Psychon Bull Rev DOI 10.3758/s13423-013-0558-1
BRIEF REPORT
Experiencing extinction within a task makes nonextinguished information learned within a different task context-dependent Rodolfo Bernal-Gamboa & Juan M. Rosas & José E. Callejas-Aguilera
# Psychonomic Society, Inc. 2013
Abstract In two experiments with rats, we analyzed the effect of experiencing extinction in one task on the context specificity of a new association learned within a different task. Rats were trained to run in a runway for water in Task 1, and received taste aversion conditioning in Task 2 (the tasks were reversed in Exp. 2). Half of the rats received conditioning and extinction of Task 1 in Context A, whereas the other half received no extinction. Then all animals received training in the alternate task in Context B, prior to testing in Context C. When they were tested in Context C, Task 2 performance was attenuated if Task 1 had been extinguished prior to Task 2. These results are similar to those we have reported in humans, and consistent with the idea that extinction prompts attention to contexts, regardless of whether or not the contexts were involved in extinction. Keywords Attention . Context processing . Extinction . Runway . Taste aversion . Rats A context change between extinction and testing results in a partial renewal of conditioned responding (e.g., Bouton & Bolles, 1979). To illustrate this concept in a conditioned taste aversion (CTA) procedure, Bernal-Gamboa et al. (2012) first paired a sucrose conditioned stimulus (CS) with a lithium chloride unconditioned stimulus (US) in Context A. Next, Groups AAA and AAB received extinction in Context A, R. Bernal-Gamboa National Autonomous University of Mexico, Mexico City, México J. M. Rosas : J. E. Callejas-Aguilera University of Jaén, Jaén, Spain J. M. Rosas (*) Departamento de Psicología, Universidad de Jaén, Paraje de las Lagunillas s/n, 23071 Jaén, Spain e-mail: [email protected]
whereas Groups ABA and ABC received extinction in a different context (Context B). CS intake decreased after conditioning and steadily increased as extinction proceeded, regardless of the extinction context. Switching contexts did not affect conditioning base performance, a common result in the renewal literature (e.g., Rosas, Todd, & Bouton, 2013). Following extinction, Group AAA received a test with sucrose in Context A, the same context in which extinction took place, whereas Groups AAB, ABA, and ABC received the test in Contexts B, A, and C, respectively; all of these were different from the context in which extinction took place. The change of context between extinction and testing led to similar renewals of the taste aversion among the latter three groups. The most influential explanation for the renewal effect was provided by Bouton (1993). He assumed the formation of CS– US excitatory associations during conditioning. During extinction, new, inhibitory associations are formed between the CS and the US (see, e.g., Konorski, 1948). Bouton assumed that contextual cues during extinction would activate an intermediate node that acts like an AND gate (Estes, 1976). This inhibition would only be retrieved when the extinction context and the CS were presented together. Otherwise, the CS–US excitatory association would be expressed without interference. Bouton (1993) suggested that the context specificity of extinction relies on the fact that extinction is both inhibitory and the second-learned information about the CS. Subsequent experiments led Nelson (e.g., 2009) to conclude that only secondlearned associations are affected by contextual changes, regardless of whether the information is excitatory or inhibitory. Bouton (1997) explained the context dependence of secondlearned information by suggesting that the ambiguity produced by extinction raises animals’ attention to the context, making ambiguous information context-dependent. Building on this idea, Rosas, Callejas-Aguilera, Ramos-Álvarez, and Abad (2006), in their attentional theory of context processing (ATCP), suggested that once the animal pays attention to the
Psychon Bull Rev
context, any information learned becomes context-specific, regardless of whether or not this information is ambiguous (e.g., Rosas & Callejas-Aguilera, 2006). According to ATCP, the context dependency of information is not related to the type of information that is learned, but to whether the information is learned in a context that is being attended (cf. Bouton, 1993, 1997). ATCP’s predictions were initially explored by Rosas and Callejas-Aguilera (2006). In a predictive-learning task, participants learned whether a customer who consumed specific foods (cues) in a restaurant (context) would experience gastric malaise (outcome). When a cue was conditioned and then extinguished, the acquisition of a relationship between a different cue and the outcome became context-dependent, regardless of whether it had been learned in the same extinction context or not (see also Rosas & Callejas-Aguilera, 2007). In Experiments 3 and 4, they found that the effect of the extinction of a cue on context dependence transferred to a different cue learned about in a different task. Although this result should be taken with caution, given that the two tasks used were quite similar (both were predictive-learning tasks, conducted on the same computer and in the same room), BernalGamboa, Callejas-Aguilera, Nieto, and Rosas (2013), using CTA and runway running in rats, found that extinction in one task favors forgetting over time in a nonextinguished cue in a different sequentially learned task. In the present work we explored, in two experiments, whether the extinction of a cue within a given task would favor the context dependence of a nonextinguished association learned within a different task, to extend Rosas and Callejas-Aguilera’s (2006) results to nonhuman animals, and those of Bernal-Gamboa et al. (2013) to the use of physical contexts. The designs are presented in Table 1. Rats were sequentially trained on two different tasks (runway running and CTA, counterbalanced as Task 1 and Task 2 across experiments). Groups E received conditioning and extinction of Task 1, whereas Groups NE received either only conditioning (Exp. 1) or the same experience with the cue and the outcome, but separately (Exp. 2) in Context A. Note that Group NE in
Experiment 1 received more training on Task 1, which could have enhanced stimulus control, making it difficult to discard the influence of incubation on potential differences at testing. However, the designs of Experiments 1 and 2 should be evaluated together, to guarantee that the expected effect of extinction depended on Groups E receiving both acquisition and extinction. All rats then received simple acquisition training with Task 2 in Context B, followed by a test of extinction in Context C. According to ATCP, performance in the different context in the final test of Task 2 should be worse after experiencing extinction in Task 1.
