Choice between food and cocaine or fentanyl reinforcers under fixed and variable schedules in female and male rhesus monkeys

Abstract

Rationale

Illicit drugs may be unpredictable in terms of the time and effort required to obtain them, and this can be modeled with variable- (VR) vs. fixed-ratio (FR) schedules. In a recent experiment (Zamarripa et al. 2022), the potency of cocaine to maintain choice was greatest under a VR (compared with a FR) when food was available under a FR schedule.

Objectives

The goal of the current study was to extend prior choice results with VR vs. FR schedules to a more efficient procedure with cocaine or fentanyl vs. food. Furthermore, the FR schedule of food delivery was manipulated to determine whether increased drug choice under a VR (compared with a FR) schedule depends on the size of the schedule of nondrug reinforcement.

Methods

Adult female (n=2) and male (n=4) monkeys chose between cocaine (0-30 μg/kg/injection) or fentanyl (0-1.0 μg/kg/injection) and food (2 pellets/delivery) under a 5-component procedure. In different conditions, food was available under a FR 25, 50, or 100 and cocaine or fentanyl were available under FR or VR 100 schedules.

Results

Cocaine’s potency to maintain choice was greatest under a VR 100 (compared with FR 100) when food was available under a FR 50 or 100, and fentanyl’s potency to maintain choice was generally greatest under a VR 100 (compared with FR 100) when food was available under a FR 25 or 100. However, outcomes between FR and VR schedules with fentanyl were less robust compared with cocaine.

Conclusion

Variability in the time and effort required to obtain illicit drugs could contribute to excessive allocation of behavior toward drug use at the expense of more predictable nondrug alternatives, supporting treatment or policies aimed at making drug access more predictable through agonist medications or a safe supply. The impact of variable requirements on drug choice may be reduced if nondrug reinforcers are relatively less costly, supporting the use of low-cost reinforcers in behavioral therapies like contingency management.

Introduction

A well-established finding in preclinical research is that drug vs. nondrug choice can be altered by manipulating environmental factors like reinforcer cost or magnitude or the delay to reinforcer delivery (e.g., Campbell & Carroll 2000; Huskinson et al. 2015, 2016; Maguire et al. 2013; Nader & Woolverton 1991, 1992; Negus 2003; Woolverton & Anderson 2006). Furthermore, there is good concordance between drug vs. food choice outcomes in preclinical settings and drug vs. money choice in the human laboratory (e.g., reduced drug choice and intake) or in contingency management (e.g., increased percent of drug-negative urines; e.g., Greenwald & Steinmiller 2009; Higgins et al. 1994; Lile et al. 2016; Packer et al. 2012; Silverman et al. 1999; Stoops et al. 2012; Toegel et al. 2020). However, laboratory investigations of drug vs. food choice typically use the same type of schedule for both alternatives, and this may not mirror real-world scenarios.

We and others have argued that nondrug reinforcers like a paycheck or other goods and services are likely to be available under relatively predictable time and effort requirements whereas illicit drugs like cocaine or fentanyl may be relatively unpredictable (see Doyle & Huskinson 2023; Huskinson 2020; Lagorio & Winger 2014 for recent discussions). Unpredictable time and effort can be modeled experimentally using variable-ratio (VR) or random-ratio (RR) schedules and can be compared with more predictable, fixed-ratio (FR) schedules. Under a VR schedule, the response requirement for each reinforcer varies throughout a session, and the average across the session equals a predetermined value. Comparisons of behavior maintained by VR or RR schedules to behavior maintained by FR schedules have largely been done with nondrug reinforcers (e.g., Madden et al. 2005). For example, compared with FR schedules, VR schedules result in high-rate behavior with less pausing between response bouts and after reinforcer delivery (e.g., Ferster & Skinner 1957), greater overall responding and intake (i.e., less elastic demand; Madden et al. 2005), and when given a choice between nondrug reinforcers under VR vs. FR schedules, subjects generally choose the VR-associated reinforcer over the FR-associated reinforcer (e.g., Fantino 1967; Field et al. 1996; Madden & Hartman 2006). The latter effect occurs even in cases when the VR value is greater than the FR value (Ahearn et al. 1992; Goldschmidt & Fantino 2004; Johnson et al. 2011, 2012). Only recently have some of these effects been extended to behavior maintained by a drug reinforcer (Huskinson et al. 2017; Lagorio & Winger 2014; Zamarripa et al. 2022).

