The role of the response–outcome association in the nature of inhibitory Pavlovian-instrumental transfer in rats

Abstract

Inhibitory stimuli can reduce animals’ reward seeking in an outcome-specific manner or outcome-general manner. However, we do not understand the factors that determine which of these effects are produced. To address this, we carried out three experiments which examined whether instrumental training with one or multiple outcomes determined the nature of subsequently observed Pavlovian-instrumental transfer (PIT). Rats underwent Pavlovian training to produce inhibitors and excitors for two outcomes using a feature-negative procedure. In Experiment 1, these stimuli were tested for their effects on a single response trained with one of those outcomes in a PIT procedure. Here, stimuli trained as inhibitors and excitors were found to produce outcome-general effects on reward seeking (in addition to an outcome-specific effect for excitors). In Experiment 2, we trained two responses, one for each of the Pavlovian outcomes, and tested the effect of the stimuli on each response individually. This design also produced outcome-general inhibitory and excitatory PIT effects. Experiment 3 followed the procedure of Experiment 2, except for implementation of a shorter Pavlovian training phase and an additional choice test, where both responses were concurrently available. This procedure produced putative inhibitory effects that were also outcome-general. However, outcome-specific excitatory effects were observed, indicating that the general inhibitory results may not be attributable to the duration of Pavlovian training. Overall, this study suggests that variations in the number of response–outcome contingencies experienced by animals do not readily determine the specificity of putative inhibitors.

Keywords

Pavlovian-instrumental transfer conditioned inhibition outcome-specific learning motivation rats

Animals use environmental information about the presence or absence of rewards to guide reward-seeking behaviour. This process can be studied through the Pavlovian-instrumental transfer (PIT) paradigm (Holmes et al., 2010). This paradigm can take various forms, but in the simplest case, subjects first receive Pavlovian training where a stimulus, such as a tone, predicts a rewarding outcome, such as food pellets. Second, the subjects are trained to make an instrumental response for that same outcome. Finally, the effect of the stimulus on instrumental responding in the absence of any outcomes is assessed. In this final test phase, reward-associated stimuli often increase instrumental responding compared with baseline periods where no stimuli are present. This effect is not limited to reward-related learning; presentation of a stimulus that predicts an aversive outcome, such as shock, can enhance a previously established avoidance response (e.g., Rescorla & Lolordo, 1965; Solomon & Turner, 1962) and suppress instrumental responding for appetitive outcomes such as food (as seen in the well-established conditioned emotional response [CER] paradigm; Estes & Skinner, 1941). We will focus here on reward-related examples for simplicity.

The associative bases of PIT effects are not fully understood, but two types of explanations have been put forward. First, the stimulus could elevate performance of reward-related behaviours in general by activating appetitive arousal, and here, the stimulus would be expected to invigorate any response linked to a congruent motivational system and the observed increase in the lever-press response available in this example reflects this activation (Rescorla & Solomon, 1967). Alternatively, stimulus presentation could activate a representation or expectation of the particular outcome that it predicts and consequently cue behaviours associated with that unique outcome (Trapold & Overmier, 1972). Importantly, however, on this account, the stimulus would not increase responses associated with other distinct outcomes, even if they are relevant to the same motivational system. Effects compatible with both motivational and expectancy-based accounts have been observed following a range different training conditions (Corbit & Balleine, 2016). Thus, it appears that multiple forms of PIT can be generated, and an important question is what training factors determine the type of PIT observed.

A series of experiments by Holland (2004) suggests that nature of the PIT effect depends on how many outcomes have been trained and whether they are associated with distinct responses (see also Corbit & Balleine, 2005). In one experiment, rats were trained to associate two stimuli (S1 and S2) with food pellets or sucrose solution as outcomes (O1 and O2) such that S1 → O1 and S2 → O2. This was followed by instrumental training where two responses (R1 and R2) were reinforced with O1 but underwent either minimal or extended training. In a subsequent extinction test, both stimuli, whether paired with the same or different outcome as the responses, produced increases in responding above baseline. Furthermore, the transfer effect was greater for the response that had undergone more extended training. The observation that both stimuli produced a similar elevation in responding, regardless of trained outcome, is consistent with a PIT effect based on the motivational significance of the outcome rather than its unique properties. Such effects are often referred to as general PIT.

Interestingly, another group in the same experiment (Holland, 2004) was trained with the same Pavlovian contingencies, S1 → O1 and S2 → O2, but in the instrumental phase, they were trained to perform two responses (R1 and R2) that now produced distinct outcomes (R1 → O1, R2 → O2). In the test phase, each stimulus had a greater effect on the response that had been associated with the same outcome (i.e., S1 increased R1 more than S2, and S2 increased R2 more than S1). The selective effects observed here are not readily explained by motivational accounts alone as both pellets and sucrose, and stimuli that predict these outcomes, should tap into a similar motivational state: hunger. These results suggest that when multiple outcomes are earned by distinct responses, stimuli invigorate responding via an expectation of a particular outcome and cue responses also associated with those outcomes. Such effects are often referred to as specific PIT.

While the above examples describe excitatory effects of stimuli on responding, inhibitory stimuli have also been seen to produce, PIT effects (e.g., Alarcón & Bonardi, 2016; Delamater et al., 2003; Laurent & Balleine, 2015; Quail et al., 2017). As for excitatory stimuli, the suppressive effects of inhibitory stimuli could arise when stimuli and responses activate incongruent motivational states and the stimuli thereby produce interference with instrumental responses in general. For example, a signal for the absence of food may decrease appetitive arousal and thus decrease responses earning outcomes that satisfy hunger, or a signal for shock might activate fear, which suppresses lever-pressing for food. Alternatively, a stimulus might signal that a specific outcome will not occur, thus reducing responses also associated with that particular outcome. Indeed, both general and outcome-specific inhibitory transfer effects have been reported, although it is difficult to make comparisons across studies that differed in multiple other factors (type of reward, methods of establishing inhibitors and possibly the nature of inhibition generated, the presence or not of discriminative stimuli, extent of training, etc.). Furthermore, no study has directly compared PIT effects produced by inhibitors on the responding of animals trained with one or two response–outcome contingencies. Thus, the aim of the following study was to test the effect of instrumental training experience on the outcome specificity of inhibitory PIT effects. All animals in this study underwent the same differential Pavlovian conditioning where two excitors predicted two distinct outcomes (A → O1 and B → O2) and two other stimuli predicted the absence of these same outcomes (X → no O1 and Y → no O2). These inhibitors were produced using a feature-negative procedure (after Laurent & Balleine, 2015). In Experiment 1, animals were trained with a single response–outcome relationship (R1 → O1), whereas in Experiment 2, animals were trained with two distinct response–outcome relationships (R1 → O1, R2 → O2). It was expected that if this variation in instrumental training influenced the effects of inhibitors in the same way as it impacts the nature of PIT generated by excitors, then inhibitory PIT would be general in Experiment 1 and outcome-specific in Experiment 2.

Experiment 1

Pavlovian training was conducted using a feature-negative design similar to that reported by Laurent and Balleine (2015) wherein two stimuli became conditioned inhibitors (e.g., A+, AX−). Rats then underwent instrumental training reinforced with one of the Pavlovian outcomes. This instrumental training involving only one of the outcomes from the Pavlovian phase was expected to produce general excitatory PIT as in Holland (2004). It is somewhat more complex to demonstrate evidence of conditioned inhibition than conditioned excitation (Rescorla, 1969). As such, two measures were used to test for inhibitory PIT. First, simple tests of the effects of both inhibitors compared response rates during the stimuli with baseline periods where no stimuli were present. Second, compound stimulus trials were assessed following the same logic as summation tests of conditioned inhibition. Specifically, we compared the effects of compounds of inhibitors and excitors with the effect of excitors alone on instrumental response rate where relatively less responding in the presence of the compound compared with the excitor alone provides evidence of inhibition. Furthermore, the use of compound stimulus trials, as used in Laurent and Balleine (2015), allowed us to test the effects of inhibitors in compounds congruent with feature-negative training or in novel compounds that were incongruent with training. In this experiment, incongruent compounds were of interest because they provide an additional test of any potential outcome-general inhibitory PIT. If inhibitory PIT effects are not specific to the outcome with which the stimulus is trained, we might expect to see evidence of inhibition in novel compound trials, where the inhibitor is paired with an excitatory stimulus that predicts a different outcome.

