use disorders. This article also provides a procedural overview and discusses key methodological issues that must be considered when designing and conducting a human drug-discrimination study. Although drug discrimination is unequivocally one of the most sophisticated and useful behavioral assays to investigate the underlying neuropharmacology of drugs in vivo, enthusiasm for its use has steadily declined in the last decade and a half. We conclude by commenting on the current state of drug-discrimination research and suggest potential avenues for future drug-discrimination research.

Introduction

Drug-discrimination procedures empirically evaluate the control that internal drug states exert over behavior and provide a highly selective method to investigate the neuropharmacological underpinnings of the interoceptive effects of drugs. As a result, drug discrimination has been one of the most widely used assays in the field of behavioral pharmacology. Since the publication of one of the earliest studies that demonstrated the control of behavior by the presence or absence of the interoceptive-stimulus effects of alcohol in rats (Conger, 1951), there has been substantial work investigating the discriminative-stimulus effects of drugs spanning more than four decades (e.g., Porter & Prus, 2009a). A PubMed search using the quoted search phrase “drug discrimination” yields 1,303 peer-reviewed publications dating back to the mid 1940s (i.e., Jellinek, 1946). Those concerning human drug discrimination comprise approximately 16% of the total published drug-discrimination literature (i.e., 215 reports). Figure 1 shows the total number of drug-discrimination publications per year since 1973 and the relative proportion of those concerning human drug discrimination.

Figure 1.

Number of published drug-discrimination reports per year from 1973 to early 2016. The total numbers of drug-discrimination publications are presented in light gray bars. The relative number of published reports relevant to human drug-discrimination studies is shown in dark gray bars. X-axis: Publication Year. Y-axis: Number of Publications.

Since the adaptation of drug-discrimination procedures for use with humans, a number of reviews have been published. These prior reviews centered on: (a) the relationship between the discriminative-stimulus and subjective effects of drugs (e.g., Preston & Bigelow, 1991; Schuster & Johanson, 1988; Schuster, Fischman, & Johanson, 1981); (b) the concordance between preclinical and human drug-discrimination findings (Kamien, Bickel, Hughes, Higgins, & Smith, 1993); and (c) the neuropharmacological selectivity of drug-discrimination procedures relative to subjective drug-effect questionnaires (Kelly, Stoops, Perry, Prendergast, & Rush, 2003). Although the present review provides general discussion of some of these previously reviewed topics, it differs from earlier reviews in that it primarily focuses on methodological issues and expands upon previous discussion of these issues when assessing the underlying neuropharmacology of amphetamines (Rush, Vansickel, & Stoops, 2011). To create some context and establish a basis for our discussion of these key methodological considerations, we first attempt to place drug discrimination within a broader context by briefly describing its history and theoretical foundations, advantages and limitations, and relevance to substance-use disorders. Second, we provide a general overview of a human drug-discrimination experiment, describe some notable procedural variations in human drug-discrimination research, and give a basic primer on the interpretation of drug-discrimination data. Third, we highlight several key methodological factors that must be considered when designing and conducting a human drug-discrimination study. Lastly, we conclude by briefly commenting on the current state and potential future of drug-discrimination research in humans. This review is not intended to describe every study that has used human drug-discrimination procedures nor provide a comprehensive overview of the discriminative-stimulus effects of all drugs or drug classes that have been investigated in humans. Instead, this review focuses on studies in humans that employed drugs from pharmacological classes that include commonly used illicit drugs (e.g., stimulants, sedatives, opioids, cannabis) for illustrative purposes.

Historical Background and Theoretical Foundations

Contemporary drug-discrimination procedures were developed, in part, out of early theories of drug-induced, state-dependent learning, which refers to the process by which the internal state produced by a drug can come to exert “stimulus control” over behavior (Overton, 1991). Perhaps the earliest recorded instance of drug-induced, state-dependent learning appeared in a report by Combe (1830) that described the case of an Irish porter for whom the internal state produced by alcohol appeared to serve as a determinant of memory retrieval. Briefly, Combe (1830) wrote that, while in a sober state the porter was unable to retrieve memories for events that took place when he was under the influence of alcohol, but that these memories became accessible again when he was in a state of alcohol intoxication. In other words, memory retrieval was dependent upon being in a state of alcohol intoxication. It was later proposed that the control of memory retrieval by internal stimuli, like those produced by alcohol, was mediated through an “organic sensation” mechanism (Ribot, 1891; but see Overton, 1991). This idea is consistent with what is now called interoception, or the ability of an organism to perceive its internal physiological state (see Craig, 2002 for review). Interoception is a basic physiological/perceptual mechanism that allows internal stimuli to come to control behavior much like that of external stimuli in the environment (e.g., Hölzl, Erasmus, & Möltner, 1996; Naqvi & Bechara, 2010).

The interoceptive-stimulus effects of drugs arise from pharmacological interactions between drug molecules (or ligands) and molecular targets (e.g., receptors or transporters) located throughout the central and/or peripheral nervous system that, in turn, modulate the functional state of endogenous neurotransmitter systems (Holtzman & Locke, 1988). Because drugs differ in their primary mechanisms of action, selectivity for various molecular targets, and relative potency of effect, these pharmacological actions produce a relatively unique set of interoceptive stimuli. These cues may then function to disambiguate a set of differential reinforcement contingencies (Colpaert, Kuyps, Niemegeers, & Janssen, 1976). In other words, the interoceptive-stimulus effects of a drug may come to function as a discriminative stimulus (S^D) such that one response is reinforced only when the internal stimuli produced by the drug are present and a different response is reinforced only when those interoceptive stimuli are absent. Once a differential pattern of responding has been established based upon the presence or absence of these internal drug stimuli, the interoceptive drug stimulus is considered to have gained control of behavior. The pharmacological selectivity of behavioral control obtained with drug-discrimination procedures is quite high and the results of these studies also correlate well with the pharmacodynamics of drugs at the molecular level (e.g., Holtzman & Locke, 1988). As a result, drug-discrimination research has contributed significantly to our understanding of the underlying neuropharmacology of many centrally acting drugs including nicotine and tobacco (Stolerman, 1991; Hoffman & Evans, 2013), caffeine (Griffiths et al., 1990; Oliveto et al., 1992), synthetic cathinones (Iversen, White, & Treble, 2014) antipsychotic drugs (Porter & Prus, 2009b), hallucinogens (Mori, Yoshizawa, Shibasaki & Suzuki, 2012), benzodiazepines (Lelas, Spealman, & Rowlett, 2000), stimulants (Nielsen & Scheel-Krüger, 1988), and opioids (Woods, Bertalmio, Young, Essman, & Winger, 1988).

As noted above, the interoceptive-stimulus effects of drugs and the ensuing stimulus control of behavior have been studied extensively in non-human laboratory animals and drugs from a diverse array of pharmacological classes have been shown to exert discriminative control over behavior in a number of animal species using drug-discrimination procedures (Glennon, Järbe, & Frankenheim, 1991). Drug-discrimination procedures have also been adapted for use with humans and are conceptually similar to preclinical drug-discrimination techniques (e.g., Preston, 1991). In a typical human drug-discrimination experiment, a behavior is differentially reinforced contingent on the presence or absence of a specific interoceptive drug stimulus. For example, following the administration of the training dose of 10 mg oral d-amphetamine, responses on one of two response options are reinforced (e.g., exchanged for money). Following placebo administration, responses allocated to the alternate response option are reinforced.

