AND and AND-OR drug mixture discriminations in rats: generalization to single drugs and drug mixtures

Rationale: Studies of the discriminative stimulus effects of drug mixtures provide an approach to polydrug abuse and studies on single drugs with multiple effects. Objective: This study was designed to investigate whether the use of the AND-OR procedure increases the specificity of drug mixture discriminations. Methods: Rats were trained to discriminate a mixture of amphetamine (0.4 mg/kg) plus pentobarbitone (10 mg/kg) from saline (AND-discrimination, n = 8) or to discriminate the same mixture from its component drugs alone (AND–OR discrimination, n = 9). The studies used two-lever operant procedures with a tandem variable interval 1-min fixed ratio 10 schedule of food reinforcement. Results: Under AND-discrimination conditions, there was partial generalization to nicotine and midazolam when each drug was administered singly, and there was no generalization to cocaine, caffeine or ethanol. With the AND-OR discrimination, there was no generalization to any of the preceding drugs administered singly. In “single substitution” tests, nicotine or midazolam was co-administered with the training doses of pentobarbitone and amphetamine, respectively; there was full generalization in the AND-discrimination and partial generalization under AND-OR conditions. Cocaine co-administered with pentobarbitone generalized fully under both procedures, but the dose of cocaine needed was much larger in the AND-OR than in the AND-discrimination. In “dual substitution” tests, mixtures of two novel substances were tested. Mixtures of either nicotine plus midazolam or caffeine plus ethanol produced very marked generalization under AND-discrimination conditions, but were without significant effect in the AND-OR procedure. Throughout the studies, in every instance where comparisons were made, generalization was greater or occurred at lower doses under AND- than under the AND-OR discrimination. Conclusions: The study yielded extensive evidence supporting the hypothesis that the AND-OR discrimination procedure increases the specificity of discriminations based on drug mixtures. Received: 6 May 1998/Final version: 4 October 1998     

Midazolam cue in rats: Generalization tests with anxiolytic and other drugs

Some characteristics of the discriminative stimulus (cue) effects of midazolam, a short-acting benzodiazepine, have been determined in rats. A standard two-bar operant conditioning procedure with food reinforcers delivered on a tandem schedule was used. The 0.4 mg/kg dose of midazolam used for training was well discriminated, typically yielding at least 95% correct responding. Other benzodiazepines increased the percentage of drug-appropriate responding in a dose-related manner and were generalized at doses which had little effect on the overall rate of responding. Doses of pentobarbitone which greatly reduced the overall rate of responding were also generalized with midazolam. Amphetamine, oxotremorine, picrotoxin, morphine, nicotine, quipazine and Ro 15-1788 were not generalized, even at doses which severely suppressed overall response rates. The midazolam cue possesses a considerable degree of specificity and provides a potentially useful assay for drug action at the benzodiazepine receptor complex.     

Discriminative stimulus effects of nicotine in rats trained under different schedules of reinforcement

There have been few comparisons between different schedules of reinforcement for establishing drugs as discriminative stimuli. Fixed-ratio (FR) 10 and tandem variable-interval 1-min FR-10 schedules have been compared directly in a conventional, nicotine-saline discrimination paradigm with food reinforcement in rats. The discrimination was acquired rapidly under both schedules, with stimulus control by nicotine (0.1 mg/kg SC) being very slightly superior under the FR schedule. In 5-min extinction tests with nicotine, rats maintained under the FR schedule yielded a clear dose-response curve with a bar-selection (quantal) index; in these rats, discrimination of nicotine appeared generally poor, and dose-response curves were shallow, when the percentage of drug-appropriate responding (quantitative index) was calculated. In contrast, rats under the tandem schedule yielded clear dose-response data with both indices. In tests with (+)-amphetamine full generalization was obtained with both schedules, and with both quantitative and quantal indices. Tests of generalization to morphine were negative regardless of the training schedule or index employed. In rats under the FR-10 schedule, overall response rates declined both within and across extinction tests; the relatively high rates of responding maintained by the tandem schedule were more sensitive to the response rate-decreasing effects of morphine and amphetamine. The results confirm that orderly data may be obtained with either a FR or a tandem schedule provided that an appropriate index of discriminative response is employed. The results generally support the validity of current practices, and there will probably be no marked differences between conclusions depending on which schedule is used.     