Method

Subjects

Twelve experimentally naïve male Long Evans rats (ARC, Perth) were used in this experiment, consistent with previous PIT studies (e.g., Holland, 2004: 12 per group). They were 7 weeks old on arrival and weighed 260–310 g. They were housed in groups (four to a cage) throughout the experiment in a colony room with a reversed 12-hr light cycle, with lights on at 21:00 hr, starting 4 weeks before the beginning of the experiment. All animals were handled daily, and 2 days prior to training, food access was restricted to 15 g of chow per rat per day. Rats had free access to water in the home cages throughout the experiment. Experimental events occurred daily between the hours of 11:00 and 16:30, and rats were fed and weighed 30–60 min after training sessions. All procedures were in accordance with the recommendations of the Australian code for the care and use of animals for scientific purposes (8th edition, 2013), and were approved by the Animal Ethics Committee at the University of Sydney (Protocol 589).

Apparatus

All experimental procedures occurred in 12 Med Associates (East Fairfield, VT, USA) operant chambers which were isolated from external light and sound by individual housings. A magazine recessed into the wall of each chamber could receive 45 mg food pellets (Bio-Serv, Flemington, NJ, USA; grain-based formula) or 0.1 mL deliveries of 20% (w/v) sucrose solution. The chambers could deliver auditory stimuli, including white noise and a 5 Hz click produced by a solenoid, both emitted at 80 dB with 60 dB of background noise produced by a ventilation fan. A 3 W, 24 V houselight located in the top centre of the wall opposite the magazine was used as a visual stimulus, as were two 75 W, 125 V key lights located either side of the magazine. The chambers were dark apart from these stimuli. Beneath each of the key lights was a retractable lever.

Procedure

Preexposure

On the day preceding training, rats were placed in the operant conditioning chambers where they received context habituation for 2 min. They were also exposed to sugar solution and grain-based pellet reinforcers in their home cages to familiarise them with these outcomes and encourage consumption during the first training session.

Pavlovian conditioning

Stimuli A and B (noise or clicker) were associated with Outcomes 1 and 2 (pellets or sucrose). In addition, A and B were presented in nonreinforced compounds with X and Y (houselight or key lights), respectively. This generated four types of trial: A → O1, AX → nothing, B → O2 and BY → nothing (see Table 1 for a design summary). All stimulus–outcome pairings were counterbalanced.

Table 1.

Design of Experiment 1.

Pavlovian training	Instrumental training	PIT Test 1	PIT Test 2
A → O1	R1 → O1	R1; Inhibitors (X, Y) Excitors (A, B)	R1; Excitors (A, B) Congruent (AX, BY) Incongruent (AY, BX)
AX → nothing
B → O2
BY → nothing

Note. PIT = Pavlovian-instrumental transfer. A, B = white noise or clicker; X, Y = houselight or key lights; O1, O2 = sucrose or pellets; R1 = left or right response lever. During PIT tests, each of the stimuli indicated for the test was presented 2 times in a set order.

Rats were trained once daily for 20 days in 85-min sessions. Each session included four 60-s presentations of each trial type with an intertrial interval (ITI) of 2–6 min (4 min on average). For each trial, the rate of magazine entries was recorded, both during the stimulus period and in the 60 s preceding stimulus onset. This prestimulus period was used to indicate the baseline response rate. Reinforcement occurred on average 3 times throughout the duration of A and B trials according to a random time 20-s schedule. All sessions began with presentations of each of the reinforced stimuli, followed by each of the nonreinforced compounds. The order of trial types was varied pseudo-randomly for the remainder of the session.

On the last day of training, one presentation of each of stimulus A and B went unreinforced. These probe trials assessed the rate of responding produced by the stimuli independently of magazine entries resulting from outcome consumption.

Instrumental conditioning

The day after Pavlovian conditioning, a lever was made available and lever presses were reinforced with Outcome 1 (grain pellets or sucrose solution, counterbalanced). During the first session, reinforcement occurred on a continuous basis. Thereafter, lever presses were reinforced according to increasing variable interval (VI) schedules of 10, 20, 30, and 60 s, with additional VI 60-s sessions from Days 6–9 (consistent with previous studies; notably, Holland, 2004; see also Delamater et al., 2003). Sessions ended when rats acquired 60 reinforcements or when the session reached 60 min in duration. On Days 1–5, if rats did not receive 45 or more reinforcements during a session, they received an additional day of training on that schedule before advancing to the next schedule. After instrumental conditioning, an additional Pavlovian conditioning session was given, as described above.

PIT Test 1 and retraining

This test assessed the ability of cues to influence instrumental responding for Outcome 1. Cues were either trained as inhibitors or excitors for the same outcome as the instrumental response (O1), or a different outcome (O2). Therefore, X and A were labelled as “Same” and Y and B as “Different.”

During the test, the response lever was available throughout, and stimuli were presented periodically. No reinforcement was delivered. One minute after the lever was extended, 60-s trials occurred with 2-min intertrial intervals (ITIs). An initial block of trials tested effects of the “Inhibitor” stimuli as follows: houselight, key lights, key lights, houselight, experienced as either X, Y, Y, X, or Y, X, X, Y for different animals. A second block of trials followed immediately and tested the effects of the excitors: noise, click, click, noise, thus A, B, B, A or B, A, A, B.

This block structure was designed to take into account within-session extinction of the baseline instrumental response rate. High baseline response rates have been shown to preclude excitatory PIT effects (Colagiuri & Lovibond, 2015), while low response rates could reasonably be expected to preclude inhibitory PIT. Therefore, we presented X and Y during the period expected to have the highest baseline of instrumental responding, that is, at the start of the session. We then presented excitatory stimuli during the potentially lower baseline occurring during the second half of that test session. Furthermore, as the main comparison of interest was between putative inhibitors associated with the same or different outcome as the lever, this comparison was not affected by the block structure.

To reestablish associations following the first extinction test, a further seven sessions of Pavlovian conditioning were given, as described above. These were followed by three more sessions of instrumental training, one of VI 30 s and two of VI 60 s.

PIT Test 2

The ability of “Inhibitor” stimuli to moderate the effect of “Excitor” stimuli was tested on the day following retraining. With the response lever available in extinction conditions, A, B, and the compounds AX, BY, AY, and BX were presented in a unique order for each rat to balance any effects of order across rats. Each stimulus was presented once to avoid floor effects due to extinction of lever-pressing throughout the session. As in Test 1, A and B were labelled “Same” and “Different”, respectively, as were compounds which included those stimuli. Compounds were labelled as either Congruent or Incongruent based on whether they matched original training (AX, BY) or were novel configurations different to those used in training (AY, BX). These latter trials assessed whether any inhibitory effects of X and Y were tied to the particular stimulus (and predicted outcome) with which they were trained.

Discrimination of visual stimuli

As nonspecific effects of the stimuli were predicted, this result could be confounded by a lack of discrimination between stimuli. We therefore assessed whether rats could perceptually discriminate between the houselight and the key-light stimuli which are not commonly used in specific PIT tasks (unlike the auditory stimuli used; for example, Corbit & Balleine, 2005; Corbit et al., 2007). We therefore reinforced the houselight over seven daily sessions. The outcome used for reinforcement was the same outcome that the houselight was initially trained to inhibit. Each session included eight presentations of the houselight, each reinforced 3 times on average, at random intervals over 60 s.

A test comparing responding with the houselight and key-light stimuli included six trials of reinforced houselight presentations intermixed with eight trials of nonreinforced key-light presentations. An additional two houselight presentations were unreinforced for direct comparison with the key light.