Advantages and Limitations of Human Drug-Discrimination Procedures

Drug-discrimination procedures in general, and human drug-discrimination procedures in particular, offer a number of advantages relative to other assays commonly used in behavioral pharmacology. First, drug discrimination is dose-dependent, with larger doses of the training drug increasing drug-appropriate responding. Typically, the range of test doses of the training drug includes at least one dose that does not occasion drug-appropriate responding, one dose that produces near maximal drug-appropriate responding, and intermediate doses that usually engender moderate levels of drug-appropriate responding. This most often results in a graded, sigmoidal dose-effect function that may then be used to detect shifts in potency and/or efficacy as a function of a pharmacological manipulation. Second, drug discrimination is also pharmacologically selective in that drugs from the same pharmacological class as the training drug generally increase drug-appropriate responding as a function of dose, whereas drugs from different classes generally produce placebo or “not drug” responding (see Glennon et al., 1991). Third, the level of experimental control in drug-discrimination experiments is enhanced because the participants have a similar training history and recent history of drug exposure. Together, these points support the assertion that drug-discrimination procedures are one of the most sophisticated behavioral assays to investigate the underlying neuropharmacology of drugs. Further, these aspects of human drug-discrimination studies are particularly noteworthy because some human laboratory measures (e.g., subjective effects) do not necessarily have these same strengths. In addition to these advantages, the relationship between the subjective- and discriminative-effects of drugs may be directly evaluated in human drug-discrimination studies (e.g., Reynolds, Bolin, Stoops, & Rush, 2013).

Despite these notable strengths, human drug-discrimination procedures also have several potential limitations that warrant consideration. First, drug-discrimination procedures require additional training before testing can begin and require a considerable investment of time and resources on the part of both volunteers and investigators. An offsetting strength, however, is that, due to the large effect sizes associated with drug-discrimination results, fewer participants are needed in order to achieve adequate statistical power in drug-discrimination studies relative to laboratory procedures that rely more heavily on subjective-effects measures. Second, drug-discrimination tasks provide a relatively limited amount of information (i.e., typically a single outcome measure such as discrimination accuracy) as compared to other behavioral measures that provide information across an array of dimensions (e.g., subjective-effects measures; Kelly et al., 2003). However, because of the more quantal nature of drug discrimination, the interpretation of the resulting findings is less complicated because conclusions may be drawn directly from performance on the discrimination task. The likelihood of Type I errors is also decreased because drug-discrimination procedures rely on a single primary-outcome measure. Third, drug-discrimination performance is relatively insensitive to changes in circulating levels of drug across the time-course of drug effects. In other words, the allocation of responses to the drug-appropriate option does not typically decrease as a function of decreasing plasma levels of the drug (e.g., Kelly, Emurian, Baseheart, & Martin, 1997). Fourth, the investigation of the specific role(s) of various molecular sites of action (e.g., transporters, receptor systems, and specific receptor subtypes) in the discriminative-stimulus effects of drugs in humans is relatively limited because medications that are approved for use with humans by the U.S. Food and Drug Administration are typically used in human drug-discrimination studies. Finally, in the context of the study of substance-use disorders, drug-discrimination procedures lack the face validity of other experimental approaches such as drug self-administration (e.g., McMahon, 2015). Although the drug-discrimination paradigm may lack a certain degree of external validity, it has strong predictive validity with respect to the underlying neurobiological and neuropharmacological mechanisms of drugs and determination of the abuse potential of novel compounds (e.g., Colpaert, 1999; Brauer, Goudie, & de Wit, 1997; Holtzman & Locke, 1988; Huskinson, Naylor, Rowlett, & Freeman, 2014; Kelly et al., 2003).

Relevance to Substance-Use Disorders

Substance-use disorders are a significant public-health concern. In 2014, an estimated 26.9 million Americans reported some form of illicit substance use in the past month (Substance Abuse and Mental Health Services Administration [SAMHSA], 2015). The problematic use of these substances will likely remain an unremitting concern as reports of illicit drug use, both lifetime and current, have remained stable or slightly increased in recent years (SAMHSA, 2015). In addition, approximately 7.1 million individuals had a history of substance dependence or abuse in the past year (SAMHSA, 2015) and relapse rates in drug-dependent patients are discouragingly high. An important question is what factors precipitate and/or contribute to the resumption of problematic drug use in persons suffering from substance-use disorders?

At least one explanation is that the interoceptive cues produced by drugs of abuse may contribute to the occurrence of relapse. Anecdotally, individuals with substance-use disorders report that when they sample a small amount drug after a period of abstinence (i.e., a lapse), the likelihood of relapse increases (de Wit, 1996). Although the exact behavioral mechanisms involved in the resumption of drug-taking behavior are not clearly understood, some theories propose that a “priming dose” of a drug may function as a discriminative stimulus that signals further drug availability and promotes subsequent drug-taking behavior (Bickel & Kelly, 1988; DeGrandpre & Bickel, 1993). Furthermore, discriminative-stimulus effects operate simultaneously with, and independently of, the reinforcing effects of the drug. Because ethical concerns preclude empirical investigation of this phenomenon in treatment-seeking individuals who have achieved drug abstinence, this theory is largely based on findings from studies with non-human laboratory animals in which drugs that have similar discriminative-stimulus effects with a previously self-administered drug restore previously extinguished drug-seeking behavior (de Wit, 1996).

The drug-reinstatement procedure is perhaps the most commonly used preclinical model to investigate the behavioral underpinnings of relapse (reviewed in Carroll & Comer, 1996; de Wit, 1996; Shaham, Shalev, Lu, de Wit, & Stewart, 2003, Shalev, Grimm, & Shaham, 2002; Spealman, Barrett-Larimore, Rowlett, Platt, & Khroyan, 1999). In this procedure, stable drug self-administration behavior is established and then extinguished by substituting saline or vehicle for the active drug (Stretch & Gerber, 1973). A stimulus is then presented to determine if it reinstates operant responding that was previously maintained by drug self-administration. The administration of a priming dose of the previously self-administered drug is most pertinent to this review because this priming dose may function as a discriminative stimulus that signals the availability of drug. Using this procedure, a non-contingent injection of cocaine or methamphetamine reliably reinstates responding that previously resulted in drug administration (Boctor, Martinez, Koek, & France, 2007; de Wit, 1996; Spealman et al., 1999; Yan et al., 2007). The high face validity of the drug reinstatement procedure is responsible, in large part, for its widespread use to investigate the behavioral and neurobiological underpinnings of relapse (Katz & Higgins, 2003; Shaham & Miczek, 2003).