Method Subjects and apparatus A group of 32 three-month-old, experimentally naive female Wistar rats (average weight 203.4 g) were used (16 in each experiment, n =8 per group). They were individually housed in standard Plexiglas cages inside a room maintained on a 12:12-h light:dark cycle. Rats were maintained with ad libitum access to food, but liquid access was restricted to two daily 15-min sessions (10 am and 5 pm) throughout the experiment. The experimental procedures were always conducted at 10 am. During the 5 pm sessions, rats received distilled water. One runway was provided by a straight 120×11×14 cm (L×W×H) box, divided into three sections separated by wood guillotine doors. The start and goal boxes measured 20 cm each, and the running section measured 80 cm. The walls and floor of the runway were made of painted wood (green). Two guillotine doors separated the start and goal boxes from the running section when closed. The entire length of the runway apparatus was covered by clear Plexiglas. A second runway was a T-Maze with one arm blocked with a wood door. Otherwise, the resulting runway was identical to that described above. The two runways were located in different rooms and counterbalanced as Contexts A and B in Experiment 2; the first-described runway served as Context A
Table 1 Experimental designs Experiment
Group
Task 1–Phase 1
Task 1–Phase 2
Task 2–Training
Task 2– Test
1
E NE E NE
A:8R-O A:8R-O A: 1X+ A: 1X–/1+
A: 4R– A: 4R-O A: 3X– A: 3X–
B: 1X+ B: 1X+ B: 8R-O B: 8R-O
C: 4X– C: 4X– C:3R– C:3R–
2
Task 1 and Task 2 stand for two different tasks (Task 1 was the runway task and Task 2, conditioned taste aversion in Exp. 1, and vice versa in Exp. 2). The group names stand for the treatments received by the animals during the first task (E, conditioned and extinguished; NE, not extinguished— conditioned in Exp. 1 and unpaired in Exp. 2). A, B, and C stand for the contexts in which each phase was conducted. Numbers correspond to the numbers of sessions conducted on each phase (sessions with R involved six trials each, whereas sessions with X involved a single trial). R-O means that running down the runway was reinforced with water access. R– means that the bottle localized at the end of the alley did not contain water. X was a 15 % sucrose solution. + indicates an LiCl injection (0.3 molar, 0.5 % body weight), and – means no injection
Psychon Bull Rev
in Experiment 1. In both, distilled water could be made available through the ceiling of the goal section. At the end of each daily session, the first runway was cleaned with a solution of 90 % ethanol, and the second was cleaned with Windex. The time to run through a runway was manually registered using a stopwatch. For the CTA procedures, a solution of 15 % sucrose diluted in distilled water was used as CS X. The US was a single intraperitoneal injection of LiCl, at 0.3 M, 0.5 % of bodyweight. Fluids were administered in 150-ml bottles with a standard spout. CTA trials were conducted in two different sets of Plexiglas cages (23×23×14 cm, L×W×H) that were placed in two different laboratory rooms. In one set, the walls of the cages were covered by dark green paper, and the floor was covered by a standard fiber-paper egg-tray. In the other set, the walls of the cages were covered with squared pattern paper (red and white 7-mm squares) and the floor was not covered. The cages were wiped with a solution of 90 % ethanol, and the egg-trays were changed after each daily session. These boxes were counterbalanced as Contexts B and C in Experiment 2, and the box with red squares served as Context A in Experiment 1. A constant noise provided by a standard fan was present in the background. Procedure Experiment 1 Task 1, Phase 1 Rats received 5 days of acclimation to the drinking schedule in their home cages. Rats were assigned to Groups E or NE on Day 5, which were matched on their consumption during the acclimation period. Sessions were conducted on consecutive days. Rats were transported to the experimental room in groups of four (two from each group) and run individually beginning on Day 6. Within each set, rats’ distribution was random with respect to the group that they belonged to. On Day 6, rats received five 2-min accesses to the runway, spaced approximately 7 min apart. Each was placed in the start box with doors open and no bottle in the goal box. On Day 7, the procedure was repeated, but was followed by three 30-s periods of free access to distilled water while the subjects were confined in the goal box. On Day 8, the rats were confined in the goal box with free access to water for three periods of 30 s. Training started on Day 9. Eight 6-trial sessions were conducted with each rat. On each trial, the animal was placed in the start box with the door closed. The door was raised after 5 s, the stopwatch was started, and the rat was allowed to run down the runway to obtain 30 s of free access to distilled water at the goal box. When all four of the rat’s paws were within the goal box, the stopwatch was stopped and the latency recorded. A maximum time of 20 s was allowed for the rat to complete the trial. If the rat did not reach the goal box within the time limit, it was
gently pushed down the runway by the experimenter and a latency of 20 s was recorded for that trial. The rat was removed from the goal box and placed back in its home cage within the experimental room. The intertrial interval was about 7 min. Then, 20 min after the last trial, rats received 12 min of free access to distilled water in their home cages at the colony room. Task 1, Phase 2 Rats received four 6-trial sessions of training during this phase. The rats in Group NE continued as in Phase 1, whereas the water bottle was removed from the goal box for the rats in Group E. Task 2, training Rats received distilled water over two days, one day in each context (order counterbalanced). The next day, all rats received 15 min of free access to X in Context B, followed by an injection of LiCl. Task 2, test After a recovery day in the home cage, the rats received one daily trial with X in extinction in Context C for four consecutive days. Experiment 2 Task 1, Phase 1 On Day 6, all of the rats received free access to X in Context A during the morning session. In Group E, this consumption was immediately followed by an injection of LiCl. On the next day, rats received distilled water during the morning session. The rats in Group NE immediately received an injection of LiCl. A recovery day followed, in which the rats had free access to distilled water in both morning and evening sessions. Task 1, Phase 2 Every rat received exposure to X in the morning session on each of the next three days in Context A. Task 2, training The instrumental training was identical to that of Task 1, Phase 1 in Experiment 1, except where noted. Habituation to the runways was conducted as described in Experiment 1, with the only exception being that identical sessions were conducted in Contexts B and C each day, with the order counterbalanced across subjects. After the three habituation days, eight 6-trial sessions were conducted with each rat in Context B. Task 2, test All of the rats received three 6-trial sessions identical to the ones described above in Task 2, Training, except that the bottle at the goal box was empty and the test was conducted in Context C. Data analysis Rats were labeled with numbers, and no indication of group membership was provided. At testing, rats were brought by
Psychon Bull Rev
hand to the experimenter with no other indication, so that the experimenter was blind to the rats’ group allocations at test. Liquid consumption was recorded by weighing the bottles before and after each trial. Consumption was used as the dependent variable during the taste aversion task, whereas latency to reach the goal box was used as the dependent variable in the instrumental task. To control for any possible differences at the end of acquisition on the tests, performance at the end of training was introduced as a covariate into the analysis. An initial data screening revealed significant differences in the variances of the data over trials; thus, MANCOVA was chosen as the analysis tool (see, e.g., Maxwell & Delaney, 2004). To keep the same analytic structure, MANOVAs were used to analyze the rest of the data. The rejection criterion was set at p < .05.