In an important extension of prior choice research, choice between reinforcers associated with FR vs. VR schedules was evaluated using a drug vs. nondrug choice procedure (Zamarripa et al. 2022). In female and male rhesus monkeys, cocaine choice was greatest when available under a VR and food under a FR schedule, and cocaine choice was lowest when available under a FR and food under a VR schedule. This effect was most robust with a relatively large response requirement (FR or VR 200) compared with a smaller requirement (FR or VR 100), indicating that a VR schedule results in increased behavioral allocation towards drug reinforcers (and increased intake) at the expense of nondrug reinforcers compared with a FR schedule.

Zamarripa and colleagues (2022) were the first to evaluate choice between drug and nondrug reinforcers under FR vs. VR schedules, and a rigorous experimental approach was taken. Each condition was conducted until responding was stable, followed by lever reversals in which each lever-reinforcer pairing was reversed, conditions were conducted with a single dose per session (between-session dosing), and a return to the Fixed-Control baseline was conducted between each condition of variability, so each experimental condition followed the same control baseline. This design boosts confidence that the results were reliable, however, it resulted in an experiment that took several years to complete. The goal of the current study was to extend prior drug vs. nondrug choice with FR vs. VR schedules to a within-session dosing procedure originally described by Negus (2003) and used successfully over the past 20 years to evaluate several environmental and pharmacological manipulations on choice between food and drug reinforcers from multiple drug classes (e.g., Banks et al. 2013; Czoty & Nader 2021; Negus 2003; Townsend et al. 2021). Using a within-session dosing procedure allows complete dose-response functions to be obtained on a within-session basis, and therefore, more quickly than a between-session procedure. Another goal of the current study was to extend prior results with cocaine to fentanyl, and the size of the schedule of nondrug reinforcement was manipulated to determine whether the size of the food FR would impact the ability of VR vs. FR schedules to result in greater drug choice.

Materials and Methods

All procedures were approved by the University of Mississippi Medical Center’s (UMMC) Institutional Animal Care and Use Committee and were conducted in accordance with the National Research Council’s Guide for Care and Use of Laboratory Animals (8^th edition, 2011).

Subjects, Apparatus, and Surgery

A total of 2 female (subject weights ranged 6.8-9.4 kg) and 4 male (subject weights ranged 12.0-14.6 kg) adult rhesus monkeys (Macaca mulatta) served as subjects. All subjects had experimental histories with food and drug self-administration. One female (425-2003) and one male (1356) served in food vs. cocaine choice in our prior, between-session dosing experiment with FR and VR schedules (Zamarripa et al. 2022). The same female (425-2003) and two males (0342, 321-2009) served in the food vs. food portion of Zamarripa et al. (2022). Subject 321-2009 also completed a study that involved food-maintained behavior and acute, experimenter-administered benzodiazepines, (+)amphetamine, and morphine (unpublished data). The other female (40066) had experience self-administering remifentanil and food vs. food choice (unpublished data), and one male (9512) had experience self-administering oxycodone and cocaine, alone and combined with kappa-opioid agonists, under a progressive-ratio procedure (Zamarripa et al. 2020) and in drug vs. nondrug choice (unpublished data). Menstrual cycles were estimated in female subjects via vaginal swabbing to detect the presence or absence of menstruation. One female (425-2003) was mostly acyclic and had only one observed episode of menstruation during the study, and the other female had relatively normal menstruation, occurring approximately every 25-35 days.

All subjects were housed individually in stainless-steel, enrichment-style cages (0.76 m X 0.76m X 0.86m; Carter2 Systems, Inc., Beaverton, OR) that were changed every two weeks. A custom-designed operant panel (Carter2 Systems, Inc.) was mounted on the side of each cage. Each operant panel contained two response levers (Med Associates, Inc.) with a red, green, and white stimulus light above each lever, a food receptacle between the two levers, 15-rpm syringe pump (flow rate=0.18 mL/s), a pellet dispenser that delivered 1-g Bio-serv flavored pellets, and a Med Associates connection panel. Each pellet dispenser contained equal proportions of a mixture of very berry, chocolate, marshmallow, banana, and pina colada flavored pellets. Personal computers with Med Associates interfaces and software were used to control experimental events and record data. Subjects were given unlimited access to water. Food was mildly restricted to maintain stable body weights by food earned during sessions (1-g BioServ flavored pellets) as well as supplemental feeding (Teklad 2050 20% protein primate diet, Harlan/Teklad, Madison, WI) that occurred at least 30 min following the end of the session. As part of an enrichment program, subjects were provided access to toys and mirrors within home cages, visual and auditory stimulation (i.e., TV, radio), daily fresh fruit/vegetables and foraging material (e.g., dried fruits, vegetables, nuts), and thrice weekly multivitamins. Additional enrichment was provided when needed based on behavioral assessments conducted by veterinary staff. Lights were maintained on a 14/10 h light/dark cycle with lights on at 0600 h.