The use of reinstatement procedures in preclinical research has provided valuable information about how interoceptive drug cues may contribute to relapse in humans (Alleweireldt, Weber, & Neisewander, 2001; Banks, Czoty, & Nader, 2007; McFarland & Ettenberg, 1997; Odum & Shahan, 2004). Although the reinstatement of drug-seeking behavior following administration of a priming dose in non-human animals appears outwardly similar to relapse, it is important to note that relapse in humans is a highly complex multifactorial phenomenon. The administration of an exogenous substance such as cocaine is often insufficient to explain the occurrence of relapse in humans. For example, the administration of an agonist medication (e.g., d-amphetamine) under clinical supervision does not appear to precipitate relapse and, in many cases, paradoxically reduces illicit drug use and increases retention rates in clinical trials (Grabowski, Shearer, Merrill, & Negus, 2004; Herin, Rush, & Grabowski, 2010). The contribution of other factors that may affect the likelihood of relapse should also be considered. These factors include, but are not limited to, environmental stressors, the influence of exteroceptive stimuli such as drug-related cues, and social influences (see Hunter-Reel, McCrady, & Hildebrandt, 2009; McKay, 1999; Sinha & Li, 2007 for reviews). Reinstatement procedures with non-human animals also do not necessarily account for or model the influence of other contingencies that may support continued drug abstinence in humans (e.g., positive social or familial outcomes, greater financial stability, health benefits).

Human Drug Discrimination Methodological Overview

To provide the foundation for our discussion of key methodological considerations in the design and conduct of human drug-discrimination studies, this section provides a general overview of the necessary experimental materials and highlights the basic procedural elements of human drug-discrimination procedures. We also describe several notable procedural variations between human drug-discrimination studies, more complex drug-discrimination procedures, and then provide some basic information on the interpretation of drug-discrimination data.

Participant Recruitment and Selection

Potential participants are typically recruited through formal advertisements in local newspapers, online classified ads (e.g., Craigslist), flyers posted in public areas, and by word-of-mouth referral. Volunteers who may qualify upon initial screening complete a rigorous in-person screening that includes a complete medical history, physical health screen, and psychiatric assessment. Volunteers also provide basic demographic information (e.g., age, sex, socioeconomic status) and complete a battery of questionnaires that assess drug-use history and severity as well as symptomology for other clinically relevant conditions such as depression and attention-deficit hyperactivity disorder. Responses on these instruments are used to determine whether volunteers satisfy the study inclusion criteria or meet criteria that would exclude them from participation (e.g., active disease process, psychiatric disorder, prescribed medication(s) contraindicated with the study medication). A physician reviews all screening materials to determine whether the volunteer is physically and psychologically eligible for participation. Thorough physical and mental health screening is absolutely imperative to ensure participant safety in any study involving the administration of pharmacological agents to humans. Given the substantial time commitment required by human drug-discrimination studies, another important consideration is whether a potential participant is able to dedicate the time necessary to complete the study. To this end, potential participants also provide information about their availability during intake screening and as part of the informed consent process.

Test Environment and Experimental Materials

The test environment and experimental materials required to conduct a human drug-discrimination experiment generally consists of a test room containing a desk, chair, a computer with a mouse, numeric keypad and programming to present the drug-discrimination task and record the data, and equipment that is used to monitor participants’ vital signs. Although the use of a computer is more typical, pen and paper could also be used for task presentation and data collection. The room may also be equipped with a television and other recreational materials (e.g., magazines, books, games, craft supplies) available to volunteers when they are not engaged in experimental activities.

In one example of a typical two-choice drug-discrimination task, the volunteer is presented with two response options (e.g., Drug A and Not Drug A) on the computer screen and instructed to respond on the Drug A option if they experience the effects of the training drug and to respond on the Not-Drug-A option if they experience no drug effect (i.e., placebo) or drug effects that are different from those produced by the training drug. They are then instructed to indicate which drug condition that they think they received by distributing responses (e.g., 100 points) between the two options using the numeric keypad. For example, if a volunteer is confident that they received Drug A, they would allocate 100 points to the Drug A option and 0 points to the Not Drug A option. Volunteers complete the drug-discrimination task multiple times at regular intervals throughout the session: usually every 30 minutes to an hour depending on the pharmacokinetics of the drug(s) under study. The total number of responses allocated to the correct response option (i.e., drug-appropriate responding) out of all possible points is exchanged for money at a constant rate. Points have been exchanged for money at rate of $0.04–$0.08 USD per point in previous drug-discrimination studies conducted in our laboratory (Rush, Kelly, Hays, & Wooten, 2002; Sevak, Stoops, Hays, & Rush, 2009). Participants can earn $20–$40 per session, although the specific rate with which points are exchanged for money (i.e., $0.04 vs. $0.08) does not appear to significantly alter performance on the task (Rush et al., 2002; Sevak et al., 2009).

The use of money as the reinforcer in human drug-discrimination studies is a primary difference from preclinical drug-discrimination studies, which typically reinforce responding with food. Further, in preclinical studies, animals are often food restricted so that food reinforcers effectively maintain behavior. Another notable difference between preclinical and human drug-discrimination studies is that some human studies often do not provide immediate reinforcement of responding, with reinforcement being withheld until the end of the session or the study. In contrast, responding by animals occurs under a schedule of reinforcement (e.g., fixed-ratio) with delivery of the maintaining event occurring upon completion of each response requirement.

Procedural Overview

As noted above, human drug-discrimination methods are conceptually similar to those used in preclinical drug-discrimination experiments. Unless otherwise noted, all drugs were administered via the oral route in the studies described in this review. Although a standardized human drug-discrimination procedure has not been established, these experiments often proceed in a fixed order across three phases: (1) Sampling, (2) Acquisition, and (3) Test.

Sampling phase

Participants complete several experimental sessions during the sampling phase to familiarize them with the interoceptive-stimulus effects of the training dose. The training dose is usually identified to participants by a specific code (e.g., Drug A). Participants may also receive placebo during this phase, which is identified by its own unique code (e.g., Not Drug A). During these sessions, participants are verbally instructed to attend to the effects of the drug because the amount of monetary compensation in future sessions is contingent upon correctly identifying the drug that they received.

Acquisition phase

Following the sampling phase, an acquisition phase (sometimes referred to as the Test-of-Acquisition or Control Phase) is carried out to determine whether participants have learned the discrimination. To this end, the training dose and placebo are administered several times, under blinded conditions, in randomized order during each session. For several hours after drug administration, participants periodically complete the drug-discrimination task and other study measures (e.g., subjective drug-effect questionnaires). Although participants are asked to identify which treatment they received on the drug-discrimination task throughout the session, the participant is not informed what the correct treatment code (i.e., Drug A vs. Not Drug A) was until the conclusion of the session. The percentage of correct responses (i.e., correct identification of the treatment) is then converted to money and the participant is told immediately how much money they earned based on their discrimination performance. A performance criterion for is often established to operationalize successful acquisition of the discrimination (e.g., 80% accuracy during four sessions), and participants who meet this performance criterion during the required number of sessions proceed to the next phase. This degree of training is relatively unique to human drug-discrimination procedures compared to other human laboratory procedures (e.g., drug self-administration) and provides participants with similar recent behavioral and pharmacological histories, which is thought to reduce within- and between-subject variability.

