I T O L D Y O U S O :
A C O M P A R I S O N O F
F E A R A C Q U I S I T I O N V I A I N S T R U C T I O N S
W I T H – O R W I T H O U T D I R E C T E X P E R I E N C E .
tessa blanken1
supervised by: angelos-miltiadis krypotos1 & tom beckers1, 2 july 14, 2014
contents
1 Introduction 3 2 Methods 5 2.1 Participants . . . 5 2.2 Apparatus . . . 5 2.3 Procedure . . . 7 2.4 Data Analyses . . . 9 3 Results 11 3.1 Confirmatory Analyses . . . 11 3.2 Exploratory Analyses . . . 164 Conclusion and Discussion 21 5 References 23 6 Appendix A: Deviations from Original Proposal 25 6.1 Experiments . . . 25
6.2 Data Analyses . . . 25
7 Appendix B: Adjusted Time Schedule 26 8 Appendix C: Original Proposed Second Experiment 27 9 Appendix D: Original Proposed Data Analyses 28 9.1 Confirmatory Analyses . . . 28
9.2 Exploratory Analyses . . . 28
Contents 2
abstract
A common laboratory model to study the pathogenesis of anxiety symptomatology is fear conditioning in which an initially neutral stimulus (e.g., tone) is paired with an aversive cue (e.g., shock). Prior research indicates that the associations between those events can be achieved via experience, instructions, or both. However, none of these studies measured all components of fear (i.e., subjective apprehension, physiological arousal, and action tendencies). In the current experiment we tested whether fear re-sponses towards innocuous cues can be induced by mere instructions, or whether ad-ditional direct experience is needed for fear responses to develop. Of importance, we measured all three indices of fear. In addition, we tested the results within a Bayesian framework; allowing us to accumulate evidence in favour of no effect, which is some-thing previous studies were unable to do. The results showed that physiological and subjective fear responses acquired via instructions alone were similar to fear responses acquired via a combination of instructions and direct experience. There was no effect of fear conditioning on action tendencies. Our results suggest that instructions alone could be sufficient to acquire fear. Clinical implications are discussed.
1
University of Amsterdam
2
introduction 3
1
introduction
Experiencing fear in reaction to dangerous situations is an adaptive emotional re-sponse. However, individuals may sometimes experience fear in absence of a real threat. In these situations, fear can take the form of an anxiety disorder or phobia. With a lifetime prevalence rate of approximately 29%, anxiety disorders, as defined by the Diagnostic and Statistical Manual of mental disorders (American Psychiatric Association, 2000), are the most common diagnosed disorders (Kessler et al., 2005). This high prevalence stresses the clinical importance of understanding the etiology of fear in order to both prevent and treat anxiety disorders.
A laboratory model to study the pathogenesis of fear is fear conditioning (Pavlov, 1927). In Pavlovian fear conditioning, a neutral stimulus (conditioned stimulus, CS; e.g., tone) is associated with an intrinsically aversive outcome (unconditioned stim-ulus, US; e.g., shock). When the conditioned stimulus reliably predicts the aversive outcome (i.e., CS-US contingency), the conditioned stimulus comes to elicit fear (con-ditioned response, CR) in the absence of a US.
To date, theories suggest that conditioned responses (e.g., fear) are acquired via associative learning (i.e., CS-US pairings). These conditioned responses can be learned not only via direct experience of the CS-US contingency, but also via instructions about the CS-US relation (Rachman, 1977). Being afraid of dogs, for example, can be the result of being attacked by a dog (learning via direct experience) or someone telling you about being attacked by a dog (learning via instructions).
This notion of multiple pathways to fear acquisition is supported by retrospective studies on childhood phobias in which instructions were, next to direct experience, reported as origins of childhood phobias (King, Eleonora, & Ollendick, 1998). Sub-sequent, experimental studies have also shown that both learning pathways lead to similar levels of skin conductance responses (SCR), which is a measure of fear (Olsson & Phelps, 2004).
More recently, it has been studied whether a combination of different pathways (e.g., instructions about an aversive stimulus followed by the actual experience of that aversive stimulus) would further enhance fear acquisition. Comparing fear responses acquired via instructions alone with fear responses acquired via a combination of in-structions and direct experience showed that direct experience did add to self-reported fear (subjective measure), but there was no added effect of experience on SCR (phys-iological) and US-expectancy (subjective) measures (Raes, De Houwer, De Schryver, Brass, & Kalisch, 2014).
In experimental studies, fear is assessed using different measurements. These mea-surements fall within different emotional components, as defined by the componen-tial model of emotions. According to this model, emotions consist of physiological responses, subjective experience, and action tendencies (Mauss & Robinson, 2009). While most studies assessed fear in terms of physiological responding (e.g., SCR as in Olsson & Phelps, 2004) and/or subjective experience (e.g., US-expectancy as in Raes et al., 2014), no previous study measured the third component, action tendencies. This is a serious limitation since it is even argued that emotions are primarily action tendencies (Frijda, 2010; Lang, 1995). According to this view emotions are driven ulti-mately by the appetitive and the aversive motivational systems and could therefore be characterized as motivationally states of readiness (action dispositions) (Lang, 1995). Thus, in order to obtain a coherent view about fear emotion responses it is preferred to measure all three indices of fear.
introduction 4
Furthermore, the different emotional components cannot assumed to be compati-ble. That is, the different components of emotions, and the measurements thereof, are not necessarily coherent. Even multiple measures of the same overarching emo-tional component can deviate. In the study of Raes et al. (2014), for example, different measures of subjective experience (US-expectancy and self-reported fear) indicated different levels of fear acquisition. Other studies have shown that also multiple mea-sures of physiological responding (i.e., SCR and fear potentiated startle (FPS)) are affected differently by fear conditioning (Sevenster, Beckers, & Kindt, 2012, 2014). In sum, to obtain a coherent view about fear emotion responses, it is preferred not only to measure all three indices of fear, but in addition include multiple measures of the same emotional component.
In this study we investigated whether instructions about CS-US contingencies lead to similar fear responses towards conditioned cues compared to fear responses ac-quired via a combination of instructions and direct experience. Following the afore-mentioned componential model, we tried to obtain a coherent view of the fear emotion responses by measuring all three emotional components and, in addition, included multiple measures of subjective experience (i.e., US-expectancy and self-reported fear) and physiological responding (i.e., SCR and FPS).
We presented pictures of three geometrical objects (i.e., a cube, cone and cylinder) as CSs. To establish immediate fear learning instructions indicated which one of these objects (e.g., cube) was going to be paired with an aversive electrical shock (the rein-forced stimulus), and which objects (e.g., cone and cylinder) would not be followed by a shock (the unreinforced stimuli). After fear acquisition, all participants received in-structions that indicated that one of the previous unreinforced stimuli (e.g., picture of a cone) would now be followed by a shock. Crucially, during the test phase, half of the participants (instructed acquisition group) encountered the stimulus that was instructed to be followed by a shock (e.g., cone), while the other half (combined acquisition group) were presented with the stimulus that was both instructed and previously experienced to be followed by a shock (e.g., picture of a cube).