Reversal of overshadowing in a drug mixture discrimination in rats

Previous studies have suggested that in some circumstances, learning processes such as overshadowing may determine the effects that one drug has upon the response to another. The experiments described here examined overshadowing in rats trained to discriminate mixtures of nicotine plus midazolam in two-lever operant procedures with food reinforcement. After training for 60 sessions, midazolam (0.32 mg/kg SC) overshadowed nicotine (0.32 mg/kg SC) so that the discriminative stimulus effect of nicotine seen in control rats trained with nicotine alone was abolished (n=8–10). In the next phase of the study, the discriminative response to midazolam in one group of mixture-trained rats was devalued by means of an extinction procedure which weakened the relationship between administration of midazolam and the response that was reinforced. Dose-response determinations then showed that the devaluation procedure had indeed attenuated the response to midazolam, whereas the previously overshadowed response to nicotine was restored. Post-session injections of drugs were used to equate the pharmacological histories of the groups and the effects seen were therefore attributable to training with the drugs and not simply to repeated exposure to them. Additionally, in the control rats trained with nicotine only (with midazolam given post-session), midazolam markedly reduced response rates, whereas in the three groups of rats trained with the mixture, midazolam had little response rate-depressant effect; this observation suggests that behaviourally contingent tolerance had developed to the response rate-reducing effect of midazolam. Application of devaluation procedures in studies of the discriminative stimulus effects of single drugs with multiple effects may provide a means for manipulating the characteristics of the discriminations obtained and for identifying individual elements of the drug-produced stimulus complex.     

Discriminative stimulus effects of a nicotine-midazolam mixture in rats

Rats were trained to discriminate the effects of nicotine (0.4 mg/kg SC) plus midazolam (0.2 mg/kg SC) from those of saline in a two-bar operant conditioning procedure involving a tandem schedule of food reinforcement. After discrimination training, the component drugs of the mixture produced very considerable amounts of drug-appropriate responding when given separately. Mecamylamine and Ro 15-1788 only slightly attenuated the discriminative response to the mixture when given separately, but completely blocked the response when administered together. In different groups of rats trained to discriminate nicotine or midazolam separately from saline, neither drug appreciably altered the dose-response curve for the other, suggesting a minimal role for pharmacological interactions when effects of mixtures were assessed. The results suggest that the two components of a compound drug-produced stimulus can be perceived separately rather than being blended into a homogenous entity. Knowledge of the characteristics of compound drug-produced stimuli may aid interpretation of the discriminative effects of single drugs with wide spectra of action.     

Drug discrimination studies in rats with caffeine and phenylpropanolamine administered separately and as mixtures

The discriminative stimulus effects of mixtures of caffeine and phenylpropanolamine (PPA) have been investigated because these drugs have been abused together. Rats were trained to discriminate caffeine (20 mg/kg), PPA (20 mg/kg), or a mixture of both drugs, from saline in a two-bar operant conditioning procedure with food reinforcers presented on a tandem VI-FR schedule. Discriminations of mixture, caffeine alone and PPA alone were 90% accurate after 40 sessions. Generalisation to both PPA and caffeine was weak (25–47%) at the doses used in the training mixture, although there was almost complete generalisation to larger doses of PPA. Under these conditions, there was a possible synergistic interaction between caffeine and PPA because the discriminative effect of the mixture could not be fully explained by the combined effects of its component drugs. However, in rats trained on caffeine, PPA had no effect on the dose-response relationship for caffeine; similarly, in rats trained on PPA, caffeine had no effect on the dose-response relationship for PPA (no synergism or antagonism). Generalisation to (+)-amphetamine and cocaine was weakest in rats trained on caffeine, was partial in rats trained on the mixture, and was complete in rats trained on PPA; thus, the mixture of caffeine and PPA was not more like cocaine or amphetamine than PPA alone. The results are in agreement with reports that caffeine and PPA may interact in a complex manner, but do not support the view that the interaction enhances their resemblance to highly abused stimulants such as amphetamine and cocaine.     

