ARTICLE IN PRESS Behaviour Research and Therapy 43 (2005) 323–336 www.elsevier.com/locate/brt Return of fear in a human differential conditioning paradigm caused by a return to the original acquistion context Debora Vansteenwegen1,, Dirk Hermans, Bram Vervliet, Geert Francken, Tom Beckers1, Frank Baeyens, Paul Eelen Department of Psychology, Centre for the Psychology of Learning and Behaviour Therapy, University of Leuven, Tiensestraat 102, Leuven B-3000, Belgium Received 17 September 2003; received in revised form 10 November 2003; accepted 21 January 2004 Abstract In a differential human fear conditioning paradigm evidence for ABA-renewal was obtained manipulating the lighting in the experimental room. During acquisition in either a dark or illuminated room, one neutral slide was sometimes paired with a loud aversive noise whereas another slide was not. Subsequently, extinction took place in the opposite lighting context. When afterwards the participants were tested again in the original acquisition context, measurements revealed a recovery of the conditioned electrodermal response and an increase in the retrospective verbal US-expectancy ratings. No response recovery was obtained in an AAA-group that received acquisition, extinction and test trials in one and the same context. Several theoretical explanations for this type of return of fear as well as implications for clinical practice are discussed. r 2004 Elsevier Ltd. All rights reserved. Keywords: Human fear conditioning; Context-specificity; Exposure; Return of fear; Skin conductance Corresponding author. Tel: +32-16-32-61-34; fax: +32-16-32-59-24. E-mail address: deb.vansteenwegen@psy.kuleuven.ac.be (D. Vansteenwegen). Debora Vansteenwegen and Tom Beckers are postdoctoral researchers of the Fund for Scientific Research FWO-Vlaanderen. 1 0005-7967/$ - see front matter r 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.brat.2004.01.001 ARTICLE IN PRESS 324 D. Vansteenwegen et al. / Behaviour Research and Therapy 43 (2005) 323–336 1. Introduction Powerful therapeutic tools exist for the treatment of specific phobias. Most often the behavioural treatment involves repeated and systematic exposure to the fear-provoking stimulus (e.g. Öst, 1997). Although such exposure-based treatment is highly effective, there is substantial evidence in the clinical literature that fear may return with the passage of time. Rachman and colleagues (e.g. Rachman, 1989) have studied the phenomenon of return of fear extensively and explored (post)treatment and individual difference variables that predispose individuals to the return of fear. Although findings across various studies are rather mixed and it is too early to make any firm predictions, several suggestions have been made in literature. Research indicates that individual difference variables such as mood during exposure (Salkovskis & Mills, 1994) and initial level of heart rate (Wood & McGlynn, 2000; Grey, Rachman, & Sartory, 1980) may correlate with levels of return of fear. There are also indications that for example the strength of covariation bias immediately after treatment (De Jong, Vandenhout, & Merckelbach, 1995), the incompleteness of fear reduction at the end of exposure (Rachman, Robinson, & Lopatka, 1987) and the speed of fear reduction can be predictive for a return of fear. Furthermore, research indicates that distraction during exposure can have a facilitative effect on return of fear (Kamphuis & Telch, 2000). Treatment variables that might reduce the return of fear are the use of varied stimuli instead of the same stimulus during exposure (Rowe & Craske, 1998a) and the use of an expanding-spaced instead of massed exposure schedule (Rowe & Craske (1998b) though see Lang & Craske (2000), for a lack of replication). The post-treatment variable of interest for the research presented in this manuscript is contextchange after exposure. There is good evidence from clinical studies that return of fear may be facilitated by entering a new context. Mineka, Mystkowski, Hladek, and Rodriguez (1999) showed a return of self-reported fear one week after a one-session exposure-based treatment when treated spider-phobics were tested in a different room. Rodriguez, Craske, Mineka, and Hladek (1999) showed a return in the level of heart rate responding for one specific stimulus two weeks after exposure treatment when the incidental (room) and meaningful (therapist) context were changed. Finally, Mystkowski, Craske, and Echiverri (2002) manipulated the contexts in a more naturalistic way and used an inside and an outside context. They showed a clear return of fear in self-report data one week after a one-session exposure-based therapy when participants were tested in the other context. The results of these clinical studies show a striking resemblance to the substantial experimental evidence from the animal conditioning literature that conditioned responses return after a context change. These studies demonstrate a context-dependency of extinction, analogous to the contextdependency of exposure treatment. It is assumed that exposure therapy involves processes analogous to extinction: repeatedly presenting the conditioned stimulus (CS) without the unconditioned stimulus (US) after acquisition most often leads to a decrease in conditioned responding. The most frequently observation reported is ABA-renewal. Bouton and colleagues (e.g. Bouton & King, 1983; Rosas & Bouton, 1997; Bouton & Swartzentruber, 1986) demonstrated that conditioned responding is renewed when after extinction in a context different from the acquisition context, animals are again tested in the original acquisition context. The effect was not only demonstrated in traditional animal fear conditioning paradigms such as conditioned suppression (Bouton & King, 1983), but also in appetitive conditioning (Bouton & ARTICLE IN PRESS D. Vansteenwegen et al. / Behaviour Research and Therapy 43 (2005) 323–336 325 Peck, 1989) and in a taste aversion paradigm (Rosas & Bouton, 1996). Similar effects are obtained when after extinction in a different context, transfer of extinction is tested in a third context (ABC-renewal, Bouton & Brooks, 1993; Bouton & Swartzentruber, 1986). The clinical studies described above can best be compared with the latter type of renewal: exposure therapy most often takes place in a context different from the original acquisition context, and return of fear is then observed in a new, third context. Finally, AAB-renewal was also demonstrated in animals (Bouton & Ricker, 1994). In such a procedure acquisition and extinction take place in an identical context but testing is executed in a new context. Classical conditioning theory and the findings from animal conditioning studies might be very helpful in clarifying the context-specificity effect of exposure. They may provide insight in the learning mechanisms that are responsible for return of fear. Recent conditioning theories about extinction (for a review see Rescorla, 2001) suggest that extinction should not be equated with unlearning. Instead extinction would involve learning about additional context-specific information (Bouton, 1988, 1994, 2000). As a result the extinction context acquires a modulatory role and helps to disambiguate between the old (acquisition) and the new (extinction) information. Translating this idea to a clinical context, this would suggest that during exposure the original association of the fearful object with fear is not subjected to change. Rather, one learns that sometimes the fearful object is not to be feared. This additional information will only come to expression when the therapy context is present. According to this approach, expectations with regard to the generality and permanence of the effect of exposure are rather pessimistic. In the animal conditioning literature, this contextual theory of extinction has received a lot of attention. Nonetheless, there are two other mechanisms that were traditionally used to explain the context-specificity of extinction. The first mechanism is the formation of a direct inhibitory association between the extinction context and the US. In the clinical literature this inhibitory context is called a safety signal. It predicts the non-occurrence of the US and as such protects the CS from extinction (Lovibond, Davis, & O’Flaherty, 2000). When leaving the therapy context, the inhibitory power of the context/therapist is no longer present and fear will return. The second mechanism received less attention in the literature. It is possible that the fearful object is perceived differently in the exposure context than in the acquisition context. In this respect, exposure might have involved a (partially) wrong stimulus and fear will return when the original acquisition stimulus is presented. In the animal conditioning literature, this mechanism is called generalisation decrement. From an a priori point of view, these mechanisms can be at work in ABA- and ABC-renewal demonstrations in animals as well as in the clinical studies. However, Bouton and his colleagues showed that in their animal conditioning preparations the context did not become inhibitory during extinction and excluded this explanation for their renewal effects (e.g. Bouton & King, 1983, Bouton & Swartzentruber, 1986). Moreover, they also showed no generalisation decrement between acquisition and extinction, excluding the second alternative mechanism. Hence they made a very strong case in favour of the modulatory mechanism in their conditioning preparations. Because contexts and stimuli in different conditioning preparations might serve different functions due to differences in salience, meaning, distance, etc., these animal conditioning findings cannot immediately be generalised to human fear conditioning studies. All three mechanisms can in principle play a role in human ABA-renewal demonstrations. A replication of the renewaleffect in a human conditioning study with experimentally induced fear might be a first step ARTICLE IN PRESS 326 D. Vansteenwegen et al. / Behaviour Research and Therapy 43 (2005) 323–336 towards unravelling these mechanisms. Additionally, it might complement the existing evidence from clinical studies. Despite the loss of some ecological validity, the strength of such a human study with experimentally induced fear is that it brings the theoretical assumptions to their essence and allows studying the phenomenon under strictly controlled circumstances. First of all, in a clinical context one does not have control over the acquisition of fear. Therefore, studying the equivalent of an ABA-renewal effect for instance, is rather difficult. Often the original acquisition context is not known and even if it was known, recreating this context is often practically/ethically impossible. Moreover, in clinical studies it is very difficult to know what the differences are between the acquisition and the extinction context, information that might be useful when one wants to differentiate between different mechanisms. Secondly, not only acquisition but also, and importantly, extinction is strictly under control of the experimenter. This might be important if one wants to make an analysis of the underlying mechanisms. When studying return of fear after exposure-based treatment, this treatment involves often more than what is included in a simple extinction-procedure, such as modelling, counterconditioning, social reinforcement, etc. In sum, it might be worthwhile to fill the gap between clinical experiments and animal conditioning studies. This was the main intention of our research. As a first step in disentangling the possible underlying mechanisms, we wanted to