While SCR and US-expectancy have shown to be directly affected by explicit CS-US contingency awareness, FPS did not (Sevenster et al., 2012). Therefore we expected the instructions to directly lead to higher SCR and US-expectancies, but not higher FPS, for the "instructed stimulus" compared with the unreinforced stimulus. In line with these predictions we did not expect an added effect of direct experience for SCR and US-expectancies. Action tendencies were measured in a speeded approach / avoid re-action time task (AAT). In these tasks rere-action times are influenced by the congruency between the response (approach vs. avoid) and the hedonic nature of the stimulus (appetitive vs. aversive); participants are faster on congruent trials (e.g., approach ap-petitive stimulus) relative to incongruent trials (e.g., approach aversive stimulus). We predicted that participants would be faster to approach the unreinforced CS and avoid the reinforced CS than vice versa. Specifically, we expected this effect to be the same when the reinforced stimulus constituted of the stimulus that was just instructed to be followed by an electric stimulus (the instructed acquisition group) compared to when the reinforced stimulus constituted of the the stimulus that was both instructed and experienced to be followed by an electric stimulus (the combined acquisition group).
Of importance, we will analyze our data within a Bayesian framework by calculat-ing Bayes factors. The Bayes factor quantifies evidence provided by the data in favour of one model as opposed to another model (Kass & Raftery, 1995), see the Data Anal-yses section for more information about the calculation of Bayes factors. Therefore, analyzing the data within a Bayesian framework allows us to accumulate evidence
methods 5
both in favour of the testable hypotheses, as well as in favour of the null hypothesis of no effect (Dienes, 2011). This is crucial given that we expect similar results for both ex-perimental groups. While using frequentist analyses we are only able to conclude that there are no differences, the Bayesian analyses will enable us to actually gain evidence that the results are similar.
2
methods
2.1 Participants
participants Fourty (26 females) healthy adults participated in the study, ranging in age between 18 and 29 years (M = 22.1, SD = 2.8). Participants received either course credits or a monetary reward (€10) for their participation. All participants gave informed consent. Participants were equally and randomly assigned to either the instructed acquisition group or the combined acquisition group. The study has been approved by the UvA Ethics Committee, 2014-CP-3566.
2.2 Apparatus
stimuli Pictures of four different viewpoints (50 mm × 50 mm) of a cube, a cylinder, and a cone, served – counterbalanced across participants – as either CS+, CS−/+, or CS−.
For the AAT two of the geometrical objects, depending on the condition of the participant, were used. The objects were placed within a surrounding frame that was landscape (100 mm × 55 mm) or portrait (55 mm × 100 mm) orientated. Furthermore, a white manikin figure (71 mm × 95 mm) was presented.
The US consisted of a 2-ms electrical stimulation delivered through two Ag elec-trodes to the wrist of the non-preferred hand.
skin conductance response Skin conductance responses (SCR) were assessed following the procedures of Sevenster et al. (2012). To measure electrodermal activity we attached two Ag/AgCl electrodes of 20 × 16 mm to the medial phalanges of the ring– and index finger of the non-preferred hand. The SCR signal was sampled at 1000 Hz. Electrodermal activity was measured using a sine-shaped excitation voltage (1V peak – peak) input device of 50 Hz. The signal from this input device was led through a signal-conditioning amplifier, and the analogue output was digitalized at 100 Hz by a 16-bit AD converter. Electrodermal activity was recorded with the software program VSSRP98. Phasic electrodermal responding to the CS was calculated by subtracting the baseline (mean SCR during the 1 s period before stimulus onset) from the maximum score (determined at 1 s intervals during the first 7 s following CS onset).
fear potentiated startle Fear potentiated startle (FPS) was assessed following the procedures of Sevenster et al. (2012). FPS was measured through electromyogra-phy (EMG) of the right orbicularis oculi muscle. We placed three 5 mm Ag/AgCl elec-trodes filled with conductive gel (Signa, Paker); two were positioned approximately 1 cm under the pupil and 1 cm below the lateral canthus, respectively; the last was placed on the forehead as a ground electrode, 1 cm below hairline (Blumenthal et al., 2005). A 40-ms-duration noise burst (104dB) with a rise/fall time shorter than 1 ms
methods 6
served as the startle probe. The startle probe was delivered binaurally through head-phones (Sennheiser, model HD 25 − 1 II). The EMG signal was sampled at 1000 Hz and amplified in two stages. The input stage had a resistance of 10 MΩ, a frequency response of DC-1500 Hz, and an amplification factor of 200. To reduce interference of the main noise we used a 50-Hz notch filter. During the second stage the signal was integrated and amplified with a variable amplification factor of 0 − 100×. To obtain the cleanest possible data without affecting response amplitude the raw EMG data were bandpass filtered (28 − 500 Hz). Maximum FPS was determined in a 0 − 250 ms interval following probe onset.
online us-expectancy ratings US-expectancy was measured continuously dur-ing each conditiondur-ing phase. Mean US-expectancy was determined durdur-ing the 8 s each stimulus was presented. Participants rated their US-expectancy using a slider: when participants placed the slider to the extreme left side, ten red lights were shown through which participants indicated that they were absolutely certain that no electric stimulus would follow; when they placed the slider in the middle, a single orange light was shown and participants indicated that they were uncertain whether an electric stimu-lus would follow; when they placed the slider to the extreme right side, ten green lights were shown and participants indicated that they were absolutely certain that an electric stimulus would follow.
action tendencies Action tendencies were measured via an approach/avoidance reaction time task. For a more detailed description of the task see the procedure below. subjective measures After each conditioning phase participants rated their fear towards the CSs ("How much fear did you experience when looking at this picture") on a continuous scale ranging from −5 (none at all) to 5 (very much).
Participants rated US pleasantness (from −5, unpleasant, to +5, pleasant), US inten-sity (weak, moderate, intense, enormous, unbearable), and valence of the CSs (from −5, unpleasant, to +5, pleasant).
Trait and state anxiety were measured with the Trait Anxiety Inventory and the State Anxiety Inventory, respectively (STAI-T, STAI-S; Spielberger, Gorsuch, & Lushene, 1970). Anxiety sensitivity was measured with the Anxiety Sensitivity Index (ASI; Peterson & Reiss, 1993).
methods 7
2.3 Procedure
Habituation Acquisition Test AAT Instructed 2CS+ 6CS+ 2CS−/+ CS−/+ 2CS−/+ 6CS−/+ 2CS− CS− 2CS− 6CS− 2NA 2NA 6NA Combined 2CS+ 6CS+ 2CS+ CS+ 2CS−/+ 6CS−/+ 2CS− CS− 2CS− 6CS− 2NA 2NA 6NA
Table 1:Experimental Design
Table 1 gives a schematic overview of the different experimental phases: habituation, acquisition, test, and AAT.
After participants read the information brochure and gave informed consent, they completed the STAI-S. Upon completion the SCR, EMG, and shock electrodes were attached. US intensity level was then individually determined by gradually increasing shock intensity until the participant indicated the shock to be "uncomfortable though not painful" (see, Sevenster et al., 2012).