Results and discussion Experiment 1 Figure 1 shows both groups’ mean latencies to reach the goal box during the eight Phase 1 sessions on Task 1 and the four Phase 2 sessions on Task 2 (left panel). The mean consumption of X during the Task 2–Training session and the four sessions of testing of Task 2–Test are shown at the right for Groups E and NE. A 2 (Group) × 8 (Session) MANOVA conducted on the Task 1–Phase 1 data revealed a main effect of session, Wilks’s λ = 0.07, F (7, 8) = 14.41, p = .001, η 2 = .927, and a significant Group × Session interaction, Wilks’s λ = 0.21, F(7, 8) = 4.40, p = .027, η 2 = .794. One simple effect of group was reliable at p = .049, η 2 = .249, on Session 6, F(1, 14) = 4.64, MSE = 0.25, but it would not be significant after any correction for the number of unplanned comparisons made.
Fig. 1 Mean latencies (in seconds) to reach the goal box on each of the eight sessions of Task 1–Phase 1 and the four sessions of Task 1–Phase 2 in Context A (left panel), and mean consumption (in milliliters) of X
A 2 (Group) × 4 (Session) MANCOVA conducted with the Task 1–Phase 2 data and using the data from the last session of Task 1–Phase 1 training as a covariate showed a significant Group × Session interaction, Wilks’s λ = 0.25, F(3, 11) = 11.14, p = .001, η 2 = .752. No effects or interactions involving the covariate were significant, Fs < 1. Latencies declined over trials in Group E, Wilks’s λ = 0.12, F(3, 5) = 12.38, p = .009, η 2 = .881, and did not change in Group NE, Wilks’s λ = 0.35, F(3, 5) = 3.13, p = .125, η 2 = .653. Group NE reached the goal box faster than Group E in every session, Fs(1, 14) ≥ 15.63, MSE = 3.21, p ≤ .001, η 2 ≥ .528. The most interesting results corresponded to performance on Task 2, in which both groups received the same treatment. Groups E and NE did not differ in their consumption of X in the Task 2–Training session, F < 1. However, a 2 (Group) × 4 (Trial) MANCOVA on the test data using the last session of Task 2–Training as a covariate revealed a significant Group × Trial interaction, Wilks’s λ = 0.37, F(3, 11) = 6.14, p = .01, η 2 = .626. No effects or interactions involving the covariate were significant [largest F(1, 13) = 2.55, MSE = 14.15, p = .134, η 2 = .164]. The simple effect of group was significant on Trials 1 and 2, Fs(1, 14) ≥ 7.05, MSE = 5.87, p ≤ .019, η 2 ≥ .335, but not on Trials 3 and 4, Fs < 1. Since testing took place in a context (C) different from the conditioning context (B), and the groups only differed in whether they received extinction during Task 1– Phase 2 training, it may be concluded that extinction in Task 1 made extinction of Task 2 faster when the test took place in a context different from the acquisition context. Experiment 2 Figure 2 presents the mean consumption of X during Task 1– Phase 1 training and Task 1–Phase 2 exposure to X (left panel). The mean latencies to reach the goal box during the eight conditioning sessions of Task 2–Training in Context B
during the single session of Task 2–Training in Context B and the four sessions of Task 2–Test in Context C (right panel) for Groups E and NE in Experiment 1. Error bars denote standard errors of the means
Psychon Bull Rev
Fig. 2 Mean consumption (in milliliters) of flavor X during Task 1– Phase 1 and Task 1–Phase 2 training in Context A (left panel), and mean latencies (in seconds) to reach the goal box on each of the eight sessions
of the Task 2–Training phase in Context B and the three sessions in extinction of Task 2–Test 2 in Context C (right panel) for Groups NE and E in Experiment 2. Error bars denote standard errors of the means
and the three sessions of Task 2–Testing in Context C are at right. Rats’ intake of X did not differ across groups in Task 1– Phase 1, F < 1. A 2 (Group) × 3 (Trial) MANCOVA conducted with the Task 1–Phase 2 data and consumption of X at the end of Task 1–Phase 1 as the covariate showed a significant effect of the covariate, F (1, 13) = 8.48, MSE = 5.72, p = .012, η 2 = .395. However, the Group × Trial interaction was still significant, Wilks’s λ = 0.16, F (2, 12) = 32.67, p < .001, η 2 = .845. The simple effect of group was significant on every Task 1–Phase 2 trial, Fs(1, 14) ≥ 6.61, MSE = 3.43, p < .022, η 2 ≥ .321. The simple effect of trial was significant in Group E, Wilks’s λ = 0.06, F(2, 6) = 50.12, p < .001, η 2 = .944, but not significant in Group NE, Wilks’s λ = 0.67, F (2, 6) = 1.50, p = .296, η 2 = .334. Group performance was appropriate to the treatment received: Consumption decreased in Group E after conditioning and recovered across extinction sessions. Consumption in Group NE remained high throughout. The most relevant results are presented in the right panel of Fig. 2. A 2 (Group) × 8 (Session) MANOVA conducted with the Task 2–Training data only showed a significant main effect of session, Wilks’s λ = 0.01, F(7, 8) = 98.16, p < .001, η 2 = .988. These data were analyzed with a 2 (Group) × 3 (Session) MANCOVA with the data from the last session of Task 2– Training as a covariate. No effects or interactions involving the covariate were found, but the Group × Session interaction was significant, Wilks’s λ = 0.60, F(2, 12) = 4.01, p = .046, η 2 = .401. The simple effect of group was significant on Session 1, F (1, 14) = 12.17, MSE = 4.07, p = .004, η 2 = .465, but not on Sessions 2 and 3, F s < 1. Greater latencies were found in Group E than in Group NE when the groups were tested in a third context, C. As in Experiment 1, this result suggests greater susceptibility to the effects of context change after previous extinction of a different cue in a different task.
The design of these experiments did not allow us to determine whether the test differences were due to a decrease in control in Group E or to an increase in Group NE with the context change. However, we can safely conclude that the extinction training in one task led to an impairment of performance in a new task when the test was conducted outside the training context.