Subjects were catheterized and fit with a mesh jacket and tether as described previously (Huskinson et al. 2019; Zamarripa et al. 2022). Prior to surgery, subjects were injected with atropine sulfate (0.04 mg/kg, i.m.) and ketamine hydrochloride (HCl; 10-20 mg/kg, i.m.) followed by inhaled isoflurane, preoperative antibiotics (cefazolin; 20-25 mg/kg, i.m.), and analgesics (carprofen, 2-4 mg/kg, s.c. and buprenorphine SR, 0.05 mg/kg, s.c.). Under aseptic conditions, a double lumen, silicone or polyvinyl chloride catheter was implanted in a major vein (femoral, jugular) with the tip terminating near the right atrium. The distal end of the catheter was passed subcutaneously to the mid-scapular region, where it exited the subject’s back. The catheter was threaded through a tether and connected to a double-lumen swivel (Lomir Biomedical, Inc.) that attached to a custom-designed operant panel (Carter2 Systems, Inc.). Postoperative analgesics (carprofen 4 mg/kg, p.o) were given daily for 3 days, and antibiotics (usually Keflex, 22.2 mg/kg, p.o. or i.m.) were given when recommended by veterinary staff. Catheters were flushed daily with heparinized saline (40-100 U/ml) at least 30 min after the session ended.

General Procedure

Except on cage-change days and some holidays, sessions were conducted daily, starting at 0930 h and were similar to the within-session dosing procedure originally described by Negus (2003). Sessions consisted of 5, 30-min components, with 12 choices/component and a 5-min inter-component interval. If subjects completed all 12 choices prior to the end of the 30-min component, a timeout ensued for the remainder of the component. In all components, red stimulus lights above both levers were illuminated, signaling that the consequences associated with both levers were available. A response to either lever made the opposite lever inactive and engaged the requirement on the active lever. This feature has been used by us and others examining choice between reinforcers associated with VR and FR schedules (Johnson et al. 2011, 2012; Huskinson et al. 2017; Zamarripa et al. 2022) and prevents switching between levers once an initial choice has been made. The first response counted toward the response requirement in effect, and if the requirement was 1, both sets of stimulus lights darkened, and the outcome associated with the chosen lever was delivered. If the response requirement was larger than 1, the stimulus light(s) above the opposite lever darkened after the first response. Once the response requirement was completed, the stimulus light(s) above the active lever darkened, and the outcome associated with the chosen lever was delivered. Completion of the requirement also resulted in a 30-s timeout that included the duration of reinforcer delivery. During reinforcer delivery, all stimulus lights darkened, and food or drug associated with the lever that was pressed was delivered. In all components, completion of the requirement on the left lever resulted in 2 food pellets/delivery under a FR 25, 50, or 100 schedule (in different conditions). Completion of the requirement on the right lever under a FR or VR 100 schedule (in different conditions) resulted in no drug delivery in component 1 and an increasing drug dose across components 2-5, achieved by increasing the injection volume and corresponding injection durations (0, 0.1, 0.3, 1, and 3 s) across components. The corresponding red light above the drug-associated lever remained constant, and a green stimulus light corresponded to increasing drug doses by changing the flash frequency in 3-s cycles as follows: component 1, no green stimulus light; component 2, 0.1 s on, 2.9 s off; component 3, 0.3 s on, 2.7 s off; component 4, 1 s on, 2 s off; component 5, 2.9 s on, 0.1 s off. All conditions were conducted at least 10 sessions and until choice was stable, or a maximum of 30 sessions. Stability required: (1) ≥80% completion of choice trials in each component, (2) preferred-lever choice was within 15% of the mean for 5 consecutive sessions, and (3) no upward/downward trends in choice. In conditions when subjects met the maximum 30-session criterion prior to reaching stability, the last 10 sessions were used in the analysis to account for between-session variability. Table 1 indicates with asterisks when the 30-session criterion was met for individual subjects.

Table 1.

Order of conditions by subject. * indicate conditions in which the 30-session maximum was met.