In a drug discrimination procedure isolation-reared rats generalize to lower doses of cocaine and amphetamine than rats reared in an enriched environment

Rats with different behavioral histories, defined by rearing and housing in either an enriched condition (EC) or an isolation condition (IC), were trained in a two-lever operant procedure to discriminate 5.0 mg/kg cocaine from saline. In cocaine dose-generalization tests, the IC rats exhibited an ED₅₀ (1.01 mg/kg) significantly lower than the EC rats (ED₅₀:1.55 mg/kg). The cocaine-appropriate responding was emitted when the rats were treated withd-amphetamine, and for thed-amphetamine test doses the ED₅₀ (0.19 mg/kg) was again significantly lower for the IC rats compared to the ECs (ED₅₀:0.33 mg/kg). These data suggest that IC rats are more sensitive to the stimulus properties of indirect dopaminergic agonists than EC rats and highlight the importance of environmental variables in governing an organism's response to the stimulus properties of abused drugs.Supported by DA 05310.     

Midazolam cue in rats: Effects of Ro 15-1788 and picrotoxin

The discriminative stimulus effect of midazolam, a short-acting benzodiazepine, was used for testing the effects of drugs thought to act as antagonists at different sites in the proposed benzodiazepine receptor complex. Rats were trained in a standard two-bar operant conditioning procedure with food reinforcers delivered on a tandem schedule. The 0.4 mg/kg dose of midazolam used for training was well discriminated, typically yielding at least 95% correct responding. The benzodiazepine receptor antagonist Ro 15-1788 blocked the discriminative effect of midazolam but did not influence generalization to pentobarbitone (7.5 mg/kg). The indirect GABA antagonist picrotoxin attenuated both generalization to pentobarbitone and its response rate-reducing effect. Picrotoxin had no effect on the discriminative effect of midazolam at 0.4 mg/kg but it blocked the effect of 01 mg/kg. Even in doses which reduced overall response rates, nicotine did not block discrimination of midazolam (0.4 mg/kg). The results are consistent with models which postulate a GABA-linked ion channel which is a site of action for barbiturates and which is downstream of the benzodiazepine receptor itself.     

The discriminative control of operant behavior by intravenous administration of drugs in rats

Rats were trained to discriminate the stimulus properties of amphetamine and saline. Food reinforcer was given after lever pressing following intravenous amphetamine infusions but not following saline infusions. Subsequent tests under extinction conditions showed that rats pressed a lever at a high rate following infusions of amphetamine, methamphetamine and cocaine, but at a low rate following saline, ethanol, epinephrine or sodium pentobarbital. Similar procedures indicated that rats could also discriminate between ethanol and saline. These findings confirm earlier results indicating that intravenously administered drug can act as discriminative stimuli in controlling operant behavior.     

Effects of d-amphetamine,morphine, naloxone,and drug combinations on visual discrimination in rats

The effects of d-amphetamine, morphine, and naloxone on visual discrimination were investigated using a two-choice discrete-trial procedure in which rats were trained to discriminate the position of a lightflash. Morphine (0.3–5.6 mg/kg) but not amphetamine (0.1–1.0 mg/kg) caused a significant dose-dependent disruption in discriminative performance. Both amphetamine and morphine increased response latencies. Naloxone (1.0 mg/kg) prevented the disruption of any aspect of performance by up to 100 mg/kg morphine. Performance after naloxone/amphetamine co-administration was not significantly different from that observed after amphetamine alone. Naloxone alone (0.3–10 mg/kg) had no effect on discrimination, spatial bias or response latencies. These results suggest that morphine and amphetamine affect different components of discrimination performance. Offprint requests to: S.G. Holzman     

Tolerance to the depressant effects of diazepam in the drug discrimination paradigm

Seven groups of rats (n = 35) were run in operant drug experiments. All groups were trained on a Fixed Ratio 10 schedule to discriminate diazepam from saline. Two groups (n = 7, n = 6), after extensive drug discrimination training (doses of 2.0 and 3.0 mg/kg diazepam), were submitted to generalization experiments with various doses of the training drug. Two additional groups, (n = 6, n = 8) in the initial phase of drug discrimination, were trained on intermediate and high doses of diazepam (i.e., 5.0 and 10.0 mg/kg). The development of tolerance to the depressant effects of diazepam for these two groups was compared to the low dose sophisticated rats. Of the above-mentioned groups, two groups were given tests after a waiting period in drug discrimination training. In this test the two groups were compared to an additional group (n = 8) in its initial phases of drug discrimination training. The results show that a large number of low doses (i.e., doses below 3.0 mg/kg) is not able to induce any tolerance to the depressant effects of diazepam in this particular paradigm. Intermediate doses of diazepam (i.e., 3.0 mg/kg), administered in a large number, induced some tolerance to the depressant effects, while another intermediate dose (5.0 mg/kg) and a high dose (10.0 mg/kg) rapidly induced a significant tolerance. Once developed, the tolerance persisted for 51 days.     