demonstrate the ABA-renewal effect in a human differential fear-conditioning paradigm with visual stimuli as CSs and a loud aversive noise as US (Vansteenwegen, Crombez, Baeyens, & Eelen, 1998). Contexts were manipulated by switching the central lighting in the experimental room on or off. One group (ABA) either received acquisition and test in the dark context and extinction in the illuminated context or vice versa. For the other group (AAA) acquisition, extinction and test were conducted in the same context (all dark or all illuminated). In the first place, we aimed to demonstrate the ABA-renewal effect using a psychophysiological index namely electrodermal responding. The evidence for renewal in clinical studies is mainly based on verbal self-reports. If one wants to study the phenomenon of the return of a real emotional response such as fear, it can be important not to rely solely on verbal indices of fear, as bodily responses are an integrative part of emotional responding. Moreover, correlations between verbal, physiological and behavioural indices are often low (Ohman, 1987). In this respect, demonstrating ABA renewal with skin conductance is not without relevance. The psychophysiological index was complemented with US-expectancy ratings in order to see whether we could replicate the already existing clinical evidence for a return in verbal indices under the more strictly controlled conditions of this human fear conditioning study. However, we took the option to formulate these ratings in a retrospective in order not to interfere with the natural development of the electrodermal responses. 2. Method 2.1. Participants Forty first-year psychology students participated in order to fulfil course requirements. Twenty participants were assigned to the ABA-group and twenty to the AAA-group. They all gave informed consent and were informed that they could decline to participate at any time. ARTICLE IN PRESS D. Vansteenwegen et al. / Behaviour Research and Therapy 43 (2005) 323–336 327 2.2. Apparatus Two clearly distinct line drawings of pictorial faces served as conditioned stimuli. For half of the participants, one picture was sometimes followed by the US (CS+) and the other stimulus was not (CS), for the other half of the participants this assignment was reversed. The slides were projected with a Kodak Carousel slide projector and shutter, controlled by Labtech Notebook programme and an IBM computer, which also controlled the stimulus sequence, the presentation duration and the intertrial intervals. Slides were projected at eye level at 40 cm 60 cm. A 95 dB (A) (Bruël and Kjaer, type 2107) burst of white noise with instantaneous rise time presented binaurally for 500 ms through headphones (Sony dynamic stereo, MDR-CD270), served as US. Electrodermal activity was recorded with Fukuda standard Ag/AgCl electrodes (1 cm diameter) filled with a Unibase electrolyte and attached to the hypothenar palm of the left hand which was cleaned with tap water. The inter-electrode distance was 2.5 cm. The Coulbourn skin conductance coupler (S71-22) provided a constant 0.5 V across electrodes. The analog signal was passed through a 12 bit AD-converter and digitised at 10 Hz from 4 s prior to conditional stimulus onset until 4 s after conditional stimulus offset. Participants were seated in an armchair (screen distance 1.5 m) in a sound attenuated experimental room, adjacent to the experimenter’s room. Verbal communication was possible through an intercom system. The experimenter could manipulate the central lighting of the experimental room from the outside. When the central lighting in the room was turned off, a very low intensity dimming light was still present. 2.3. Procedure At the beginning of the experiment the participants were instructed that from time to time a loud aversive noise would be presented and that they had to watch the slides attentively. They were instructed to focus on the moments when the loud aversive noise occurred. No further instructions were given with regard to the contingencies between CSs and the US. During acquisition, extinction and test, slides were presented for 8 s. Intertrial intervals varied between 16 and 24 s (average 20 s) and electrodermal activity was recorded from 4 s before CS onset to 4 s after CS offset. During acquisition CS+ and CS were presented 10 times in a semi-randomised order. No more than two consecutive trials of CS+ or CS were presented. Eight out of 10 CS+ presentations were immediately followed by the loud noise. Within each block of five trials one CS+ was not followed by the US. The US never followed the ten presentations of the CS. During extinction, CS+ and CS were again presented 10 times each, but now both without noise, in a semi-randomised order. Finally during test, CS+ and CS were presented three times each, in a semi-randomised order. For half of participants of the ABA-group, acquisition and test occurred with the central lighting in the experimental room switched off, whereas during extinction the lighting was switched on. For the other half of the participants of the ABA-group, acquisition and test took place in a dark room whereas for extinction the central lighting was switched off. Half of the participants of the AAA-group were trained and tested in dark, the other half with the central lighting on. ARTICLE IN PRESS 328 D. Vansteenwegen et al. / Behaviour Research and Therapy 43 (2005) 323–336 After the test phase, the recording electrodes were removed and the participants were taken to the adjacent experimenter’s room. The participants were asked to complete a graph representing the evolution of their US-expectancies during the just finished experiment. The experimenter explained the meaning of the two axes. The Y-axis of the graph depicted an 11-point USexpectancy scale starting at 0 (never) and ending