In all conditioning trials the CS was presented for 8 s (see Figure 1). The startle probe was delivered 7 s after stimulus onset, followed by the US after another 500 ms (on reinforced trials only). Intertrial intervals (ITIs) were 15, 20, or 25 s with an average of 20 s. Presentation of CSs and noise alone (NA) trials was semi-random (no more than two consecutive trials of the same CS or NA).
To assess baseline startle responding the startle probe alone (noise alone, NA) was presented during ITIs. After each conditioning phase, participants rated self-reported CS fear.
habituation phase Each stimulus (CS+, CS−/+, CS−) and NA trials were pre-sented twice to acquire baseline responding. None of the stimulus presentations was followed by a US.
acquisition phase All three CSs and NAs were presented six times each. One of the pictures (CS+) was paired with a shock on five out of six trials, whereas the other
pictures were never paired with a shock (CS−/+and CS−).
Before the start of the acquisition block, the stimuli were presented on-screen and it was indicated, both verbally and on-screen, which object (CS+) would be most of the time followed by a shock, and which objects (CS−/+, CS−) would never be followed by a shock. It was explained that participants should learn the contingencies between the different CSs and the US.
test phase After fear acquisition the stimuli were shown on-screen and all partic-ipants were instructed, both verbally and on-screen, that now both the previous rein-forced stimulus (CS+) and one of the previous unreinforced stimuli (CS−/+) would
methods 8
1
0
2
3
4
5
6
7
8
seconds
stimulus
SCR
FPS
US-exp.
Figure 1:A reinforced conditioning trial. Picture duration was 8 s. Participants continuously rated their US-expectancy during all conditioning trials, and mean US-expectancy ratings were calculated during stimulus presentation. The startle probe was delivered after 7, followed by the US after 500 ms. SCR was calculated within 0 − 7 s interval following CS onset.
sometimes be followed by a shock. It was explained that the other unreinforced stim-ulus (CS−) would still never be followed by a shock. The instructions specifically
indicated which of the objects would be followed by a shock, and what object would not be followed by a shock.
Crucially, during the test phase, participants in the instructed acquisition group were presented with the previous unreinforced pictures (i.e., CS−/+, CS−), while participants in the combined acquisition group were presented with the originally reinforced picture (CS+) and one of the originally unreinforced pictures (CS−). Note
that for participants in the instructed group the CS−/+ now served as the reinforced
stimulus, while for participants in the combined group the reinforced stimulus still constituted of the CS+. The unreinforced stimulus was the same in both groups (i.e., CS−). The reinforced stimulus during test phase (either CS−/+ or CS+) will be re-ferred to as Final CS+. The CSs and NA trials were presented twice and none of the
trials was reinforced.
approach / avoidance task Participants were presented with the stimuli they en-countered during test phase: participants in the instructed acquisition group were presented with CS−/+and CS−, while participants in the combined acquisition group were presented with the CS+and the CS−.
The AAT consisted of two blocks of 20 trials each (4 practice trials followed by 16 test trails). For practice trials each stimulus was presented from two different viewpoints, randomly selected out of the four possible viewpoints. For test trials, each CS was presented eight times in semi-random order (no more than two consecutive trials of the same type).
On each trial, a white manikin figure was shown, centered on either the bottom or top half of a black computer screen. The CS appeared after 1, 500 ms, centered on the opposite side of the screen. Participants were instructed, both verbally and
methods 9
on-screen, to move the manikin according to the orientation of the white frame (i.e., toward landscape and away from portrait or vice versa), with reversed instructions between blocks. The instructions emphasized speed and accuracy.
In order to move the manikin upward or downward participants had to press the "Y" (labeled "↑") or "B" (labeled "↓") key, respectively, on a standard computer keyboard. The manikin started moving when a response was made and disappeared after 500 ms. In case of an incorrect response, a red cross followed at the manikin’s starting position for 500 ms; there was no feedback after a correct response. The ITI was 2, 000 ms. The time between CS picture onset and response (i.e., reaction time) was measured as a dependent variable.
exit interview and questionnaires Upon completion, participants evaluated the CSs and the US, they answered exit questions concerning their contingency awareness about the CS-US relationships and they filled out the Trait Anxiety Invetory (STAI-T) and the Anxiety Sensitivity Index (ASI).
2.4 Data Analyses
2.4.1 Confirmatory Analyses
First, we checked whether all participants were aware about the CS-US contingencies during the experimental phases. This was assessed by: (1) participants must have had correctly indicated during exit interview which CS was followed by a shock and which were not, and (2) expectancy ratings during the last two acquisition trials had to be at least 2-points higher for the Final CS+ compared to the CS−
For all statistical analyses we set the alpha-level to .05. STAI-T and STAI-S scores, ASI scores, and US and CS characteristics ratings were analyzed with separate anal-yses of variance (ANOVAs) with group (instructed vs. combined) as the between-subject factor.
To reduce heteroscedasticity we standardized the SCR and FPS values by taking the square root. Due to a high rate of negative SCR values (35.38%) we standardized the negative SCR values by taking the square-root of the absolute value and then give these values a negative sign (Milad et al., 2006). For FPS there were only a few negative values (< .1%), and thus we replaced these values with zero before taking the square root. Mean scores of US-expectancy ratings, SCR, FPS, and self-reported CS fear were calculated per stimulus and per conditioning block.
To test for baseline differences between stimuli during habituation phase, we per-formed separate ANOVAs on mean SCR, FPS, and US-expectancies during habitua-tion with a 3 (stimulus: CS+, CS−/+, CS−) × 2 (group: instructed vs. combined)
design, with stimulus as the within-subject factor and group as the between-subject factor.
To test whether fear conditioning, during the acquisition phase, led to differences in fear responses for the different CSs we analyzed mean US-expectancy ratings, SCR, and FPS during acquisition with separate 3 (stimulus: CS+, CS−/+, CS−) × 2 (group:
instructed vs. combined) ANOVAs, with stimulus as the within-subject factor and group as the between-subject factor.
To test whether instructions led to similar levels of differential responding compared to a combination of instructions and experience in the test phase, we performed sepa-rate 2 (stimulus: Final CS+vs. CS−) × 2 (group: instructed vs. combined) ANOVAs
methods 10
on mean US-expectancy ratings, SCR, and FPS during test-phase, with stimulus as the within subject-factor and group as the between subject factor.
For the AAT, median RTs of each participant was calculated for each stimulus (Final CS+, CS−) by response type (approach, avoid) combination. RTs were analyzed with a 2 (stimulus: reinforced vs. unreinforced) × 2 (response: approach vs. avoid) × 2 (group: instructed vs. combined) ANOVA, with stimulus and response as within-subject factors and group as between-within-subject factor.
Greenhouse-Geisser corrections were performed in case the assumption of spheric-ity was violated.