Discussion The goal of these experiments was to extend the findings of Rosas and Callejas-Aguilera (2006) with humans to animals and physical contexts (see also Bernal-Gamboa et al., 2013). The experiments evaluated whether the context dependence of an association would be favored by prior extinction of a different association learned in a different task. Both Experiments 1 and 2 showed that extinction made subsequent learning in a different task context-specific. These results add to those reported in humans using fictitious restaurants and farms as contexts (Rosas & CallejasAguilera, 2006), as well as those reported in animals with retention intervals (Bernal-Gamboa et al., 2013). All of these results suggest that once extinction is experienced, subsequent learning becomes context-dependent, regardless of whether it takes place in the same context in which extinction had previously taken place (Rosas & Callejas-Aguilera, 2007), in a different context (Rosas & Callejas-Aguilera, 2006), or, as reported here, within an apparently quite different task. These results are consistent with the principles of ATCP proposed by Rosas et al. (2006), which state that extinction raises animals’ attention to contexts, so that all information learned within them becomes context-specific. However, as we have noted elsewhere, the idea that extinction raises attention to contexts
Psychon Bull Rev
is speculative, given that only a single experiment has provided an independent measure of attention being aroused outside of its supposed effect on contextual control (Nelson, Lamoureux, & León, 2013). Pursuing independent measures of attention to contexts should be a goal for research within this field before we will be fully able to accept ATCP as an explanation of context-switch effects in the literature. Finally, it is worth briefly discussing that analytical challenges have been raised to ATCP’s central premise that attention to contexts is aroused by ambiguity (or any of the factors discussed by Rosas et al., 2006). For example, ATCP could suggest that, contrary to what has been assumed since Bouton’s (1993) review, context specificity should be the rule, rather than the exception, in human and nonhuman animal learning. Our lives are filled with uncertainty from the very moment we are born. If contextual control applies to all things being learned, rather than just to those that are producing the uncertainty, then most learning should be context-specific. Such a conclusion is challenged by many occasions in which such context specificity is not observed (see Bouton, 1993; Rosas et al., 2013). A number of factors are likely to mediate these results. For instance, León, Abad, and Rosas (2011) found that simple acquisition may be context-specific at the beginning of training, with that contextual control disappearing with increased training (see also Hall & Honey, 1990). Most tests for contextual dependence are conducted once training has reached asymptote, which would explain why the context specificity of acquisition often goes undetected in the literature (e.g., Bouton, 1993). It could also be true that the attention to contexts raised by ambiguity wanes rapidly as time passes. Author note The research presented here was made possible by Grant No. PSI2010-15215 from the Spanish Ministry of Science and Innovation and by Grant No. HUM-642 from the government of Andalucía. We thank Manuel M. Ramos-Álvarez and James B. Nelson for their comments to a previous version of the manuscript. The participation of R.B.-G. in this project was partially supported by Grant PAPIIT (IN307113).
References Bernal-Gamboa, R., Callejas-Aguilera, J. E., Nieto, J., & Rosas, J. M. (2013). Extinction makes conditioning time-dependent. Journal of Experimental Psychology: Animal Behavior Processes, 39, 221– 232. doi:10.1037/a0032181
Bernal-Gamboa, R., Juárez, Y., González-Martín, G., Carranza, R., Sánchez-Carrasco, L., & Nieto, J. (2012). ABA, AAB and ABC renewal in taste aversion learning. Psicológica, 33, 1–13. Bouton, M. E. (1993). Context, time, and memory retrieval in the interference paradigms of Pavlovian learning. Psychological Bulletin, 114, 80–99. doi:10.1037/0033-2909.114.1.80 Bouton, M. E. (1997). Signals for whether versus when an event will occur. In M. E. Bouton & M. S. Fanselow (Eds.), Learning, motivation, and cognition: The functional behaviorism of Robert C. Bolles (pp. 385–409). Washington DC: American Psychological Association. doi:10.1037/10223-019 Bouton, M. E., & Bolles, R. C. (1979). Contextual control of the extinction of conditioned fear. Learning and Motivation, 10, 445–466. doi:10.1016/0023-9690(79)90057-2 Estes, W. K. (1976). Structural aspects of associative models of memory. In C. N. Cofer (Ed.), The structure of human memory (pp. 31–53). San Francisco: Freeman. Hall, G., & Honey, R. C. (1990). Context-specific conditioning in the conditioned-emotional-response procedure. Journal of Experimental Psychology: Animal Behavior Processes, 16, 271–278. doi:10.1037/ 0097-7403.16.3.271 Konorski, J. (1948). Conditioned reflexes and neuron organization. Cambridge: Cambridge University Press. León, S. P., Abad, M. J. F., & Rosas, J. M. (2011). Context–outcome associations mediate context-switch effects in a human predictive learning task. Learning and Motivation, 42, 84–98. doi:10.1016/j. lmot.2010.10.001 Maxwell, S. E., & Delaney, H. D. (2004). Designing experiments and analyzing data: A model comparison approach (2nd ed.). Belmont: Wadsworth. Nelson, J. B. (2009). Contextual control of first- and second-learned excitation and inhibition in equally ambiguous stimuli. Learning & Behavior, 37, 85–94. doi:10.3758/LB.37.1.95 Nelson, J. B., Lamoureux, J. A., & León, S. P. (2013). Extinction arouses attention to the context in a behavioral suppression method with humans. Journal of Experimental Psychology: Animal Behavior Processes, 39, 99–105. doi:10.1037/a0030759 Rosas, J. M., & Callejas-Aguilera, J. E. (2006). Context switch effects on acquisition and extinction in human predictive learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 32, 461–474. doi:10.1037/0278-7393. 32.3.461 Rosas, J. M., & Callejas-Aguilera, J. E. (2007). Acquisition of a conditioned taste aversion becomes context dependent when it is learned after extinction. Quarterly Journal of Experimental Psychology, 60, 9–15. doi:10.1080/17470210600971519 Rosas, J. M., Callejas-Aguilera, J. E., Ramos-Álvarez, M. M., & Abad, M. J. F. (2006). Revision of retrieval theory of forgetting: What does make information context-specific? International Journal of Psychology and Psychological Therapy, 6, 147–166. Rosas, J. M., Todd, T. P., & Bouton, M. E. (2013). Context change and associative learning. Wiley Interdisciplinary Reviews: Cognitive Science, 4, 237–244. doi:10.1002/wcs.1225
BRIEF REPORT