	Food FR25				Food FR50				Food FR100
	Cocaine		Fentanyl		Cocaine		Fentanyl		Cocaine		Fentanyl
Subject	FR100	VR100	FR100	VR100	FR100	VR100	FR100	VR100	FR100	VR100	FR100	VR100
40066	12	11	9	10	4	3*	1	2	5*	6*	8*	7*
425-2003	9	10	12*	11*	5*	6	8*	7	4*	3	1	2
1356	12	3	1	2	5	4	6	7	10	11	9	8*
0342	2	1	3	4	10	9	11	12	7	8	6	5
9512	1	2	12	11	6	5*	4*	3	7	8	9	10
321-2009	-	-	-	-	7	8	6	5	2	1*	3*	4*

In Fixed-Control conditions, food (2 pellets/delivery) vs. cocaine (0-0.03 mg/kg/injection) or fentanyl (0-0.001 mg/kg/injection) choice was examined under a FR 25, 50, or 100 schedule of food and a FR 100 schedule of cocaine or fentanyl delivery. Cocaine and fentanyl doses were determined individually by selecting a dose range that resulted in 100% drug choice in component 5, and an attempt was made to select a range that also resulted in ≥80% completion of choice trials. However, in some cases, the dose that resulted in 100% drug choice in component 5 also reduced completed choice trials to less than 80%. In Variable-Drug conditions, food was available as in Fixed Control, and cocaine or fentanyl were available under a VR 100. Three values constituted the VR schedule: 1, 100, and 199. These values were selected to allow for smaller, equal, or larger requirements per reinforcer compared to the FR schedule, with the same total possible number of drug-lever responses and is similar to that used in previous experiments (Huskinson et al. 2017; Johnson et al. 2011, 2012; Zamarripa et al. 2022). If the option associated with the VR lever was chosen, values were presented pseudorandomly to ensure that the same value did not occur more than 4 trials in a row. Obtained VR values were close to 100 and are presented for each subject and VR drug condition within each individual-subject panel (Figures 1, 2, and 3). Table 1 shows the order of conditions experienced by each subject. One male subject (321-2009) did not complete the FR 25 food conditions. All other subjects completed all conditions.

Fig. 1. FR 25 Food vs. FR/VR 100 Cocaine or Fentanyl.

%Drug Choice is shown as a function of drug dose (μg/kg/injection) for individual subjects for Fixed Cocaine (open symbols), Variable Cocaine (red or shaded circles), Fixed Fentanyl (filled triangles), and Variable Fentanyl (yellow or shaded triangles). Data points represent the average of the final stable sessions for each condition, +/− one standard error of the mean (SEM).

Fig. 2. FR 50 Food vs. FR/VR 100 Cocaine or Fentanyl.

Data are shown in an identical manner as Figure 1.

Fig. 3. FR 100 Food vs. FR/VR 100 Cocaine or Fentanyl.

Data are shown in an identical manner as Figures 1 and 2.

Data Analysis

Percent drug choice was expressed as drug trials completed divided by the total number of drug and food trials completed multiplied by 100. Percent values were plotted for each subject as a function of dose for the sessions in which responding was considered stable for each food FR condition (25, 50, 100), schedule of drug reinforcement (FR or VR 100), and drug type (cocaine, fentanyl). Changes in choice across FR vs. VR conditions occurred in component 3 for some subjects and in component 4 for other subjects. Furthermore, percent drug choice reached a ceiling in component 4 for some subjects and in component 5 for other subjects. To account for these individual differences across doses and components, group averages were plotted for percent drug choice in relation to the Max dose, defined as the smallest dose that maintained >80% drug choice during Fixed Control. Specifically, percent drug choice was plotted as a function of Dose 0, when no drug was available (Component 1), the dose 1/2 log lower than the Max dose (Max −1/2 log), the dose 1 log lower than the Max dose (Max −1 log), and the Max dose. Trials completed were plotted as a function of food, drug, and total trials (food + drug) completed, summed across components 1-5. For group analyses, separate two-way repeated-measures ANOVAs were conducted for each food FR (25, 50, 100), within each drug, and for each dependent measure (drug choice and trials completed) with schedule of drug reinforcement (FR or VR 100) and dose (Dose 0, Max −1 log, Max −1/2 log, Max dose) or trial type (food, drug, total), respectively, as within-subject variables. Planned comparisons were conducted using Bonferroni’s multiple comparisons to compare differences in percent drug choice or trials completed between FR vs. VR schedules at each dose or trial type, respectively.

In addition, ED₅₀ values were calculated for each subject in each condition using GraphPad Prism (version 9.5.1) by log transforming the x-axis and applying the log(agonist) vs. normalized response, nonlinear regression. For group analyses, separate two-way ANOVAs were conducted for cocaine and fentanyl with food FR (25, 50, 100) and schedule of drug reinforcement (FR or VR 100) as within-subject variables. Subject 321-2009 was not included in the analysis, because this subject did not complete all conditions. Planned comparisons were conducted using Bonferroni’s multiple comparisons to compare ED₅₀ values between FR vs. VR schedules at each food FR value.

