Internship Report
Individual Differences in Fear Generalization
Phillip Lino von Klipstein
Supervision: Inna Arnaudova, Tom Beckers
Fear conditioning is an influential etiological model of anxiety disorders. Yet, individual differences seem to play a much larger role in the acquisition of anxiety disorders than in acquisition of fear through fear conditioning. Fear generalization paradigms offer a promising approach to study individual differences in fear conditioning. They provide “weaker” situations through the usage of ambiguous stimuli, which allow more room for influences of individual differences. Using a generalization paradigm, the present study investigates the influence of neuroticism on fear responses. Previous research on this topic has produced heterogenic results. We argue that this heterogeneity is due to divergence between automatic and controlled fear responses and that neuroticism is only related to controlled fear responses towards ambiguous stimuli, but not to automatic responses. The study assesses automatic responses through fear-potentiated startle, skin conductance reaction and an approach-avoidance reaction time task, whereas controlled responses are assessed through US expectancy ratings and actual avoidance behavior.
Note: The following report merges a research proposal (abstract, introduction, method) and a later added results and discussion section.
INTRODUCTION
Pavlovian fear conditioning has been an etiological laboratory model of anxiety disorders for almost a century (Mineka & Zinbarg, 2006; Watson & Rayner, 1920). In the lab, it refers to the pairing of an initially neutral stimulus (e.g., a tone) with an aversive unconditioned stimulus (US, e.g., an electric shock). The neutral stimulus thereafter itself elicits a fear reaction and is referred to as a conditioned stimulus (CS). According to this model, pathological fear in the real world develops through the same mechanism: initially neutral stimuli (e.g., house spiders or crowded places) co-occur with an aversive or traumatic event (e.g., a frightened peer or a panic attack) whereby the initially neutral stimuli come to elicit excessive fear themselves (Mineka & Zinbarg, 2006).
The proposed study addresses the role of individual differences in fear conditioning and the controversy that individual differences seem to play a much larger role in the acquisition of pathological fear than in the acquisition of fear through fear conditioning in the lab (e.g., Beckers, Krypotos, Boddez, Effting, & Kindt, 2013). Fear conditioning in the lab is successful on the majority of people (Lissek et al., 2005), while in contrast only 10 to 30% of trauma survivors develop an anxiety disorder (Engelhard, van den Hout & McNally, 2008; Mineka & Zinbarg, 2006). Further, whether or not people develop an anxiety disorder is related to a number of personality traits (e.g. neuroticism, Gershunny & Sher, 1998; Jorm et al., 2000) and genetic markers (Gordon & Hen, 2004). However, investigating these factors in fear conditioning has lead to inconsistent results with most studies finding no relation to fear responses (e.g., Davidson, Payne & Sloane, 1964; Otto et al., 2007; Pineless, Vogt & Orr, 2009; but see e.g., Levey & Martin, 1981). In the current study, we will address this issue by further investigating the role of neuroticism, one proposed vulnerability factor for anxiety disorders.
Lissek, Pine and Grillon (2006) argued that most fear conditioning paradigms constitute “strong situations” that leave little room for interindividual differences (see also Beckers et al., 2013). Strong situations are defined as situations with unambiguous stimuli of a clear hedonic value that produce uniform appetitive or aversive responses (Lissek et al., 2006). In most differential fear conditioning studies the threat stimulus (CS+) and safety stimulus (CS-) unambiguously predict an aversive stimulus and the absence of an aversive stimulus, respectively, and thus produce uniform responses. More ambiguous situations in fear conditioning experiments should be better suited for the study of interindividual variation.
Fear generalization paradigms might offer such ambiguity (Lissek et al. 2008; Lommen, Engelhard & van den Hout, 2010). In such procedures, generalization stimuli (GS) provide ambiguity, because they are perceptually similar, yet distinguishable, to CS+ or CS-. For example, Lommen et al. (2010) used black and white circles as CS+ and CS- and tested reactions towards circles of intermediate shades of grey. If individual differences in fear conditioning exist, they should emerge in reactions towards GS as these stimuli provide a weaker situation that leaves more room for the effect of individual differences. Regarding neuroticism, Lommen et al. (2010) did find such differences in avoidance behavior. The proposed study further investigates the question, whether people high in neuroticism (N+) compared to people low in neuroticism (N-) show increased fear responses towards GS in the fear generalization paradigm.