Test phase

The final phase is the test phase, during which the discriminative-stimulus effects of different doses of the training drug, novel drugs, or a combination of drugs are investigated. Any session in which participants receive a different dose of the training drug, a different drug, or a combination of drugs is considered a “test session”. Participants are not told the purpose of test sessions, nor do they know when these sessions are scheduled until completing the session. As is the case in preclinical studies, there is not a correct response during these sessions. Therefore, participants either receive the maximum amount of money available, regardless of whether or not they correctly identified the drug condition that was administered, or are compensated based upon their average prior discrimination accuracy. Test-of-acquisition sessions that are identical to those in the acquisition phase are also carried out periodically during the test phase to verify that participants continue to accurately identify the training dose relative to placebo. If a participant fails to correctly identify the drug condition that they received in a test-of-acquisition session during the test phase, additional sessions are carried out to re-establish accurate discrimination. The number of test-of-acquisition sessions conducted during this phase may vary but is typically fewer than the total number of test sessions.

In general, there are two strategies in the choice of drug conditions administered in the test phase with the goal of elucidating the neuropharmacological mechanisms that mediate the discriminative-stimulus effects of the training drug. The first is the use of substitution procedures, in which a range of doses of other drugs is tested to determine if they share discriminative-stimulus effects with the training drug. In other words, the test drug of interest is administered in place of (or “substituted” for) the training drug. If the neuropharmacology of the test drug(s) is known, inferences regarding the neuropharmacological mechanisms that mediate the effects of the training drug can be made based on which test drug(s) significantly increase drug-appropriate responding. The reverse is true if the underlying neuropharmacology of the training drug is known but that of test drug is not. The second approach is to determine a dose-effect function for the training drug alone and in combination with pharmacologically selective compounds. These compounds can be administered concurrently with the training drug or one given as a pretreatment to the other, depending on the pharmacokinetic profiles of the training and test drugs. Inferences are made regarding the neuropharmacological mechanisms that mediate the discriminative-stimulus effects of the training drug based on the mechanism of action of the test drugs that shift the training-drug dose-response curve. Although pretreatment tests with selective antagonists are commonly carried out in preclinical drug-discrimination studies, this test strategy has been used less frequently in drug discrimination studies with humans. One limitation and potential explanation for the underutilization of this approach in humans is that antagonist compounds that are available for use in humans to elucidate the underlying neuropharmacological mechanisms of stimulants that primarily target monoamine systems (e.g., risperidone; Rush, Stoops, Hays, Glaser, & Hays, 2003) typically have actions at multiple receptors and/or neurotransmitter systems. The limited selectivity of these pharmacological agents may restrict conclusions about the specific mechanisms responsible for the discriminative-stimulus effects of these drugs in humans. However, this is not a uniform limitation across all drug classes. For example, selective antagonists are available for opioid drugs and have been used in previous human drug-discrimination studies (e.g., Oliveto, Sevarino, McCance-Katz, & Feingold, 2002; Strickland, Rush, & Stoops 2015).

Notable Procedural Variations

The majority of human drug-discrimination research has used two-choice drug-discrimination procedures. However, responses on the option not associated with the training drug could reflect the absence of drug effects or the presence of interoceptive stimuli that are dissimilar to those of the training drug (Overton, 1974; Young, 1991), and in fact, human drug-discrimination studies often explicitly instruct participants to respond on the non-training-drug option in these circumstances. Consequently, interpretation of results, in particular with regard to phenomena such as partial generalization, is challenging (e.g., Colpaert, 1991; Stolerman, 1991). One procedural variation to address concerns about interpretation of responses on the non-training-drug response option utilizes the typical drug and placebo training conditions, along with the presentation of a third “novel-response” option (e.g., Bickel, Oliveto, Kamien, Higgins, & Hughes, 1993; Oliveto, Bickel, Kamien, Hughes, & Higgins, 1994).

In the seminal study that employed the novel-response option procedure, participants first learned to discriminate between oral triazolam (0.32 mg/70 kg) and placebo using a traditional two-option task (i.e., Drug A versus Drug B; Bickel et al., 1993). After acquiring the discrimination, a range of triazolam doses was tested (0, 0.1, 0.24 and 0.32 mg/70 kg). Next, participants learned the novel-response procedure, in which they were instructed to respond on the third option if the effects of the drug administered during the experimental session were unlike either of the other drugs (i.e., Drug A or Drug B). The same range of triazolam doses was then tested again. Under the two-response condition, the training dose of triazolam engendered 100% drug-appropriate responding whereas the other test doses occasioned 100% placebo responding. Interestingly, when the novel-response option was included, a lower dose (0.24 mg/70 kg) and the training dose of triazolam significantly increased drug-appropriate responding.

Another procedural variation that is intended to improve the selectivity with which the interoceptive effects of drugs are classified is the three-choice drug-discrimination procedure (e.g., Preston, Bigelow, Bickel, & Liebson, 1987). Briefly, three-choice drug-discrimination procedures are implemented in three phases (i.e., Sampling, Acquisition, and Test) much in the same way as two-choice procedures. The primary difference between three-choice procedures compared to typical two-choice or novel-response procedures is that participants are exposed to three different drugs during the sampling phase and each is identified with a unique code (e.g., Drug A, B, and C). Participants must then accurately identify the three different compounds during the acquisition phase before moving on to the test phase. These procedures have been theorized to provide a more accurate characterization of the discriminative-stimulus effects of a drug and to better elucidate the complex underlying neuropharmacology of drugs that interact with multiple neurotransmitter systems (e.g., Johanson, Kilbey, Gatchalian, & Tancer, 2006) or that have mixed agonist-antagonist actions (e.g., France & Woods, 1985; White & Holtzman, 1981, 1983). In addition, three-choice drug-discrimination procedures could theoretically be used to investigate the discriminative-stimulus effects of drugs that are frequently used in combination (e.g., Schechter, 1998), but there are currently no published studies that have trained a drug discrimination using a drug combination in humans.

Interpretation of Drug-Discrimination Data

Before discussing specific methodological considerations that may impact drug-discrimination outcomes, some general information on the interpretation of drug-discrimination data is provided. Figure 2 shows four hypothetical dose-response curves: a training drug (T) and three test drugs (A, B, and C). As is the case with other pharmacological assays, dose-effect functions in drug-discrimination studies are interpreted across two primary dimensions: efficacy (i.e., drug-appropriate responding on the y-axis) and potency (i.e., drug dose on the x-axis).

Figure 2.

Four hypothetical dose-response curves representing the dose-response functions for the training drug (T) and three test drugs (A, B, and C). X-axis: Potency (as a function of increasing drug dose). Y-axis: Efficacy (i.e., percent of drug-appropriate responding).

Efficacy in the context of drug discrimination refers to the degree of drug-appropriate responding engendered by a particular dose of a test drug relative to the training drug. In the hypothetical dose-response data shown in Figure 2, Drugs A and B have similar efficacy to the training drug because they engender similar levels of drug-appropriate responding. In contrast, Drug C engenders approximately 50% drug-appropriate responding relative to the training drug. This “downward shift” in the dose-response function for Drug C suggests that it has reduced efficacy compared to the training drug.

Potency refers to the amount of drug (i.e., dose) required to produce an effect of given intensity. The potency of a test compound in drug-discrimination studies is judged in relation to the training dose such that drugs that engender drug-appropriate responding at lower doses than the training dose are more potent and those that require higher doses than the training dose are less potent. For example, Drug A in Figure 2 produces significant drug-appropriate responding at lower doses than the training drug and, thus, the dose-response function for Drug A is shifted leftward. This leftward shift in the dose-response function of drug relative to the reference compound (i.e., training drug) is indicative of increased potency. In contrast, Drug B in Figure 2 requires higher doses than those tested for the training drug to engender significant drug-appropriate responding. This rightward shift in the dose-effect function for Drug C relative to the training drug is indicative of decreased potency.