Drugs

Cocaine HCl and fentanyl HCl were provided by the National Institute on Drug Abuse (Rockville, MD) and were prepared using 0.9% sterile saline. Injections were delivered at a rate of 0.18 ml/s. All solutions were passed through a 0.22 μm Millipore filter prior to administration.

Results

Figure 1 shows %Drug Choice for individual subjects under a FR 25 for food and a FR or VR 100 schedule for drug. Percent drug choice for fentanyl (triangles) and cocaine (circles) are plotted as a function of Drug Dose (μg/kg/injection) for Fixed Cocaine (open circles) and Fixed Fentanyl (filled triangles) and Variable Cocaine (red/shaded circles) and Variable Fentanyl (yellow/shaded triangles). For all subjects and conditions, percent drug choice increased as a function of dose. Across subjects, Fixed Control and Variable Drug with cocaine or fentanyl were similar, or slightly increased for fentanyl (subjects 40066, 0342, 9512), in terms of drug choice. In general, changing the schedule of drug reinforcement from a FR to a VR had little effect on drug choice when food was available under a FR 25 schedule. There was some between-subject variability in dose ranges for fentanyl. For one female (40066) and two males (0342, 9512), the fentanyl dose range was 0.01-0.3 μg/kg/injection across components and for one female (425-2003) and one male (1356), the dose range was 0.03-1.0 μg/kg/injection. For cocaine, dose ranges were 1.0-30 μg/kg/injection for all subjects except one male (9512) whose range was 0.3-10 μg/kg/injection.

Figure 2 shows %Drug Choice for individual subjects under a FR 50 for food and a FR or VR 100 schedule for drug, and symbols are identical to Figure 1. Percent drug choice increased as a function of dose for all subjects in all conditions. For cocaine, percent drug choice was greater under Variable Cocaine compared with Fixed Cocaine in component 4 for three subjects (425-2003, 1356, 321-2009), in component 3 for two subjects (0342, 9512), and was not different across conditions for one subject (40066). For fentanyl, percent drug choice was greater under Variable Fentanyl compared with Fixed Fentanyl in component 4 for three subjects (40066, 425-2003, 0342) and was largely unchanged across conditions for the other three subjects (1356, 9512, 321-2009). Finally, there were between-subject differences in dose ranges. The fentanyl dose range was 0.03-1.0 μg/kg/injection for both females (40066, 425-2003) and one male (1356) and was 0.01-0.3 μg/kg/injection for the other three males. For cocaine, dose ranges were identical to that described for Figure 1.

Figure 3 shows %Drug Choice for individual subjects under a FR 100 for food and a FR or VR 100 schedule for drug, and symbols are identical to Figures 1 and 2. For all subjects, percent drug choice increased as a function of dose in all conditions. For cocaine, percent drug choice was greater under Variable Cocaine compared with Fixed Cocaine in components 3 and 4 for two subjects (1356, 321-2009), in component 3 for two subjects (0342, 9512), in component 4 for one subject (425-2003), and in component 5 for one subject (40066). For fentanyl, percent drug choice was greater under Variable Fentanyl compared with Fixed Fentanyl in components 2, 3, and 4 for one subject (40066), in components 3 and 4 for one subject (0342), and in component 4 for one subject (1356). Percent drug choice was unchanged or slightly decreased under Variable compared with Fixed Fentanyl in the other three subjects (425-2003, 9512, 321-2009). Finally, dose ranges for both drugs were identical to that described for Figure 2.