Experiencing extinction within a task makes nonextinguished information learned within a different task context-dependent Rodolfo Bernal-Gamboa & Juan M. Rosas & José E. Callejas-Aguilera
# Psychonomic Society, Inc. 2013
Abstract In two experiments with rats, we analyzed the effect of experiencing extinction in one task on the context specificity of a new association learned within a different task. Rats were trained to run in a runway for water in Task 1, and received taste aversion conditioning in Task 2 (the tasks were reversed in Exp. 2). Half of the rats received conditioning and extinction of Task 1 in Context A, whereas the other half received no extinction. Then all animals received training in the alternate task in Context B, prior to testing in Context C. When they were tested in Context C, Task 2 performance was attenuated if Task 1 had been extinguished prior to Task 2. These results are similar to those we have reported in humans, and consistent with the idea that extinction prompts attention to contexts, regardless of whether or not the contexts were involved in extinction. Keywords Attention . Context processing . Extinction . Runway . Taste aversion . Rats A context change between extinction and testing results in a partial renewal of conditioned responding (e.g., Bouton & Bolles, 1979). To illustrate this concept in a conditioned taste aversion (CTA) procedure, Bernal-Gamboa et al. (2012) first paired a sucrose conditioned stimulus (CS) with a lithium chloride unconditioned stimulus (US) in Context A. Next, Groups AAA and AAB received extinction in Context A, R. Bernal-Gamboa National Autonomous University of Mexico, Mexico City, México J. M. Rosas : J. E. Callejas-Aguilera University of Jaén, Jaén, Spain J. M. Rosas (*) Departamento de Psicología, Universidad de Jaén, Paraje de las Lagunillas s/n, 23071 Jaén, Spain e-mail: [email protected]
whereas Groups ABA and ABC received extinction in a different context (Context B). CS intake decreased after conditioning and steadily increased as extinction proceeded, regardless of the extinction context. Switching contexts did not affect conditioning base performance, a common result in the renewal literature (e.g., Rosas, Todd, & Bouton, 2013). Following extinction, Group AAA received a test with sucrose in Context A, the same context in which extinction took place, whereas Groups AAB, ABA, and ABC received the test in Contexts B, A, and C, respectively; all of these were different from the context in which extinction took place. The change of context between extinction and testing led to similar renewals of the taste aversion among the latter three groups. The most influential explanation for the renewal effect was provided by Bouton (1993). He assumed the formation of CS– US excitatory associations during conditioning. During extinction, new, inhibitory associations are formed between the CS and the US (see, e.g., Konorski, 1948). Bouton assumed that contextual cues during extinction would activate an intermediate node that acts like an AND gate (Estes, 1976). This inhibition would only be retrieved when the extinction context and the CS were presented together. Otherwise, the CS–US excitatory association would be expressed without interference. Bouton (1993) suggested that the context specificity of extinction relies on the fact that extinction is both inhibitory and the second-learned information about the CS. Subsequent experiments led Nelson (e.g., 2009) to conclude that only secondlearned associations are affected by contextual changes, regardless of whether the information is excitatory or inhibitory. Bouton (1997) explained the context dependence of secondlearned information by suggesting that the ambiguity produced by extinction raises animals’ attention to the context, making ambiguous information context-dependent. Building on this idea, Rosas, Callejas-Aguilera, Ramos-Álvarez, and Abad (2006), in their attentional theory of context processing (ATCP), suggested that once the animal pays attention to the
Psychon Bull Rev
context, any information learned becomes context-specific, regardless of whether or not this information is ambiguous (e.g., Rosas & Callejas-Aguilera, 2006). According to ATCP, the context dependency of information is not related to the type of information that is learned, but to whether the information is learned in a context that is being attended (cf. Bouton, 1993, 1997). ATCP’s predictions were initially explored by Rosas and Callejas-Aguilera (2006). In a predictive-learning task, participants learned whether a customer who consumed specific foods (cues) in a restaurant (context) would experience gastric malaise (outcome). When a cue was conditioned and then extinguished, the acquisition of a relationship between a different cue and the outcome became context-dependent, regardless of whether it had been learned in the same extinction context or not (see also Rosas & Callejas-Aguilera, 2007). In Experiments 3 and 4, they found that the effect of the extinction of a cue on context dependence transferred to a different cue learned about in a different task. Although this result should be taken with caution, given that the two tasks used were quite similar (both were predictive-learning tasks, conducted on the same computer and in the same room), BernalGamboa, Callejas-Aguilera, Nieto, and Rosas (2013), using CTA and runway running in rats, found that extinction in one task favors forgetting over time in a nonextinguished cue in a different sequentially learned task. In the present work we explored, in two experiments, whether the extinction of a cue within a given task would favor the context dependence of a nonextinguished association learned within a different task, to extend Rosas and Callejas-Aguilera’s (2006) results to nonhuman animals, and those of Bernal-Gamboa et al. (2013) to the use of physical contexts. The designs are presented in Table 1. Rats were sequentially trained on two different tasks (runway running and CTA, counterbalanced as Task 1 and Task 2 across experiments). Groups E received conditioning and extinction of Task 1, whereas Groups NE received either only conditioning (Exp. 1) or the same experience with the cue and the outcome, but separately (Exp. 2) in Context A. Note that Group NE in
Experiment 1 received more training on Task 1, which could have enhanced stimulus control, making it difficult to discard the influence of incubation on potential differences at testing. However, the designs of Experiments 1 and 2 should be evaluated together, to guarantee that the expected effect of extinction depended on Groups E receiving both acquisition and extinction. All rats then received simple acquisition training with Task 2 in Context B, followed by a test of extinction in Context C. According to ATCP, performance in the different context in the final test of Task 2 should be worse after experiencing extinction in Task 1.