In tests of stimulus substitution, Drugs A and B in Figure 2 would be considered to “fully substitute” for the training drug and it would be inferred that Drugs A and B produce similar discriminative-stimulus effects to those of the training drug (T). Drug C, on the other hand, would be considered to produce “partial substitution” and the inference would be that Drug C produces discriminative-stimulus effects that differ from those of the training drug (e.g., a drug having distinct pharmacological actions) or are similar but weaker than the training drug (e.g., a partial agonist or lower dose of the training drug).

For tests in which a pharmacologically selective compound is co-administered with the training drug, a leftward shift of the dose-effect function indicates that co-administration of Drug A enhanced the discriminative-stimulus effects of the training drug. In this case, a leftward shift in the dose-response curve suggests that Drug A functions as an agonist at similar sites of action to the training drug, either directly or via downstream modulation. In contrast, a rightward shift in the dose-response curve following co-administration of Drug B suggests that Drug B reduced the effect of the training drug (i.e., antagonism). However, as shown in Figure 2, the inference might be that the antagonist effect was surmountable because significant drug-appropriate responding occurred at higher doses of the training drug. The rightward and downward shift of the dose-response function following co-administration of Drug C indicates that Drug C produced reductions in both potency and efficacy of the discriminative-stimulus effects of the training drug by a mechanism that was not surmountable by higher doses of the training drug.

Key Methodological Considerations

This section of the review highlights several key methodological factors that should be considered in the design and conduct of human drug-discrimination studies. These considerations include: 1) instructional control; 2) route of administration; 3) training dose; 4) test doses and drugs; 5) measurement issues; 6) ethical issues related to the population of interest; and 7) the use of drug discrimination versus subjective-effects ratings. For brevity, many of the examples below are derived from human drug-discrimination experiments with central nervous system stimulants. However, the concepts that these studies are used to illustrate are not specific to the pharmacological class of the drug under study and should generalize to all human drug-discrimination studies.

Instructional Control

In human drug-discrimination studies, participants typically receive verbal instructions about how to complete the drug-discrimination task. The ability to use verbal instructions is unique to drug-discrimination research with humans and has several inherent advantages. First, the number of sessions required to meet the discrimination criterion is significantly reduced because of verbal instruction. For example, only 4–11 sessions were required for participants to successfully learn the discrimination in four previous studies that investigated the discriminative-stimulus effects of dopaminergic psychostimulants (Hart, Haney, Foltin, & Fischman, 2002; Oliveto, Rosen, Woods, & Kosten, 1995; Rush, Stoops, Wagner, Hays, & Glaser, 2004; Stoops, Lile, Glaser, & Rush, 2005). In contrast, rats and non-human primates may require as many as 60 sessions to acquire the discrimination in preclinical drug-discrimination experiments (e.g., Czoty, Ramanathan, Mutschler, Makriyannis, & Bergman, 2004; Gatch, Taylor, Flores, Selvig, & Forster, 2006; Munzar, Baumann, Shoaib, & Goldberg, 1999; Powell & Holtzman, 2000). Despite the considerable investment required to train laboratory animals in drug-discrimination studies, an offsetting strength is that data collection with animals can continue for a considerably longer period of time relative to humans. Also important to note is that the number of sessions required to meet the acquisition criterion in human drug-discrimination studies might be artificially low because some participants were discharged if they were unable to acquire the discrimination in a predetermined number of sessions (e.g., 12 sessions). The number of participants who are not included in human drug-discrimination studies because they fail to learn the discrimination is typically small (i.e., 1–3 participants or 10–15% of the total number enrolled).

Another advantage of human drug-discrimination studies is that the influence of different sets of instructions on performance can be empirically determined. For example, participants in some studies have been instructed that they were to learn to discriminate between Drug A and Drug B. In other experiments, participants were instructed that they would be required to determine whether they received Drug A or a drug other than Drug A (i.e., Not Drug A). Preston and colleagues systematically compared the influence that this relatively minor difference in instructions exerted on discrimination performance (Preston & Bigelow, 2000; Preston, Liebson, & Bigelow, 1992). In those studies, volunteers who had a history of opioid abuse learned to discriminate 3 mg/70 kg intramuscular hydromorphone versus saline. Dose-response functions were then determined for several other opioid drugs (i.e., hydromorphone, butorphanol, pentazocine, nalbuphine, and buprenorphine). The intrinsic efficacy of these drugs at mu and kappa opioid receptors ranges from full agonist to antagonist (Reisine & Pasternak, 1996). Each drug fully substituted for hydromorphone when the Drug A versus Drug B instruction set was used (Preston et al., 1992). In contrast, discrimination performance was more consistent with the intrinsic efficacy of these drugs at mu and kappa receptors when the Drug A versus Not Drug A instruction set was used (Preston & Bigelow, 2000). The full mu-receptor agonists hydromorphone and buprenorphine increased drug-appropriate responding in a dose-dependent manner, and each drug fully substituted for the training dose at the highest dose tested. In contrast, the mu- and kappa-receptor mixed agonist-antagonists butorphanol and nalbuphine did not fully substitute for hydromorphone at any of the dose and an inverted U-shaped dose-response function was obtained with the kappa-receptor agonist pentazocine. These findings collectively demonstrate the impact that verbal instructions can have on discrimination performance and that the pharmacological selectivity of the discrimination is enhanced when the Drug A versus Not Drug A instruction set is used.

Route of Administration

The route of drug administration (e.g., oral, intranasal, intravenous) is another important factor that must be considered. Administering the study drugs by the same route by which they are abused in the naturalistic environment increases both the ecological and external validity of the study. On the other hand, some commonly abused drugs are known to produce peripheral effects if they are administered via naturalistic routes. Cocaine is often insufflated intranasally in the naturalistic environment, which produces local anesthetic effects (i.e., numbing of the nasal mucosa) in addition to its central effects. Peripheral drug cues can confound the interpretation of drug-discrimination findings because they can function as discriminative stimuli independent of the centrally mediated effects of the drug.

The potential contribution of peripheral cues to the discriminative-stimulus effects of intranasal cocaine has been evaluated previously (Schuh, Schubiner, & Johanson, 2000). In this study, participants (N=3) learned to discriminate intranasal cocaine (50 mg) from placebo (46 mg lactose admixed with 4 mg cocaine). Intranasal administration of low cocaine doses (e.g., 4 mg) produce a numbing sensation, but are not thought to produce discernible blood levels, and have been used as the placebo-control condition in studies with intranasal cocaine in humans for blinding purposes (e.g., Foltin & Fischman, 1988; Higgins et al., 1990, 1993). Schuh and colleagues (2000) also attempted to minimize the peripheral cues of intranasal cocaine by applying benzocaine (20%) to the nasal mucosa. Although benzocaine most likely masked the nasal numbing produced by intranasal cocaine, participants were still able to reliably discriminate cocaine from placebo in as little as 15 seconds after drug administration. Therefore, the volunteers were likely using other peripheral cues (e.g., taste) to discriminate between these conditions rather than centrally mediated effects because they were able to accurately discriminate between the doses before the onset of the centrally mediated effects of intranasal cocaine (i.e., approximately 2 minutes after administration; Lukas et al., 1996). Similarly, the discriminative-stimulus effects of orally administered alcohol and smoked marijuana also appear to be affected by peripheral cues (Chait et al., 1988; Kelly et al., 1997).