Figure 4 shows group averages for cocaine conditions. Fixed Control (open circles) and Variable Cocaine (red/shaded circles) are shown for each food FR (25, 50, 100) from left to right and for %Cocaine Choice (top row) and Total Trials Completed for Food, Drug, and Food + Drug combined (Total; bottom row). Percent drug choice is plotted as a function of Dose 0 (component 1), Max −1 log, Max −1/2 log, and Max dose where Max was defined as the lowest dose of cocaine that resulted in >80% drug choice in Fixed Control (i.e., component 4 or 5). In all conditions, and consistent with single-subject data, cocaine choice increased with increases in dose [significant main effects of dose: FR 25, F(3,12)=1,023, p<0.01; FR 50, F(3,15)=133.7, p<0.01; FR 100, F(3,15)=64.7, p<0.01]. When food was available under a FR 25, there were no effects of schedule type for cocaine (i.e., FR vs. VR 100) [F(1,4)=1.5, p=0.28], and no dose by schedule type interaction [F(3,12)=0.7, p=0.57]. In addition, there were no main effects of schedule type on trials completed (bottom left panel) [F(1,4)=0.2, p=0.65]. When food was available under a FR 50, there were significant main effects of schedule type [F(1,5)=21.7, p<0.01] and a dose by schedule type interaction [F(3,15)=31.8, p<0.01]. Cocaine choice was greater at the Max −1/2 log dose under Variable Cocaine (63.8%) compared with Fixed Control (21.8%; p<0.01). For trials completed, there were no main effects of schedule type (bottom middle panel) [F(1,5)=0.9, p=0.39]. However, there was a significant schedule type by trial type interaction [F(2,10)=20.3, p<0.01], and significantly more food trials and fewer drug trials were completed under FR compared with VR cocaine (p’s<0.01). When food was available under a FR 100, there were significant main effects of schedule type [F(1,5)=8.7, p<0.05], a significant interaction [F(3,15)=5.5, p<0.01], and cocaine choice was greater at the Max −1/2 log dose under Variable Cocaine (68.8%) compared with Fixed Control (37.1%; p<0.01). For trials completed, there were no main effects of schedule type (bottom right panel) [F(1,5)=0.9, p=0.39]. However, there was a significant schedule type by trial type interaction [F(2,10)=10.6, p<0.01], and significantly more food trials and fewer drug trials were completed under FR compared with VR cocaine (p’s<0.05).

Fig. 4. Group Means Food vs. Cocaine Choice.

Group means are shown for %Cocaine Choice (top row) and Total Trials Completed (bottom row) as a function of dose or trial type, respectively, as described in the Data Analysis section. The left column represents conditions when food was available under a FR 25 and cocaine was available under a FR or VR 100, the middle column represents conditions when food was available under a FR 50 and cocaine under a FR or VR 100, and the right column represents conditions when food was available under a FR 100 and cocaine under a FR or VR 100. Red or shaded symbols or bars represent Variable Cocaine and open symbols or bars represent Fixed Cocaine. In the bottom row, data points are shown for individual subjects; circles represent males and triangles represent females. Double asterisks represent differences between variable and fixed conditions at p<0.05 and p<0.01, respectively. Error bars are +/− one SEM.

Figure 5 shows group averages for fentanyl conditions. Fixed Control (filled triangles) and Variable Fentanyl (yellow/shaded triangles) are shown for each food FR (25, 50, 100) from left to right and for %Fentanyl Choice (top row) and Total Trials Completed for Food, Drug, and Food + Drug combined (Total; bottom row). Percent drug choice is plotted as in Figure 4. In all conditions, fentanyl choice increased with increases in dose [significant main effects of component: FR 25, F(3,12)=41.1, p<0.01; FR 50, F(3,15)=41.0, p<0.01; FR 100, F(3,15)=62.4, p<0.01]. There were main effects of schedule type (FR vs. VR) under the FR 25 food condition [F(1,4)=8.7, p<0.05] but not under the FR 50 or 100 food conditions (p’s>0.05), and there were no significant component by schedule type interactions for any food FR condition (p’s>0.05). In post-hoc comparisons, fentanyl choice was significantly greater at the Max −1/2 log dose under Variable Fentanyl compared with Fixed Control in the FR 25 and 100 food conditions (i.e., 36.3% vs. 28.7% and 55.1% vs. 39.1%, respectively; p’s<0.05). For trials completed, there were no main effects of schedule type (FR vs. VR; bottom panels, p’s>0.05) for any food FR. However, there was a significant schedule type by trial type interaction under food FR 25 [F(2,8)=8.7, p<0.01] but not under food FRs 50 or 100 (p’s>0.01).

Fig. 5. Group Means Food vs. Fentanyl Choice.

Group means are shown for %Fentanyl Choice (top row) and Total Trials Completed (bottom row) as a function of dose or trial type, respectively, as described in the Data Analysis section. Conditions are identical as in Figure 4, except with fentanyl rather than cocaine. Yellow or shaded symbols or bars represent Variable Fentanyl and filled symbols or bars represent Fixed Fentanyl. In the bottom row, data points are shown for individual subjects; circles represent males and triangles represent females. Error bars are +/− one SEM.