In order to compare our frequentists results with Bayesian results, we computed separate Bayes factors for each main effect and interaction effect. In calculating Bayes factors, probabilities are assingend to competing hypothesis. To test whether there is an effect of stimulus, for example, we have the following competing hypotheses: (1) there is no effect of stimulus (the null hypothesis, H0) and (2) there is an effect
of stimulus (the alternative hypothesis, H1) . Before we observe the data (D), each
hypothesis has a prior probability: p(H0) and p(H1). The ratio of these probabilities
is the prior odds. When the data (D) are observed, the prior odds are updated to the posterior odds, see Equation 1.
p(H0| D) p(H1| D) = p(D| H0) p(D| H1) ×p(H0) p(H1) (1) The posterior odds is the ratio of the posterior probabilities for each hypothesis given the data: p(H0|D)
p(H1|D). The change from prior odds to posterior odds is quantified
by the ratio of the conditional probabilities of the data given the hypotheses: p(D|H0)
p(D|H1),
which is called the Bayes factor. Thus, a positive Bayes factor of, for example, 20, in-dicates that the data are 20 times as likely to have occurred under H0than under H1.
Note that this is irrespective of the prior probabilities that are assigned to the compet-ing hypotheses a priori. Therefore, the Bayes factor is an objective quantification of the way the data change your beliefs under the null model (H0) versus an alternative
model (H1). Of importance, this allows us to quantify evidence in favour of both
com-peting hypotheses. Hence, unlike frequentist analyses, we can accumulate evidence in favour of the null hypothesis of no effect.
In the current experiment we tested for the main effects of stimulus and group by comparing the no-interaction model with both main effects (model with the ef-fect) to a no-interaction model with only one main effect (model without the efef-fect) (Wetzels, Grasman, & Wagenmakers, 2012). Specifically, to test for the main effect of stimulus, we compared the no-interaction model with both main effects (MS+G) to a
no-interaction model with the main effect of group (MG). Similarly, we tested for the
main effect of group by comparingMS+G toMS. To test for the effect of the
interac-tion term we compared the full model containing both main effects and the interacinterac-tion effect (MF), to the same model without the interaction effect (MS+G).
To analyze the AAT data in a Bayesian framework we fitted a Bayesian Hierarchical Diffusion Model (BHDM) to the reaction times data. Additionally, the BHDM allows us to account for individual differences and estimates underlying psychological pro-cesses that are involved in the AAT; properties we cannot account for using frequentist analyses. The Ratcliff diffusion model conceptualizes the decision process by incorpo-rating different parameters that represent different psychological processes: (1) bound-ary separation a; (2) drift rate v; (3) starting point z; and (4) non-decision time Ter
Wagenmak-results 11
ers, submitted in which we restricted the model by (i) assuming that the performance differences are due to drift rate only, (ii) assuming a symmetric start point (z = a/2), which entails that the approach and avoidance response are equally attractive a priori, and (iii) by restricting the a, z, and Terparameters to not vary across trials . We used
the Markov Chain Monte Carlo (MCMC) to estimate the Bayesian model parameters. To ensure convergence we ran three chains, each chain consisted of 120, 000 samples. We discarded the first 20, 000 samples as burn in, and then used a thinning factor of 10, making the posterior distributions based on a total of (120, 000 − 20, 000)/10 = 10, 000 samples.
2.4.2 Exploratory Analyses
Exploratory we analyzed the subjective and physiological outcome measures includ-ing trials in the analyses as a ’time’-factor. This way we investigated the rate of learn-ing for each stimulus within each conditionlearn-ing phase.
Also, to test for the effect of instructions on self-reported CS fear we performed a 2 (time: after acquisition vs. after test phase) × 2 (group: instructed vs. combined) ANOVA on self-reported CS fear towards the CS−/+. To test for differences in
self-reported CS fear between the CSs after the manipulation we performed a 3 (stimulus: CS+, CS−/+, CS−) × 2 (group: instructed vs. combined) ANOVA on the fear ratings after the test phase.
3
results
questionnaires and evaluations The groups did not differ in state or trait anx-iety, Fs < 1. There was, however, a difference in anxiety sensitivity between groups, F(1, 38) = 4.63, p = .04; participants in the combined acquisition group (M = 15.2, SD = 6.9) scored higher on the ASI than participants in the in the instructed acquisition group (M = 11.0, SD = 5.3). Note that the direction of this difference could actually only go against our hypothesis, and will therefore not impair the interpretation of the results.
Shock intensity ranged from 6 to 52 mA (M = 20.6, SD = 10.5) and did not differ between groups, F < 1. There were no differences in the participants’ evaluation of the CSs and US, Fs < 1.03.
3.1 Confirmatory Analyses
skin conductance response For SCRs during the different conditioning phases, see Figure 2. There were no baseline differences in skin conductance response (SCR) towards the CSs, see habituation in Figure 2; main effect of stimulus: F(1.51, 57.37) < 1, a result that was similar across groups; stimulus × group: F(1.51, 57.37) < 1.
There was successful acquisition of electrodermal responding, evident by higher SCR for the CS+ compared to the CS−/+ and the CS−, see Figure 2; main effect of stimulus: F(1.62, 61.51) = 40.73, p < .001, η2=.29, an effect that did not differ between groups; stimulus × group: F(1.62, 61.51) < 1.
During the test phase, SCR to the Final CS+ was higher than to the CS−, see test in Figure 2; main effect of stimulus: F(1, 38) = 5.08, p = .03, η2=.03. Of importance,
results 12
from the differential SCR in the combined acquisition group (CS+vs. CS−), indicating
similar levels of SCR in both groups; stimulus × group interaction: F(1, 38) = 1.13, p =.29, η2=.006.
Hab Acq Test
Instructed Acquisition SCR 0.0 0.2 0.4 0.6 0.8 1.0 CS+ CS− CS−/+ ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
Hab Acq Test
Combined Acquisition SCR 0.0 0.2 0.4 0.6 0.8 1.0 CS+ CS− CS−/+ ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
Figure 2:Mean skin conductance responding to the CS+, CS−/+and CS−during habituation, acquisition, and test phase for the instructed acquisition group (top) and the com-bined acquisition group (bottom).
Bayesian Analyses on SCR data Table 2 gives the calculated Bayes factors for both
main effects (stimulus and group) and the interaction effect (stimulus × group) for all conditioning phases. For the habituation phase, the Bayesian hypothesis tests favoured the null model (without main and interaction effects). That is, there is strong evidence that SCR was the same for the different CSs (no effect of stimulus), that there were no differences in SCR between groups (no effect of group), and that SCR for the different stimuli was the same between groups (no stimulus × group interaction effect). Both for the acquisition phase and the test phase, the Bayesian hypothesis tests provide strong evidence in favour of the model that includes the stimulus effect; but not in favour of the model that includes the group effect, nor for the model that includes the stimulus × group interaction effect. Thus, there is convincing evidence that there was acquisition of SCR towards the conditioned stimuli during the acquisition phase, and, that this did not differ between groups. Of importance, the analyses provide substantive evidence that during the test phase, there was a difference in SCR for the Final CS+ compared to the CS−, and that this effect was the same between groups; indicating no added effect of direct experience on SCR. These Bayesian results are in line with the frequentist results.
results 13
Main effect Main effect Interaction effect Stimulus Group Stimulus × Group Phase BFMS+G:MG BFMS+G:MS BFMSG:MS+G
Habituation 0.13 0.68 0.15 Acquisition 30.49 × 109 0.78 0.14
Test 15.54 × 105 0.71 1.53
Table 2:Bayes factors for main and interaction effects on SCR responding for every conditioning phase.
fear potentiated startle Due to a technical error the data of two participants were not saved for the habituation phase. The missing values were replaced with the mean EMG response per stimulus in habituation phase.