at 10 (always). On the X-axis four different moments were indicated. For the ABA-participants, the first moment was described as just before it became dark (/light), the second as just after it became dark (/light), the third as at the end of the dark (/light) phase and the fourth as when it became light again. For the AAA-participants similar ratings had to be given for four different moments. Because of the fact that no context-change took place in this group and hence no references to this context-change could be included in the instructions, the instructions were adjusted. The first three moments were described as the evolution (increases or decreases in expectation) across the experiment, except for the last six presentations. The rating of the fourth moment was described as the mean expectancy across these last six presentations. After the instructions, the participants of both groups saw a picture of the CS+ and were asked to give the four ratings for the CS+ first. Then they were asked to do the same for the CS. Afterwards participants had to rate the pleasantness of the US on an 11-point graphic scale (anchored: 10, unpleasant and +10, pleasant) and US-intensity on a 5-point categorical scale (weak, moderate, intense, enormous and unbearable). 3. Results 3.1. US-ratings The mean pleasantness rating for the US was 6.44 (SD=2.66), and the mean intensity score lay near the third label, namely intense (M ¼ 3:00, SD=0.59). Previous research by the authors (Vansteenwegen et al., 1998) that used this US independently of a conditioning procedure2 revealed similar ratings for the loud noise. The mean pleasantness rating was 5.833 and the mean intensity rating was 3.056. 3.2. Skin conductance Skin conductance responses were visually inspected and corrected for artefacts before they were analysed statistically. Skin conductance response amplitudes were defined as the maximal increase starting within 1–4 s after conditional stimulus onset (Prokasy & Kumpfer, 1973). Zero responses were included in all analyses in order to include the information about response frequencies as well (Dawson, Schell, & Filion, 2000). Magnitudes were range corrected using the largest (unconditioned) response (peak between 9 and 13 s during the acquisition phase) elicited by the loud aversive noise (Lykken & Venables, 1971) as the maximum range for each individual. The range-corrected response magnitudes were subjected to a square root transformation in order to normalise the distribution prior to statistical analysis. 2 In fact, the loud noise was used as a startle-eliciting probe in order to be able to measure startle eye-blink reflex modulation during picture-viewing. ARTICLE IN PRESS D. Vansteenwegen et al. / Behaviour Research and Therapy 43 (2005) 323–336 329 In Fig. 1 mean amplitudes for CS+ and CS are presented separately for the ABA-group and the AAA-group for the 10 acquisition trials, 10 extinction trials and three test trials. In this figure one can notice that both groups show acquisition and extinction, although extinction developed somewhat slower in the AAA-group than in the ABA-group. Changing the context between acquisition and extinction in the latter group did not seem to have a large impact on the learned differentiation, indicating only small, if any, generalisation decrement. In the ABA-group, after a return to the original acquisition context, the CS+/CS differential conditioned responding was clearly renewed. This renewal was not observed in the AAA-group. Statistical analyses confirm this data pattern. Mean amplitudes for CS+ and CS for five crucial moments: the first (acq1) and the last acquisition trial (acq10), the first (ext1) and the last extinction trail (ext10) and the first testtrial Fig. 1. Mean range-corrected first interval skin conductance response amplitudes during viewing CS+ and CS for ten acquisition trials (acq1–acq10), for ten extinction trials (ext1–ext10) and three test trials (test1–test3) are presented separately for the ABA-group and the AAA-group. The means are presented square rooted because this was necessary to obtain normally distributed data for statistical analyses. ARTICLE IN PRESS 330 D. Vansteenwegen et al. / Behaviour Research and Therapy 43 (2005) 323–336 (test1), for the AAA-group and the ABA-group were first analysed for each group separately. Then between-group comparisons were made within a Group (ABA/AAA) X Moment (acq1, acq10, ext1, ext10, test1) X CS-type (CS+/CS) ANOVA with repeated measurements on the last two variables.3 Only 34 participants were included in this analysis. Three participants from each group were excluded because one of the necessary data-points was not available due to technical shortcomings. Acquisition and extinction: In the ABA-group at the end of acquisition (acq 10), there was a significant CS+/CS differentiation, F ð1; 16Þ ¼ 13:63, p ¼ 0:002, MSE=0.086 that was not present at the beginning of acquisition Fð1; 16Þ41. The CS+/CS differentiation was significantly larger at acq10 than at the start of acquisition (acq1), Fð1; 16Þ ¼ 4:58, p ¼ 0:048, MSE=0.082. Furthermore, there was no significant CS+/CS differentiation left at the end of extinction, F ð1; 16Þo1 and hence a significant (CS+/CS) (acq10, ext10) was observed, F ð1; 16Þ ¼ 7:33, p ¼ 0:016, MSE=0.09, indicating a strong decrease in responding during extinction. Hence the ABA-group clearly showed acquisition and extinction. In the AAA-group, the CS+/CS differentiation at the end of acquisition was almost significant, F ð1; 16Þ ¼ 3:12, p ¼ 0:09, MSE=1.4 whereas no differentiation was obtained at the beginning of acquisition, F ð1; 16Þo1. The (acq1, acq10) CS-type interaction was not significant, F ð1; 16Þ ¼ 2:37, p ¼ 0:14, whereas the (acq1, ext1) CS-type interaction clearly was, F ð1; 16Þ ¼ 7:03, p ¼ 0:018, MSE=0.067. This is evidence for an