Investigation of extinction learning in multi-context paradigm

The exposure therapy can be stressful for participants, so they might refuse to start or continue the treatment. The modal verb “might” in this case represents a lack of information because a significant amount of people with specific phobias does not seek any treatment. For example, Seim and Spates (2009) mention that among college students with specific phobias, 30% were agreeing to receive treatment if it also provides them with extra credit, but only 18% were still interested in treatment even with an absence of any academic benefit. Among those people who came to the clinic, dropout rates notably vary from one research to another due to the high number of possible protocols of exposure therapy. For example, Garcia-Palacios et al (2007) study that included 150 participants demonstrated a high difference between in vivo and virtual exposure (dropout rates were 27% and 3% respectively) while a meta-analysis of 23 studies with 608 participants performed by Opris et al (2012) does not reveal a significant difference between dropout rates in real and virtual exposure: it was about 10% in both cases. Comparison of dropout rates of exposure therapy and other psychotherapeutic interventions also meets many methodological difficulties connected with the fact that most studies are not randomized and there could be initial dissimilarities between those who received different kinds of therapy. According to meta-analysis performed by Swift & Greenberg (2014), in most conditions dropout rates for the exposure therapy are not higher than for other interventions. Moreover, patients with anxiety disorders may estimate exposure therapy as a less stressful method comparing with its estimations made by clinicians who are trying to predict patients' expectations, and patients often prefer it to pharmacotherapy (Olatunji et al, 2009). When patients rate different psychotherapeutic approaches by “usefulness” and “likeness” scales, exposure therapy is generally marked as highly useful. Its likeness is lower than for any other technique, but the differences are relatively small: in Cox, Fergus & Swinson (1994) study, mean likeness rates for exposure therapy, breathing exercise or education about panic were 4.96, 5.86 and 6.00 points by 8-point Likert scale, respectively.

The phenomenon of fear recovery is relatively feebly marked in clinical treatment of specific phobia in comparison with laboratory experiments with fear conditioning because treatment commonly continues as long as is needed for the significant attenuation of fear. Choy et al (2007) considered 10 studies dedicated to different specific phobias and reported that the treatment outcome was maintained or even improved with the passage of time. However, follow-up measures are generally based on self-reports, and what is interesting there is that participants might overestimate the stability of their own results. To illustrate this, authors describe four studies dedicated to flight phobia. In three of them, the follow-up has been based on self-reports and patients “reported improved subjective anxiety or flying activity over time”. However, many factors remain unknown and uncontrollable: whether responders took sedative drugs before the flight? Which flights did they choose? Who flights together with them? Would they fly at all if they had an opportunity to take a train or stay home? The fourth study (Ost, Brandberg & Alm, 1997), sponsored by Scandinavian Airlines System, used a real domestic flight to estimate immediate and long-term effects of therapy. It demonstrated that, although 93% of participants were able to fly immediately after the treatment, only 64% of them agreed to perform this task in a one year after treatment.

1.2 Approaches to facilitate fear extinction

Both fear extinction procedure in laboratory conditions and exposure therapy in a clinical setting can be modified by many ways to influence the efficacy and stability of achieved results. The conditioning and extinction protocols may vary in number and nature of stimuli and contexts, sequence and duration of events, additional influences on the internal state of participants, including pharmacological or electrophysiological stimulation. This chapter does not cover all possible modifications of the procedure but provides a brief overview of these approaches which can be applied to human studies and potentially improve the clinical outcomes of the therapy.

Behavioral approach: overview.

Inhibitory learning model suggests that the fear extinction does not lead to the erasure of previous fear memory but connected with the formation of a new inhibitory association between CS and US. Importantly, this assumption not only demonstrates accordance with collected data but also allows to propose new modifications of exposure treatment which potentially could improve its efficacy. Michelle Craske, one of the leading researchers in this field, points out that anxious individuals who poorly response to typical exposure treatment may have deficits in inhibitory learning (Craske et al, 2014). From this point of view, the key goal of the therapy is not a habituation of participant to fearful stimuli, but enhancement of learning, which can be achieved via expectancy violation (unpredictable sequence of fearful and neutral stimulation), variability of stimuli and contexts, sophisticated timeline where the association between US and CS is sometimes reinforcement even on extinction stage and so on - the major priority is to maintain high coning of attention during the whole treatment session.

Weisman and Rodebaugh (2018) develop ideas proposed by Craske and her colleagues and present a high-detailed list of possible improvements of exposure therapy intended both to develop non-threat associations and enhance retrieval of newly-learned associations. They consider not only modifications of the exposure protocols itself, but also additional factors which can improve learning, such as physical activity (it is commonly accepted that both chronic and acute physical activity influence learning, for example, due to increase of BDNF synthesis; see Szuhany, Bugatti & Otto, 2015, for the review), or inducing of positive affect before the treatment which potentially can influence an emotional coloring of memory concerning presented stimuli.