The duration of tolerance to the anorexigenic effect of amphetamine in rats

Rats were given dl-amphetamine 16 mg/kg twice daily for 15 days. Complete tolerance to the anorexigenic effect of amphetamine developed from day 7–11. A single injection of 16 mg/kg amphetamine was given to the amphetamine pretreated rats and to saline pretreated controls at different time-points after withdrawal, and their food intakes were compared. Signs of tolerance were present at 16 but not 20 days after withdrawal.     

Characterizing withdrawal in rats following repeated drug administration using an amphetamine-vehicle-haloperidol drug discrimination

Rationale: Previous research using an amphetamine (AM)-haloperidol (HA) drug- drug discrimination task has shown that predominant responding on the HA-appropriate lever occurs 24 h after a single or multiple administrations of 10 mg/kg AM. Conversely, rebound responding on the AM-appropriate lever occurs following single or multiple administrations of 1 mg/kg HA. HA-appropriate responding was also observed 24 h following a single injection of AM using a three-lever, AM-vehicle-HA discrimination task. However, a single administration of HA did not produce robust rebound responding on the AM-appropriate lever. The present studies seek to clarify the discrepancy between responding following HA in the two- and three-choice tasks. Objective: Experiment 1 examined the extent of rebound responding that could be achieved following ten daily administrations of either 10 mg/kg AM or 1 mg/kg HA. Experiment 2 explored potential differences between the two- and three-choice tasks in characterizing the post-HA cue. Methods: Animals were trained to discriminate 0.35 mg/kg AM, vehicle, and 0.033 mg/kg HA. In experiment 1, animals received ten daily injections of 10 mg/kg AM, vehicle, or 1 mg/kg HA, and were tested 24 h after the final injection, and again 8, 15, and 22 days post-treatment. In experiment 2, animals were retrained and then treated daily with either vehicle or 1.0 mg/kg HA for 10 days, and then tested 24 h after the final injection, and again 5 and 11 days post-treatment, with either all three levers or with only the AM- and HA-appropriate levers available. Results: In experiment 1, multiple injections of AM produced robust HA lever responding, which is consistent with results from previous studies that used the two-choice, AM-HA discrimination task. However, multiple injections of HA did not produce predominant responding on the AM-appropriate lever. In experiment 2, animals treated with either vehicle or HA responded predominantly on the vehicle-appropriate lever when tested with all three levers present. When tested with the vehicle lever removed, however, animals treated with vehicle responded predominantly on the HA-appropriate lever, whereas those treated with HA responded predominantly on the AM-appropriate lever. Conclusions: These results suggest that the two-choice and three-choice task used here differ in how the post-HA withdrawal cue is characterized. This finding emphasizes the importance of knowing the relative locations of the agonist-, vehicle-, and antagonist-produced cues on the interoceptive stimulus continuum established by discrimination training. Received: 3 December 1997/Final version: 16 November 1998     

The serotonin<Subscript>2C</Subscript> receptor agonist Ro-60-0175 attenuates effects of nicotine in the five-choice serial reaction time task and in drug discrimination

Rationale There is evidence that serotonin_2C (5-HT_2C) receptors can modulate some behavioural effects of nicotine, but the generality of this action is not known. Objective To analyse the influence of the 5-HT_2C agonist Ro-60-0175 on responses to nicotine in the five-choice serial reaction time task (5-CSRTT) and on its discriminative stimulus effect; these procedures constitute models for attention-enhancing and subjective effects of nicotine, respectively. Materials and methods In the 5-CSRTT, rats were trained to obtain food reinforcers by detecting light stimuli and then challenged with Ro-60-0175 (0.3–0.8 mg/kg) and nicotine (0.2 mg/kg). For drug discrimination studies, rats were trained to discriminate nicotine (0.2 mg/kg) from saline in a two-lever procedure using a tandem schedule of food reinforcement. Results In the 5-CSRTT, nicotine positively influenced most response indices, confirming previous results. Ro-60-0175 increased response latencies and omission errors and reduced anticipatory responding but had little effect on response accuracy; importantly, it counteracted the effects of nicotine on response speed and omission errors. Pentobarbitone (10–14 mg/kg) also slowed performance of the 5-CSRTT but did not weaken the nicotine-induced enhancement of performance. In the drug discrimination procedure, Ro-60-0175 was not generalised with nicotine but shifted the nicotine dose–response curve to the right in a dose-related manner. Conclusions The data suggest that selective occupancy of 5-HT_2C receptors can attenuate some effects of nicotine in the 5-CSRTT and weaken the nicotine discriminative stimulus; these effects cannot be explained by a sedative action of Ro-60-0175. Electronic supplementary material The online version of this article (doi:) contains supplementary material, which is available to authorized users.     