When peripheral and other non-central cues (e.g., expectations, visual, etc.) are effectively masked, the discriminative-stimulus effects of abused drugs are generally conserved and remain consistent across various routes of administration. For example, in volunteers with a history of cocaine abuse who had learned to discriminate between 80 mg/70 kg oral cocaine and placebo, oral and intranasal cocaine produced comparable dose-related increases in cocaine-appropriate responding across a range of doses (Oliveto et al., 1995). The sedative drug triazolam (0.25 and 0.50 mg/70 kg per os) was included as a negative control and generally produced placebo-appropriate responding. These findings demonstrate that the discriminative-stimulus effects of a drug that produces perceptible peripheral cues may be effectively evaluated using a route of administration that is not typically used in the naturalistic environment.

One application of human drug-discrimination procedures is that they may be used to determine the initial efficacy of putative pharmacotherapies for stimulant-use disorders. The peripheral drug cues associated with some routes of administration common in naturalistic settings are especially problematic when using the human drug-discrimination paradigm for this purpose. The training drug may continue to engender significant levels of drug-appropriate responding during maintenance on a putative pharmacotherapy because the discrimination is based on peripheral cues. Under this scenario, the putative pharmacotherapy may be deemed ineffective even though it may have attenuated the central effects of the training drug.

Training Dose

Another key consideration is the selection of the training dose. From an ethical standpoint, the minimum necessary dose of the training drug should be used because participants will be exposed to it several times. Lower drug doses are more difficult to discriminate, however. In a series of studies conducted by Chait and colleagues (1984, 1985, 1986a, 1986b), only 37 of 72 (~51%) participants who attempted to learn to discriminate between 10 mg oral d-amphetamine and placebo successfully acquired the discrimination. In another series of studies from our laboratory, approximately 85% of participants (33 of 39) successfully learned to discriminate 15 mg oral d-amphetamine (Lile et al., 2005a; Lile, Stoops, Wagner, Glaser, & Rush, 2005b; Rush et al., 2003; Rush et al., 2004; Stoops, Lile, Glaser, & Rush, 2006). Thus, the number of participants that are able to acquire the discrimination increases following a modest increase in the training dose.

The training dose also influences subsequent discrimination performance. Preclinical studies have shown that animals trained to discriminate lower drug doses (versus higher doses) are more sensitive to the discriminative stimulus effects of the drug, resulting in a leftward shift in the dose-response curve for the training drug (e.g., Terry, Witkin, & Katz, 1994). There appear to be four published studies that have evaluated the particular impact of training dose on discrimination performance in humans and the results of these studies are consistent with the enhanced sensitivity (i.e., leftward shift) observed in preclinical studies (Kollins & Rush, 1999; Perkins et al., 1996; Perkins, Fonte, Sanders, White, & Wilson, 1999; Rush, Critchfield, Troisi, & Griffiths, 1995). In the study by Kollins & Rush (1999), participants learned to discriminate 10 mg oral d-amphetamine (i.e., low-dose group) or 20 mg oral d-amphetamine (i.e., high-dose group) and then a range of doses of oral d-amphetamine (1.25–20 mg) was tested. Participants in the low-dose group exhibited a statistically significant leftward shift in the dose-response function for discrimination performance. These findings demonstrate that participants who are trained to discriminate a lower dose of the training drug are more sensitive to the discriminative-stimulus effects of low doses of the training drug as compared to those who learn to discriminate a high dose.

Worth noting is that differences in training dose failed to affect responding on many of the subjective-effects questionnaires that were included in the studies described above (Perkins et al., 1996; Kollins & Rush, 1999; Rush et al., 1995). For example, d-amphetamine dose-dependently increased scores on the A, BG, and MBG scales (i.e., stimulant-sensitive scales) of the Addiction Research Center Inventory (ARCI; Kollins & Rush, 1999), but there was not a significant main effect of training dose or a significant training dose by d-amphetamine dose interaction. Further, although d-amphetamine significantly increased responses on 8 of 11 items from an investigator-developed Drug-Effect Questionnaire, the dose-response function was shifted significantly leftward in the low-dose versus the high-dose group on only four of these items (i.e., improved performance; like the drug; stimulated; and feel like talking or socializing). Similar effects were observed in the study that assessed the effects of training dose on the discriminative-stimulus and subject-rated effects of nicotine (Perkins et al., 1996). These findings also suggest that, to some degree, the discriminative-stimulus and subject-rated effects of drugs represent unique interoceptive experiences that can be dissociated.

Test Doses and Test Drugs

As mentioned previously, drug-discrimination performance is dose dependent in that drug-appropriate responding typically increases as a function of increasing drug dose. If a sufficient range of doses is tested, at least one active dose will produce a minimal effect whereas at least one dose will produce maximal or near-maximal drug-appropriate responding. Generally, this can be achieved in human drug-discrimination studies if a 6-fold range of doses of the training drug is included. For example, a 6-fold range of doses of oral methylphenidate (0, 5, 10, 20, and 30 mg) yielded a steep, graded dose-effect function in participants who had learned to discriminate 30 mg oral methylphenidate from placebo (Stoops et al., 2005). The lowest dose tested (5 mg) engendered placebo-like responding, whereas maximal drug-appropriate responding was obtained with the highest dose (30 mg). As mentioned earlier, this dose-dependent pattern of responding provides an ideal baseline to then determine how behavioral (e.g., influence of instructions) and pharmacological manipulations (e.g., treatment with an antagonist) modulate the discriminative-stimulus effects of the training drug. The influence of these manipulations can be quantified in terms of the magnitude and direction of the shift of the dose-effect function.

Some studies have tested a dose of the training drug that is higher than the training dose (e.g., Johanson, Lundahl, & Schubiner, 2007; Oliveto et al., 1995; Rush & Baker, 2001; Sevak et al., 2011). As noted above, inferences can be made about the neuropharmacological mechanisms that mediate the discriminative-stimulus effects of the training drug with human drug-discrimination procedures based on the mechanism of action of drugs that produce shifts in the dose-response curve of the training drug (Kelly et al., 2003). These procedures are also used to ascertain whether a putative pharmacotherapy attenuates the behavioral effects of the training drug (e.g., Johanson, 1992). Testing a dose higher than the training dose is necessary to determine if an attenuation of the discriminative-stimulus effects of the training drug is surmountable.