Data analysis

To examine responding at the end of Pavlovian training (Day 20), we planned a set of contrasts for probe trials of excitors and a randomly selected trial of each of the nonreinforced compounds. The first contrast compared the average responding over both probe trials of excitors with trials from both nonreinforced compounds. The second compared probes of each excitor and the third compared the trials of each nonreinforced compound.

For PIT tests, the average lever-pressing rate over the 1-min periods preceding each of the stimulus presentations was used as the response baseline. For PIT Test 1, baselines were calculated for each block, whereas in PIT Test 2, this was calculated for the entire test.

Single factor within-subjects analyses of variance (ANOVAs) evaluated the effects of stimulus presentations on lever-pressing rates in each test, with a separate analysis conducted for each block of PIT Test 1. We refer to the factor included in these ANOVAs as the “stimulus interval,” which includes three levels: the response rate during the ITI, “Excitors,” or “Inhibitors.” To compare responding during these three “stimulus intervals,” nonorthogonal contrasts were utilised. The test-wise error rate for planned contrasts was α = .05.

To test for inhibition in PIT Test 1, comparisons were made between the baseline response rate for Block 1 and responding during each of X and Y. Differences in the effects of X and Y were determined by comparing responding during each of these stimuli directly. Similarly, for Block 2, excitatory effects of A and B were determined by comparing each of these stimuli to baseline. Outcome-specific excitation was analysed by comparing responding during trials of A and B. No comparisons were made across blocks and so we only interpreted effects of stimuli on the baseline directly proximal to those stimuli.

PIT Test 2 tested moderation of excitation by putative inhibitors. We therefore conducted contrasts comparing the response rate during an excitor with that occurring during each of the congruent and incongruent compounds containing that excitor. To further test the specificity of responding to “Same” and “Different” excitors, we included a factorial structure in the ANOVA conducted on these stimuli. This structure had a “compound” factor with “excitor,” “congruent compound,” and “incongruent compound” levels, as well as a second factor of “stimulus type” (“Same” or “Different”).

In addition, to examine whether the specificity of excitatory PIT was consistent across each test, we conducted an ANOVA on responding during excitors, factorially comparing effects of test and of “Same” versus “Different” stimuli.

To examine discrimination between light stimuli, we compared average magazine entry rate during nonreinforced houselight probes with the average of matched key-light trials in the discrimination test. Analysed key-light trials were matched to houselight probes based on their order in the session. These response rates were also compared with the average rate in the minute preceding each of the four trials, which acted as a baseline response rate.

All ANOVAs and contrasts were calculated with partial eta-squared measures of effect size. In addition, 95% confidence intervals (CIs) were calculated for the mean difference in each planned contrast.

Results

Pavlovian and instrumental training

Rates of conditioned responding during probe trials on Day 20 are shown in Figure 1. A within-subjects ANOVA demonstrated a main effect of stimulus (F_4,44 = 24.64, p < .001, $η_{p}^{2}$ = .69). A contrast comparing the average rate during A and B with the rate during AX and BY showed that responding was greater during the excitors than during the nonreinforced compounds (F_1,11 = 67.50, p < .001, $η_{p}^{2}$ = .86, 95% CI = [7.84, 13.58]). In addition, responding during nonreinforced compounds was significantly different to the baseline response rate (F_1,11 = 9.51, p = .010, $η_{p}^{2}$ = .46, 95% CI = [0.76, 4.57]). A and B did not produce observably different rates of magazine entries (F_1,11 = 0.04, p = .843), nor did AX and BY (F_1,11 = 0.34, p = .570).

Figure 1.

Mean magazine entry rate during nonreinforced trials of A and B, and during nonreinforced compounds AX and BY on Day 20 of Pavlovian training.

Instrumental training proceeded as expected, with rats producing 6.91 ± 1.02 (M ± SEM) lever presses per minute by the end of instrumental training.

PIT Test 1

As shown in Figure 2, in Block 1 of this test, the stimuli trained as the “Same Inhibitor” and “Different Inhibitor” each suppressed responding relative to the Block 1 baseline. This description of the results was confirmed by a significant main effect of stimulus interval (F_2,22 = 8.13, p = .002, $η_{p}^{2}$ = .43). Contrasts indicated that during the “Same Inhibitor,” response rate significantly decreased from baseline (F_1,11 = 15.25, p = .002, $η_{p}^{2}$ = .58, 95% CI = [−3.00, −0.84]). Responding was also decreased during the “Different Inhibitor” (F_1,11 = 7.12, p = .022, $η_{p}^{2}$ = .39, 95% CI = [−2.97, −0.29]). There was no significant difference in responding between the putative inhibitory stimuli (F_1,11 = 0.48, p = .501; Figure 2).

Figure 2.

Instrumental responding during PIT Test 1. “Inhibitors” for the “Same” outcome (X) and the “Different” outcome (Y) show decreased responding compared with the baseline for Block 1. Excitors for the “Same” outcome (A) and for the “Different” outcome (B) show increased responding compared with baseline for Block 2. The excitor predicting the same outcome as the response had a greater effect on responding than one predicting a different outcome.

In Block 2, there was also an effect of stimulus interval (F_2,22 = 29.19, p < .001, $η_{p}^{2}$ = .73). As shown in Figure 2, contrasts on Block 2 responding showed a significant increase from baseline during the “Same Excitor” (F_1,11 = 54.94, p < .001, $η_{p}^{2}$ = .833, 95% CI = [4.22, 7.78]). An increase was also seen during the “Different Excitor” (F_1,11 = 48.08, p < .001, $η_{p}^{2}$ = .81, 95% CI = [2.13, 4.12]). However, responding during the “Different Excitor” was significantly less than that during the “Same Excitor” (F_1,11 = 8.33, p = .015, $η_{p}^{2}$ = .43, 95% CI = [0.68, 5.07]).

PIT Test 2

Results of this test are shown in Figure 3; an ANOVA revealed a significant effect of stimulus (F_6,66 = 5.76, p < .001). As was found in the first test, contrasts showed significant increases from baseline during the “Same Excitor” (F_1,11 = 36.73, p < .001, $η_{p}^{2}$ = .77, 95% CI = [3.42, 7.33]) and during the “Different Excitor” (F_1,11 = 14.48, p = .003, $η_{p}^{2}$ = .57, 95% CI = [1.32, 4.93]). However, the difference between excitors was marginal in this test (F_1,11 = 4.22, p = .064).

Figure 3.

Mean lever presses/minute during each stimulus or compound presented in PIT Test 2.

Both stimuli trained as inhibitors acted to reduce the effect of excitors when presented as compounds. Planned contrasts showed that responding was reduced in the “Same Congruent Compound” (AX) compared with the “Same Excitor” (A) (F_1,11 = 8.99, p = .012, $η_{p}^{2}$ = .45, 95% CI = [−7.08, −1.09]). The “Same Incongruent Compound” (AY) also produced reduced responding compared with the “Same Excitor” (A) (F_1,11 = 10.00, p = .009, $η_{p}^{2}$ = .48, 95% CI = [−5.65, −1.01]). Similarly, responding during the “Different Congruent Compound” (BY) was less than during the “Different Excitor” (B) (F_1,11 = 6.15, p = .031, $η_{p}^{2}$ = .36, 95% CI = [−5.03, −0.30]), as was responding during the “Different Incongruent Compound” (BX) (F_1,11 = 13.28, p = .004, $η_{p}^{2}$ = .55, 95% CI = [−4.28, −1.06]). Factorial analysis of Same and Different stimuli across levels of compound (no compound, incongruent, and congruent compounds) showed no interaction of stimulus type with compound (F_2,22 = 0.37, p = .700, $η_{p}^{2}$ = .03). In contrast to the difference seen in excitors alone, no main effect of stimulus type was seen over all compound types (F_1,11 = 3.29, p = .097, $η_{p}^{2}$ = .23).

In addition, analysis of responding during excitors in this test and the previous test indicated no main effect of test (F_1,11 = 1.67, p = .223, $η_{p}^{2}$ = .13) or interaction of test and stimulus identity (F_1,11 = 0.38, p = .550, $η_{p}^{2}$ = .03). In contrast to the marginal effect occurring within this test, the main effect of stimulus identity across both tests was significant, with greater responding occurring during the “Same excitor” than during the “Different Excitor” (F_1,11 = 7.83, p = .017, $η_{p}^{2}$ = .42).