However, both reviews mentioned above are focused on applied goals of cognitive behavioral therapy and implicitly suggest high variability between participants, their fears and their behavior between treatment sessions. From the purely scientific point of view, more promising findings can be obtained from better controlled laboratory experiments both with animals and humans. For example, it allows more thoroughly consider optimal time intervals between fear conditioning and extinction. In the experiments conducted by Johnson, Escobar, and Kimble (2010), rats received extinction training either in 12 minutes or in 24 hours after fear acquisition. Then different groups of animals were tested for spontaneous recovery of the fear after different amounts of time. It has been demonstrated that those who were tested in 48 hours after the extinction training demonstrated better suppress of fear if there was a delay between fear acquisition and fear extinction. Conversely, those who were tested in 7 days after the training elicit less spontaneous recovery of fear in case if the extinction training immediately followed the fear acquisition. Hence, authors conclude that delayed extinction is beneficial on a short-term, whereas immediate extinction demonstrates better results in the long run. However, they mention other studies which do not support this implication. Accordingly, the relationship between delays and efficacy of fear extinction may be more complex than it is suggested at a current stage.

Another question important both for laboratory studies and clinical applications is the efficacy of massive extinction. This approach supposed that increasing the number or frequency of conditioned stimulus during extinction may lead to deeper extinction learning. However, results of different studies in this area contradict each other. According to Laborda and Miller (2013), massive extinction with a high number of presented stimuli leads to significant attenuation of the return of fear. According to Urcelay, Wheeler, and Miller (2009), increasing of time spacing between stimuli leads to better fear reduction than massive extinction. According to the review written by Fitzgerald, Seemann, and Maren (2014) which consider these articles and number of others exploring efficacy of massed extinction in comparison with temporally spaced presentation of conditioned stimulus, it is difficult to make any conclusion at the current stage, because results of different experiments widely vary depending on subjects and particular parameters of stimulation.

It has been already mentioned that learning can be improved by physical exercise, but even more important factor is sleep, which is crucial for memory consolidation and many other aspects of learning (for a review, see Rasch & Born, 2013). To test whether sleep facilitates fear extinction, Pace-Schott at al. (2009) conducted a research where participants either were sleeping or were awaking after fear acquisition and extinction. There were two conditioned stimuli (colored lamps) while extinction procedure included only one of them. After 12 hours participants were tested again. This particular study does not reveal better extinction learning among sleep group, but brings even more promising results: sleep group has demonstrated generalization of fear extinction from the one to another colored lamp. Later Pace-Schott and his colleagues have also demonstrated on spider-fearful participants that sleep facilitate retention of the fear extinction, and, again, leads to better generalization (Pace-Schott, Verga, Bennet & Spencer, 2012). However, Pace-Schott has also demonstrated that such effects might be connected not only with sleep, but also with the time of a day, and particularly with circadian changes in hormone levels (Pace-Schott et al., 2013), so it is important to take these factors into consideration if the study connected with sleep.

Behavioral approach: focus on multi-context paradigm.

It is generally agreed that exposure to a conditioned stimulus in several contexts improves generalization of the extinction and leads to attenuated fear renewal while testing occurs in unfamiliar context (Craske et al, 2014). Importantly, even on early stages of exploration of this phenomenon, it had been already obvious that it has the high potential for improving the clinical treatment. One of the first articles focused on multiple contexts extinction (Gunther, Denniston and Miller, 1998) already includes a recommendation for clinicians to conduct exposure therapy in multiple settings. The article itself described two experiments in rats. In the first experiment, animals were deprived of water but during the fear acquisition phase, conducting in a context A, received a foot shock combined with a conditioned stimulus, a white noise, while they attempt to drink. After conditioning, one group received no extinction at all, the second went through fear extinction in a context B, and the third in three contexts B, C and D. The contexts were different cages with different combinations of lights (houselight and flashing light) and with the presence or absence of odor cue (methyl salicylate, an organic ester typically presented in essential oils produced by several plants; it smells like a mint chewing gum). During the test, which occurs on the 15th day after extinction in a new context E, researchers measure how long does it take for the thirsty rat to dare drinking for 5 seconds. In accordance with their hypothesis, those who had gone through exposure in three different contexts were notably braver than those who experienced extinction in only one context, who, in turn, were braver than those who did not received extinction at all. In the second experiment, the conditions were generally the same. There were two groups of rats. The first received fear acquisition training in a single context and then received fear extinction training in three different contexts. The second has gone through the fear acquisition in three different contexts and then was exposed to conditioned stimulus alone in three other contexts. As it had been expected, for the second group it takes more time to drink water in the test conditions. Hence, authors also pointed that difficulties in exposure therapy might be explained by the phobia acquisition in multiple contexts, and the further increase of the number of therapeutic contexts can be useful.