Previous research regarding this question has also produced null results (e.g., Torrents-Rodas et al., 2013) leading to an overall mixed empirical picture. We propose that these differences in results might be explained by differences between automatic and controlled fear responses. According to Evers et al. (2013) automatic responses are considered relatively fast, efficient and unconscious, whereas controlled responses are considered relatively effortful, deliberate and conscious. Thereby, for example, physiological measures constitute automatic responses, while subjective ratings and behavioral measures are considered controlled. Evers et al. (2013) supported this distinction through evidence that convergence between measures within automatic and controlled response systems is much higher than between measures of different response systems.
We argue that differences between N+ and N- participants in fear reactions towards ambiguous stimuli will only occur on controlled but not on automatic measures because research in this area provides more positive results for controlled measures than for automatic measures. More specifically, studies on physiological measures such as skin conductance reaction (SCR; Boddez et al., 2012; Grillon & Ameli, 2001; Torrents-Rodas et al., 2013) and fear-potentiated startle (FPS; Kindt & Soeter, 2014; Torrents-Rodas et al., 2013) have almost exclusively produced null results. However, exceptions for FPS can be found in the studies by Grillon and Ameli (2001) and Craske et al. (2009). In contrast, on more controlled measures Baas, van Ooijen, Goudriaan and Kenemans (2008), Boddez et al. (2012) and Kindt and Soeter (2014) found relations to N for US expectancies and as mentioned above Lommen et al. (2010) found relations for avoidance behavior. However, while there are relatively more positive results for controlled measures, a number of null results have also been found (for US
expectancies: Lommen et al., 2010; Torrents-Rodas et al., 2013; for avoidance behavior: van Meurs, Wiggert, Wicker & Lissek, in press). Frijda (2010) argue that the differentiation between automatic and controlled responses also applies to behavior. Yet, research on the effect of neuroticism on fear responses towards ambiguous stimuli has not investigated automatic behavior. Krieglmeyer, De Houwer and Deutsch (2013) suggest that approach-avoidance tendencies constitute such automatic behavior and the present study therefore incorporates an approach-avoidance tendency measure.
Further support for the idea of a connection between neuroticism and controlled fear responses comes from ample evidence that N+ individuals show cognitive processing biases towards threatening stimuli, such as attentional bias (Bar-Haim, Lamy, Pergamin, Bakermans-Kranenburg & van IJzendoorn, 2007), interpretation bias (the tendency to interpret ambiguous stimuli as threatening; Calvo & Castillo, 2001; Eysenck, McLeod & Mathews, 1987) and judgmental bias (the tendency to overestimate the likelihood and costs of negative outcomes; Butler & Mathews, 1983; Eysenck & Derakshan, 1997). As these biases constitute distortions of deliberate evaluations they should influence responses on controlled measures, but leave automatic responses unaffected.
To investigate the suggested differences between automatic and controlled fear responses the present study assesses multiple controlled and automatic measures in the paradigm used by Lommen et al. (2010). We hypothesize that N+ participants relative to N- participants show increased responses towards GSs on controlled but not on automatic measures. Specifically, we expect such group differences in US expectancies and avoidance behavior but not for approach-avoidance tendencies, FPS and SCR.
METHOD
Sample characteristicsParticipants will be recruited through an online advertisement and screened by phone. Participants between 18 and 50 years old, with proficient Dutch language skills, no major general health problems, no psychiatric disorder, no anxiety disorder, no post-traumatic stress disorder, no ADHD, no epilepsy, no heart problems, no pacemaker, not on medication that can influence attention, reactions or memory, no colorblindness, no hearing problems and (for women) that are not pregnant are eligible for the study. The intended sample size is N = 65. Participants will receive course credit or financial compensation for their efforts.
Stimuli and apparatus
Conditioning and generalization stimuli. Following the procedure of Lommen et al. (2010), a set of 10 colored circles (56 mm diameter) will be used as conditioning and generalization stimuli (see Figure 1). For a random half of the participants, the circles will be assigned to stimulus type in opposite order, so that the two darkest circles serve as CS+.
Electric stimulus. The US will be a mild 2-ms electric shock, with an intensity set to a “not painful, but quite uncomfortable” level in a work-up procedure (Orr et al., 2000), delivered to the dorsal surface of the forearm of non-dominant hand.
Fear potentiated startle. FPS will be measured through electromyography of the right orbicularis oculi muscle in response to a 40-ms burst of white noise (104 dB) delivered through headphones. Two 6-mm sintered Ag/AgCl electrodes filled with a conductive gel will be positioned approximately 1 cm under the pupil and 1 cm below the lateral canthus, respectively; a ground electrode will be placed on the forehead, 1 cm below hairline (Blumenthal et al., 2005).