Discrimination of visual stimuli

A two-way ANOVA indicated that responding during the houselight increased significantly across training (Figure 4a). There was a significant main effect of trial type such that the magazine entry rate during baseline periods was significantly less than that occurring during houselight trials (F_1,11 = 37.83, p < .001, $η_{p}^{2}$ = .78). There was also a main effect of training day (F_6,66 = 6.65, p < .001, $η_{p}^{2}$ = .38), and a Training Day × Trial Type interaction (F_6,66 = 5.95, p < .001, $η_{p}^{2}$ = .35).

Figure 4.

(a) Acquisition of magazine entry response during the houselight. (b) Test responding is the average magazine entry rate during two houselight probe trials, two matched key-light trials, and the average of the minutes preceding these trials as a baseline.

There was significant variation in responding during different stimuli in the discrimination test (F_2,22 = 32.95, p < .001, $η_{p}^{2}$ = .75). Responding during houselight probes was significantly greater than during matched key-light trials (F_1,11 = 15.61, p = .002, $η_{p}^{2}$ = .59, 95% CI = [1.90, 6.68]). Responding was greater than baseline during both the houselight (F_1,11 = 64.33, p < .001, $η_{p}^{2}$ = .85, 95% CI = [6.68, 11.74]) and during the key light (F_1,11 = 17.66, p = .001, $η_{p}^{2}$ = .62, 95% CI = [2.34, 7.49]) (Figure 4b).

Discussion

In PIT Test 1, both putative inhibitors reduced the instrumental response rate. Notably, the stimulus trained to inhibit Outcome 2 led to a reduction in instrumental responding for Outcome 1. Correspondingly, the excitor for Outcome 2 increased instrumental responding for Outcome 1, although to a lesser extent than the excitor paired with Outcome 1. Therefore, both putative inhibitors and excitors generated reductions in responding for either outcome, consistent with general PIT effects. The elevation of the response in the presence of both excitatory stimuli is consistent with Holland’s (2004) findings that when animals are trained to respond for only one of the Pavlovian outcomes, excitatory stimuli produced general PIT effects. Another result of interest, however, was the finding that the “Same Excitor” produced greater responding than the “Different Excitor.” Holland (2004) observed no evidence of outcome specificity in the effects of excitors on instrumental responding, and so our data differ in this regard.

A novel finding was that the excitors’ effects on responding in PIT Test 2 were reduced by joint presentation with either the “Same Inhibitor” or “Different Inhibitor,” indicating that the effect of putative inhibitors was not tied to the particular outcome or stimulus with which the “Inhibitor” was trained. This pattern of behaviour conforms to the hypothesised general PIT effect. Furthermore, prior studies have not tested whether an outcome-general excitatory PIT effect can be moderated by the presentation of an inhibitory stimulus; our result may therefore for be the first such observation.

The absence of a significant difference in responding during X− and Y− could be explained by a failure to discriminate between the two visual stimuli. Further training of the houselight after PIT tests resulted in differential conditioned responding to the houselight and key light at test. This indicates that the animals could treat the stimuli as perceptually distinct events. Although a difference in magazine entry rates was observed between the stimuli, the key light, which had not received any training, also produced elevations in magazine entry rates from baseline. This may have been due to partial generalisation of excitatory conditioning from the houselight. The precise extent of generalisation cannot be determined as the amount of conditioned responding produced by this stimulus before retraining of the houselight was not assessed. Nonetheless, there is evidence that rats can discriminate between these stimuli.

Experiment 2

In Experiment 1, where only one response–outcome contingency was trained, we did not observe any evidence of outcome-specific inhibitory PIT. This corresponds to Holland’s (2004) findings with excitatory stimuli: stimuli predicting the occurrence or absence of Outcome 2 influenced the response trained with Outcome 1. In a second experiment, Holland (2004) also observed that training a separate response earning Outcome 2, in addition to an independent response earning Outcome 1, changed the nature of the PIT observed, leading to outcome-specific excitatory PIT. To examine whether this additional instrumental contingency would increase the specificity of any potential inhibitory PIT in our procedure, we trained animals with the Pavlovian contingencies used in Experiment 1 (A → O1, B → O2, AX → no O1 and BY → no O2), as well as two separate responses for Outcome 1 and Outcome 2 (R1O1, R2O2). We predicted that stimuli associated with these outcomes would now show specific PIT affecting only the response associated with a common outcome. Assessing the effects of all four stimuli on each response individually already necessitated two tests and so we did not carry out additional summation tests as in Experiment 1. We thereby aimed to avoid excessive extinction or learning about nonreinforced test scenarios affecting results.

Method

Subjects and apparatus

Twelve male experimentally naive Long Evans rats were utilised and were housed under the same conditions as those in Experiment 1. At the start of food deprivation, rats weighed 509 (SD = 19) g. Apparatus was as for Experiment 1.

Procedure

Preexposure

Pavlovian training and the Pavlovian reminder session were as described in Experiment 1. The design of Experiment 2 is summarised in Table 2.

Table 2.

Experiment 2 design summary.

Pavlovian	Instrumental	PIT test
A → O1	R1 → O1	R1; Same Inhibitor (X), Different Inhibitor (Y) Same Excitor (A), Different Excitor (B)
AX → nothing	R2 → O2
B → O2		R2;Same Inhibitor (Y), Different Inhibitor (X)Same Excitor (B), Different Excitor (A)
BY → nothing

Note. PIT = Pavlovian-instrumental transfer. A, B = white noise or clicker; X, Y = houselight or key lights; O1, O2 = sucrose or pellets; R1, R2 = left or right lever.

Instrumental training

Two instrumental conditioning sessions occurred daily for 9 days, one for each response–outcome pair. The order of sessions was varied across days and there was a delay of at least 90 min between sessions. Response 1 was reinforced with Outcome 1 and Response 2 with Outcome 2. Right and left levers served as the responses, counterbalanced between rats. Outcomes were either sucrose solution or pellets and were also counterbalanced. The schedules of reinforcement were the same as in Experiment 1.

PIT test

The effects of A, B, X and Y on each lever were assessed in two single-lever tests. Single-lever tests were used to provide a clear assessment of the impact of each stimulus on each response without possible response competition between the two levers (e.g., excitatory effects on R1 could decrease performance of R2 but this would not necessarily reflect inhibition). These tests occurred over 2 days and the response tested on each day was counterbalanced. Each test occurred as for PIT Test 1 of Experiment 1.

Data analysis

Pavlovian training data were treated as previously described while the addition of a second response to the design necessitated a different treatment of test data. As each of the stimuli in this design had symmetrical relationships with responses, responding was collapsed within these relationships. Therefore, we created names for these categories of stimuli, which refer to the training of the stimuli intended to generate excitors or inhibitors, as well as whether the response outcome was the same as or different to the outcome the stimuli were trained with. Note that these labels used for convenience reflect the training history of the stimuli and we later discuss the evidence as to whether the stimuli have in fact taken on excitatory or inhibitory properties. Responding on R1 during A was collapsed with responding on R2 during B and categorised as “Same Excitor.” R1 response rates during B and R2 rates during A were collapsed and categorised as “Different Excitor.” R1 during X was collapsed with R2 during Y forming the “Same Inhibitor” category, and R1 during Y was collapsed with R2 during X as the “Different Inhibitor” category.

As for PIT Test 1 of Experiment 1, responding during all cue categories was compared with baselines for the relevant block to indicate whether they resulted in inhibition or excitation. The “Same” and “Different Inhibitor” cues were compared, as were “Same” and “Different excitor” cues, to indicate the outcome specificity of cue effects. These comparisons were carried out as described in Experiment 1.