In humans, multiple contexts paradigm also leads to reducing fear renewal. Importantly, its efficacy has been demonstrated not only in experiments with conditioned fear (for example, Glautier, Elgueta and Nelson, 2013), but also in samples with specific phobias. Vansteenwegen et al. (2007) worked with spider-fearful participants who were exposed to a set of movies depicted rooms (basement, bathroom, kitchen and living room) with or without a spider moving across a piece of furniture. There were three groups of participants. Initially, all of them watched one-minute movie with a spider. Then control group watched 11 movies with the same room and without a spider, single exposure group watched 11 movies with the same room and with a spider, and multiple exposure group watched 11 movies with three different rooms with a spider. In a post-test, all three groups watched again the first video, and for the new test trial, all three groups watched a video with a spider in a new context. Level of fear was measured by skin conductance response and self-reports. Both exposure groups demonstrated an equal level of fear extinction during the post-test. In a new test, as predicted, single context group demonstrated some fear renewal while the multiple context group did not.

Similar results have been obtained in another sample of spider phobic participants using virtual reality exposure (Shiban et al, 2012). The multiple context exposure group observed a spider in four virtual rooms differ by color while control group observed a spider four times in the same room. In the renewal test, all participants observed a spider in a new room. Fear was measured by skin conductance level, self-reports in 60-seconds intervals during all stages of the experiment and by behavioral approach tests administered before and after the procedure. All measurements demonstrated that multiple exposure attenuates fear renewal and lead to better results in post-test BAT than in single context group.

To the best of our knowledge, there are no studies of multi-context paradigm where pictorial stimuli had been used. It is presumably connected with the fact that the multi-context paradigm itself is relatively novel and most studies still intended to explore its efficacy in general, so researchers prefer more realistic types of stimulation to achieve a stronger response. On the other hand, specifically in case of spider phobia usage of pictures is also quite rare and connected not with the exposure per se but with some auxiliary studies of neural responses to fearful objects (e.g. Wendt, Lotze, Weike, Hosten & Hamm, 2008; Leutgeb, Schafer & Schienle, 2009), because this fear is common and well characterized, and researchers are always able to obtain videotapes or virtual reality, or just living spiders for the more effective exposure. In case of our study, there is also no intention to propose usage of pictures of spiders as a solitary instrument to treat spider phobia. Although we suggest that it might be useful in early stages of treatment for those participants who reject common methods of exposure, but mainly we consider this well-characterized fear as a model to explore more general properties of multi-context images presentation.

Pharmacological approach.

There is growing number of research demonstrates that fear extinction can be facilitated by pharmacological interventions. Dozens of substances have been tested on rodents, and several of them have already revealed promising results in humans. According to the review written by Fitzgerald et al (2014), the main categories of pharmacological agents which may enhance extinction are amino acid receptor modulators (for example, D-cycloserine, partial agonist of N-methyl-D-aspartate receptor); monoamine modulators (for example, propranolol, в adrenergic antagonist); cholinergic, cannabinoid, and peptide modulators (including cannabidiol, non-psychotomimetic cannabinoid); steroid hormone modulators (such as cortisol, agonist of steroid receptors); and several others. Mentioned examples are the substances which had been tested in humans (however, it is not a whole list of such substances).

Probably the most extensively studied fear extinction facilitator is the D_cycloserine. From the middle of the XX century, it has been used in tuberculosis therapy, which helped to implement it into the clinical practice relatively fast after the discovery of its effects on the brain, without extended safety research (besides, doses which are used in psychotherapy are significantly lower than those used in tuberculosis treatment).

D_cycloserine is a partial agonist of N-methyl-D-aspartate (NMDA) glutamate receptor. As it has been already mentioned, NMDA receptor is remarkable because its activation occurs in case of coincidence of two events: glutamate signal coming from a presynaptic terminal and postsynaptic depolarization (see Purves et al, 2012). This, it registered neural signals which came simultaneously from two directions. In case of this coincidence, it allows calcium to pass into the neuron, which, in turn, initiates molecular cascades associated with long-term potentiation, and, generally, with learning and memory formation.

It is generally accepted that D-cycloserine interacts with the glycine-binding site of the NMDA receptor and facilitates its function during extinction. Administered before the extinction training, it leads to a notable increase in the efficacy of extinction in both animals and humans (Norberg, Krystal & Tolin, 2008; Schade & Paulus, 2016). Today it is extensively studied as a potential component of treatment for schizophrenia, depression, autism and even Alzheimer's disease, but the most widely studied area is the enhancing of an exposure therapy.

Generally, meta-analyses which include animal studies demonstrated notably higher effect sizes than those which framed only by humans (Cohen's d=1.19 vs .42, according to a comparison made by Norberg et al, 2008). This difference can be explained by contamination effects of other simultaneously taken drugs and comorbid diseases (Rodrigues et al, 2014). It is also possible to suggest that human samples are generally more heterogeneous, and studies in humans are generally less controllable and influenced by a number of unexpected confounders.

It is important to mention that the Cochrane collaboration is skeptical about D_cycloserine. In 2015, it reviewed 21 randomized clinical trials dedicated to the facilitation of cognitive and behavioral therapies for PTSD, OCD, specific phobias and several other diseases and came to a conclusion that at the time of a publication there were no significant evidence of the superiority of D_cycloserine comparing to placebo. Authors mentioned that despite some promising data from the individual studies, the overall quality of trials is low, clinical reports are incomplete, sample sizes are small and it the moment they cannot recommend D_cycloserine as a treatment of choice (Ori et al, 2015).