Skin conductance response. Electrodermal activity will be measured using two Ag/AgCl electrodes of 20 by 16 mm attached with adhesive tape to the medial phalanges of the first and third fingers of the non-dominant hand.
Procedure
Upon arrival participants will receive information about the study and provide informed consent. Throughout the study the experimenter will give instructions in English, whereas instruction on the screen will be provided in Dutch. At the beginning of the study the electrodes for physiological measurement will be attached and an appropriate strength of the electric shock will be determined. Participants will go through five experimental stages: an acquisition stage, an approach-avoidance reaction time task (AART task), an actual avoidance stage, a test stage and an assessment stage.
Acquisition stage. The acquisition stage will start with 10 startle probes for habituation. Thereafter, 24 trials will incorporate 4 trials of each CS and 8 trials of noise alone (NA; only startle probe). Trial order will be randomized in blocks, so that the same stimulus is presented no more than twice in a row. In all stimulus trials, first a stimulus is presented for 8 s along with a rating scale for US expectancy on which participants will indicate whether they expect the current stimulus to be followed by an electric shock. The scale will range from -5 “certainly no electric stimulus” to 5 “certainly an electric stimulus”. Participants will have 5 s after stimulus onset to give their rating. 7 s after trial onset the startle probe will be presented. On CS+ trials the electric shock will be presented 7.5 s after trial onset. Trials are separated by inter-trial intervals (ITI) of average 20 s (15 s, 20 s or 25 s). The acquisition stage will end with a 3 min memory consolidation break where nothing will be asked of participants.
AART task. Before the AART task all electrodes will be removed. The AART task will consist of a first phase investigating responses towards the CS+ and the CS- and a second phase investigating responses towards GS2 and GS5. Each phase will include two blocks of 20 trials (4 practice, 16 test) each. Half of practice and test trials will show CS+/GS2, half will show CS-/GS5. The order of stimuli will be semi-randomized (i.e., not more than two consecutive stimuli of the same type). At the beginning of each trial a manikin figure (36 mm x 21 mm) will appear in the middle of the bottom or top half of the screen. After 1.5 s a stimulus will be presented in a 59 mm x 106 mm white rectangle, whose orientation (horizontal or vertical) will determine whether participants will have to make the manikin move towards or away from the stimulus by pressing the Y-button (labeled ) or the
B-button (labeled ), respectively. Both written and oral instructions will emphasize accuracy and speed of responses. After a response, the manikin will move in the indicated direction for 2 s. In case of a false response, a red cross will be presented for 0.5 s where the manikin started, whereas correct responses will not receive feedback. The next trial will then start after a 2 s ITI. The time between stimulus onset and participants’ response will be measured as the dependent variable. The order in which horizontal versus vertical shapes are approached will be randomized between participants
Actual avoidance stage. Before the actual avoidance stage the shock electrodes will be reattached. Participants will see each of the 10 stimuli once for 8s in randomized order. Trials will be separated by a 1 s ITI. Participants will be instructed that they can avoid a potential electric shock by pressing the SPACE-button for all those stimuli that show a caption below the circle saying “The SPACE-button is now activated”. The caption will be presented for all stimuli but CS+, which will be followed by an electric shock. Instructions will emphasize to only press the SPACE-button if the participant expects an electric shock. The number of SPACE-button presses will be recorded as the dependent variable.
Test stage. Before the test stage, electrodes for SCR and FPS will be reattached. The test stage will begin with 10 habituation startle probes. Thereafter, 6 trials consisting of 2 CS+, 2 CS- and 2 NA will serve as a reminder phase that refreshes participants’ memory of CS-US contingencies. The following 24 actual test trials will include two blocks of one trial of each CS and GS and two NA trials. Trials will be presented in a semi-randomized order, so that blocks of 6 trials include one NA plus one stimulus of each of the following pairs: the two CS+, GS1/GS2, GS3/GS4, GS5/GS6, the two CS-. CS+ presentations will be followed by an electric shock. Participants will again be asked to give US expectancy ratings on the same rating scale as in the acquisition stage and to base them “on what you have learned in the first stage when all electrodes where attached”. Stimulus timing will match the acquisition phase. Assessment stage. In the assessment stage participants will fill out the neuroticism scale of the Eysenck Personality Questionnaire (EPQ-N; Eysenck & Eysenck, 1975) on the computer. This questionnaire consists of 22 items that are answered with yes (=1) or no (=0). Additionally, participants will rate how pleasant each of the 10 stimuli and the electric stimulus was on a scale from -5 “unpleasant” to 5 “pleasant”. Also, they will rate on a scale from 1 to 5 how intense and how startling the electric shock was.