Results

Pavlovian and instrumental training

Pavlovian training produced conditioned responding in probe trials as shown in Figure 5. Differences were detected across stimuli, as indicated by a within-subjects ANOVA (F_4,44 = 64.24, p < .001, $η_{p}^{2}$ = .85). Planned contrasts indicated that responding collapsed across excitors A and B was greater than the rate occurring across nonreinforced compounds AX and BY (F_1,11 = 87.78, p < .001, $η_{p}^{2}$ = .89, 95% CI = [11.79, 19.04]). Responding to nonreinforced compounds was not significantly different to the baseline response rate (F_1,11 = 4.75, p = .052).

Figure 5.

Mean (±SEM) magazine entry rate during probe trials on Day 20 of Pavlovian training for Experiment 2. Rates for excitors (A and B), nonreinforced compounds (AX and BY), and baseline responding are shown.

Instrumental training produced 19.74 ± 4.54 lever presses per minute on R1 and 14.19 ± 1.36 lever presses per minute on R2.

PIT test

The PIT Test resulted in responding (depicted in Figure 6) which showed a significant effect of stimulus interval during Block 1 (F_2,22 = 17.15, p < .001, $η_{p}^{2}$ = .61). Contrasts on Block 1 responding showed that during the “Same Inhibitor,” the response rate was less than baseline (F_1,11 = 22.55, p = .001, $η_{p}^{2}$ = .67, 95% CI = [−8.34, −3.06]), as it was during the “Different Inhibitor” (F_1,11 = 31.39, p < .001, $η_{p}^{2}$ = .74, 95% CI = [−8.72, −3.80]). There were no observable differences in responding between “Inhibitor” stimuli (F_1,11 = 0.21, p = .656).

Figure 6.

Mean lever-pressing rate during stimuli in Test 1 of Experiment 2. “Same Inhibitor” stimuli are X for R1 and Y for R2, and “Different Inhibitor” stimuli are Y for R1 and X for R2. “Same” excitors are A for R1 and B for R2. “Different” excitors are B for R1 and A for R2.

Similarly, in Block 2, an effect of stimulus interval was observed (F_2,22 = 24.24, p < .001, $η_{p}^{2}$ = .69). Contrasts on Block 2 responding showed increased responding during the “Same Excitor” compared with baseline (F_1,11 = 47.82, p < .001, $η_{p}^{2}$ = .81, 95% CI = [5.06, 9.78]) as well as the “Different Excitor” (F_1,11 = 25.65, p < .001, $η_{p}^{2}$ = .70, 95% CI = [2.79, 7.08]). The difference between “Same” or “Different” excitors was marginal (F_1,11 = 4.30, p = .062).

Discussion

Our observation of significant reductions from baseline responding during both stimuli trained as inhibitors and the lack of difference between them was consistent with an outcome-general PIT effect, which we had predicted to occur in Experiment 1. That this nonspecific effect of putative inhibitors occurred despite variations in instrumental training between experiments indicates that the number of response–outcome associations that are trained, and in particular, whether each of the outcomes associated with Pavlovian conditioned stimuli (CSs) is also earned by a distinct response may not be an important determinant of the specificity of inhibitory PIT effects. However, an aspect of the current data that limits this conclusion is the fact that we did not observe a robust outcome-specific excitatory PIT effect. Both “Same” and “Different” excitors elevated responding and only a marginal difference was seen between their effects. We expected outcome-specific PIT based on Holland’s (2004) finding that two instrumental response–outcome contingencies caused specific PIT and other similar demonstrations with excitatory stimuli in the literature (Corbit et al., 2001; Holland, 2004). The lack of outcome-specific excitatory PIT may indicate the influence of some unexpected factor in our design that increases general PIT effects.

Experiment 3

Adding an additional response–outcome contingency to our design in Experiment 2 was expected to increase specificity of transfer effects based on the findings of Holland (2004). Outcome-specific inhibition would also be predicted following Laurent and Balleine (2015). In their study, specific inhibitory PIT effects indicate that inhibitory training had generated learning about the absence of specific outcomes and removed the need for controls for nonassociative effects of the relevant stimuli. Here, without the predicted specific PIT effects, we do not have any direct evidence that the stimuli trained as inhibitors had acquired inhibitory properties, as outcome-general reductions in instrumental responding may have been due to disruptive processes. Consequently, we have qualified all references to inhibition throughout this article to acknowledge this lack of direct evidence of inhibition although the labels “Same Inhibitor” and “Different Inhibitor” are still used to reflect the treatment of different stimuli in training. Nonetheless, in an attempt to address the unexpected differences to the results of Laurent and Balleine (2015), Experiment 3 tested procedural differences that may have accounted for the lack of outcome-specific inhibitory PIT in our findings.

One way our methodology diverged from Holland (2004) was that we implemented many more Pavlovian training sessions (20 vs. the 8 days used by Holland). This training was implemented to ensure that animals fully discriminated reinforced stimuli from nonreinforced compounds. However, increased Pavlovian training has been shown to occlude PIT effects by generating competition between Pavlovian and instrumental responses (Holmes et al., 2010), and to the extent that this competition was greater following our training, it could interfere with responding in a general manner, whereas more limited training, and less interference, may reveal outcome-specific inhibitory effects of the stimuli. Experiment 3 sought to determine whether the design utilised in Experiment 2 could produce outcome-specific transfer after only 10 sessions of Pavlovian training which is more typical of PIT experiments in the literature and more consistent with Holland (2004).

A further variable that differed from past findings of outcome-specific inhibitory PIT was the type of test used. Our experiments used single-lever tests while other studies utilised choice tests (Laurent & Balleine, 2015). In a choice test, any competition from the Pavlovian magazine should be equivalent for the two instrumental responses and so this type of test may be more sensitive in detecting outcome-specific effects. Furthermore, choice tests allow for competition between instrumental responses and include the stimuli inherent in the presence of the additional response lever, both factors which could cause variation from results observed in single-lever tests. Although it has been argued that response competition does not account for the results of particular studies (e.g., Quail et al., 2017, in reference to Laurent & Balleine, 2015), PIT generated in single-lever versus choice tests has not been directly compared. Therefore, the present experiment included two tests, one single-lever and one choice test, to allow for comparison between them.

Method

Subjects and apparatus

As in Experiments 1 and 2, 12 experimentally naive male Long Evans rats were the subjects. Rats initially weighed 411 g on average (SD = 8.94). All animals were handled daily for 7 days prior to training and animals were otherwise housed and maintained in the same conditions as rats in Experiments 1 and 2. Apparatus was also as for prior experiments.

Procedure

All procedures are summarised in Table 3.

Table 3.

Experiment 3 design summary.

Pavlovian	Instrumental	PIT Test 1	PIT Test 2 (choice)
A → O1	R1 → O1	R1; X, Y, A, B	R1, R2; X, Y, A, B
AX → nothing	R2 → O2	R2; X, Y, A, B
B → O2		For R1: Same Inhibitor = X, Different Inhibitor = YSame Excitor = A, Different Excitor = B
BY → nothing		For R2: Same Inhibitor = Y, Different Inhibitor = XSame Excitor = B, Different Excitor = A

Note. PIT = Pavlovian-instrumental transfer. A, B = white noise or clicker; X, Y = houselight or key lights; O1, O2 = sucrose or pellets; R1, R2 = left or right lever. Categorisation of stimulus–response relationships is shown as pertains to both PIT tests.

Pavlovian training

Training parameters were retained from Experiments 1 and 2; however, only 10 training sessions were carried out, half of that used in previous experiments.

Instrumental training

Responding on two separate levers for Outcomes 1 and 2 (sucrose or pellets) was established as in Experiment 2. After instrumental training, a day of Pavlovian training was carried out as a reminder session.

PIT Test 1: single levers

The effect of stimuli A, B, X, and Y on each response was measured across two single-lever tests using the same procedure as Experiment 2.

Retraining

Four days of Pavlovian retraining were completed. This was approximately half that used in Experiment 1.

PIT Test 2: two-lever choice

This test examined the effect of the stimuli tested in PIT Test 1 under choice test conditions. Stimulus presentations followed the blocked structure of “Inhibitor” and “Excitor” stimuli used in PIT Test 1. The order of stimuli was changed to key light, houselight, houselight, key light for Block 1 and click, noise, noise, click for Block 2. Only one session of this test was completed in which both response levers were available. The responding on each lever was measured throughout the session. Other parameters were retained from previous tests.