Extinction and reacquisition of differential responding in rats trained to discriminate between chlordiazepoxide and saline

In order to assess the resistance of drug discriminative responding to prolonged reinforcement omission, rats were trained to discriminate between either 6.0 mg/kg PO or 30.0 mg/kg PO. CDP and saline, using a food reinforced (VI40-FR10) operant procedure. Dose generalization tests were conducted for both groups. Sessions were then run without reinforcement while drug (D) and saline (S) administrations were continued (extinction phase). After a maximum of 30 sessions without reinforcement, or when the rats emitted less than ten responses on either lever during three successive sessions (extinction criterion), reinforcement was reinstated. Finally, additional dose generalization tests with CDP were run. The discriminative responding controlled by the D and S administrations was not affected significantly by prolonged reinforcement omission in either group. For both groups, response rates were decreased and latencies to initiate responding were increased during the extinction phase. Response rate reduction occurred more rapidly for the drug condition in the high training dose group. This group also reached the extinction criterion sooner than the low training dose group. The reacquisition process occurred very rapidly. Response rates increased substantially after the first reinforcement had been obtained. After ten reacquisition sessions, response rates and latencies had reached values similar to those observed before extinction was initiated. Data revealed no differences between groups and the course of reacquisition did not differ between the S and D conditions. The generalization tests executed before the extinction phase and after the reacquisition phase yielded similar results and were in agreement with earlier findings. The major conclusion was that the resistance to extinction of the discriminative accuracy was substantial.     

Antagonism of AND and AND-OR drug mixture discriminations in rats

It has been suggested that use of the AND-OR training method may be associated with an enhancement of the pharmacological specificity of discriminations based on mixture of drugs. Rats were trained to discriminate a mixture of nicotine (0.4 mg/kg s.c.) plus midazolam (0.2 mg/kg s.c.) from saline (AND-discrimination, n = 8) or to discriminate the mixture from either drug alone (AND-OR discrimination, n = 6). The studies used two-lever operant procedures with food reinforcers presented on a tandem schedule. After discriminations were acquired to 80% accuracy, the nicotine antagonist mecamylamine (0.03–1.0 mg/kg s.c.) and the benzodiazepine antagonist flumazenil (0.32–10 mg/kg i.p.) were tested on the response to the mixture of nicotine plus midazolam. The antagonist effects of either mecamylamine or flumazenil given alone were more marked in rats trained under the AND-OR procedure than in rats trained on the AND-discrimination. Similarly, the antagonist effects of mixtures of mecamylamine plus flumazenil were much more potent under the AND-OR than under the AND-discrimination procedure. The AND-OR method reduced the dose of the antagonist mixture needed to produce complete block by a factor of about 10, as compared with the AND-discrimination. These striking differences in sensitivity to antagonists support the view that AND-OR or related procedures may enhance the pharmacological specificity of complex drug discriminations.     

Environmental modification of tolerance to morphine discriminative stimulus properties in rats

The development of tolerance to the discriminative stimulus properties of morphine was examined in rats trained to discriminate saline and 3.2 mg/kg morphine under amultiple timeout 15 min, 5 min fixed-ratio 30 schedule of food delivery. Generalization gradients were generated by administering increasing doses of morphine before successive timeout periods within the experimental session. Over the course of the study, the minimal discriminable dose (MDD) of morphine under control conditions fluctuated but did not systematically increase or decrease. Acute pretreatments of 3.2–17.8 mg/kg morphine 4–24 h before a generalization test resulted in minor changes in the MDD. To examine development of tolerance, supplemental doses of morphine (17.8 mg/kg) or saline were administered twice daily while discrimination training was either suspended or continued. Tolerance was assessed by weekly generalization tests. Greater tolerance developed to the morphine stimulus when training was suspended than when training was continued. For both training conditions, response rates during generalization tests were markedly suppressed during supplemental morphine administration, and original generalization gradients were recaptured within 2 weeks after termination of supplemental morphine administration. Supplemental saline administration did not alter the discriminative or rate-altering effects of morphine under either training condition. Thus, the magnitude of tolerance to a morphine discriminative stimulus reflected an interaction of supplemental drug treatment with the training conditions imposed during that treatment.     