Figure 6 shows average ED₅₀ (μg/kg/injection) values for Fixed Cocaine (open bars, top panel), Variable Cocaine (red/shaded bars, top panel), Fixed Fentanyl (filled bars, bottom panel), and Variable Fentanyl (yellow/shaded bars, bottom panel). While visual inspection of the ED₅₀ values for cocaine appear to show similar ED₅₀ values across increasing food FRs for Fixed Cocaine and food FR dependent reductions in ED₅₀ values for Variable Cocaine, there were no main effects of food FR or FR vs. VR schedule of drug reinforcement and no interactions. Visual inspection of the ED₅₀ values for fentanyl appear to show food FR-dependent reductions for Fixed and Variable Fentanyl, and ED₅₀ values tend to be smaller for Variable compared with Fixed Fentanyl conditions. In the group analysis for fentanyl, there was a significant main effect of FR vs. VR schedule of drug reinforcement [F(1,4)=8.8, p<0.05] but no significant effects of food FR and no food FR by FR vs. VR schedule of drug reinforcement interaction (p’s>0.05).

Fig. 6. Group Mean ED₅₀ Values.

Group mean ED₅₀ values are shown for cocaine (top panel) and fentanyl (bottom panel) as a function of food FR values. In the top panel, open bars represent Fixed Cocaine, and red or shaded bars represent Variable Cocaine. In the bottom panel, filled bars represent Fixed Fentanyl, and yellow or shaded bars represent Variable Fentanyl. Data points are shown for individual subjects, circles represent males and triangles represent females.

Discussion

In the current experiments, female and male monkeys chose between food and cocaine or fentanyl under FR and VR schedules. The schedule of food (FR 25, 50, 100) and drug (FR vs. VR 100) reinforcement both contributed to overall drug choice. Cocaine choice increased under a VR compared with a FR schedule, but only when the food schedule was relatively large (FR 50 or 100), and fentanyl choice was generally increased under a VR compared with a FR schedule at all food FRs. However, the observed increases in fentanyl choice under a VR compared with a FR schedule were less robust compared with the increases obtained with cocaine. From a translational perspective, the implication of these results is that variability in terms of the time and effort required to obtain drugs may result in greater or excessive allocation of behavior toward drug use at the expense of engaging in more predictable, nondrug alternatives. Furthermore, variability in the time and effort required to obtain cocaine in an illicit market may have less impact on drug use (compared with predictable time and effort requirements) when nondrug alternatives are less costly or more frequent (i.e., the FR 25 food condition). These findings have implications for treatments like contingency management; access to more frequent nondrug reinforcers may reduce the ability of variable drug access to enhance drug choice. Indeed, in contingency-management research, more frequent access to nondrug reinforcers (i.e., monetary vouchers) improves treatment outcomes (e.g., Griffith et al. 2000).

Another important outcome from the current experiments is that prior work in drug vs. nondrug choice with FR vs. VR schedules was systematically replicated with a different procedure (i.e., within-session dosing) and with a drug other than cocaine (i.e., fentanyl). While effects were more robust with cocaine in the current experiment compared with fentanyl, these results suggest that variability in the time and effort required to obtain fentanyl may enhance fentanyl choice compared with more predictable time and effort requirements. Furthermore, within-session dosing allows for more timely completion of full dose-response functions. As a result, this procedure will provide a testable platform for potential treatments aimed at reducing enhanced drug taking that occurs under VR schedules or in experiments aimed at identifying the neurochemical or behavioral mechanisms underlying these outcomes.

In prior research with single-operant arrangements, relatively large response requirements resulted in less elastic demand or greater behavioral output under a RR compared with a FR schedule (Lagorio and Winger, 2014; Madden et al. 2005). Similarly, in prior choice experiments, VR-associated reinforcer choice generally is more robust with larger requirements compared with smaller ones (e.g., Fantino 1967; Field et al. 1996; Madden and Hartman 2006; Zamarripa et al. 2022). This also occurred in the current experiment with cocaine, however, the response requirement was manipulated for food rather than increasing both requirements together, as was done in prior work. Nonetheless, a relatively small food FR reduced the impact of making cocaine available under a VR compared with a FR schedule. Taken together, prior research on the size of the requirement could indicate that variable and scarce access to drugs is not likely to reduce drug use. Rather, such access is likely to increase the amount of behavior allocated toward drug-related activities at the expense of engaging in more predictable nondrug-related activities. However, if nondrug-related activities can be made available under small time and effort requirements, they could more effectively compete with variable and scarce access to drugs. Finally, it is possible that more robust outcomes would have been obtained with fentanyl had we used a larger response requirement associated with fentanyl (e.g., FR or VR 200). Conversely, more robust outcomes may have been obtained with both cocaine and fentanyl if we increased the food FR further (e.g., FR 200). Future research could explore these possibilities.