Statistical analysis
Participants that completely fail to learn CS-US contingencies (defined as US expectancy for CS+ < 0 or for CS- > 0 on last trials) will be excluded from further analyses. Following Lommen et al. (2010) participants with EPQ-N < 4 will be assigned to a N- group and participants with N-EPQ > 11 to a N+ group. However, if this method of group assignment will lead to a substantial loss of power, a median split will be performed instead. Acquisition. First, mean responses across trials will be calculated for US expectancies, SCR and FPS. Then, acquisition data will be analyzed by 2 (stimulus: CS+, CS-) x 2 (group: N+, N-) ANOVAs, using Greenhouse-Geisser correction as necessary, with stimulus as within-subject factor and group as between-within-subject factor.
AART. For the AART practice trials, trials with false responses and test trials with reaction times > 3 s will be excluded from analyses (Krypotos, Effting, Arnaudova, Kindt & Beckers, 2014). Data will be analyzed in 2 (stimulus) x 2 (behavior: approach, avoid) x 2 (group: N+, N-) ANOVAs for both CS+ vs. CS- and GS2 vs. GS5, with stimulus and behavior as within-subject factors and group as between-within-subjects factor.
Actual avoidance. Each stimulus will be assigned a color value, so that 0 represents the first CS-, 1 represents the second CS- and so forth until 8 and 9 represent the two CS+. For each participant an average color value of avoided stimuli will be calculated. Correlations between neuroticisms and average color value of avoided stimuli, as well as between neuroticism and the total number of avoided stimuli will be calculated and tested for significance.
Test. Reminder phase data for US expectancy, SCR and PFS will be analyzed in 2 (stimulus: CS+, CS-) x 2 (group: N+, N-) to check that previous tasks did not lead to extinction of fear responses. Stimuli will be categorized into categories of two, so that always two neighboring stimuli constitute one category. US expectancy, SCR and FPS data from the test stage will be subjected to two 2 (stimulus category: CS+, CS- and GS12, GS56) x 2 (group: N+, N-) ANOVAs, with stimulus category as within-subjects factor and group as between-subjects factor.
ADDITION TO METHOD
ParticipantsA total of 66 adults participated in the study. Two participants were excluded from the analysis because their physiological data was lost, resulting in a sample of 64 participants (41 females; mean age = 22.09; age range = 18-41). The remaining participants all successfully learned CS-US contingencies. 24 participants were classified as N+ and 17 as N- following the criteria of Lommen et al. (2010). All analyses on neuroticism group differences were performed using this sample of 41 participants (25 females, mean age = 21.51, age range = 18-28).
Statistical Analysis
RESULTS
No group differences were observed in the selected shock level, t(39) = 0.87, p = .39, or any of the shock perception measures, ps > .10. Mean shock perception ratings indicated that participants perceived the shock as unpleasant (M = -3.39), intense (M = 2.95) and startling (M = 3.63).
Acquisition and Reminder Stages
US expectancy. The data showed that both neuroticism groups learned CS-US contingencies and that learning remained until the reminder phase. Participants expected a shock after a CS+ and not after a CS- in both the acquisition phase, F(1, 39) = 1584.08, p < .001, η𝑝2 = .98, and the reminder phase, F(1, 39) = 1878.97, p < .001, η𝑝2 = .98. Further, there were no group differences in distinguishing stimuli in either acquisition phase, F < 1, or reminder phase, F(1, 39) = 1.64, p = .21, η𝑝2 = .04.
FPS. The same pattern of learning appeared for FPS. Participants responded more strongly towards CS+ than CS- in both the acquisition phase, F(1, 39) = 5.91, p = .02, η𝑝2 = .13, and the reminder phase, F(1, 39) = 4.81, p = .03, η𝑝2 = .11. Also, there were again no group differences in distinguishing stimuli in either acquisition phase, F(1, 39) = 1.68, p = .20, η𝑝2 = .04, or reminder phase, F < 1.
SCR. Learning also occurred on SCRs. Participants responded more strongly towards CS+ than CS- in both the acquisition phase, F(1, 39) = 9.38, p = .004, η𝑝2 = .19, and the reminder phase, F(1, 39) = 4.23, p = .05, η𝑝2 = .10. Further, there were again no group differences in distinguishing stimuli in either acquisition phase, F < 1, or reminder phase, F(1, 39) = 1.00, p = .32, η𝑝2 = .02.