acquisition-effect although the effect developed rather slow. Furthermore, at the end of extinction the CS+/CS-differentiation has disappeared (ext10), F ð1; 16Þo1. The decrease from beginning to end of extinction was not significant however, F ð1; 16Þ ¼ 2:097, p ¼ 0:35, indicating that extinction in this group was less strong. However, the between-group interactions concerning acquisition and extinction were all nonsignificant. The most important interaction Group CS-type (acq10, ext10) for example, revealed the following statistics, F ð1; 32Þ ¼ 1:25, p ¼ 0:27. Generalisation decrement: In light of the alternative accounts of renewal spelled out in the introduction, it is important to check the impact of the context-change between acquisition and extinction in the ABA-group. This context-change did not seem to have a strong impact on conditioned responding. The CS+/CS differentiation was still present at the first extinction trial, F ð1; 16Þ ¼ 4:73, p ¼ 0:044, MSE=0.08. Although the means show a smaller differentiation between CS+ and CS at the first extinction trial than at the end of acquisition, this differentiation was not significant, F ð1; 16Þ ¼ 1:86, p ¼ 0:18. As expected, in the control group— where no context-change occurred—no differences between the last acquisition and first extinction trial were obtained either, F ð1; 16Þo1. The means even show a larger instead of a smaller differentiation. When comparing both between-group effects directly; no significant differences were obtained. The Group CS-type interaction was neither significant at the last acquisition trial nor at the first extinction trial, both F ð1; 32Þo1, nor was the three-way interaction Group CS-type (acq10 and ext1) significant, Fð1; 32Þ ¼ 1:19, p ¼ 0:58. Renewal of conditioned responding: In the ABA-group, the difference between CS+ and CS was significant at test, Fð1; 16Þ ¼ 13:82, p ¼ 0:002, MSE=0.12. The increase in responding to the 3 When analyses were executed that included the dark/light manipulation as an additional between-group variable, there was no significant main effect of this variable nor were there significant interaction effects. ARTICLE IN PRESS D. Vansteenwegen et al. / Behaviour Research and Therapy 43 (2005) 323–336 331 CS+ from the last extinction trial towards test was significant, F ð1; 16Þ ¼ 19:59, po0:001, MSE=0.122. Furthermore, also the CS-type (ext10, test1) interaction was significant F ð1; 16Þ ¼ 11:17, p ¼ 0:004, MSE=0.079. Hence a return of conditioned responding was observed in the ABA-group. In the AAA-group, the difference between CS+ and CS during test was not significant, F ð1; 16Þ ¼ 2:92, p ¼ 0:107. The increase for CS+ from extinction to test and the CStype (Ext10, test1) interaction were not significant either, both F ð1; 16Þo1, so there was no evidence for a return in this group. When comparing the two between-group effects directly, responses to the CS+ at the first test trial were larger in the ABA-group than in the AAA-group, F ð1; 32Þ ¼ 9:10, p ¼ 0:006, MSE=0.17, as was the increase from the end of extinction to test for CS+, F ð1; 32Þ ¼ 8:9, p ¼ 0:006, MSE=0.10. When taking into account the changes for the CS, the between-group interactions were not significant: the CS+/CS differentiation at test did not differ significantly, F ð1; 32Þ ¼ 2:55, p ¼ 0:12 and also the Group CS-type (ext10, test1) interaction did not reach significance, F ð1; 32Þ ¼ 2:15, p ¼ 0:15. 3.3. Retrospective US-expectancy ratings In Fig. 2 mean US-expectancy ratings for CS+ and CS for the 4 retrospective test moments are presented, for the ABA-group and the AAA-group separately. In this figure one can notice that both groups show clear acquisition. In the AAA-group the CS+/CS differentiation becomes smaller towards the end of the experiment, indicating extinction. However the Fig. 2. Mean retrospective US-expectancy ratings are presented for CS+ and CS during four moments separately for the ABAgroup and the AAA-group: first block of acquisition (M1), second block of acquisition (M2), extinction (M3), and test (M4). ARTICLE IN PRESS 332 D. Vansteenwegen et al. / Behaviour Research and Therapy 43 (2005) 323–336 differentiation did not completely disappear. In the ABA-group, when comparing the second and the third moment, extinction is observed as well, however in this group a clear return of USexpectancy for the CS+ is observed at the last moment. Statistical analyses confirm this data pattern. Mean US-expectancy for CS+ and CS for the four moments were first analysed for each group separately. Then the between-group comparisons were calculated within a Group (ABA/ AAA) Moment (acq10, ext1, ext10, test1) CS-type (CS+/CS) ANOVA with repeated measurements on the two last variables. Acquisition and extinction: Planned comparisons confirm that both groups show acquisition. Note that the instructions with regard to these first three moments were different in the two groups. Hence between group comparisons will not be executed. For the acquisition rating (Moment 1), there was a significant CS+/CS differentiation in the ABA-group, F ð1; 19Þ ¼ 22:92, po0:001, MSE=6.98. In the AAA-group, this differentiation was manifest at moment 2, F ð1; 19Þ ¼ 161:21, po0:001, MSE=3.04, whereas the differentiation at moment 1 was only marginally significant, F ð1; 19Þ ¼ 3:29, p ¼ 0:085, MSE=3.50. In the ABA-group at the end of extinction (Moment 3) no differentiation between CS+ and CS was observed anymore F ð1; 19Þo1. In the AAA-group, the CS+ still elicited a higher US-expectancy than the CS at moment 3, Fð1; 19Þ ¼ 41:11, po0:001, MSE=6.11. However in both groups there was evidence for extinction when looking at the decrease in differentiation between the beginning of extinction (moment 2) and the end of extinction (moment 3), both F ð1; 19Þ44:78. Generalisation decrement: The US-expectancy