Another widely studied group of pharmacological agents are the monoamine modulators, particularly the adrenergic antagonists. They interact with adrenergic receptors and decrease adrenergic signaling. Most of the drugs from this group were initially developed to treat cardiovascular diseases since they are able to attenuate effects of adrenaline on heart and vessels. However, later it turns out that adrenergic antagonists also can influence norepinephrine signaling in the brain.

Norepinephrine, or noradrenaline, is a neurotransmitter which plays a crucial role in the formation of emotional memories, and particularly in fear processing (for the review, see Mueller & Cahill, 2010). Rodent studies demonstrate that infusion of norepinephrine into basolateral amygdala enhances memory, while the infusion of noradrenergic в receptors agonists blocks enhancement of memory consolidation. в adrenergic antagonist called propranolol administered before extinction training facilitates fear extinction in both rodents and humans (Fitzgerald et al, 2014).

The concept of memory consolidation and reconsolidation is important in a context of any learning, including fear acquisition and fear extinction. These processes require protein synthesis, and it has been manifoldly demonstrated that admission of protein synthesis inhibitors disrupts the formation of new memories (Hernandez & Abel, 2008). It is suggested that administration of propranolol also indirectly impairs protein synthesis. Although particular molecular mechanisms underlying propranolol effects on protein synthesis are still not described in full detail, the causal connection between propranolol and disruption of memory consolidation and reconsolidation is well established in multiple behavioral studies (Lonergan, Olivera-Figueroa, Pitman, & Brunet, 2013).

Effects of propranolol on fear extinction in humans enlight that the process of fear extinction itself is complex and multicomponent, and different aspects of fear memory can be more or less prone to pharmacological intervention. For example, Bos, Beckers, and Kindt (2012) demonstrated that admission of propranolol does not influence extinction on a physiological level, estimated by startle reflex and galvanic skin response, but significantly decrease the subjective fear reported by participants. More importantly, experiments with propranolol help to study a borderline between emotional and declarative memory about the same event. Participants who received propranolol shortly before or shortly after exposure to fearful object or reminder about traumatic event demonstrate a notable decrease in emotional response, although declarative memory about this thing remains completely intact (Kindt, Soeter & Vervliet, 2009; Soeter & Kindt, 2012).

Neurobiological approach.

A growing number of studies demonstrates that non-invasive brain stimulation (NIBS) can influence a wide range of cognitive functions, such as attention, learning, memory (Coffman, Clark & Parasuraman, 2014), fluid intelligence (Santarnecchi et al., 2013), cognitive control (Feeser, Prehn, Kazzer, Mungee & Bajbouj, 2013), social conformity (Klucharev, Munneke, Smidts & Fernбndez, 2011). Hence, there are no theoretical barriers to suppose that fear extinction learning also can be enhanced through transcranial stimulation.

However, there are important practical limitations. It has been already mentioned in the previous chapters that the main brain regions linked to brain acquisition and extinction are amygdala, hippocampus, ventromedial prefrontal cortex (vmPFC) and dorsal anterior cingulate cortex (dACC). No one of them is located superficially in the brain, which generates obvious difficulties for the noninvasive stimulation of these areas. The deepest stimulation can be achieved through transcranial magnetic stimulation (TMS) by the double-cone coil (Hardwick, Lesage & Miall, 2014). However, even this coil cannot efficiently and safely stimulate areas located deeper than 4 cm from the surface of a head (Deng, Lisanby & Peterchev, 2013). Hence, at the current stage of technological development, it seems impossible to directly stimulate amygdala or hippocampus without invasive electrodes. Stimulation of dACC is theoretically achievable even though there are sporadic evidence (for example, De Ridder, Vanneste, Kovacs, Sunaert & Dom, 2011). This approach is technically challenging and quite irrelevant to studies of fear extinction because this area is predominantly associated with a fear acquisition. Transcranial magnetic stimulation of vmPFC is also technically challenging comparing with stimulation of more superficial areas, but achievable and described in some publications (for example, Lev-Ran, Shamay-Tsoory, Zangen & Levkovitz, 2012). Thus, it might potentially be a target area for investigation of fear extinction by brain stimulation. Also, it is important to take into account that those brain areas which are located too deep to be achieved by TMS or any electrical stimulation directly, still can be affected through the transsynaptical spreading of excitatory stimulation effects (Pascual-Leone et al, 1998). Thus, the indirect facilitation of activity in the amygdala and hippocampus can be considered.