There were no baseline differences in startle responses towards the CSs, see habitu-ation in Figure 3; main effect of stimulus: F(2, 76) = 1.56, p = .22, η2 =.006, and this did not differ between groups; stimulus × group: F(2, 76) < 1.
Hab Acq Test
Instructed Acquisition EMG 0 5 10 15 20 25 CS+ CS− CS−/+ ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
Hab Acq Test
Combined Acquisition EMG 0 5 10 15 20 25 CS+ CS− CS−/+ ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
Figure 3:Mean fear potentiated startle to the CS+, CS−/+and CS−during habituation, acqui-sition, and test phase for the instructed acquisition group (top) and the combined acquisition group (bottom).
FPS was higher for the CS+compared to the CS−/+and the CS−, indicating success-ful acquisition of EMG responses, see acquisition in Figure 3; main effect of stimulus: F(1.45, 55.15) = 25.69, p < .001, η2 = .06. This effect did not differ between groups;
results 14
During the test phase, FPS was higher for the Final CS+compared to the CS−, see
test in Figure 3; main effect of stimulus: F(1, 38) = 22.46, p < .001, η2 = .108. Of importance, differential FPS was similar for the combined acquisition group (CS+vs. CS−) and the instructed acquisition group (CS−/+vs. CS−), indicating similar EMG fear responses in both groups; stimulus × group: F(1, 38) < 1.
Bayesian Analyses on EMG data Table 3 shows the Bayes factors for the main and
interaction effects on the EMG data for all conditioning phases. For the habituation phase the Bayesian hypothesis tests favoured the null model; without the main ef-fects of stimulus and group and the stimulus × group interaction effect. This means that there were no differences in startle response towards the CSs during habituation, and this was the same in both groups. For the acquisition phase, there is substantive evidence for an effect of stimulus on FPS, indicating successful acquisition of EMG responses. There is no evidence for an effect of group, nor for a stimulus × group interaction effect, indicating acquisition of FPS to be the same across groups. The Bayesian hypothesis tests showed the same pattern for the test phase: there is con-vincing evidence for an effect of stimulus, but not for an effect of group, nor for a stimulus × group interaction effect. Of importance, this indicates that during test phase there was an effect of stimulus, and of importance, that this effect was the same across groups, which suggests that there was no added effect of experience. These Bayesian results were in concordance with the frequentist results.
Main effect Main effect Interaction effect Stimulus Group Stimulus × Group Phase BFMS+G:MG BFMS+G:MS BFMSG:MS+G
Habituation 0.31 1.03 0.23 Acquisition 14.67 × 105 0.68 1.63
Test 15.46 × 105 0.71 1.61
Table 3:Bayes factors for main and interaction effects on EMG responding for every condition-ing phase.
us-expectancy ratings One participant indicated during the exit interview that he interpreted the US-expectancy scale the other way around. Therefore his responses were reversed. The analyses were also performed excluding this one participant; this did not change the effects that were found.
During acquisition phase, participants learned to expect a US after the reinforced stimulus (CS+) but not after the unreinforced stimuli (CS−/+ and CS−), evident by higher US-expectancy ratings for the CS+ compared to the CS−/+ and the CS−, see acquisition in Figure 4; main effect of stimulus: F(1.02, 38.66) = 30.61, p < .001, η2 = .331. This effect did not differ across groups; stimulus × group: F(1.02, 38.66) < 1.
During the test phase, US-expectancy was higher for the Final CS+ compared to
the CS−, see test in Figure 4; main effect of stimulus: F(1, 38) = 21.31, p < .001, η2 = .198. Of importance, there were no differences in differential US-expectancy ratings between the combined condition (CS+vs. CS−) and the instructed condition (CS−/+ vs. CS−), indicating similar levels of US-expectancies in both groups; stimulus ×
results 15 Acq Test Instructed Acquisition Expectancy −4 −2 0 2 4 CS+CS− CS−/+ ● ● ● ● ● ● ● ● ● ● Acq Test Combined Acquisition Expectancy −4 −2 0 2 4 CS+CS− CS−/+ ● ● ● ● ● ● ● ● ● ●
Figure 4:Mean US-expectancy to the CS+, CS−/+and CS−during acquisition, and test phase for the instructed acquisition group (top) and the combined acquisition group (bot-tom).
Bayesian Analyses on US-expectancy data Bayes factors for main and interaction
ef-fects on the US-expectancy ratings in the acquisition and the test phase are presented in Table 4. For the acquisition phase, the Bayesian hypothesis test favour the model that includes the stimulus effect, indicating that participants successfully differenti-ated in expectancies between the CSs. There is no evidence for the effect of group, nor for the stimulus × group interaction effect, which showed that this effect was the same between groups. For the test phase the same model is preferred: including a main effect of stimulus, but not the main effect of group nor the stimulus × group interaction effect. This indicates that there is strong evidence for an effect of stimulus on the US-expectancy ratings, and that this effect was the same across groups; sug-gesting no added effect of direct experience. These results were again in line with the frequentist results.
results 16
Main effect Main effect Interaction effect Stimulus Group Stimulus × Group Phase BFMS+G:MG BFMS+G:MS BFMSG:MS+G
Acquisition 29.49 × 108 0.39 0.16
Test 28.69 × 108 0.38 0.15
Table 4:Bayes factors for main and interaction effects on US-expectancy ratings for acquisition phase and test phase.
approach avoidance task Statistical analysis showed that participants were not faster to approach the CS− and avoid the Final CS+ than vice versa, see Figure 5;
stimulus × response interaction: F(1, 38) < 1. Furthermore, there were no differences between experimental groups; stimulus × response × group interaction F(1, 38) < 1.
Instructed Acquisition 600 650 700 750 800
Response Times (msec)
Approach Avoidance CS− CS+ Combined Acquisition 600 650 700 750 800
Response Times (msec)
Approach Avoidance
CS− CS+
Figure 5:Mean median RT in ms in approaching and avoiding the CS−(left) and CS+(right).
3.2 Exploratory Analyses
For the exploratory analyses we included the trials in the statistical analyses to analyze the course of learning.
skin conductance response The results showed no baseline differences in SCR towards the different CSs in habituation phase, F(1.51, 57.37) < 1, and this did not differ between groups F(1.51, 57.37) < 1. There was, however, a significant stimu-lus × time × group interaction during the habituation phase, F(2, 76) = 3.46, p = .04, η2 = .019, indicating that respondents in the instructed acquisition group show a different SCR over time (from trial 1 to trial 2) for the CSs, than respondents in the combined acquisition group, see habituation trials 1 − 2 in Figure 6. Subgroup analyses indicated that the stimulus × time interaction was not significant for the instructed acquisition group (F(2, 38) = 2.34, p = .11), nor for the combined
acquisi-results 17
tion group (F(2, 38) = 1.41, p = .26). Thus, although there were differences between groups, this only concerns the pattern of results, and therefore has no implications on the interpretation of the results.