ratings do not allow us to draw strong conclusions with regard to the generalisation decrement caused by the context-change based on between-group comparisons. It was impossible to create a reference point for the AAA-control group that was comparable with the obvious context-change present in the ABA-group. For the ABA-group however, no significant decrease in CS+/CS differentiation was observed after the context-change, F ð1; 19Þo1, and the CS+/CS differentiation at the beginning of extinction (moment 2) was highly significant, F ð1; 19Þ ¼ 33:89, po0:001, MSE=5.20. Renewal of conditioned responding: In the ABA-group the difference between CS+ and CS was significant at test, F ð1; 19Þ ¼ 34:63, po0:001, MSE=4.79. The increase for the CS+ from the last extinction trial towards the test was significant, F ð1; 19Þ ¼ 10:54, p ¼ 0:004, MSE=7.20. Furthermore, also the CS-type (ext10, test) interaction was significant, F ð1; 19Þ ¼ 10:65, p ¼ 0:004, MSE=5.24, all indicating renewed conditioned responding after the context change. In the AAA-group, the difference between CS+ and CS during test was also significant, F ð1; 19Þ ¼ 7:01, p ¼ 0:015, MSE=2.69, indicating that also in the control group some conditioned responding was still present. However this can not be taken as evidence for a return because the US-expectancy for the CS+ diminished from extinction to test, F ð1; 19Þ ¼ 26:63, po0:001, MSE=3.96. Also when taking into account the changes for CS, a significant CStype (Ext10, test) interaction was obtained, F ð1; 19Þ ¼ 18:88, po0:001, MSE=3.43 indicating a continuing decrease. When comparing the two between-groups directly, responses to CS+ at test were larger in the ABA-group than in the AAA-group, F ð1; 39Þ ¼ 12:55, po0:001, MSE=7.28 and also the increase from the end of extinction to the test for CS+ was significantly larger for the ABA-group, F ð1; 39Þ ¼ 32:12, po0:001, MSE=5.60.When taking into account the changes for the CS as well, significant between-group interactions were obtained: The CS+/CS differentiation at test ARTICLE IN PRESS D. Vansteenwegen et al. / Behaviour Research and Therapy 43 (2005) 323–336 333 was stronger in the ABA-group than in the AAA-group, F ð1; 39Þ ¼ 9:18, p ¼ 0:001, MSE=3.77, and also the Group CS-type (ext10, test) interaction was significant, F ð1; 39Þ ¼ 27:66, po0:001, MSE=4.36, indicating a return of responding in the ABA-group that was not observed in the AAA-group. 4. Discussion It was the intention of this study to demonstrate a return of conditioned electrodermal responding after extinction when returning to the original acquisition context. The experimental as well as the control group showed clear acquisition. Acquisition developed somewhat slower in the AAA-group as on the last acquisition trial no significant CS+/CS differentiation was observed yet. However, the data from the first extinction trial clearly show that these participants have learned the differentiation. This trial can be considered as a valid assessment of acquisition because the measurement of the skin conductance took place before the US-omission and no context-change took place between acquisition and extinction in this group. Extinction took place in both groups: At the end of extinction no differentiation between CS+ and CS was left anymore. In contrast to the ABA-group, in the AAA-group we did not obtain a significant decrease in differentiation from the beginning to the end of extinction. This suggests that extinction effects in this group were not really strong. However, it was clear from the data that in this group at the end of extinction and during test no significant CS+/CS differentiations were observed anymore. Given the evidence for acquisition and extinction of differential electrodermal responding, we can start interpreting the data with respect to our main hypothesis regarding a return of electrodermal responding after a context-change. The data demonstrate a clear return of conditioned responding in the ABA-group, whereas in the AAA-group no such return was obtained. There was in the former group a significant difference between CS+ and CS at test that was not observed in the AAA-group; also the difference in responses to the CS+ between the end of extinction and the test was significantly larger for the ABA-group than for the AAA-group. We consider this as good evidence for ABA-renewal. It demonstrates that extinction-effects are context-specific and that electrodermal conditioned responding may recover after a return to the original acquisition context. One might have noticed that no significant between-group interactions were obtained when the CS+/CS differentiation was taken into account, however they were significant when only CS+ was taken into account. Significant effects based on a CS+/ CS differentiation are known to be more difficult to obtain for two reasons. First, these comparisons only take into account influences that are specific to associative conditioning and exclude all alternative influences such as familiarity of the context that might influence CS+ as well as CS. In this respect, they are stricter than some other conditioning procedures. Second, it is known that in these human conditioning paradigms the unreinforced presentation of the CS+ during test can have immediate impact on the responses to the subsequent presentation of CS and vice versa (e.g. Lovibond, 2003). The retrospective US-expectancy ratings corroborate the skin conductance data pattern. Both groups showed acquisition as well as extinction. As expected, the control group showed a steady decline in differentiation towards the end of the experiment, whereas the ABA-group showed an ARTICLE IN PRESS 334 D. Vansteenwegen et al. / Behaviour Research and Therapy 43 (2005) 323–336 increase in differentiation from the third to the fourth moment, indicating