As stated previously, drug-appropriate responding generally increases as an orderly function of dose if the test drug is from the same pharmacological class as the training drug, but does not if the test drug belongs to a different pharmacological class. At least two experimental controls are necessary to determine the pharmacological selectivity of the discriminative-stimulus effects of a drug. The first is a positive control condition, which should be pharmacodynamically and pharmacokinetically similar to the training drug (e.g., methylphenidate in d-amphetamine-trained participants). A positive control would be predicted to engender drug-appropriate responding and fully substitute for the training dose at a minimum of one dose tested. If a positive control is not included, the possibility that participants will only emit drug-appropriate responses after receiving the training drug cannot be ruled out. In addition, the rate of onset of the interoceptive effects of the drug can also function as a discriminative stimulus. To avoid this potential confound, it is important to consider of the pharmacokinetics of the positive control. A second necessary control is a negative control condition. The negative control compound should be pharmacologically distinct from the training drug (e.g., triazolam in d-amphetamine-trained participants) and would not be expected to produce drug-appropriate responding. Because the effects of this drug are expected to differ from that of the training drug, other behavioral indices (e.g., subject-rated drug-effect questionnaires or psychomotor-performance measures) may be necessary to ensure that it produced measurable behavioral or physiological effects. If the negative control compound does not affect other behavioral or physiological outcomes, the most parsimonious explanation would be that biologically active doses were not tested. Positive and negative pharmacological control conditions are critically important in studies designed to determine the behavioral and/or pharmacological effects (e.g., abuse potential) of novel compounds relative to the training drug (e.g., Lile, Stoops, Durell, Glaser, & Rush, 2006; Rush et al., 1998).

Measurement Considerations

Data from human drug-discrimination studies are generally expressed in one of two ways. One approach is to express the number of participants that identified the test condition as the training dose as a percentage. This method typically yields a graded dose-response curve for the training drug (e.g. Johanson, 1991a, 1991b; Johanson et al., 2006, 2007) and is pharmacologically selective in that the number of participants who identify the test dose as the training condition is increased by compounds similar to the training drug but not by those that are pharmacologically distinct. However, this approach dichotomizes drug-appropriate responding (i.e., a participant did or did not identify a test dose as the training dose) which necessitates the use of non-parametric inferential statistics that are generally less powerful than parametric statistical tests.

In the second approach, drug-appropriate responding is allowed to vary along a continuum. For example, participants distribute 100 points between two response options (e.g., Drug A or Not Drug A; Drug A or Drug B) depending on their level of certainty about the identity of the administered drug at the time the task is presented (e.g., Bickel, Bigelow, Preston, & Liebson, 1989; Preston, Bigelow, & Liebson, 1990). Under this arrangement, the responses can vary from 0–100, making drug-appropriate responding a continuous variable. Because drug-appropriate responding is expressed as a continuous variable, this approach supports the use of parametric inferential statistics (e.g., analysis of variance) and is similar to measuring subject-rated drug effects using a 100-mm visual analog scale.

Population of Interest and Ethical Considerations

The way(s) in which the specific population under study might affect experimental outcomes is of critical importance when designing a human drug-discrimination experiment. Most notably, the discriminative-stimulus effects of drugs may differ in participants that vary in their substance-use history and current abuse or dependence status. Although the discriminative-stimulus effects of various drugs have been assessed in normal healthy volunteers (e.g., Rush et al., 1995; Silverman & Griffiths, 1992), drug-dependent individuals (e.g., Lile et al., 2011; Oliveto et al., 2013), and individuals with a history of drug dependence who are currently abstinent/detoxified (e.g., Preston et al., 1989), there are no published studies in which the discriminative-stimulus effects of particular drugs have been prospectively compared between these populations. Several factors should be taken into consideration when selecting the most appropriate population of experimental participants given the specific research question(s) and the primary aim(s) of the study. For example, participants with an extensive history of substance abuse may be most appropriate in the context of testing whether a novel compound has potential for abuse itself or may effectively attenuate the discriminative-stimulus effects of a drug with known abuse potential. An important caveat is that their extensive drug-use history may complicate interpretation of the results because of differences in expectancies, conditioning history, and tolerance (Brauer et al., 1997). Concerns about alterations in their ability to detect interoceptive changes as a result of chronic drug use may also raise questions about the validity of drug-discrimination research with individuals who have been diagnosed with a drug-use disorder (e.g., Paulus & Stewart, 2014; Verdejo-Garcia, Clark, & Dunn, 2012). Although there are advantages and disadvantages to using various populations, research in individuals with and without histories of substance abuse is necessary to gain a more complete understanding of the neuropharmacological mechanisms that underlie the discriminative-stimulus effects of drugs (Brauer et al., 1997).

Drug Discrimination versus Subjective-Effects Ratings

The relationship between subjective drug effects and discriminative-stimulus effects has long been a topic of debate, particularly when attempting to draw comparisons between preclinical and human laboratory findings (e.g., Schuster et al., 1981; Schuster & Johanson, 1988; see Brauer et al., 1997 for review). In a sense, subjective-effects ratings are similar to drug-discrimination performance because they are also used to measure interoceptive drug effects (see Brauer et al. 1997 for a review). Subjective-effects questionnaires are either standardized (e.g., the ARCI; Haertzen, Hill, & Belleville, 1963) or are investigator developed to focus on particular aspects of the subjective experience produced by a drug (e.g., positive or negative drug effects, stimulant- or sedative-like effects, somatic effects). In studies that employ these measures, participants periodically complete a battery of subjective-effect questionnaires to quantify the qualities and magnitudes of subjective experience that occur over time following drug administration. However, attempting to make direct comparisons between drug-discrimination performance and ratings of subjective drug effects may be problematic for several reasons. First, subjective drug effects represent internal events and experiences that cannot be observed or independently measured by an outside observer. Second, descriptions of subjective drug effects are based on individual experiences and conditioning history, which vary across individuals. Third, because subjective drug effects are compound stimuli, they represent a range of co-occurring internal sensations that vary across multiple dimensions (e.g., good effects, euphoria, bad effects, stimulant effects, sedative effects). It follows that the discriminative-stimulus effects of a drug may only reflect a subset of the array of internal sensations engendered by a drug. Although they may not reflect identical internal processes, the discriminative-stimulus and subject-rated effects of drugs are correlated. For example, increases in drug-appropriate responding on a drug-discrimination task are typically accompanied by increases on ratings of subjective drug effects, and correspondence between subjective-effects ratings and drug-discrimination performance has been demonstrated with drugs from diverse pharmacological classes (e.g., opioids, sedatives, and stimulants; Oliveto et al., 1994; Preston, Bigelow, Bickel, & Liebson, 1989; Rush et al., 2011).

As reviewed previously (e.g., Kelly et al., 2003; Preston & Bigelow, 1991; Schuster & Johanson, 1988; Schuster et al., 1981), one advantage of studies with human participants is that this relationship between drug discrimination and ratings of subjective effects can be assessed directly. In a recently published retrospective analysis, the relationship between the discriminative-stimulus and subjective effects of d-amphetamine was assessed using Pearson correlational analyses (Reynolds et al., 2013). Data from six experiments conducted in our laboratory that used similar procedures and measures were combined to evaluate the relationship between the discriminative-stimulus and subjective effects of d-amphetamine. d-Amphetamine dose-dependently increased drug-appropriate responding and produced a constellation of prototypical stimulant-like subjective effects (e.g., increased ratings of Good Effects and Like Drug). Significant correlations with discrimination performance were observed on 15 of 20 items from the Drug-Effect Questionnaire across a range of dimensions (e.g., Willing to Pay For and Active-Alert-Energetic) but the magnitude of these relationships was small to moderate (r values < 0.52). These findings, along with other reports, suggest that there is a moderate degree of concordance between the discriminative-stimulus and subjective effects of drugs.