Data analysis

Training data were examined as described for prior experiments, as were PIT test data. Responding in each test was collapsed into four categories based on the factors of “Inhibitor” versus “Excitor” and “Same” versus “Different” outcome (see Table 3 for more detail). Statistical analyses were consistent with Experiment 2, for both tests in this experiment. Responding was collapsed across response levers for both single-lever and choice tests. As in Experiment 1, the consistency of specific excitatory PIT across both tests was analysed using an ANOVA with test and stimulus type (“Same” vs. “Different”) as factors.

Results

Pavlovian and instrumental training

Probes on the final day of Pavlovian training showed significant differences in response rates during the various stimuli (F_4,44 = 9.72, p < .001, $η_{p}^{2}$ = .47) as shown in Figure 7. Responding to probe trials of both excitors was significantly greater than that for trials of both unreinforced compounds (F_1,11 = 11.77, p = .006, $η_{p}^{2}$ = .52, 95% CI = [2.09, 9.58]). In addition, responding to nonreinforced compounds was significantly greater than during baseline responding (F_1,11 = 29.88, p < .001, $η_{p}^{2}$ = .73, 95% CI = [6.75, 15.84]).

Figure 7.

Magazine entry rate (mean ± SEM) during probe trials on Day 10 of Pavlovian training for Experiment 3. Rates for excitors (A and B), nonreinforced compounds (AX and BY), and baseline responding are shown.

PIT Test 1: single levers

Response rates during Block 1 of the first PIT test are shown in Figure 8 and indicated a significant effect of stimulus interval (F_2,22 = 9.12, p = .001, $η_{p}^{2}$ = .45). Contrasts on the first block of trials in this test indicated that responding was reduced below baseline by both the “Same Inhibitor” (F_1,11 = 10.42, p = .008, $η_{p}^{2}$ = .49, 95% CI = [−2.19, −0.41]) and the “Different Inhibitor” (F_1,11 = 14.55, p = .003, $η_{p}^{2}$ = .57, 95% CI = [−2.45, −0.66]). However, there was no difference in response rate during each stimulus (F_1,11 = 0.49, p = .500).

Figure 8.

Response rates to stimuli in PIT Test 1 of Experiment 3. “Inhibitor” stimuli for the “Same” outcome as responding are X for R1 and Y for R2, while those for “Different” outcomes are Y for R1 and X for R2. Excitors for the “Same” outcome as responding are A for R1 and B for R2, those for the other “Different” outcome are B for R1 and A for R2.

There were also significant differences across stimulus intervals during the second block of the test (F_2,22 = 68.15, p < .001, $η_{p}^{2}$ = .86). During the second block of trials, responding was increased above baseline during the “Same Excitor” (F_1,11 = 89.75, p < .001, $η_{p}^{2}$ = .89, 95% CI = [5.32, 8.54]) and during the “Different Excitor” (F_1,11 = 92.64, p < .001, $η_{p}^{2}$ = .89, 95% CI = [4.25, 6.77]). Furthermore, responding during the “Same Excitor” was significantly greater than during the “Different Excitor” (F_1,11 = 6.34, p = .029, $η_{p}^{2}$ = .37, 95% CI = [0.18, 2.66]).

PIT Test 2: two-lever choice

There was evidence of effects of stimulus interval in Block 1 of this test (F_2,22 = 28.32, p < .001, $η_{p}^{2}$ = .72; data shown in Figure 9). “Inhibitor” stimuli were presented in the first block of trials, and reductions from baseline response rate were shown through contrasts for both the “Same Inhibitor” (F_1,11 = 28.84, p < .001, $η_{p}^{2}$ = .72, 95% CI = [−3.14, −1.32]) and the “Different Inhibitor” (F_1,11 = 33.36, p < .001, $η_{p}^{2}$ = .75, 95% CI = [−3.22, −1.44]). Response rates during each of these stimuli were not observably different (F_1,11 = 0.33, p = .576).

Figure 9.

Response rates to stimuli in Test 2 of Experiment 3. “Inhibitor” stimuli for the “Same” outcome as responding are X for R1 and Y for R2, while those for “Different” outcomes are Y for R1 and X for R2. “Excitor” stimuli for the response outcome are A for R1 and B for R2, those for the different outcome are B for R1 and A for R2.

Responding over stimulus intervals also varied in Block 2 (F_2,22 = 30.03, p < .001, $η_{p}^{2}$ = .73). Responding was increased from baseline by both the “Same Excitor” (F_1,11 = 129.21, p < .001, $η_{p}^{2}$ = .92, 95% CI = [3.63, 5.37]) and during the “Different Excitor” (F_1,11 = 13.57, p = .004, $η_{p}^{2}$ = .55, 95% CI = [0.85, 3.36]). Responding was significantly greater for the “Same Excitor” than for the “Different Excitor” (F_1,11 = 10.84, p = .007, $η_{p}^{2}$ = .50, 95% CI = [0.79, 4.0]). Analysis of excitors across Test 1 and Test 2 indicates that responding during the “Same excitor” was significantly higher than the “Different excitor” (F_1,11 = 14.84, p = .003, $η_{p}^{2}$ = .57). The interaction effect did not show any variation in this difference across tests (F_1,11 = 1.34, p = .271, $η_{p}^{2}$ = .11). However, responding over both stimuli had reduced significantly in Test 2 compared with Test 1 (F_1,11 = 24.12, p < .001, $η_{p}^{2}$ = .69).

Discussion

We consistently observed evidence of specific excitatory PIT in this experiment, with responding during “Same” excitors greater than during “Different” excitors. However, responding during “Different” excitors was still greater than baseline, indicating that some degree of general transfer also occurred. Despite specific excitatory effects, putative inhibitors still produced general reductions in response rate, as was seen in previous experiments. These effects were observed in both choice and single-lever tests, indicating that our findings are not simply the result of this design feature.

General discussion

All stimuli trained as inhibitors in this study were observed to decrease instrumental responding for rewards. This effect occurred regardless of whether the stimuli were trained to predict the absence of the same or different outcome as was earned by the response(s) being tested. Putative inhibitors also reduced the effects of excitors when these stimuli were presented together as compounds, and again, this reduction was observed regardless of whether the “Inhibitor” was tested with the stimulus it was paired with in training or with the alternate excitor that predicted a distinct outcome. A possible interpretation of these results is that stimuli trained as conditioned inhibitors acquired general inhibitory properties that allowed suppression of instrumental responding for reward regardless of the specific identity of the reward earned by the lever or predicted by the excitatory stimulus. Thus, contrary to our predictions, at least with the parameters used here, training with one versus two R-O contingencies did not alter the nature of PIT observed.

When animals were trained with a single R-O relationship, as predicted, we observed that stimuli paired with either the same or different outcome as the lever elevated responding. However, despite general excitatory effects of both excitors, the stimulus paired with the same outcome as the response had a yet greater effect on responding than the stimulus that predicted a different outcome, evidence of specific excitatory PIT which was not expected. Training with two R-O contingencies did not change this pattern; we observed (now predicted) specific excitatory effects (on top of a similar overall excitatory effect of both excitors) and only outcome-general effects of the putative inhibitors. Thus, in our hands, the number of R-O contingencies trained does not alone account for whether observed PIT is outcome-specific or outcome-general.

Our results for excitatory stimuli are not entirely consistent with those of Holland (2004); the PIT we observed had both outcome-specific and general excitatory properties that changed very little according to whether one or two R-O contingencies were trained. Of note, we observed behaviour consistent with general inhibitory PIT both when animals were trained with a single response–outcome contingency and when they were trained with two distinct response–outcome contingencies. Therefore, we did not observe the same pattern of results described by Holland (2004) for excitatory stimuli, where general PIT effects occurred in animals trained with only one response–outcome contingency but the specificity of PIT increased with the addition of a second response–outcome association. In fact, counter to our predictions, the specificity of the excitatory stimuli was not clearly defined by the number of response–outcome contingencies trained; in Experiment 1, where a single R-O contingency was trained, despite robust transfer generated by the “Different” stimulus, the “Same” stimulus produced yet higher responding. Furthermore, in Experiment 2, where two R-O contingencies were trained, the Same-Different comparison was only marginal, although reliable effects were observed in both tests from Experiment 3, where two R-O contingencies were trained but the duration of the Pavlovian training was shortened.