that the return to the original acquisition context caused a recovery effect. Note that these ratings were taken retrospectively and should be interpreted with caution. However, a recovery-effect was obtained despite the fact that the subjects in fact experienced no loud noises in the test phase. This makes clear that the participants were able to interpret the graph and the US-expectancy scale in a proper way, also it makes it less probable that the ratings exclusively reflected demand-effects. Nevertheless, these data can only be suggestive for effects that one would have obtained with online verbal ratings. With regard to the mechanisms that may underlie the return of fear in humans we can only draw some preliminary conclusions. The fact that we have no evidence for a generalisation decrement between acquisition and extinction in the skin conductance data, at least makes one of the three mechanisms less plausible. It seems not to be the case that the CS was perceived differently in the extinction context than in the acquisition context. The two other mechanisms can still play a role. First, participants may have learned that the extinction context is safe. A direct inhibitory association may have developed between the context and the US, protecting the CS from extinction. Second, the context might have come to play a modulatory role, arbitrating between situations in which the CS is followed by the US and situations in which the CS is not followed by the US. It is reasonable to assume that also in a clinical context both mechanisms can play a role. On the one hand, the therapy room and the presence of the therapist can become safety signals, paradoxically protecting the fear-object from extinction. On the other hand, one might learn an exception to the rule such as that the fear-object is no longer dangerous in the therapy context or in the presence of the therapist, but remains dangerous once outside. One of the aims for further research will be to try to disentangle these two mechanisms and see whether both or only one of these mechanisms is responsible for the effects obtained. Disentangling the mechanisms is only interesting for the clinical practice when these mechanisms suggest different methods for preventing the return of fear. In general terms, one could state that in the case of a modulatory role of the context, the advice would be to equate as much as possible the exposure context with the context wherein the patient has to function in the future. Equating the exposure context with the relevant contexts for the future would also be beneficial in case the context became directly inhibitory and the context-dependency effect is based on the second mechanism. On the other hand, focusing on an equation of the exposure context with the original acquisition context is only beneficial when one wants to prevent that the context and the therapist become safety-signals. In the animal conditioning literature, two additional methods are described in order to reduce return of fear after a context change. It seems that these methods are effective in reducing return of fear that is based on either of the two mechanisms described above. A first method is to vary the contexts or stimuli used during exposure therapy. There is evidence from two animal conditioning experiments that conducting extinction in different contexts might be effective in reducing the response recovery when entering a new context (Chelonis, Calton, Hart, & Schachtman, 1999; Gunther, Denniston, & Miller, 1998). Also in the clinical study of Rowe and Craske (1998a, b) there is evidence that using varied stimuli is effective in reducing the return of fear when confronted with a new stimulus. Indeed, using multiple contexts as well as using different stimuli might prevent that the context protects the CS from extinction. Moreover, it also enhances the likelihood that the additional association formed during extinction becomes active in ARTICLE IN PRESS D. Vansteenwegen et al. / Behaviour Research and Therapy 43 (2005) 323–336 335 several new contexts. A second method is to make use of retrieval cues that help to reactivate the extinction information. Brooks and Bouton (1994) (see also Brooks, Palmatier, Garcia, & Johnson, 1999, for the effect on spontaneous recovery) demonstrated in an animal conditioning experiment that presenting a cue during extinction as well as during test overcomes part of the recovery of the conditioned responses due to context-change. As long as retrieval cues contribute to the retrieval of the extinction information, they may be helpful in preventing return of fear. Either when the direct context–US association is responsible or when the CS–noUS association is under control of the context, making the extinction information more prominent might be beneficial. Regardless of the underlying mechanism that might be responsible for the effects, the present data are quite convincing evidence for the context-specificity of extinction in a human fear conditioning paradigm using electrodermal responding and retrospective US-expectancy ratings as indices of learning. This study fills the gap between the demonstrations in the animal conditioning literature on the one hand and the clinical demonstrations of a return of fear due to a context change after successful treatment. References Bouton, M. E. (1988). Context and ambiguity in the extinction of emotional learning: Implications for exposure therapy. Behaviour Research and Therapy, 26, 137–149. Bouton, M. E. (1994). Conditioning, remembering, and forgetting. Journal of Experimental Psychology: Animal Behavior Processes, 20, 219–231. Bouton, M. E. (2000). A learning theory perspective on lapse, relapse and the maintenance of behavior change. Health Psychology, 19, 57–63. Bouton, M. E., & Brooks, D. C. (1993). Time and context effects