AART
AART reaction times towards CS+ and CS- are displayed in Figure 1. Participants did not exhibit the critical stimulus x response interaction, F < 1, suggesting no difference between CS+ and CS- in reaction time patterns. Further, neuroticism group did not have any influence, Fs < 1. The absence of any differences between CS+ and CS- in approach-avoidance patterns is problematic and questions whether the AART task worked in the present study.
Figure 2 shows AART reaction times towards GS2 and GS5. Responses showed that there was a significant stimulus x response interaction, F(1, 39) = 6.53, p = .01, η𝑝2 = .14. More specifically, participants approached both GS2 and GS5 faster than avoiding them, t(40) = 2.50, p = .02 and t(40) = 5.67, p < .001,
while approaches were made faster towards the GS5 than towards the GS2, t(40) = 2.37, p = .02, but no difference between stimuli was found for avoidance responses, t < 1. Thereby, contrary to the pattern for CS+ and CS-, responses towards GS2 and GS5 exhibited differences in approach-avoidance patterns between stimuli, suggesting that the AART task worked for these stimuli. Further, in line with our hypothesis, the critical stimulus x response x neuroticism group interaction remained non-significant, F(1, 39) = 0.59, p = .45, η𝑝2 = .01, indicating that there were no group differences in approach avoidance tendencies. This suggests that neuroticism does not have any influence on approach-avoidance tendencies towards ambiguous stimuli.
Fig. 1. Mean RT towards CS+ and CS-
when approaching or avoiding the stimulus in the AART (error bars represent the standard error of the mean).
Actual avoidance
Figure 3 shows the number of actual avoidance responses towards stimuli. Participants displayed avoidance responses towards all stimuli and the more a stimulus resembled the CS+ the more frequently participants avoided it (except for CS+ itself). Also, some participants tried to avoid shocks following the CS+ despite a signal that shocks could not be avoided. Contrary to our hypothesis neither the mean color value of avoided stimuli, nor the number of avoided stimuli correlated with neuroticism, r = .02, t(62)
= 0.15, p = .44 and r = .06, t(62) = 0.47, p = .32, respectively. Thus, there seemed to be no relation between neuroticism and actual avoidance responses.
Fig. 2. Mean reaction times in ms towards GS2 and GS5 for
low neuroticism (left panel) and high neuroticism individuals (right panel) in the AART (error bars represent the standard error of the mean).
Fig. 3. Number of participants that showed an
Test
US expectancy. Participants showed higher US expectancies for CS+ than for CS-, F(1, 39) = 2355.00, p < .001, η𝑝2 = .98. Further, neuroticism group did not influence US expectancies towards CSs by itself or in interaction with stimulus, Fs < 1. The same pattern emerged for GSs. As illustrated in Figure 4, US expectancies were higher for GS12 than for GS56, F(1, 39) = 57.82, p < .001, η𝑝2 = .60. Critically, there was no stimulus x neuroticism group interaction,
F(1, 39) = 0.01, p = .91, η𝑝2 = .0003. Thereby, contrary to our hypothesis, neuroticism did not have an influence on US expectancies towards GSs.
FPS. FPS towards CSs was stronger for CS+
than for CS-, F(1, 39) = 21.43, p < .001, η𝑝2 = .35, while neuroticism group did not have any influence,
Fs < 1. Again, the same pattern emerged for GSs. As
Figure 5 illustrates, FPS was stronger for GS12 than for GS56, F(1, 39) = 7.19, p = .01, η𝑝2 = .16. The critical stimulus x neuroticism group interaction remained non-significant, F(1, 39) = 0.49, p = .49, η𝑝2 = .01. Thereby, in line with our hypothesis, neuroticism did not have an influence on FPS towards GSs.
SCR. SCR towards CSs showed no effect of stimulus, F(1, 39) = 1.54, p = .22, η𝑝2 = .04, neuroticism group, F(1, 39) = 1.87, p = .18, η𝑝2 = .05, or their interaction, F < 1. SCR towards GSs are displayed in Figure 6. The same pattern emerged for
GSs. More specifically, there was no effect of stimulus, F(1, 39) = 1.86, p = .18, η𝑝2 = .05, or of neuroticism group, F < 1. Critically, in line with our hypothesis there was no stimulus x
Fig. 4. Mean US expectancy ratings for
GS12 and GS56 in the test phase (error bars represent the standard error of the mean).