Method Subjects and apparatus A group of 32 three-month-old, experimentally naive female Wistar rats (average weight 203.4 g) were used (16 in each experiment, n =8 per group). They were individually housed in standard Plexiglas cages inside a room maintained on a 12:12-h light:dark cycle. Rats were maintained with ad libitum access to food, but liquid access was restricted to two daily 15-min sessions (10 am and 5 pm) throughout the experiment. The experimental procedures were always conducted at 10 am. During the 5 pm sessions, rats received distilled water. One runway was provided by a straight 120×11×14 cm (L×W×H) box, divided into three sections separated by wood guillotine doors. The start and goal boxes measured 20 cm each, and the running section measured 80 cm. The walls and floor of the runway were made of painted wood (green). Two guillotine doors separated the start and goal boxes from the running section when closed. The entire length of the runway apparatus was covered by clear Plexiglas. A second runway was a T-Maze with one arm blocked with a wood door. Otherwise, the resulting runway was identical to that described above. The two runways were located in different rooms and counterbalanced as Contexts A and B in Experiment 2; the first-described runway served as Context A
Table 1 Experimental designs Experiment
Group
Task 1–Phase 1
Task 1–Phase 2
Task 2–Training
Task 2– Test
1
E NE E NE
A:8R-O A:8R-O A: 1X+ A: 1X–/1+
A: 4R– A: 4R-O A: 3X– A: 3X–
B: 1X+ B: 1X+ B: 8R-O B: 8R-O
C: 4X– C: 4X– C:3R– C:3R–
2
Task 1 and Task 2 stand for two different tasks (Task 1 was the runway task and Task 2, conditioned taste aversion in Exp. 1, and vice versa in Exp. 2). The group names stand for the treatments received by the animals during the first task (E, conditioned and extinguished; NE, not extinguished— conditioned in Exp. 1 and unpaired in Exp. 2). A, B, and C stand for the contexts in which each phase was conducted. Numbers correspond to the numbers of sessions conducted on each phase (sessions with R involved six trials each, whereas sessions with X involved a single trial). R-O means that running down the runway was reinforced with water access. R– means that the bottle localized at the end of the alley did not contain water. X was a 15 % sucrose solution. + indicates an LiCl injection (0.3 molar, 0.5 % body weight), and – means no injection
Psychon Bull Rev
in Experiment 1. In both, distilled water could be made available through the ceiling of the goal section. At the end of each daily session, the first runway was cleaned with a solution of 90 % ethanol, and the second was cleaned with Windex. The time to run through a runway was manually registered using a stopwatch. For the CTA procedures, a solution of 15 % sucrose diluted in distilled water was used as CS X. The US was a single intraperitoneal injection of LiCl, at 0.3 M, 0.5 % of bodyweight. Fluids were administered in 150-ml bottles with a standard spout. CTA trials were conducted in two different sets of Plexiglas cages (23×23×14 cm, L×W×H) that were placed in two different laboratory rooms. In one set, the walls of the cages were covered by dark green paper, and the floor was covered by a standard fiber-paper egg-tray. In the other set, the walls of the cages were covered with squared pattern paper (red and white 7-mm squares) and the floor was not covered. The cages were wiped with a solution of 90 % ethanol, and the egg-trays were changed after each daily session. These boxes were counterbalanced as Contexts B and C in Experiment 2, and the box with red squares served as Context A in Experiment 1. A constant noise provided by a standard fan was present in the background. Procedure Experiment 1 Task 1, Phase 1 Rats received 5 days of acclimation to the drinking schedule in their home cages. Rats were assigned to Groups E or NE on Day 5, which were matched on their consumption during the acclimation period. Sessions were conducted on consecutive days. Rats were transported to the experimental room in groups of four (two from each group) and run individually beginning on Day 6. Within each set, rats’ distribution was random with respect to the group that they belonged to. On Day 6, rats received five 2-min accesses to the runway, spaced approximately 7 min apart. Each was placed in the start box with doors open and no bottle in the goal box. On Day 7, the procedure was repeated, but was followed by three 30-s periods of free access to distilled water while the subjects were confined in the goal box. On Day 8, the rats were confined in the goal box with free access to water for three periods of 30 s. Training started on Day 9. Eight 6-trial sessions were conducted with each rat. On each trial, the animal was placed in the start box with the door closed. The door was raised after 5 s, the stopwatch was started, and the rat was allowed to run down the runway to obtain 30 s of free access to distilled water at the goal box. When all four of the rat’s paws were within the goal box, the stopwatch was stopped and the latency recorded. A maximum time of 20 s was allowed for the rat to complete the trial. If the rat did not reach the goal box within the time limit, it was
gently pushed down the runway by the experimenter and a latency of 20 s was recorded for that trial. The rat was removed from the goal box and placed back in its home cage within the experimental room. The intertrial interval was about 7 min. Then, 20 min after the last trial, rats received 12 min of free access to distilled water in their home cages at the colony room. Task 1, Phase 2 Rats received four 6-trial sessions of training during this phase. The rats in Group NE continued as in Phase 1, whereas the water bottle was removed from the goal box for the rats in Group E. Task 2, training Rats received distilled water over two days, one day in each context (order counterbalanced). The next day, all rats received 15 min of free access to X in Context B, followed by an injection of LiCl. Task 2, test After a recovery day in the home cage, the rats received one daily trial with X in extinction in Context C for four consecutive days. Experiment 2 Task 1, Phase 1 On Day 6, all of the rats received free access to X in Context A during the morning session. In Group E, this consumption was immediately followed by an injection of LiCl. On the next day, rats received distilled water during the morning session. The rats in Group NE immediately received an injection of LiCl. A recovery day followed, in which the rats had free access to distilled water in both morning and evening sessions. Task 1, Phase 2 Every rat received exposure to X in the morning session on each of the next three days in Context A. Task 2, training The instrumental training was identical to that of Task 1, Phase 1 in Experiment 1, except where noted. Habituation to the runways was conducted as described in Experiment 1, with the only exception being that identical sessions were conducted in Contexts B and C each day, with the order counterbalanced across subjects. After the three habituation days, eight 6-trial sessions were conducted with each rat in Context B. Task 2, test All of the rats received three 6-trial sessions identical to the ones described above in Task 2, Training, except that the bottle at the goal box was empty and the test was conducted in Context C. Data analysis Rats were labeled with numbers, and no indication of group membership was provided. At testing, rats were brought by
Psychon Bull Rev
hand to the experimenter with no other indication, so that the experimenter was blind to the rats’ group allocations at test. Liquid consumption was recorded by weighing the bottles before and after each trial. Consumption was used as the dependent variable during the taste aversion task, whereas latency to reach the goal box was used as the dependent variable in the instrumental task. To control for any possible differences at the end of acquisition on the tests, performance at the end of training was introduced as a covariate into the analysis. An initial data screening revealed significant differences in the variances of the data over trials; thus, MANCOVA was chosen as the analysis tool (see, e.g., Maxwell & Delaney, 2004). To keep the same analytic structure, MANOVAs were used to analyze the rest of the data. The rejection criterion was set at p < .05.