Although the discriminative-stimulus and subjective effects of drugs are moderately correlated, they are not isomorphic. There are instances in which drugs produce similar discriminative-stimulus effects but dissimilar subjective effects and vice versa. In one prior study, participants learned to discriminate between oral d-amphetamine (10 mg) and placebo (Chait, Uhlenhuth, & Johanson, 1986a). During substitution tests, high doses of the anorectic medications phenylpropanolamine and mazindol occasioned near maximal drug-appropriate responding (i.e., ≥80%). d-Amphetamine and phenylpropanolamine produced a similar pattern of prototypical stimulant-like subjective effects (e.g., increases in ratings of Vigor, Arousal, and Stimulated) whereas mazindol failed to increase ratings on these subjective-effects measures. There are also instances in which drugs produce a similar profile of subjective effects but dissimilar discriminative-stimulus effects. In one study where participants learned to discriminate 20 mg oral d-amphetamine from placebo (Rush, Kollins, & Pazzaglia, 1998), bupropion (i.e., 50–400 mg) produced minimal d-amphetamine-appropriate responding (i.e., less than 40%) but engendered dose-dependent increases in ratings of prototypical stimulant-like subjective effects (e.g., Active-Alert-Energetic, Elated, and Vigorous) similar to those of d-amphetamine.

Drug-discrimination data may also provide information about the abuse potential of drugs. If a novel drug shares discriminative-stimulus effects with a drug with known abuse potential, the novel drug may also have increased potential for abuse. However, drug-discrimination studies require a considerable investment of time and resources on the part of both investigators and research volunteers. Subjective effects measures provide an efficient and relatively inexpensive means of assessing abuse potential. Similar to drug discrimination, inferences about abuse potential can be made by comparing the subjective responses produced by a novel drug to those produced by a reference compound with known abuse potential (Fischman & Foltin, 1991). Subjective-effects measures have been used extensively to assess abuse potential in humans and, in certain cases, the use of subjective-effects measures alone may be sufficient for the assessment of abuse potential (e.g., Reynolds et al., 2013).

Drug-discrimination procedures may be better suited to assess the underlying neuropharmacology of drugs compared to subjective-effects questionnaires. Preclinical studies have implicated central monoamine systems, namely dopamine (DA), in the discriminative effects of amphetamines (e.g., Tidey & Bergman, 1998). The results of human drug-discrimination studies also suggest that central monoamine systems are involved the discriminative-stimulus effects of drugs. In one such study, the discriminative-stimulus and subjective effects of a range of doses of d-amphetamine (i.e., 0, 2.5, 5, 10, and 15 mg) were assessed alone and following pretreatment with the DA and serotonin (5-HT) blocker risperidone (i.e., 0 and 1 mg) in participants who had learned to discriminate 15 mg d-amphetamine from placebo (Rush et al., 2003). Risperidone pretreatment significantly attenuated the discriminative-stimulus effects of d-amphetamine and some, but not all, of the positive and stimulant-like subjective effects of d-amphetamine. In contrast, studies that rely primarily on subjective-effects measures generally do not suggest that subjective-effects ratings are attenuated by the administration of monoamine antagonists (Brauer & de Wit, 1996, 1997; Brauer, Rukstalis, & de Wit, 1995; reviewed in Brauer et al., 1997). For example, the subjective effects of 20 mg of methamphetamine were not altered following treatment with 0.75 mg of risperidone or 3 mg of the non-selective DA receptor antagonist haloperidol (Wachtel, Ortengren, & de Wit, 2002). The rate of compensation in drug-discrimination studies is based upon correct identification of the training drug whereas payment in studies that only use subjective effects is not contingent on accurate performance. Although the influence of contingencies on subjective-effects ratings has not been prospectively evaluated, it is possible that placing a contingency on the correct identification of a drug in drug-discrimination studies also affects subjective-effects ratings. These findings collectively suggest that the results of drug-discrimination studies tend to be more consistent with the hypothesized neuropharmacology of drugs and, as a result, may be better suited to investigate the underlying pharmacological mechanisms of drugs compared to subjective-effects questionnaires.

Although the discriminative-stimulus and subjective effects of drugs are related, the use of these procedures should not be considered mutually exclusive or duplicative. Subjective-effects ratings provide qualitative information (e.g., index of positive or negative drug effects) about the interoceptive cues engendered by drugs and, depending on the research question, may be the most practical method for the assessment of abuse potential in humans (Rush et al., 2011). Whereas, if the goal of a study is to determine the underlying neuropharmacology of a drug, then drug-discrimination procedures may be more informative. Overall, subjective-effects and drug-discrimination measures provide complementary information about the nature of a drug’s effects, abuse potential, and the neuropharmacological mechanisms that mediate the interoceptive-stimulus effects of a drug. The concomitant use of subjective-effects and drug-discrimination procedures is quite practical and is therefore likely to be more informative compared to the use of either of these measures in isolation (Reynolds et al., 2013).

Conclusions

Although the extant literature firmly establishes human drug discrimination as a sophisticated, versatile, and useful behavioral assay of in vivo neuropharmacology, the interest in, and use of, human drug-discrimination research, and drug-discrimination research in general, has waned since its peak in the late 1990s. One factor that has potentially led to the decrease in enthusiasm for drug-discrimination studies in substance-abuse research is that the role of discriminative-stimulus effects in substance abuse may be less apparent relative to behavioral processes that are the focus of other experimental approaches. McMahon (2015) articulates a particularly poignant example regarding the downward trend in the publication of drug-discrimination compared with the continued increase in the publication of drug self-administration research. Specifically, that article states that drug discrimination lacks the face validity of drug self-administration because operant behavior maintained by a drug reinforcer more closely resembles the observable behavioral phenomenon of substance abuse (McMahon, 2015). Although behavioral models that have face validity are intuitively appealing, an argument could be made that the predictive validity of a model is more important. The validity of the drug-discrimination paradigm for identifying the underlying neuropharmacology of centrally acting drugs in whole organisms is unparalleled. Less research has centered on the role of discriminative-stimulus effects in substance abuse but, as outlined above, they may play a particularly important role in relapse and the resumption of problematic drug use.

Although the use of human drug-discrimination procedures in the future is uncertain, the emergence and growing popularity of synthetic cathinones (i.e., bath salts), synthetic marijuana (i.e., spice or K2), and devices that are used to vaporize nicotine (e.g., e-cigarettes) and cannabis may create renewed opportunities for future drug-discrimination research. Furthermore, creative thinking about the application of human and laboratory animal drug-discrimination procedures to the investigation of interoceptive events that may contribute to substance abuse (e.g., drug withdrawal, anxiety, stress, etc.) may also provide opportunities for the use of these procedures to investigate the abuse-related behavioral effects of drugs in addition to underlying neuropharmacology.

Public Health Statement.

Substance-use disorders are an unrelenting public health concern. Human drug-discrimination procedures provide valuable information about the ability of drugs to exert control over behavior, the underlying neuropharmacology of drugs, and the abuse potential of drugs. In addition these procedures provide a tool to screen novel medications for substance-use disorders. Overall, human drug-discrimination has significantly contributed to our understanding of the effects of commonly abused drugs and has helped to identify pharmacological and behavioral mechanisms that are targets for substance-use disorder intervention development.