Fig. 5. Mean raw startle EMG in μV towards
GS12 and GS 56 (error bars represent standard error of the mean).
neuroticism group interaction, F(1, 39) = 0.44, p = .51, η𝑝2 = .01. Thereby, in line with our hypothesis, neuroticism did not have an influence on SCR towards GSs.
Fig. 6. Mean raw SCR in μS towards GS12
and GS56 (error bars represent standard errors of the mean).
DISCUSSION
The present study investigated whether high neuroticism individuals show stronger fear responses towards ambiguous stimuli in a fear generalization paradigm than low neuroticism individuals. To account for mixed results in the past, we hypothesized that such differences would only occur for controlled measures (i.e., US expectancy and actual avoidance) and not for automatic measures (i.e., FPS, SCR and approach-avoidance tendencies). The results indicate that neuroticism did not have an influence on fear responses towards GSs for any of the included measures, thereby contradicting our hypotheses for controlled measures and confirming our hypotheses for automatic measures. While our hypotheses were thus met for automatic measures, our central hypothesis to find differences between automatic and controlled measures was not met. Our most striking finding is surely the uniform absence of an effect of neuroticism across measures.
This study provides a very comprehensive test of differences in fear conditioning between people high and low on neuroticism. It covers fear acquisition, later responses towards CSs, and most importantly responses towards stimuli that are perceptually similar to the CSs. The dimensional character of these stimuli allows a direct look at the relation between similarity to the CSs and fear responses. In the past, the fear generalization paradigm has shown to be able to find individual differences in fear generalization (e.g., Lissek et al., 2010, 2013). Further, the present study covers a large spectrum of fear responses including automatic and controlled measures, as well as measures of behavioral, cognitive and physiological responses. Despite these strengths of the study, some of our results question its functionality on certain measures. More specifically, participants showed no different responses between CS+ and CS- in the AART task and in SCR at test. However, a re-analysis of AART data (see Appendix) and the observed differences between GSs suggest that the AART task in fact worked. Further, SCR did show differences between CSs at acquisition and reminder.
Evaluating our findings in light of other studies that used a fear generalization paradigm reveals a rather clear pattern, with most studies finding no effect of neuroticism. More specifically, including the present study, ten out of eleven investigated connections between neuroticism and a measure of fear response towards GSs did not find an effect (Lommen et al., 2010; Torrents-Rodas et al., 2013; van Meurs et al., in press). Only Lommen et al. (2010) found an effect of neuroticism on actual avoidance. The present study provides a
close replication of their design, but failed to replicate this finding1. Since the fear generalization paradigm, at least in its current application, uniformly does not find any effect of neuroticism across measures, it also does not provide a meaningful test of differences between automatic and controlled fear responses. Therefore, no conclusions are warranted from the present results regarding this distinction.
Besides the fear generalization paradigm, studies have used other paradigms to investigate fear responses towards ambiguous stimuli. Boddez et al. (2012) found an effect of neuroticism in a blocking paradigm, as well as Grillon and Ameli (2001), and Kindt and Soeter (2014) in fear inhibition paradigms. Importantly though, these paradigms likely tap into different processes than the fear generalization paradigm. More specifically, the blocking paradigm tabs into selectivity of threat appraisal (Boddez et al., 2012); that is, does a person attribute threat to all stimuli that are present during an aversive event or just to those that are actually predictive of the aversive event. This process is likely not involved in the fear generalization paradigm.
Distinguishing inhibition paradigms from generalization paradigms is more complex because generalization paradigms theoretically involve both excitatory and inhibitory learning. Excitatory learning describes the formation of CS-threat association (e.g., white circle – shock), whereas inhibitory learning describes the formation of CS-safety association (e.g., black circle – no shock) (Bouton, 1993; LeDoux, 1995). The competition between these associations is considered to determine the strength of fear responses towards a stimulus (Myers & Davis, 2004). In the generalization paradigm responses towards GSs should depend on generalized excitatory associations from the CS+ and generalized inhibitory associations from the CS-. If such generalization is the only determinant of fear responses in our study and if excitatory and inhibitory associations generalize with equal strength one would expect an antisymmetric shape of generalization gradients. As Figure A2-A5 (in the Appendix) show, generalization gradients do not match such a shape, mostly because responses towards GS34 are too low. These gradients suggest that the inhibitory associations generalized stronger, a conclusion that seems unlikely since research on the generalization of extinction indicates that generalization of inhibitory associations is not very powerful2 (Vervliet, Vansteenwegen,
1
Next to our correlational analysis, we replicated their analysis using group comparisons for the number of avoided stimuli and the mean color-value of avoided stimuli. Participants did not show any differences, ts < 1.