During the acquisition phase, SCR was higher for CS+compared to the CS−/+and the CS−, see acquisition trials 1 − 6 in Figure 6; main effect of stimulus: F(1.62, 61.51) = 40.73, p < .001, η2=.158. The main effect of stimulus did not differ over time; stimulus
× time: F(6.72, 255.51) = 1.05, p = .40, η2=.009, nor between groups; stimulus × time
× group: F(6.72, 255.51) = 1.08, p = .38, η2=.009.
As can be seen at the test trials 1 − 2 in Figure 6, SCR for the Final CS+was higher than for the CS−; main effect of stimulus: F(1, 38) = 5.08, p = .03, η2=.021. The effect of stimulus did not differ over trials; stimulus × time: F(1, 38) = 1.30, p = .26, η2 = .003. Furthermore, the course of SCR did not differ between groups; stimulus × time × group: F(1, 38) = 2.85, p = .10, η2=.007. 0.0 0.2 0.4 0.6 0.8 1.0 Instructed Acquisition SCR ● ● ● ● ● ● ● ● ● ● 1 2 1 2 3 4 5 6 1 2 Trials ● CS+ CS− CS−/+
Habituation Acquisition Test
0.0 0.2 0.4 0.6 0.8 1.0 Combined Acquisition SCR ● ● ● ● ● ● ● ● ● ● 1 2 1 2 3 4 5 6 1 2 Trials ● CS+ CS− CS−/+
Habituation Acquisition Test
Figure 6:Mean skin conductance to the CS+, CS−/+and CS−per trial during habituation, ac-quisition, and test phase for the instructed acquisition group (top) and the combined acquisition group (bottom).
results 18
fear potentiated startle Initially, FPS did not differ for the CSs, see habituation trials 1 − 2 in Figure 7; main effect of stimulus: F(2, 76) = 1.56, p = .22, η2 = .005. During habituation phase, FPS for the CSs did not differ over time and group; stimulus × time × group: F < 1.
As can be seen at acquisition trials 1 − 6 in Figure 7, FPS was higher for the CS+ compared to the CS−/+ and the CS−; main effect of stimulus: F(1.45, 55.16) =
25.69, p < .001, η2 = .045. This effect did not differ over trials; stimulus × time:
F(7.28, 276.50) = 1.84, p = .08, η2 =.008. Furthermore, the course of FPS towards the CSs did not differ between groups; stimulus × time × group: F(7.28, 276.50) < 1.
During test phase, see test trials 1 − 2 in Figure 7, FPS was higher for the Final CS+ compared to the CS−; main effect of stimulus: F(1, 38) = 2.25, p < .001, η2 = .093. Taking into account the different trials it is seen that differential FPS (Final CS+ vs.
CS−) differs over time; stimulus × time interaction: F(1, 38) = 7.52, p = .009., η2 =.02.
The course of FPS for the CSs over time did not differ between experimental groups; stimulus × time × group: F(1, 38) < 1.
10 12 14 16 18 20 Instructed Acquisition EMG ● ● ● ● ● ● ● ● ● ● ● CS+ CS− CS−/+ NA 1 2 1 2 3 4 5 6 1 2 Trials
Habituation Acquisition Test
10 12 14 16 18 20 Combined Acquisition EMG ● ● ● ● ● ● ● ● ● ● ● CS+ CS− CS−/+ NA 1 2 1 2 3 4 5 6 1 2 Trials
Habituation Acquisition Test
Figure 7:Mean fear potentiated startle to the CS+, CS−/+ and CS− per trial during habitu-ation, acquisition, and test phase for the instructed acquisition group (top) and the combined acquisition group (bottom).
results 19
us-expectancy ratings As can be seen from acquisition trials 1 − 6 in Figure 8 US-expectancy ratings for the CS+was higher than for the CS−/+and the CS−; main effect of stimulus: F(1.02, 38.65) = 29.68, p < .001, η2 = .294. This effect of stimulus did not differ over time; stimulus × time: F(4.48, 170.07) = 1.21, p = .31, η2 = .002. Furthermore, the course of US-expectancy ratings for the CSs did not differ between groups; stimulus × time × group: F(4.48, 170.07) = 1.27, p = .28, η2=.002.
During the test phase, US-expectancy ratings for the Final CS+were higher than for
the CS−, main effect of stimulus: F(1, 38) = 21.31, p < .001, η2 =.188. This difference in US-expectancy for the CSs did not differ over trials; time × stimulus: F(1, 38) < 1. Furthermore, there were no differences between groups in US-expectancy ratings for the CSs over time; time × stimulus × group: F(1, 38) < 1.
−10 −5 0 5 10 Instructed Acquisition Expectancy ● ● ● ● ● ● ● ● 1 2 3 4 5 6 1 2 Trials ● CS+ CS− CS−/+ Acquisition Test −10 −5 0 5 10 Combined Acquisition Expectancy ● ● ● ● ● ● ● ● 1 2 3 4 5 6 1 2 Trials ● CS+ CS− CS−/+ Acquisition Test
Figure 8:Mean US-expectancy to the CS+, CS−/+and CS− per trial during acquisition, and test phase for the instructed acquisition group (top) and the combined acquisition group (bottom).
fear ratings We compared self-reported CS fear towards the CS−/+ before and after instructions were given. Note that both groups received these instructions, but
results 20
Hab Acq Test
Instructed Acquisition Ratings −500 −400 −300 −200 −100 0 100 200 CS+ CS− CS−/+ ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
Hab Acq Test
Combined Acquisition Ratings −500 −400 −300 −200 −100 0 100 200 CS+ CS− CS−/+ ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●
Figure 9:Mean fear ratings to the CS+, CS−/+and CS− after each conditioning block for the instructed acquisition group (top) and the combined acquisition group (bottom).
only the instructed acquisition group encountered this specific stimulus afterwards. The instructions led to an increase in self-reported CS fear for the CS−/+, see Figure 9; main effect of time: F(1, 38) = 48.35, p < .001, η2=.301. The increase in fear ratings was the same in boht groups; time × group interaction: F(1, 38) = 1.27, p = .27, η2 = .011, indicating the effect of instructions to be the same, regardless of whether this stimulus is encountered later on.
Subsequent we tested whether actual experience of an aversive shock in addition to instructions would increase self-reported fear (as found in Raes et al., 2014). The results showed that there was a difference in self-reported CS fear towards the differ-ent CSs (main effect of stimulus: F(2, 76) = 50.43, p < .001, η2 =.357), but, unlike the results in (Raes et al., 2014), this effect did not differ between the two groups; stimulus × group interaction: F(2, 76) = 1.27, p = .29, η2 = .014. Follow-up analyses showed
that there was no difference in self-reported CS fear towards the stimulus that was actually paired with a shock (CS+) and the stimulus that is only told to be paired with a shock (CS−/+), t(38) = −0.52, p = .60.
conclusion and discussion 21
4
conclusion and discussion
In this study we investigated whether there is a difference in fear responses acquired via instructions alone compared to fear responses acquired via a combination of in-structions and direct experience. Participants were instructed that during test phase two stimuli (CS+and CS−/+) would be followed by a shock (US). However, in the
pre-ceding training phase, only the CS+ was actually paired with the US. During the test
phase, participants in the instructed acquisition group encountered the originally un-reinforced CS−/+, while participants in the combined acquisition group encountered the previous reinforced CS+.