It is possible that the block structure of our tests, where the excitors were tested in the second half of the test where baseline responding was particularly low, made it relatively easy to detect excitatory effects of the “Different” stimuli. Alternatively, the unexpected effects of “Different” excitors may have been due to partial generalisation across either the stimuli or outcomes trained for the “Same” excitor. Any degree of such generalisation could have produced apparently general PIT but due to imperfect discrimination rather than recruitment of a separate arousal process. This may have occurred to a greater extent in our study than in Holland (2004), as our stimuli were trained in a single daily session with trials of stimuli intermixed. Contrastingly, Holland (2004) carried out separate training sessions for each stimulus.

An important limitation to our observations of inhibitors’ effects is that we did not provide independent evidence of inhibitory conditioning. Comparison of the effects of stimuli X− and Y− used here with those of an unreinforced control stimulus either in summation or a retardation of acquisition test as described by Rescorla (1969) would allow for direct observation of whether these stimuli had acquired inhibitory properties. Furthermore, inclusion of novel or briefly preexposed (or other form of uninformative control) stimulus in the PIT procedures of this study would provide an opportunity to demonstrate that X and Y could reduce behaviour to a greater extent than a stimulus lacking a history of inhibitory conditioning (as seen in Alarcón & Bonardi, 2016; Quail et al., 2017). This would evidence the role of the inhibitory association in generating a behavioural outcome, beyond the that of the disruptive effect of a nonexcitatory stimulus.

A control stimulus would not have been needed for the interpretability of our results, had we seen the expected outcome-specific inhibitory PIT in Experiments 2 and 3. Specific PIT would have indicated that the content of inhibitory learning was affecting behaviour. In addition, although we removed compound PIT tests from Experiments 2 and 3 to avoid additional extinction of instrumental behaviour, such tests would have allowed for observation of specific inhibitory PIT effects of X− and Y− in compound form. Such findings would indicate that reductions in behaviour were not due to the disruptive effects of novel presentations of inhibitory stimuli on their own, as well as ruling out potential learning that stimulus compounds were not reinforced. Any such learning could have caused the similar responding to congruent and incongruent compounds seen in Experiment 1. Although had animals relied on such a rule, this would not explain why stimuli trained as inhibitors in compounds consistently produced effects when presented alone in all three experiments.

Our results for the stimuli trained as inhibitors contrast those of Laurent and Balleine (2015) in several ways. Across three experiments, we consistently observed that these stimuli suppressed instrumental responding below baseline and did so whether the action had been trained with the same or a different outcome than that which the stimulus predicted the absence of. In addition, we failed to see any evidence that the putative inhibitors promoted performance of alternative actions. We also observed that the “Inhibitor” stimuli reduced responding to the excitors when presented in compound with either the stimulus they were originally trained with (e.g., AX−; congruent condition) or with the alternate stimulus (BX−; incongruent condition) providing further evidence of outcome-general effects.

The differences between the current results and those of Laurent and Balleine (2015) are somewhat surprising as our training procedures were very similar. However, a methodological difference may account for these findings. In Pavlovian training, Laurent and Balleine (2015) used a 3 kHz pure tone and houselight as inhibitors, whereas we trained the houselight and key lights as inhibitory stimuli out of concern over how two auditory stimuli would be perceived in compound. It is possible that the tone and houselight were better discriminated leading to more specific effects. Data displayed in Figure 4 suggest that animals were able to discriminate the houselight from the key lights, but this parameter would be one to explore in future experiments.

The ability to detect outcome-specific effects may have been limited by a behavioural floor if the effect were to mirror that seen with the excitors; if outcome-specific reductions in responding should be expected to be in addition to an outcome-general effect, this might have been difficult to detect. In addition, for Experiment 1 and 3, there was evidence that responding during unreinforced compounds was greater than baseline which may indicate that inhibition was incomplete. Therefore, any inhibitory effects on behaviour would not be equivalent in magnitude to those of excitors and may therefore require more power to detect any specific PIT effects.

Another possible explanation for the lack of outcome-specific inhibition is that animals generalised between stimuli. However, behaviour consistent with general inhibitory effects was observed under conditions where specific excitatory effects were observed; while in all tests, the so-called “Different” stimulus elevated responding from baseline indicating some excitatory properties, potentially due to generalisation, the “Same” stimulus consistently had a yet greater excitatory effect on responding. Thus, animals must have encoded the distinct properties of the different outcomes and discriminated between at least some of the stimuli (i.e., the excitors), yet this did not appear to be reflected in the effects of the stimuli trained as inhibitors. While we demonstrated (Figure 4) that the animals can discriminate between the visual stimuli, they may not have done so during the initial training or PIT testing.

A further explanation could be the nature of encoding during inhibitory learning. This was evident in Quail et al. (2017) who carried out a PIT study in humans, meaning that individuals could be asked to describe the Pavlovian contingencies they experienced in training. On the basis of what they reported, participants could be split into groups of specific learners (individuals reporting that inhibitors predicted that specific outcomes would be absent, for example, X predicted no M&Ms) and general learners (those reporting that inhibitors predicted the absence of any outcome, for example, X predicted no reward). Although the excitatory PIT effects were similar and specific in both groups, the inhibitory PIT effects differed with specific inhibitory PIT only observed for individuals who had reported specific inhibitory relationships following Pavlovian training. The general learners demonstrated general inhibitory PIT. Therefore, if rats in the current series had associated putative inhibitors with the absence of outcomes in general, then this learning would have been expected to produce general inhibitory PIT. Future animal experiments could address this possible explanation by assessing whether Pavlovian conditioned responses to excitors are affected specifically by inhibitors that were trained with the same outcome. Such a method follows the summation test procedure described by Rescorla (1969).

It should also be noted that the literature examining the outcome specificity of conditioned inhibition itself is small and reports mixed results. Some of these studies used conditioned suppression as their measure and so in that regard are not unlike PIT but the motivational systems supporting the Pavlovian and instrumental learning varied dramatically and rarely were multiple instrumental contingencies trained. In one example, Rescorla and Holland (1977) observed that the inhibitory effects of stimulus X trained with stimulus A (in an A+/AX− design) transferred to another excitor B that had been trained with the same unconditional stimulus (US). Such transfer of inhibition was not seen if A and B were trained with aversive versus appetitive USs. Thus, inhibition transferred across different stimuli predicting the same US but not if the stimuli predicted USs from different domains (appetitive vs. aversive). That study did not examine transfer across stimuli trained with different USs within a single domain. A study by Nieto (1984) did this by examining transfer of conditioned inhibition across different aversive USs and found good evidence of transfer of inhibition (see also Pearce et al., 1981). However, an early experiment by Lolordo (1967) found panel pressing to avoid shock was promoted by a CS+ for either shock or a loud noise, whereas only a CS− predicting the absence of shock reduced avoidance responding and a CS− for the incongruent noise US was without effect providing some evidence of outcome-specific inhibition. Thus, it appears that conditioned inhibition can vary in its specificity and specific PIT can only be expected if the original inhibitory learning is outcome-specific.

As noted in the rationale for Experiment 3, a further factor that may have been relevant to our observation of general PIT effects was the duration of Pavlovian training in Experiments 1 and 2. In Experiment 3, we found that a reduced number of Pavlovian training sessions produced an outcome-specific transfer effect for the excitors, but a general effect was still observed for the stimuli trained as inhibitors. It is worth noting that Laurent and Balleine (2015) observed outcome-specific effects with the duration of Pavlovian conditioning that we implemented in Experiments 1 and 2. Furthermore, decreasing the duration of Pavlovian training in Experiment 3 did not eliminate the effect of “different excitors” on responding, an effect not observed by Laurent and Balleine (2015). Therefore, other factors likely contribute to our findings.