Due to above mentioned technical problems, there is a limited number of research dedicated to the enhancement of fear extinction through NIBS in humans. A recent review on brain stimulation for facilitating fear extinction (Marin, Camprodon, Dougherty & Milad, 2014) mostly focused on invasive techniques implemented for the treatment of severe OCD and on animal studies. In a chapter dedicated to non-invasive brain stimulation of fear extinction, they mentioned only two studies, both dedicated to PTSD and stimulation of dlPFC as an area which abnormalities are associated with PTSD development and which is easier to achieve with respect to vmPFC. The first study (Osuch et al., 2009) combined TMS with exposure therapy and revealed improvement of hyperarousal symptoms; but there were only nine participants and the sample was quite heterogeneous. The second study administered TMS sessions alone, without any additional behavioral task, and measured dynamic of symptoms according to questionnaires. Results were also promising, and also authors demonstrated that high-frequency repetitive TMS (20 Hz) is not associated with any negative effect in patients with PTSD (Boggio et al., 2010).

Despite these promising results, the number of studies dedicated to TMS application in the treatment of fear-related disorders is unexpectedly growing too slow. It is possible to suggest that this phenomenon is connected with the impressive success of pharmacological enhancement. Every particular researcher may prefer to invest resources in the area where experiments are easier and results are stronger. Neuroenhancement is a risky area; results of the highly labor-consuming studies are sometimes vague or reverse to anticipated, like in a recent tES study where stimulation actually increased the fear response (Abend et al, 2016).

It is possible to suggest, however, that promising improvements can be found out at the junction of neuroenhancement and pharmacological enhancement, as shown in animal studies (Nasehi, Khani-Abyaneh, Ebrahimi-Ghiri & Zarrindast, 2017).

In the final count, it does not matter which method of facilitation of fear extinction turns out to be the best. What is important is to find a way not to be afraid anymore.

2. Research accomplishment

We have exposed spider-fearful participants to images of spiders in one or several contexts and compared level of fear decreasing after the exposure using overt (self-reports, BAT) and covert (SGR) responses. The results partially confirm our initial hypothesis which suggests that those who have seen spiders in different contexts will demonstrate a more prominent decrease of fear.

2.1 Materials and methods

Participants were recruited via massive online survey. On the first page of the survey, they have been asked to estimate their fear of spiders using the 7-points Likert scale. There were also masking questions about other phobias to prevent people from exaggerating their level of fear of spiders in case is they are too interested in the research. Those who estimate their fear of spiders at 6 or 7 points were redirected to the page with the 18-items Fear of Spiders Questionnaire (FSQ; Russian adaptation of the questionnaire developed by Szymanski & O'Donohue, 1995). If they get at least 40 points in this questionnaire (which represents a moderate level of fear) and were acceptable by other criteria (right-handed, age between 18 and 35, healthy), they were invited in the laboratory. We have invited only females, because the number of males who are afraid of spiders (or, at least, admit it) is significantly less than a number of females and we decide to safe them all for the next stage of the research. There were 1442 people who participated in the survey, 140 of them satisfied our criteria, 47 of them have visited the laboratory. 24 participants were randomly assigned to the Multiple contexts group and 23 to the single context group. Mean age of participants was 27.45 years old, mean FSQ score was 75.13 (according to Muris & Merckelbach (1996), the mean FSQ score for spider-phobic participants is 89.1(±19.6)).

We have chosen Fear of Spiders Questionnaire as a self-report measure because it is widely used worldwide (for example, Booth, Peker & Oztop, 2016) and considered as the best questionnaire to estimate the moderate level of fear (Muris & Merckelbach, 1996). It includes 18 statements (for example, “I would feel very nervous if I saw a spider now”; “Spiders are one of my worst fears”), and the level of agreement with each statement is estimated by participants by 7-points Likert scale from 0 to 6. Adaptation of the FSQ for Russian participants had been performed in accordance with the recommendations from the literature (Benet-Martinez, 2007). Firstly, we performed translation-back-translation procedure, then test both versions of the questionnaire on a set of bilingual participants (N = 40), then reformulate those questions for which the correlations between English and Russian versions of the item were the lowest, and then test it again on another set of bilingual participants (N = 42), delete outliers which had been obviously connected with inattentive reading of the questions because of the completely opposite answers for the same questions in Russian and English, and calculate inter-items correlations and Cronbach's alphas (which reflects an internal consistency) for English and Russian version. The mean inter-item correlation between English and Russian questionnaires was 0.844, and Cronbach's alpha for Russian adaptation was 0.97. The final version of the questionnaire can be found in Appendix 1. We have prepared an article about this process of adaptation, and it is currently under review.

We have also used BAT with a living spider as a behavioral measurement of fear. We worked with commonly used (Smits, Telch & Randall, 2002; Soeter & Kindt, 2015) baby tarantula (Brachypelma smithi), presented in Figure 1. We called it Ivan Petrovich in honor of Pavlov, although we don't know whether it is male or female, because even breeders unable to guarantee it before the spider grows up. It is considered as a non-dangerous spider which practically never bite humans. Limited anecdotal evidence from those spider breeders who have such experience suggests that it is comparable with the bite of a bee. We have also found one case study of skin urticaria caused by prolonged contact with urticating hair of tarantula (Cooke, Grover, Miller & Duffy, 1973). Hence, in our form of informed consent we have a sentence “During the experiment you will have an opportunity to touch a spider if you want to do so. It generally does not bite humans, but in rare case if it happens it experienced as a bite of a bee. Also in rare cases human may develop a skin allergic response on the urticating hair at spider's abdomen. So please do not touch a spider if you have any suggestion that you may be prone to allergic reactions”. In a lab, we had antiseptic ointment and antihistamine drugs to apply them in case of a bite (but there were no such cases). The procedure had been approved by the HSE ethics committee.