The results of our study showed, first of all, that instructions about CS-US con-tingencies resulted in increased fear responses towards the conditioned stimulus, in terms of physiological responding (SCR and FPS) as well as subjective experience (US-expectancies and self-reported CS fear). Second, fear responses towards the con-ditioned stimuli did not differ for the instructed acquisition group compared to the combined acquisition group, indicating that there are no differences in fear acquired via a combination of instructions and direct experience compared to fear acquired via instructions alone.
Our study has some meaningful additions to other studies on fear conditioning. First, unlike previous studies, we tested our results within a Bayesian framework, enabling us to accumulate evidence in favour of the "null hypothesis" of no effect; conclusions we would otherwise not have been able to make. The Bayesian analyses were in line with the frequentist results, and, in addition, showed substantial support for our hypothesis that the elicited fear responses after instructions alone did not differ from fear responses after a combination of instructions and direct experience.
Second, we measured multiple indices of physiological and subjective components of fear and, of importance, all assessments of fear showed coherent emotional respond-ing. This is in contrast to earlier research on fear conditioning in general (Sevenster et al., 2012, 2014), and research on the added effect of direct experience on top of in-structions in particular (Raes et al., 2014). Compared with inin-structions alone, Raes et al. (2014) showed that there was an added effect of direct experience on self-reported fear, but not on SCR and US-expectancy. The authors provided different explanations for this restricted additive effect of experience on fear responses and argued that their results might be an underestimation of the actual effect. Our results, however, suggest that the added effect of direct experience on top of instructions in fear acquisition is rather limited, and perhaps even negligible.
Our finding, that mere instructions suffice to acquire fear responses, is in line with the propositional account of learning. According to propositional models, associative learning is mediated by the formation and evaluation of propositions (i.e., qualified statements) about relations in the world (e.g., "A causes B") (De Houwer, 2009). These models claim that propositions can be formed both via direct experience of the CS-US relation as well as via instructions about the CS-US relation. Our results are further supported by the notion of multiple pathways to fear conditioning (Rachman, 1977). Moreover, our results indicate that a combination of multiple pathways does not nec-essarily enhance fear acquisition.
The finding that instructions alone seem to be sufficient to elicit fear has important clinical implications. First, our results empirically support the notion that instructions alone can be the origin of childhood phobias (as suggested by the retrospecitve study of King et al., 1998). That is, it seems possible that psychopathologies can be acquired
conclusion and discussion 22
without directly experiencing the aversive outcome. Second, our results do also open up possibilities regarding the prevention of fear acquisition in children. By raising awareness, for instance among parents and teachers, about the possible effects that fearful instructions can have on fear acquisition, we might be able to limit the situ-ations in which children are at risk for developing childhood phobias solely on the basis of fearful instructions.
However, it should be noted that the aforementioned conclusions only pertain to the physiological and subjective components of fear. Action tendencies, the third emo-tional component of fear, were measured using an approach/avoidance reaction time task (AAT). The results showed that participants were not slower on incongruent trials (i.e., approach the conditioned fear stimulus or avoid the conditioned safety stimulus) than on congruent trials (i.e., avoid conditioned fear stimulus or approach conditioned safety stimulus). Thus, although conditioning – either via instructions alone or via a combination of instructions and direct experience – led to fear acquisition in terms of physiological responding and subjective experience, we were unable to show that fear conditioning led to fear responses in terms of action tendencies. Note however, that this was the same in both groups, which again indicates that there were no differences in fear acquisition between groups.
A possible explanation for the absence of an effect of fear conditioning on avoid-ance tendencies is the number of stimuli that the participants encountered. It has been argued that using AAT to measure action tendencies towards conditioned cues is primarily designed to differentiate between two conditioned stimuli (Arnaudova, Kry-potos, Effting, Kindt, & Beckers, in preperation). To incorporate this idea we choose a between-subject design so that participants only encountered two stimuli during the AAT. However, during the actual conditioning phase, the participants were presented with three stimuli and hence three CS-US pairings were learned. It could be argued that because the participants learned three CS-US contingencies, this might influenced the suitability of the AAT in measuring action tendencies.
In the current experimental design we did not control for the effect of latent in-hibition, which might resulted in an underestimation of the effect. During training all participants were explicitly instructed and later experienced that the CS−/+ was
not followed by a shock. Therefore, this stimulus might be seen as a "safety" signal, indicating a "safe" period in which no shock would be administered (Lubow, 1973). Subsequent instructions indicated that the same "safe" stimulus would now be some-times followed by a shock. It has been shown that it is more difficult to condition a "safety stimulus" than it is to condition a neutral stimulus (Moscovitch & LoLordo, 1968). Thus, it can be argued that in reality, instructions concerning "non-safety sig-nals" might even yield higher levels of fear than we found in the present study. Future research could test this possibility by, for example, masking the stimulus presenta-tions during fear acquisition. Subsequent instrucpresenta-tions can inform participants which particular stimulus was followed by a shock during acquisition, and afterwards fear responses towards this stimulus can be measured. In this way it is possibly prevented that the stimulus becomes a safety signal during acquisition.
Another possible limitation of the design is the possibility of US-generalization. In the current design, all participants experienced the aversive outcome. Therefore, it is possible that the instructions only lead to fear responses because the US information (i.e., "a shock will follow") referred to a situation that had been experienced (Raes et al., 2014). To overcome this problem, no shocks should be administered to partici-pants in the instructed acquisition group. However, in this case, participartici-pants in the
references 23
instructed group might doubt the credibility of the given instructions, resulting in another confounding factor.
In conclusion, we showed that physiological as well as subjective fear responses ac-quired via instructions alone are similar to fear responses acac-quired via a combination of instructions and direct experience. Our study suggests that instructions alone could be sufficient to acquire fear, which could have great clinical implications.
5
references
American Psychiatric Association. (2000). DSM-IV-TR: Diagnostic and statistical manual of mental disorders, text revision. American Psychiatric Association.
Arnaudova, I., Krypotos, A.-M., Effting, M., Kindt, M., & Beckers, T. (in preperation). Selective conditioning of cue competition: Evidence across measures.
Blumenthal, T. D., Cuthbert, B. N., Filion, D. L., Hackley, S., Lipp, O. V., & Van Boxtel, A. (2005). Committee report: Guidelines for human startle eyeblink electromyo-graphic studies. Psychophysiology, 42, 1–15.
De Houwer, J. (2009). The propositional approach to associative learning as an alter-native for association formation models. Learning & Behavior, 37, 1–20.
Dienes, Z. (2011). Bayesian versus orthodox statistics: Which side are you on? Per-spectives on Psychological Science, 6, 274–290.