An alternative account of our results is that a factor in our experiment led animals to associate stimuli with features common to sucrose and pellets. Such features could include their relevance to the motivational state of hunger (Konorski, 1967). This is significant, as the motivational content of an outcome can determine PIT effects produced by stimuli associated with that outcome. For example, Balleine (1994) trained stimuli to predict either sucrose solution or food pellets in hungry animals. Following this, animals were shifted to state of thirst rather than hunger and an instrumental response was trained with water as the outcome. Half of the animals were tested hungry and the other half were tested thirsty. Under hunger, both stimuli produced PIT (an outcome-general effect). However, only the stimulus paired with sucrose solution increased responding under thirst. These results were interpreted as stemming from sucrose and pellet’s common relevance to hunger (due to their nutritive value) and their contrasting relevance to thirst (due to only sucrose solution containing water). Therefore, PIT tests in distinct motivational states can be used to draw out the learned motivational relevance of Pavlovian stimuli.

Manipulations of motivational state could therefore determine whether outcome-general PIT effects are the result of reinforcers’ shared motivational properties. Both outcomes in the current experiment (sucrose solution and food pellets) were relevant to hunger. However, if training and/or tests had been carried out under water deprivation, based on the findings of Balleine (1994), only sucrose-associated cues would be expected to elevate responding under thirst. Such an experiment could help determine whether outcome-general effects in this study were based on learning about motivational properties of outcomes and distinguish outcome-general PIT effects from generalisation between the stimuli themselves. It would also provide a novel test of whether inhibitory PIT effects reflect relevance to particular motivational states. Furthermore, general PIT, which on this account relies on activation of a motivational state, has been shown to be reduced when that motivation is reduced; in a three-stimulus paradigm that generates both outcome-specific and general PIT, a shift from hunger to satiety reduced general PIT but outcome-specific effects remained (Corbit et al., 2007). A similar motivational manipulation could be imposed on the design used here to determine whether a motivational shift removes the apparently general inhibitory influence of stimuli.

Finally, it should be noted that our conclusions about the specificity of apparent inhibition effects applies only to conditioned inhibition established using the feature negative (A+/AX−) paradigm employed here. Other procedures such as backwards conditioning (Delamater et al., 2003) or discriminative conditioning may produce different learning that would transfer in a different way.

Although there are inconsistencies between this study and some previous demonstrations of outcome-specific inhibitory PIT, only a small number of studies have specifically investigated the ability of inhibitors to produce PIT. The varied results indicate that animals, including humans, can use information about the absence of outcomes in a number of ways, including to reduce behaviours earning that particular outcome, to invigorate responses earning alternative outcomes, or as in the present study, to reduce reward-seeking behaviours in general. Future research should determine whether the stimulus’s associations with the motivational properties of outcomes led to this finding.

Footnotes

Acknowledgements

We acknowledge the assistance of Robert A. Boakes in editing an earlier version of this article.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Lindsay J Kemp

References

Alarcón

Bonardi

(2016). The effect of conditioned inhibition on the specific Pavlovian-instrumental transfer effect. Journal of Experimental Psychology: Animal Learning and Cognition, 42(1), 82–94. https://doi.org/10.1037/xan0000087

Balleine

B. W.

(1994). Asymmetrical interactions between thirst and hunger in Pavlovian-instrumental transfer. The Quarterly Journal of Experimental Psychology Section B, 47(2), 211–231. https://doi.org/10.1080/14640749408401357

Colagiuri

Lovibond

P. F.

(2015). How food cues can enhance and inhibit motivation to obtain and consume food. Appetite, 84, 79–87.

Corbit

L. H.

Balleine

B. W.

(2005). Double dissociation of basolateral and central amygdala lesions on the general and outcome-specific forms of Pavlovian-instrumental transfer. Journal of Neuroscience, 25(4), 962–970. https://doi.org/10.1523/JNEUROSCI.4507-04.2005

Corbit

L. H.

Balleine

B. W.

(2016). Learning and motivational processes contributing to Pavlovian-instrumental transfer and their neural bases: Dopamine and Beyond. Current Topics in Behavioral Neurosciences, 27, 259–289.

Corbit

L. H.

Janak

P. H.

Balleine

B. W.

(2007). General and outcome-specific forms of Pavlovian-instrumental transfer: The effect of shifts in motivational state and inactivation of the ventral tegmental area. European Journal of Neuroscience, 26(11), 3141–3149. https://doi.org/10.1111/j.1460-9568.2007.05934.x

Corbit

L. H.

Muir

J. L.

Balleine

B. W.

(2001). The role of the nucleus accumbens in instrumental conditioning: Evidence of a functional dissociation between accumbens core and shell. Journal of Neuroscience, 21(9), 3251–3260. https://doi.org/10.1523/jneurosci.21-09-03251.2001

Delamater

A. R.

Lolordo

V. M.

Sosa

(2003). Outcome-specific conditioned inhibition in Pavlovian backward conditioning. Learning & Behavior, 31(4), 393–402.

Estes

W. K.

Skinner

B. F.

(1941). Some quantitative properties of anxiety. Journal of Experimental Psychology, 29(5), 390–400.

10.

Holland

P. C.

(2004). Relations between Pavlovian-instrumental transfer and reinforcer devaluation. Journal of Experimental Psychology: Animal Behavior Processes, 30(2), 104–117. https://doi.org/10.1037/0097-7403.30.2.104

11.

Holmes

N. M.

Marchand

A. R.

Coutureau

(2010). Pavlovian to instrumental transfer: A neurobehavioural perspective. Neuroscience & Biobehavioral Reviews, 34(8), 1277–1295. https://doi.org/10.1016/j.neubiorev.2010.03.007

12.

Konorski

(1967). Integrative activity of the brain: An interdisciplinary approach. University of Chicago Press.

13.

Laurent

Balleine

B. W.

(2015). Factual and counterfactual action-outcome mappings control choice between goal-directed actions in rats. Current Biology, 25(8), 1074–1079. https://doi.org/10.1016/j.cub.2015.02.044

14.

Lolordo

V. M.

(1967). Similarity of conditioned fear responses based upon different aversive events. Journal of Comparative and Physiological Psychology, 64(1), 154–158.

15.

Nieto

(1984). Transfer of conditioned inhibition across different aversive reinforcers in the rat. Learning and Motivation, 15(1), 37–57.

16.

Pearce

J. M.

Montgomery

Dickinson

(1981). Contralateral transfer of inhibitory and excitatory eyelid conditioning in the rabbit. Quarterly Journal of Experimental Psychology, 33(1), 45–61.

17.

Quail

S. L.

Laurent

Balleine

B. W.

(2017). Inhibitory Pavlovian–instrumental transfer in humans. Journal of Experimental Psychology: Animal Learning and Cognition, 43(4), 315–324.

18.

Rescorla

R. A.

(1969). Pavlovian conditioned inhibition. Psycho-logical Bulletin, 72(2), 77–94. https://doi.org/10.1037/h0027760

19.

Rescorla

R. A.

Holland

P. C.

(1977). Associations in Pavlovian conditioned inhibition. Learning and Motivation, 8(4), 429–447.

20.

Rescorla

R. A.

Lolordo

V. M.

(1965). Inhibition of avoidance behavior. Journal of Comparative and Physiological Psychology, 59(3), 406–412.

21.

Rescorla

R. A.

Solomon

R. L.

(1967). Two-process learning theory: Relationships between Pavlovian conditioning and instrumental learning. Psychological Review, 74(3), 151–182. https://doi.org/10.1037/h0024475

22.

Solomon

R. L.

Turner

L. H.

(1962). Discriminative classical conditioning in dogs paralyzed by curare can later control discriminative avoidance responses in the normal state. Psychological Review, 69(3), 202–218.

23.

Trapold

M. A.

Overmier

J. B.

(1972). The second learning process in instrumental learning. In Black

A. H.

Prokasy

W. F.

(Eds.), Classical conditioning II: Current theory and research. Appleton-Century-Crofts.