Figure 1. The spider which had been used during BAT placed on the author's hand

BAT procedure has been based on the one used by Soeter & Kindt (2015) with some minor modifications. It was preceded with an oral instruction explaining that we were going to propose a sequence of behavioral tasks which suggests an interaction with a real spider. However, the important point was that participant was completely free to reject and stop the process at any moment. After participant confirmed that she understands that she is not obliged to continue BAT and agreed to try it, we started to explain the first step. Then if participant completed it, we explained the next one and so on. If participant rejected in the middle of the step, it brings her 0.5 points for this step. After the first refusal, we have asked if participant sure that she does not want to continue, and the procedure stops. The steps were the following:

1. Look at the spider in a closed jar from a 1.5-meter distance for 10 seconds.

2. Look at the spider in a closed jar from 0.5-meter distance for 10 seconds.

3. Put the palm of the dominant hand on a top of closed jar for 10 seconds.

4. Open the jar.

5. Hold an open jar with the spider under the table (stiff-armed) for 10 seconds.

6. Hold an open jar with the spider under your knees for 10 seconds.

7. Direct the spider's movement with a pencil for 10 seconds.

8. Put the spider from a jar to a bigger plastic box.

9. Direct the spider's movement with a bare finger for 10 seconds.

10. Hold spider in your hand for 10 seconds.

Original procedure proposed by Soeter and Kindt contains 8 steps. We added two different distances in steps 1-2 and steps 5-6 based on subjective responses of spider-phobic participants during the pilot experiments because they underline that there is a big subjective difference between short and long distance. In case of steps 1-2, Soeter and Kindt initially starts from a smaller distance, but we added the longer one due to logistics reason: it allows to hide a jar with a spider under a plastic box on a far end of the table and then suddenly demonstrate it when the participant agreed to start BAT. The key difference between our procedure and original one is that we have measured skin galvanic response during the BAT. It is very uncommon, because the BAT needs free hands, so we attached electrodes on the participant's foot, which is acceptable for the measurement but relatively unusual (Boucsein et al, 2012). The figure 2 represents an experimental setup during the explanation before the step 7 (participant has allowed to use this picture).

Electrodermal activity has been recorded with Ag/AgCl electrodes filled with TD-246 isotonic paste and placed on the inner side of the foot over the abductor hallucis muscle with a 3 centimeters inter-electrode distance. We have used exosomatic measurement with direct current (0.5 V) via BrainVision recorder and software.

Figure 2. Experimental setup. Red arrows represent a jar with a spider and electrodes (placed on the foot) for measurement of skin galvanic response during BAT

Stimuli were presented on a 17 inches computer screen at 70 centimeters distance from the participant's head (all participants had normal or corrected-to-normal vision). For the exposure, we have used two sets of pictorial stimuli prepared by my colleague, PhD student Vlad Muravyev. A multi-context group have seen 5 different spiders in 6 different contexts. A single-context group have seen 5 spiders, 6 times each, in only one context. So there were 30 images of spiders, each of them presented for 5 seconds. Intervals between pictures were randomized from 6 to 10 seconds (which is enough to separate skin galvanic responses elicited by stimuli presentations). Both groups have also seen 30 pictures with pure context (several or one, respectively) without a spider which we have used as a baseline for the comparison of the SGR amplitude elicited by spider or just a neutral picture. At the renewal stage, all participants have seen 10 completely new spiders in completely new contexts. Examples of stimuli are presented on the figure 3.

Figure 3. Examples of stimuli. A - exposure for multiple context group, B - exposure for single context group, C - renewal (same for both groups)

The whole procedure was the following. Participants completed the first FSQ online and then were invited to the laboratory. Here they read and signed the informed consent and also completed the Spielberger anxiety scale (Russian adaptation by Hanin, 1976) to provide us with an opportunity to exclude those who are too anxious independently of spiders (we did not reveal any influence of anxiety to our results, so these data were not included in the final analysis). Then all participants went through the behavioral approach test (BAT). After the BAT we gave them a distraction task (a Stroop test) to decrease the fear. Then participants were randomly assigned to the multiple or single context group and saw the set of images with spiders and the set of images with pure contexts without spiders (order was counterbalanced; those who have seen spiders only in one context have seen the same one pure context, and those who have seen spiders in several contexts have seen the same several pure contexts; the number of pictures and duration of the presentation is identical). It was the exposure part, the only one which was different between groups. Then after 10 minutes break they went through the fear renewal stage where all participants saw the set of completely novel spiders in completely novel contexts. After that they participated in second BAT, completed FSQ for the second time, received a small monetary reward (250 rubles, approximately $4) and went home. The follow up is the third FSQ completed online in a week after the experiment. The SGR has been measured during first BAT, exposure, renewal and second BAT.