Results and discussion Experiment 1 Figure 1 shows both groups’ mean latencies to reach the goal box during the eight Phase 1 sessions on Task 1 and the four Phase 2 sessions on Task 2 (left panel). The mean consumption of X during the Task 2–Training session and the four sessions of testing of Task 2–Test are shown at the right for Groups E and NE. A 2 (Group) × 8 (Session) MANOVA conducted on the Task 1–Phase 1 data revealed a main effect of session, Wilks’s λ = 0.07, F (7, 8) = 14.41, p = .001, η 2 = .927, and a significant Group × Session interaction, Wilks’s λ = 0.21, F(7, 8) = 4.40, p = .027, η 2 = .794. One simple effect of group was reliable at p = .049, η 2 = .249, on Session 6, F(1, 14) = 4.64, MSE = 0.25, but it would not be significant after any correction for the number of unplanned comparisons made.
Fig. 1 Mean latencies (in seconds) to reach the goal box on each of the eight sessions of Task 1–Phase 1 and the four sessions of Task 1–Phase 2 in Context A (left panel), and mean consumption (in milliliters) of X
A 2 (Group) × 4 (Session) MANCOVA conducted with the Task 1–Phase 2 data and using the data from the last session of Task 1–Phase 1 training as a covariate showed a significant Group × Session interaction, Wilks’s λ = 0.25, F(3, 11) = 11.14, p = .001, η 2 = .752. No effects or interactions involving the covariate were significant, Fs < 1. Latencies declined over trials in Group E, Wilks’s λ = 0.12, F(3, 5) = 12.38, p = .009, η 2 = .881, and did not change in Group NE, Wilks’s λ = 0.35, F(3, 5) = 3.13, p = .125, η 2 = .653. Group NE reached the goal box faster than Group E in every session, Fs(1, 14) ≥ 15.63, MSE = 3.21, p ≤ .001, η 2 ≥ .528. The most interesting results corresponded to performance on Task 2, in which both groups received the same treatment. Groups E and NE did not differ in their consumption of X in the Task 2–Training session, F < 1. However, a 2 (Group) × 4 (Trial) MANCOVA on the test data using the last session of Task 2–Training as a covariate revealed a significant Group × Trial interaction, Wilks’s λ = 0.37, F(3, 11) = 6.14, p = .01, η 2 = .626. No effects or interactions involving the covariate were significant [largest F(1, 13) = 2.55, MSE = 14.15, p = .134, η 2 = .164]. The simple effect of group was significant on Trials 1 and 2, Fs(1, 14) ≥ 7.05, MSE = 5.87, p ≤ .019, η 2 ≥ .335, but not on Trials 3 and 4, Fs < 1. Since testing took place in a context (C) different from the conditioning context (B), and the groups only differed in whether they received extinction during Task 1– Phase 2 training, it may be concluded that extinction in Task 1 made extinction of Task 2 faster when the test took place in a context different from the acquisition context. Experiment 2 Figure 2 presents the mean consumption of X during Task 1– Phase 1 training and Task 1–Phase 2 exposure to X (left panel). The mean latencies to reach the goal box during the eight conditioning sessions of Task 2–Training in Context B
during the single session of Task 2–Training in Context B and the four sessions of Task 2–Test in Context C (right panel) for Groups E and NE in Experiment 1. Error bars denote standard errors of the means
Psychon Bull Rev
Fig. 2 Mean consumption (in milliliters) of flavor X during Task 1– Phase 1 and Task 1–Phase 2 training in Context A (left panel), and mean latencies (in seconds) to reach the goal box on each of the eight sessions
of the Task 2–Training phase in Context B and the three sessions in extinction of Task 2–Test 2 in Context C (right panel) for Groups NE and E in Experiment 2. Error bars denote standard errors of the means
and the three sessions of Task 2–Testing in Context C are at right. Rats’ intake of X did not differ across groups in Task 1– Phase 1, F < 1. A 2 (Group) × 3 (Trial) MANCOVA conducted with the Task 1–Phase 2 data and consumption of X at the end of Task 1–Phase 1 as the covariate showed a significant effect of the covariate, F (1, 13) = 8.48, MSE = 5.72, p = .012, η 2 = .395. However, the Group × Trial interaction was still significant, Wilks’s λ = 0.16, F (2, 12) = 32.67, p < .001, η 2 = .845. The simple effect of group was significant on every Task 1–Phase 2 trial, Fs(1, 14) ≥ 6.61, MSE = 3.43, p < .022, η 2 ≥ .321. The simple effect of trial was significant in Group E, Wilks’s λ = 0.06, F(2, 6) = 50.12, p < .001, η 2 = .944, but not significant in Group NE, Wilks’s λ = 0.67, F (2, 6) = 1.50, p = .296, η 2 = .334. Group performance was appropriate to the treatment received: Consumption decreased in Group E after conditioning and recovered across extinction sessions. Consumption in Group NE remained high throughout. The most relevant results are presented in the right panel of Fig. 2. A 2 (Group) × 8 (Session) MANOVA conducted with the Task 2–Training data only showed a significant main effect of session, Wilks’s λ = 0.01, F(7, 8) = 98.16, p < .001, η 2 = .988. These data were analyzed with a 2 (Group) × 3 (Session) MANCOVA with the data from the last session of Task 2– Training as a covariate. No effects or interactions involving the covariate were found, but the Group × Session interaction was significant, Wilks’s λ = 0.60, F(2, 12) = 4.01, p = .046, η 2 = .401. The simple effect of group was significant on Session 1, F (1, 14) = 12.17, MSE = 4.07, p = .004, η 2 = .465, but not on Sessions 2 and 3, F s < 1. Greater latencies were found in Group E than in Group NE when the groups were tested in a third context, C. As in Experiment 1, this result suggests greater susceptibility to the effects of context change after previous extinction of a different cue in a different task.
The design of these experiments did not allow us to determine whether the test differences were due to a decrease in control in Group E or to an increase in Group NE with the context change. However, we can safely conclude that the extinction training in one task led to an impairment of performance in a new task when the test was conducted outside the training context.