Frijda, N. H. (2010). Impulsive action and motivation. Biological Psychology, 84, 570– 579.
Kass, R. E., & Raftery, A. E. (1995). Bayes factors. Journal of the American Statistical Association, 90, 773–795.
Kessler, R. C., Berglund, P., Demler, O., Jin, R., Merikangas, K. R., & Walters, E. E. (2005). Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the National Comorbidity Survey Replication. Archives of General Psychiatry,
62, 593–602.
King, N. J., Eleonora, G., & Ollendick, T. H. (1998). Etiology of childhood phobias: Current status of Rachman’s three pathways theory. Behaviour Research and Ther-apy, 36, 297–309.
Krypotos, A.-M., Beckers, T., Kindt, M., & Wagenmakers, E.-J. (submitted). A Bayesian hierarchical diffusion model decomposition of performance in approach-avoidance tasks.
Lang, P. J. (1995). The emotion probe: Studies of motivation and attention. American Psychologist, 50, 372–385.
Lubow, R. E. (1973). Latent inhibition. Psychological Bulletin, 79, 398–407.
Mauss, I. B., & Robinson, M. D. (2009). Measures of emotion: A review. Cognition and Emotion, 23, 209–237.
Milad, M. R., Goldstein, J. M., Orr, S. P., Wedig, M. M., Klibanski, A., Pitman, R. K., & Rauch, S. L. (2006). Fear conditioning and extinction: Influence of sex and menstrual cycle in healthy humans. Behavioral Neuroscience, 120, 1196–1203. Moscovitch, A., & LoLordo, V. M. (1968). Role of safety in the Pavlovian backward
fear conditioning procedure. Journal of Comparative and Physiological Psychology,
66, 673–678.
Olsson, A., & Phelps, E. A. (2004). Learned fear of “unseen” faces after Pavlovian, observational, and instructed fear. Psychological Science, 15, 822–828.
References 24
Peterson, R. A., & Reiss, S. (1993). Anxiety sensitivity index revised test manual. Wor-thington, OH: IDS Publishing.
Rachman, S. (1977). The conditioning theory of fear acquisition: A critical examination. Behaviour Research and Therapy, 15, 375–387.
Raes, A. K., De Houwer, J., De Schryver, M., Brass, M., & Kalisch, R. (2014). Do cs-us pairings actually matter? A within-subject comparison of instructed fear conditioning with and without actual cs-us pairings. PloS One, 9, e84888. Ratcliff, R. (1978). A theory of memory retrieval. Psychological Review, 85, 59.
Sevenster, D., Beckers, T., & Kindt, M. (2012). Instructed extinction differentially affects the emotional and cognitive expression of associative fear memory. Psy-chophysiology, 49, 1426–1435.
Sevenster, D., Beckers, T., & Kindt, M. (2014). Fear conditioning of SCR but not the startle reflex requires conscious discrimination of threat and safety. Frontiers in Behavioral Neuroscience, 8, 32.
Spielberger, C. D., Gorsuch, R. L., & Lushene, R. E. (1970). Manual for the state-trait anxiety inventory. Palo Alto, CA: Consulting Psychologists Press.
Wetzels, R., Grasman, R. P., & Wagenmakers, E.-J. (2012). A default Bayesian hypoth-esis test for ANOVA designs. The American Statistician, 66, 104–111.
appendix a: deviations from original proposal 25
6
appendix a: deviations from original proposal
6.1 Experiments
Originally we proposed to perform two experiments. As we indicated in our dis-cussion section, the current design does not control for the effect of latent inhibition. Therefore we wanted to perform a second experiment (see Appendix C) in which we control for this effect. However, due to time constraints we were unable to start the second experiment.
6.2 Data Analyses
The performed analyses differ somewhat from the proposed analyses (see Appendix D). Since during test phase all participants only got to see two of the three stimuli, and because the final reinforced stimulus was different between groups, we could not execute the analyses as we initially said we would.
6.2.1 Bayesian Hierarchical Diffusion Model
We wanted to fit a Bayesian hierarchical diffusion model on the AAT reaction time data. Unfortunately we were unable to perform these analyses due to time constraints.
appendix b: adjusted time schedule 26
7
appendix b: adjusted time schedule
Month Activities Hours
October – 1. Read background literature 40hours January 2. Weekly meetings
February 1. Read background literature 60hours (part-time) 2. Start on proposal
3. Weekly meetings
March 1. Proposal 160hours (full-time) 2. Weekly meetings
April 1. Data collection 100hours (21
2 weeks full-time)
2. Start on data analyses script 3. Weekly meetings
May 1. Data collection 120hours (part-time) 2. Finish data analyses script
3. Start working on final report 4. Weekly meetings
June 1. Data analyses 60hours (part-time) 2. Final report
3. Weekly meetings
July 1. Finish final report 60hours (11
2 weeks full-time)
2. Weekly meetings
appendix c: original proposed second experiment 27
8
appendix c: original proposed second
experi-ment
Habituation Acquisition Test AAT Instructed 2CS+ 12 masked
pre-sentations 2CS+ CS+ 2CS− (6 shocks) 2CS− CS− 2NA 6NA 2NA Combination 2CS+ 6CS+ 2CS+ CS+ 2CS− 6CS− 2CS− CS− 2NA 6NA 2NA
Table 6:Experimental Design of Experiment 2
The procedure will be similar to that of the current experiment, with a few modifi-cations, see Table 6 for the experimental design. Two geometrical pictures (50 mm × 50 mm) of a cube, and a cylinder serve as CS+ and CS− (counterbalanced across
participants). Furthermore, a virtual curtain will serve as the masked stimulus. habituation phase This will be identical to the current experiment
acquisition phase The acquisition phase will be identical to that of the current experiment for the combined acquisition group (except that there are now only two CSs (CS+ and CS−). In the instructed acquisition group however, there will be 12
presentations of the masked stimulus, of which 6 trials will be reinforced. Which trials will be reinforced will be determined semi-randomly (with no more than two consecutive reinforced trials).
test phase. The test phase will be the same as in the current experiment, except that before the test phase, participants in the instructed-acquisition group will be informed that behind the virtual curtain there was a specific picture (CS+) when they received a shock, and another picture (CS−) when they did not.
appendix d: original proposed data analyses 28
9
appendix d: original proposed data analyses
Here the original proposed data analyses are given, but only for the analyses that changed in the actual analyses of the results.
9.1 Confirmatory Analyses 9.1.1 Test
To test the first two hypotheses, we will perform a 2 (group: instructed vs. combina-tion) × 3 (stimulus: CS+, CS−1, CS−2) × 2 (trial: mean in acquisition vs. mean in test) ANOVA on US-expectancy ratings, SCR, and FPS, with stimulus and trial as within subject-factors and group as the between subject factor.
9.2 Exploratory Analyses 9.2.1 Fear Ratings
Also, we will compare self-reported CS fear in both experiments by performing a 2 (group: instructed vs. combination) × 2 (stimulus: CS+ vs. CS−) × (phase: post acquisition vs. post test-phase) ANOVA on self-reported CS fear.