2.2 Results

As it described above, in our study we had one independent variable (exposure to spider images in multiple context vs single context) and used three measurements of fear: FSQ score, BAT score and electrodermal activity. We have revealed that the number of contexts does not significantly influence behavioral outcome and self-report scores, but influences electrodermal activity during the fear renewal stage and during BAT.

FSQ score had been measured three times: several days before the experiment, just after the experiment and one week later. Data were analyzed using 2 (multiple versus single group) Ч 3 (time of measurement: before the exposure, just after the exposure and one week follow-up) repeated measures analysis of variance, which revealed the significant main effect of time, F(2, 90) = 10.93, p<.001, but there was no significant effect of group. According to post hoc paired t-tests, the decrease itself is non-significant between FSQ1 and FSQ2, but significant between FSQ2 and FSQ3 both for the multiple contexts (M = 6.96, 95% CI [1.68, 12.23], t(23) = 2.73, p = 0.012) and single context group (M = 9.26, 95% CI [3.9, 14.6], t(22) = 3.58, p = 0.002). The Figure 4 illustrates that there is also some tendency to higher decline of the fear in the multiple contexts group (in accordance with the initial hypothesis), although it is very far from the statistical significance.

Figure 4. FSQ scores before exposure, after exposure, and 1-week follow-up (Mean scores ± SEM)

BAT scores were analyzed by 2 (group) Ч 2 (time) repeated measures analysis of variance. In the whole sample of participants, analysis has revealed significant increase of BAT score, F(1, 45) = 9.79, p = 0.003, but no main effect of group. We have also mentioned that 15 out of 47 participants have achieved the final stage of BAT even during the first test performed before the exposure, although the average FSQ1 score for these brave people (M = 71.93, SD = 18.78) did not significantly differ from others (M = 76.62, SD = 16.68), t(25)= -0.85, p=0.41. Participants mostly explained this unexpectable fact declaring that our particular spider is either bigger or smaller than the spiders they are afraid of. So we have also analyzed the BAT scores excluding those who have completed the whole sequence of tasks before the exposure, but it does not influence the results: main effect of time was significant, F(1,27) =11.46, p=0.002, while main effect of group was not. However, as in the case of FSQ scores, we observed a non-significant trend in predicted direction: average BAT score after the exposure a bit higher in multiple contexts group than in single context group (Figure 5).

Figure 5. BAT scores before and after exposure (Mean scores ± SEM)

Reverse correlation between FSQ1 scores and BAT1 scores in our sample was weak and marginally significant: r(45) = -0.29, p = 0.051. However, it becomes a bit stronger with exclusion of 15 brave people mentioned above: r(30) = -0.35, p = 0.04.

SGR amplitudes were computed as a difference between maximum and minimum response in a 5 second interval started 100 ms before stimuli onset (in case of BAT, it was a moment when participant starts to perform a suggested task). SGR analysis was made offline using BrainVision Analyzer2 software. Skin galvanic responses were corrected for artefacts during visual inspection. Each trial was baseline corrected by 100ms before the trial and 1 Hz low-pass filtered. There were 45 participants included in SGR analysis (24 from multiple contexts group and 21 from single context group) due to technical failures which lead to incomplete recordings for two participants.

Electrophysiological responses during exposure to spider images, pure landscapes images and renewal stage have been analyzed by 2 (Group) Ч 3 (Condition) two-way analysis of variance. The results presented in Table 1.

Table 1. Analysis of variance for SGR during exposure and renewal

Notes. ** p<0.01, ***p<0.001

As it can be seen from the Table 1, two-way analysis of variance revealed a significant main effect of condition and a significant interaction between condition and group. Post hoc t-test confirms that participants from multiple contexts group demonstrate significantly lower SGR at the renewal stage (M = 1.25, SD = 1.93) than participants from single context group (M = 1.95, SD = 3.63), t(287) = -2,46, p = 0,014. The results are visualized in Figure 6.

Figure 6. SGR (Mean SEM, in Microsiemens) during exposure to pure landscapes, spider images and at the renewal stage. Notes. * p<0.05

We have also analyzed changes of SGR amplitude during first versus second BAT. In our sample, all participants have completed stages 1-3 (look at the spider from two different distances and put the palm on the top of the closed jar with a spider inside), and several of them refuse to continue after that, so we have concentrated on averaged SGR from first three steps. 2(Group) Ч 2(Time) repeated measures analysis of variance revealed significant main effect of time, F(1,222) = 11.08, p = 0.001, and significant interaction between time and group, F(1,222) = 4.34, p = 0.03. There was no main effect of group. Post hoc paired t-tests demonstrated that the decrease of SGR amplitude is significant for the multiple contexts group (M = 3.09, 95% CI [0.83, 5.36], t(68) = 2.72, p = 0.008) and non-significant for single context group (M = 1.07, 95% CI [-0.58, 2.73], t = 1.29, p = 0.2). However, as it can be seen on figure 7, initial average BAT amplitude accidentally was a bit higher for the multiple contexts group than for the single context group, and the prominent decrease might be partially explained by this fact. But anyway, the difference between average group SGRs during the first BAT is non-significant (M = -1.43, 95% CI [-3.41, 0.55], t=-1.45, p=0.15), and the interaction between time and group confirmed by the analysis of variance, so we considered described results as a confirmation of our initial hypothesis.