on performance in a Pavlovian discrimination reversal. Journal of Experimental Psychology: Animal Behavior Processes, 19, 165–179. Bouton, M. E., & King, D. A. (1983). Contextual control of the extinction of conditioned fear: Tests for the associative value of the context. Journal of Experimental Psychology: Animal Behavior Processes, 9, 248–265. Bouton, M. E., & Peck, C. A. (1989). Context effects on conditioning, extinction, and reinstatement in an appetitive conditioning preparation. Animal Learning & Behavior, 17, 188–198. Bouton, M. E., & Ricker, S. T. (1994). Renewal of extinguished responding in a second context. Animal Learning & Behavior, 22, 317–324. Bouton, M. E., & Swartzentruber, D. (1986). Analysis of the associative and occasion-setting properties of contexts participating in a Pavlovian discrimination. Journal of Experimental Psychology: Animal Behaviour Processes, 12, 333–350. Brooks, D. C., & Bouton, M. E. (1994). A retrieval cue for extinction attenuates response recovery (renewal) caused by a return to the conditioning context. Journal of Experimental Psychology: Animal Behaviour Processes, 20, 366–379. Brooks, D. C., Palmatier, M. I., Garcia, E. O., & Johnson, J. L. (1999). An extinction cue reduces spontaneous recovery of a conditioned taste aversion. Animal Learning & Behavior, 27, 77–88. Chelonis, J. J., Calton, J. L., Hart, J. A., & Schachtman, T. R. (1999). Attenuation of the renewal effect by extinction in multiple contexts. Learning and Motivation, 30, 1–14. Dawson, M. E., Schell, A. M., & Filion, D. L. (2000). The electrodermal system. In J. T. Cacioppo, L. G. Tassinary, & G. G. Berntson (Eds.). Handbook of psychophysiology (pp. 200–224). Cambridge: Cambridge University Press. De Jong, P. J., Vandenhout, M. A., & Merckelbach, H. (1995). Covariation bias and the return of fear. Behaviour Research and Therapy, 33, 211–213. Grey, S. J., Rachman, S., & Sartory, G. (1980). Tonic level in heart-rate and return of fear. Biological Psychology, 11, 268. ARTICLE IN PRESS 336 D. Vansteenwegen et al. / Behaviour Research and Therapy 43 (2005) 323–336 Gunther, L. M., Denniston, J. C., & Miller, R. R. (1998). Conducting exposure treatment in multiple contexts can prevent relapse. Behaviour Research and Therapy, 36, 75–91. Kamphuis, J. H., & Telch, M. J. (2000). Effects of distraction and guided threat reappraisal on fear reduction during exposure-based treatments for specific fears. Behaviour Research and Therapy, 38, 1163–1181. Lang, A. J., & Craske, M. G. (2000). Manipulations of exposure-based therapy to reduce return of fear: A replication. Behaviour Research and Therapy, 38, 1–12. Lovibond, P. F. (2003). Causal beliefs and conditioned responses: Retrospective revaluation induced by experience and by instruction. Journal of Experimental Psychology: Learning, Memory and Cognition, 29, 97–106. Lovibond, P. F., Davis, N. R., & O’Flaherty, A. S. (2000). Protection from extinction in human fear conditioning. Behaviour Research and Therapy, 38, 967–983. Lykken, D., & Venables, P. H. (1971). Direct measurement of skin conductance: A proposal for standardization. Psychophysiology, 8, 656–672. Mineka, S., Mystkowski, J. L., Hladek, D., & Rodriguez, B. I. (1999). The effects of changing contexts on return of fear following exposure treatment for spider fear. Journal of Consulting and Clinical Psychology, 67, 599–604. Mystkowski, J. L., Craske, M. G., & Echiverri, A. M. (2002). Treatment context and return of fear in spider phobia. Behaviour Therapy, 33, 399–416. Ohman, A. (1987). The psychophysiology of emotions: An evolutionary-cognitive perspective. Advances in Psychophysiology, 2, 79–127. Öst, L. G. (1997). Rapid treatment of specific phobias. In G. Davey (Ed.). Phobias: A handbook of theory, research and treatment (pp. 227–246). Chichester: Wiley. Prokasy, W. F., & Kumpfer, K. L. (1973). Classical conditioning. In W. F. Prokasy, & D. C. Raskin (Eds.). Electrodermal activity in psychological research (pp. 157–203). Academic Press: New York. Rachman, S. (1989). The return of fear: Review and prospect. Clinical Psychology Review, 9, 147–168. Rachman, S., Robinson, S., & Lopatka, C. (1987). Is incomplete fear-reduction followed by a return of fear?. Behaviour Research and Therapy, 25, 67–69. Rescorla, R. A. (2001). Experimental extinction. In R. R. Mowrer, & S. B. Klein (Eds.). Handbook of contemporary learning theories (pp. 119–154). Mahwah, NJ: Lawrence Erlbaum Associates. Rodriguez, B. I., Craske, M. G., Mineka, S., & Hladek, D. (1999). Context-specificity of relapse: Effects of therapist and environmental context on return of fear. Behaviour Research and Therapy, 37, 845–862. Rosas, J. M., & Bouton, M. E. (1996). Spontaneous recovery after extinction of a conditioned taste aversion. Animal Learning & Behavior, 24, 341–348. Rosas, J. M., & Bouton, M. E. (1997). Renewal of a conditioned taste aversion upon return to the conditioning context after extinction in another one. Learning and Motivation, 28, 216–229. Rowe, M. K., & Craske, M. G. (1998a). Effects of an expanding-spaced vs massed exposure schedule on fear reduction and return of fear. Behaviour Research and Therapy, 36, 701–717. Rowe, M. K., & Craske, M. G. (1998b). Effects of varied-stimulus exposure training on fear reduction and return of fear. Behaviour Research and Therapy, 36, 719–734. Salkovskis, P., & Mills, I. (1994). Induced mood, phobic responding and the return of fear. Behaviour Research and Therapy, 32, 439–445. Vansteenwegen, D., Crombez, G., Baeyens, F., & Eelen, P. (1998). Extinction in fear conditioning: Effects on startle modulation and evaluative self-reports. Psychophysiology, 35, 729–736. Wood, B. S., & McGlynn, F. D. (2000). Research on posttreatment return of claustrophobic fear arousal, and avoidance using mock diagnostic imaging. Behavior Modification, 24, 379–394.