2
One might also argue that presenting GSs without a shock in a previous stage of the experiment caused low responses towards GSs. However, such an influence should be of equal strength for all GSs, which would preserve an antisymmetric shape of gradients at least among GSs. Also, this argument is only applicable to data obtained in the test phase.
Baeyens, Hermans & Eelen, 2005; Vervliet, Vansteenwegen & Eelen, 2004). Further, one might even expect generalization of excitatory associations to be stronger because it is more important evolutionary. Generalization of excitatory associations protects individuals from potential threats, whereas generalization of inhibitory associations only reduces false alarms. Since generalized associations thus do not explain the results convincingly, another influence is likely present. We suggest that this influence is participants’ explicit understanding of the pattern of shock delivery. When participants notice that they are being tested on stimuli along a dimension, they likely use this knowledge as a heuristic for whether to expect a shock. This reduces responses towards GSs that are easily distinguishable from the CS+ (i.e., GS34 and GS56), which possibly produces a floor effect that makes differences in generalization of inhibitory associations undiscoverable. Thereby, the fear generalization paradigm likely does not capture fear inhibition processes adequately and should be distinguished from fear inhibition paradigms.
While different underlying processes might account for differences in past results, it is also important to note that generalization studies have only investigated generalization of well-established threat associations. Past research as well as the present study made sure of a thorough fear acquisition through high CS-US contingencies and the number of acquisition trails used (e.g., Lommen et al., 2010). We suggest turning attention towards the generalization of less well-established threat associations. This holds the advantage of creating an even weaker situation and thereby increasing the room for individual differences. Further, it is highly adaptive to generalize fear towards established threats and an appropriate strength of such generalization increases fitness, thus creating a uniform strength of generalization across individuals. Such a unifying influence of natural selection might arguably be lower for generalization of ambiguous threats as such generalization is less critical for survival. Therefore, one might expect more individual differences in generalization of fear towards ambiguous threats. Further, using such ambiguous threats would enable future research to turn towards the question whether individual differences might lie in the threshold of fear that is necessary for the fear to generalize, rather than in the strength of fear generalization itself.
In summary, the uniform absence of an effect of neuroticism in the fear generalization paradigm and the fact that studies investigating different processes have been more successful in finding effects, suggests that the fear generalization paradigm simply does not capture the processes involved in causing the vulnerability of N+ individuals to develop anxiety
disorders. It is therefore questionable whether the primary process covered by the paradigm, namely generalization of excitatory associations, is underlying this vulnerability. However, future research should address generalization of less well-established threats to further understand this process. Finally, fear generalization research could so far not account for the paradox that fear conditioning outcomes are mostly unrelated to neuroticism, yet other paradigms and processes, like fear inhibition (Kindt & Soeter, 2014), constitute promising avenues for future efforts.
REFERENCES
Baas, J. M. P., van Ooijen, L., Goudriaan, A., & Kenemans, J. L. (2008). Failure to condition to a cue is associated with sustained contextual fear. Acta Psychologica, 127(3), 581– 92.
Bar-Haim, Y., Lamy, D., Pergamin, L., Bakermans-Kranenburg, M. J., & van IJzendoorn, M. H. (2007). Threat-related attentional bias in anxious and nonanxious individuals: A meta-analytic study. Psychological Bulletin, 133, 1–24.
Barlow, D. H. (2000). Unraveling the mysteries of anxiety and its disorders from the perspective of emotion theory. American Psychologist, 55, 1247- 1263.
Beckers, T., Krypotos, A.-M., Boddez, Y., Effting, M., & Kindt, M. (2013). What’s wrong with fear conditioning? Biological Psychology, 92(1), 90–96.
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.
Boddez, Y., Vervliet, B., Baeyens, F., Lauwers, S., Hermans, D., & Beckers, T. (2012). Expectancy bias in a selective conditioning procedure: trait anxiety increases the threat value of a blocked stimulus. Journal of Behavior Therapy and Experimental
Psychiatry, 43(2), 832–837.
Bouton, M. E. (1993). Context, time, and memory retrieval in the interference paradigm of Pavlovian learning. Psychological Bulletin, 114, 80–99.
Butler, G., & Mathews, A. (1983). Cognitive processes in anxiety. Advances in Behaviour
Research and Therapy, 5, 51–62.
Calvo, M. G., & Castillo, M. D. (2001). Selective interpretation in anxiety: Uncertainty for threatening events. Cognition and Emotion, 15, 299–320.