Discussion The goal of these experiments was to extend the findings of Rosas and Callejas-Aguilera (2006) with humans to animals and physical contexts (see also Bernal-Gamboa et al., 2013). The experiments evaluated whether the context dependence of an association would be favored by prior extinction of a different association learned in a different task. Both Experiments 1 and 2 showed that extinction made subsequent learning in a different task context-specific. These results add to those reported in humans using fictitious restaurants and farms as contexts (Rosas & CallejasAguilera, 2006), as well as those reported in animals with retention intervals (Bernal-Gamboa et al., 2013). All of these results suggest that once extinction is experienced, subsequent learning becomes context-dependent, regardless of whether it takes place in the same context in which extinction had previously taken place (Rosas & Callejas-Aguilera, 2007), in a different context (Rosas & Callejas-Aguilera, 2006), or, as reported here, within an apparently quite different task. These results are consistent with the principles of ATCP proposed by Rosas et al. (2006), which state that extinction raises animals’ attention to contexts, so that all information learned within them becomes context-specific. However, as we have noted elsewhere, the idea that extinction raises attention to contexts
Psychon Bull Rev
is speculative, given that only a single experiment has provided an independent measure of attention being aroused outside of its supposed effect on contextual control (Nelson, Lamoureux, & León, 2013). Pursuing independent measures of attention to contexts should be a goal for research within this field before we will be fully able to accept ATCP as an explanation of context-switch effects in the literature. Finally, it is worth briefly discussing that analytical challenges have been raised to ATCP’s central premise that attention to contexts is aroused by ambiguity (or any of the factors discussed by Rosas et al., 2006). For example, ATCP could suggest that, contrary to what has been assumed since Bouton’s (1993) review, context specificity should be the rule, rather than the exception, in human and nonhuman animal learning. Our lives are filled with uncertainty from the very moment we are born. If contextual control applies to all things being learned, rather than just to those that are producing the uncertainty, then most learning should be context-specific. Such a conclusion is challenged by many occasions in which such context specificity is not observed (see Bouton, 1993; Rosas et al., 2013). A number of factors are likely to mediate these results. For instance, León, Abad, and Rosas (2011) found that simple acquisition may be context-specific at the beginning of training, with that contextual control disappearing with increased training (see also Hall & Honey, 1990). Most tests for contextual dependence are conducted once training has reached asymptote, which would explain why the context specificity of acquisition often goes undetected in the literature (e.g., Bouton, 1993). It could also be true that the attention to contexts raised by ambiguity wanes rapidly as time passes. Author note The research presented here was made possible by Grant No. PSI2010-15215 from the Spanish Ministry of Science and Innovation and by Grant No. HUM-642 from the government of Andalucía. We thank Manuel M. Ramos-Álvarez and James B. Nelson for their comments to a previous version of the manuscript. The participation of R.B.-G. in this project was partially supported by Grant PAPIIT (IN307113).
References Bernal-Gamboa, R., Callejas-Aguilera, J. E., Nieto, J., & Rosas, J. M. (2013). Extinction makes conditioning time-dependent. Journal of Experimental Psychology: Animal Behavior Processes, 39, 221– 232. doi:10.1037/a0032181
Bernal-Gamboa, R., Juárez, Y., González-Martín, G., Carranza, R., Sánchez-Carrasco, L., & Nieto, J. (2012). ABA, AAB and ABC renewal in taste aversion learning. Psicológica, 33, 1–13. Bouton, M. E. (1993). Context, time, and memory retrieval in the interference paradigms of Pavlovian learning. Psychological Bulletin, 114, 80–99. doi:10.1037/0033-2909.114.1.80 Bouton, M. E. (1997). Signals for whether versus when an event will occur. In M. E. Bouton & M. S. Fanselow (Eds.), Learning, motivation, and cognition: The functional behaviorism of Robert C. Bolles (pp. 385–409). Washington DC: American Psychological Association. doi:10.1037/10223-019 Bouton, M. E., & Bolles, R. C. (1979). Contextual control of the extinction of conditioned fear. Learning and Motivation, 10, 445–466. doi:10.1016/0023-9690(79)90057-2 Estes, W. K. (1976). Structural aspects of associative models of memory. In C. N. Cofer (Ed.), The structure of human memory (pp. 31–53). San Francisco: Freeman. Hall, G., & Honey, R. C. (1990). Context-specific conditioning in the conditioned-emotional-response procedure. Journal of Experimental Psychology: Animal Behavior Processes, 16, 271–278. doi:10.1037/ 0097-7403.16.3.271 Konorski, J. (1948). Conditioned reflexes and neuron organization. Cambridge: Cambridge University Press. León, S. P., Abad, M. J. F., & Rosas, J. M. (2011). Context–outcome associations mediate context-switch effects in a human predictive learning task. Learning and Motivation, 42, 84–98. doi:10.1016/j. lmot.2010.10.001 Maxwell, S. E., & Delaney, H. D. (2004). Designing experiments and analyzing data: A model comparison approach (2nd ed.). Belmont: Wadsworth. Nelson, J. B. (2009). Contextual control of first- and second-learned excitation and inhibition in equally ambiguous stimuli. Learning & Behavior, 37, 85–94. doi:10.3758/LB.37.1.95 Nelson, J. B., Lamoureux, J. A., & León, S. P. (2013). Extinction arouses attention to the context in a behavioral suppression method with humans. Journal of Experimental Psychology: Animal Behavior Processes, 39, 99–105. doi:10.1037/a0030759 Rosas, J. M., & Callejas-Aguilera, J. E. (2006). Context switch effects on acquisition and extinction in human predictive learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 32, 461–474. doi:10.1037/0278-7393. 32.3.461 Rosas, J. M., & Callejas-Aguilera, J. E. (2007). Acquisition of a conditioned taste aversion becomes context dependent when it is learned after extinction. Quarterly Journal of Experimental Psychology, 60, 9–15. doi:10.1080/17470210600971519 Rosas, J. M., Callejas-Aguilera, J. E., Ramos-Álvarez, M. M., & Abad, M. J. F. (2006). Revision of retrieval theory of forgetting: What does make information context-specific? International Journal of Psychology and Psychological Therapy, 6, 147–166. Rosas, J. M., Todd, T. P., & Bouton, M. E. (2013). Context change and associative learning. Wiley Interdisciplinary Reviews: Cognitive Science, 4, 237–244. doi:10.1002/wcs.1225