Figure 7. SGR amplitude (Mean SEM, in Microsiemens) during 1-3 steps of BAT

ANOVA and other cumbersome calculations had been performed in R studio (R Core Team, 2017). Easy calculations (such as founding a mean score or standard deviation) had been mostly performed in Microsoft Excel. Graphs had been created in both programs depending on where it was faster to understand how to achieve a desirable kind of a graph. All data processing presented above had been performed by the author, but the preparation of SGR data for the analysis (subtraction, baseline correction, and filtering) had been made by Egor Levchenko (who is the first-year master student of the same program) and Vlad Muravyev.

Summary and discussion

Our initial hypothesis suggests that those spider-phobic participants who have seen images of spiders in multiple contexts will demonstrate a higher decrease of fear than those who have seen images of spiders in only one context. The results have partially confirmed this presupposition: there are no significant differences between groups in self-reports and BAT scores, but analysis of in electrodermal activity reveals significant interactions between group and condition both during fear renewal stage (F(2,3264) = 6,29, p=0.02) and second BAT (F(1,222) = 4.34, p = 0.03).

Obtained results are interesting both from the clinical and scientific points of view. Our initial intention was to explore an exposure therapy with pictorial stimuli as a gentler alternative to real-life exposure. It potentially gains the opportunity to propose it to those patients who consider the real-life exposure too stressful. Indeed, we have demonstrated that even the single session of pictorial exposure is associated to a small but significant decrease of fear in both treatment groups, so it can be hypothesized that several sessions may lead to the clinically meaningful improvement. However, there are two important limitations. First, the presented piece of work does not include a control group which has not seen any pictures of spiders at all (because we were primarily interested in the comparison between multiple contexts and single context exposure, and because it is only the preliminary part of the project which hopefully will be developed and continued). Second, we cannot attribute the fear decline purely to pictures because the protocol also includes BAT which can be considered as a small element of real-life exposure. In our sample, there were no participants who rejected BAT from the very beginning, and we cannot declare that pictures lead to the improvement separately. But it is also a hypothesis which can be easily checked in following student projects. Anyway, even the combination of pictorial exposure with small and voluntary components of life exposure, as had been used in our experiment, may be potentially useful for clinicians. We have not proved that the multi-context pictorial exposure works better than single-context regarding behavioral outcomes, but this is one more thing which can be additionally explored with enlarging a sample or length of exposure to increase statistical power.

From the scientific point of view, it is noticeable that there is a discrepancy between overt and covert manifestations of fear. It raises important questions for the next stage of this project, where it is suggested to apply transcranial magnetic stimulation to explore whether it can influence the process of fear extinction with the usage of pictorial stimuli. My colleague Vlad Muravyev, who proposed and will continue this research, presumably would stimulate vmPFC hence studies discussed in a paragraph 2.1.3 demonstrate that the level of vmPFC activation is positively correlated with fear extinction success, and available data about transcranial magnetic stimulation applied to vmPFC also confirms the suggestion that enhancement of fear extinction through vmPFC stimulation is possible (Guhn et al, 2014). The data obtained at this stage of the study emphasize the fact that fear processing is complex, and it is possible to reveal factors which influence covert manifestations of fear in a larger extent than for overt reactions. Accordingly, it can be possible to reveal neural circuits involved primarily in overt or covert fear reactions.

So this research project turned out to be even more exciting than we suggested when we initiate it, and I do hope that exploring will be continued.

We have adapted the Fear of Spiders Questionnaire for the Russian audience. The article dedicated to this sub-project is not published yet, but once it will, then our contribution would be useful for other researchers who work in Russian-speaking countries and consider fear of spiders as a central object of their interests or as a model for more general fear research.

We have established a protocol of exposure therapy based on pictorial stimuli and demonstrated that this kind of intervention is associated with the small but significant decrease of fear. This promising result should be studied more thoroughly, but potentially it can be helpful in developing more lenient approaches to exposure therapy.

We have demonstrated that exposure to fearful pictures in multiple contexts leads to the significantly more prominent decline in fear-associated electrodermal activity than single-context exposure, and at the same time does not influence behavioral measurements of fear (or at least these effects are small and undetectable on relatively small samples). This distinction between covert and overt fear reactions is important for the next stage of research where it is suggested to apply transcranial magnetic stimulation to explore the role of vmPFC in fear extinction.