Drug discrimination in rats: evidence for amphetamine-like cue state following chronic haloperidol

Two groups of male Sprague-Dawley rats were trained to discriminate which of two levers to press for milk reinforcement on a VI-20 sec schedule of reinforcement on the basis of whether they were injected intraperitoneally with d-amphetamine (0.50 mg/kg or 1.50 mg/kg) or saline 15 min prior to daily 30 min training sessions. Following acquisition of the discrimination, dose-response functions were generated for both training-dose groups during 5 min test sessions. All subjects were then injected with 1.0 mg/kg of haloperidol for ten consecutive days and retested on either saline or intermediate doses of amphetamine on days 1, 2, 4 and 7 following the final haloperidol injection. The results indicated that chronic haloperidol enhanced the discriminative stimulus properties of amphetamine in both training groups. More importantly, when tested on saline, subjects in both training groups made significantly more responses on the d-amphetamine lever than observed prior to chronic haloperidol. On the basis of linear regression analysis of the dose-response curves it was shown that rats in both groups responded as though they had been injected with 0.18 mg/kg of d-amphetamine. In a second experiment this increase in amphetamine-lever responding when animals were tested with saline following chronic haloperidol was replicated and in addition it was observed that chronic amphetamine had the opposite effect on this measure.     

Discriminative stimulus properties and brain distribution of phencyclidine in rats following administration by injection and smoke inhalation

Four male Sprague-Dawley rats were trained to discriminate IP injections of 3.0 mg/kg phencyclidine (PCP) from saline under a 2-lever fixed-ratio 32 schedule of food presentation. After reliable discriminative control of lever choice was established, other doses of injected PCP were tested resulting in dose-dependent increases in PCP-lever selection and dose-dependent decreases in rates of responding. When doses of PCP were administered by exposure to smoke from cigarettes containing PCP, a dose-dependent increase in PCP-lever responding was also observed. delta 9-Tetrahydrocannabinol administered via smoke exposure, up to doses which markedly suppressed response rates, did not result in PCP-appropriate responding, demonstrating the specificity of the PCP stimulus by the inhalation route. Brain levels and distribution of 3H-PCP were determined in rats administered doses calculated to result in 50% generalization by the IP injection or smoke inhalation routes. By both routes of administration roughly equivalent brain levels were attained and the distribution was relatively even across the seven brain areas analyzed. These results demonstrate the validity of using the injection route of administration when studying PCP experimentally, in spite of the fact that PCP is abused primarily by smoking.     

Role of training conditions in discrimination of central nervous system stimulants by rats

Amphetamine and some related compounds were compared in rats trained to discriminate (+)-amphetamine (0.4,1.0 or 1.6 mg/kg) or cocaine (10.0 mg/kg) from the nondrug condition in a standard, two-bar procedure with food reinforcement (n=5–6). Amphetamine and cocaine were generalized completely with each other, in most cases at dose levels which did not greatly reduce the overall numbers of responses. The ED₅₀ values for amphetamine and cocaine varied with the drug and dose used for training, and it was concluded that the stimuli produced by the two drugs were similar but may not be identical. There was an excellent correlation between ED₅₀ values derived from indices of barselection and percentage-responding on the drug-appropriate bar. Apomorphine was generalized with amphetamine only in the rats trained with the higher doses of amphetamine, and only when administered in doses which greatly reduced the overall number of responses. Para-hydroxyamphetamine increased responding on the drug-appropriate bar only when administered in high doses to the rats trained with the lowest dose of amphetamine (0.4 mg/kg). The results strengthen the evidence that the particular drug and dose level used for training can significantly affect the outcome of generalization tests, and challenge the notion that the discriminability of drugs is an immutable property that is amenable to absolute measurement.