Several mechanisms have been purported to underly effects obtained with VR or RR compared with FR schedules with drug and nondrug reinforcers, and these have been described by us and others in detail (Huskinson et al. 2017; Huskinson 2020; Lagorio and Winger 2014; Madden et al. 2007, 2011). A potential behavioral mechanism is that choice under VR vs. FR schedules can be explained within a delay-discounting framework (see Madden et al. 2007, 2011). With ratio schedules, some amount of time is required to complete a response requirement, and the time taken to complete the requirement can be conceptualized as a delay to reinforcer delivery (i.e., a delay from the first to the final response). With VR schedules, greater weight may be applied to the sometimes-immediate delivery that occurs, and proportionately less weight is applied to the sometimes-delayed delivery that also occurs. Conversely, the FR option always results in some relatively fixed delay to reinforcement, depending on the time to complete the requirement. Thus, excessive allocation of behavior toward reinforcers associated with VR schedules may result from the sometimes-immediate possibility of obtaining a reinforcer under a VR schedule.

Consistent with a delay-discounting framework, choice of reinforcers associated with VR over FR schedules occurs most robustly when the smallest possible ratio value is 1 and goes away as the smallest value approaches that of the FR schedule (e.g., Fantino 1967; Field et al. 1996). Conversely, the number of component ratios may not affect choice of reinforcers associated with VR schedules as similar findings are reported with two-value VR schedules and with three-, four- or nine-value VR schedules (Fantino 1967; Johnson et al. 2011; Sherman and Thomas 1968; Zamarripa et al. 2022). To our knowledge, one study has directly compared choice between food associated with fixed- and variable-interval schedules, and the number of component intervals (2, 3, or 7 values) did not affect choice (Davison 1972). More research is needed to determine confidently whether VR arrangements with more or less ratio values would differentially affect choice of drug or nondrug reinforcers associated with FR vs. VR schedules.

From a neurobiological perspective, a large body of work referred to as, “Dopamine Prediction Error,” suggests that uncertain reward delivery produces a larger, more sustained dopaminergic response compared to predictable rewards (Fiorillo et al. 2003). The dopaminergic response also can predict subjects’ choices of predictable or variable reinforcer amounts (Sugam et al. 2012; see Nasser et al. 2017; Schultz 2016 for recent reviews on Prediction Error). Similarly, when rats’ behavior was maintained by saccharin delivery, dopamine overflow in the nucleus accumbens was an increasing function of the size of the VR schedule and was greater compared with an equal FR schedule (Mascia et al. 2019). Finally, the dopaminergic response to unpredictable cocaine delivery was larger compared with predictable cocaine delivery (D’Souza and Duvauchelle 2008). A larger, more sustained dopaminergic response could be a mechanism underlying enhanced choice of reinforcers associated with VR vs. FR schedules in the current and previous experiments.

Taken together, research comparing VR with FR schedules indicates that variability in the time and effort required to obtain drugs worsens drug-related behavior compared with more predictable conditions of drug access. These findings have implications for policy and treatment aimed at finding ways to reduce variability related to illicit-drug access that occurs in the real world. Policy and treatment implications have been discussed in detail (Doyle and Huskinson 2023), but two of these are worth mentioning here. One way to reduce variable drug access is through agonist medications. The major assumption behind this approach is that the medication will alleviate withdrawal symptoms and prevent craving. While this assumption is undoubtedly true, this approach also allows an individual to dedicate more time to nondrug-related activities. FDA-approved agonist medications are available for opioids (e.g., methadone, buprenorphine). It is possible that agonist medications for other illicit substances would be similarly effective and have been proposed for stimulant-use disorder (e.g., Negus & Henningfield 2015; Stoops & Rush 2013).

Another way to reduce variability of illicit substances is somewhat controversial. Strategies aimed at eliminating substances from the illicit market do not reduce drug consumption; rather, when it becomes more difficult to procure drugs, producers resort to synthetic analogues to evade drug laws (e.g., Tamama 2021; Tyndall 2020), and the result is the emergence of synthetic substances with unknown health risks. Such unpredictability could be reduced or eliminated by creating a safer drug supply by providing substances with known contents to individuals with substance-use disorder or through legal regulation. A safer supply reduces crime and improves health outcomes (e.g., reduction in abscesses, infectious disease transmission, mortality) in those who use illicit substances (e.g., Bernstein et al. 2020; Fairgrieve et al. 2018; Flemming et al. 2020). Finally, if results from the current and prior investigations on VR vs. FR schedules translates to humans, a more predictable supply could reduce the amount of behavior allocated toward drug-related activities and perhaps create more time for nondrug-related activities.