Craske, M. G., Waters, A. M., Nazarian, M., Mineka, S., Zinbarg, R. E., Griffith, J. W., ... Ornitz, E. M. (2009). Does neuroticism in adolescents moderate contextual and explicit threat cue modulation of the startle reflex?. Biological psychiatry, 65(3), 220-226.
Davidson, P. O., Payne, R. W., & Sloane, R. B. (1964). Introversion, Neuroticism, and Conditioning. Journal of Abnormal Psychology, 68(2), 136–143.
Engelhard, I. M., van den Hout, M. A., & McNally, R. J., (2008). Memory consistency for traumatic events in Dutch soldiers deployed to Iraq. Memory, 16, 3–9.
Evers, C., Hopp, H., Gross, J. J., Fischer, A. H., Manstead, A. S. R., & Mauss, I. B. (2013). Emotion response coherence: A dual-process perspective. Biological Psychology. Eysenck, H. J., & Eysenck, S. B. G. (1975). Manual of the Eysenck personality questionnaire.
San Diego, CA: Educational and Industrial Testing Service.
Eysenck, M. W., & Derakshan, N. (1997). Cognitive biases for future negative events as a function of trait anxiety and social desirability. Personality and Individual
Differences, 22, 597–605.
Eysenck, M. W., MacLeod, C., & Mathews, A. (1987). Cognitive functioning and anxiety.
Psychological Research, 49, 189–195.
Frijda, N. H. (2010). Impulsive action and motivation. Biological psychology, 84(3), 570-579.
Gershuny, B. S., & Sher, K. J., (1998). The relation between personality and anxiety: find- ings from a 3-year prospective study. Journal of Abnormal Psychology, 107, 252–262. Gordon, J. A., & Hen, R., (2004). Genetic approaches to the study of anxiety. Annual Review
of Neuroscience, 27, 193–222.
Grillon, C., & Ameli, R. (2001). Conditioned inhibition of fear-potentiated startle and skin conductance in humans. Psychophysiology, 38(5), 807–15.
Hamm, A. O., & Weike, A. I. (2005). The neuropsychology of fear learning and fear regulation. International Journal of Psychophysiology, 57(1), 5–14.
Jorm, A .F., Christensen, H., Henderson, A. S., Jacomb, P. A., Korten, A. E., & Rodgers, B., (2000). Predicting anxiety and depression from personality: is there a synergistic effect of neuroticism and extraversion? Journal of Abnormal Psychology, 109, 145– 149.
Kindt, M., & Soeter, M. (2014). Fear Inhibition in High Trait Anxiety. PloS one, 9(1), e86462.
Krieglmeyer, R., De Houwer, J., & Deutsch, R. (2013). On the nature of automatically triggered approach–avoidance behavior. Emotion Review, 5(3), 280-284.
Krypotos, A.-M., Effting, M., Arnaudova, I., Kindt, M., & Beckers, T. (2014). Avoided by Association: Acquisition, Extinction, and Renewal of Avoidance Tendencies Toward Conditioned Fear Stimuli. Clinical Psychological Science, 2(3), 336-343.
LeDoux, J. E. (1995). Emotion: Clues from the brain. Annual Review of Psychology, 46, 209– 235.
Levey, A., & Martin, I. (1981). Personality and conditioning. In H. Eysenck (Ed.), A model
for personality (pp. 123–168). Berlin, West Germany: Springer-Verlag.
Lissek, S., Biggs, A. L., Rabin, S. J., Cornwell, B. R., Alvarez, R. P., Pine, D. S., & Grillon, C. (2008). Generalization of conditioned fear-potentiated startle in humans: Experimental validation and clinical relevance. Behaviour Research and Therapy,
46(5), 678–687.
Lissek, S., Kaczkurkin, A. N., Rabin, S., Geraci, M., Pine, D. S., & Grillon, C. (2013). Generalized Anxiety Disorder Is Associated with Overgeneralization of Classically Conditioned Fear. Biological Psychiatry, 1–7.
Lissek, S., Pine, D. S., & Grillon, C. (2006). The strong situation: A potential impediment to studying the psychobiology and pharmacology of anxiety disorders. Biological
Psychology, 72(3), 265–270.
Lissek, S., Rabin, S., Heller, R. E., Lukenbaugh, D., Geraci, M., Pine, D. S., & Grillon, C. (2010). Overgeneralization of conditioned fear as a pathogenic marker of panic disorder.
The American Journal of Psychiatry, 167(1), 47–55.
Lommen, M. J. J., Engelhard, I. M., & van den Hout, M. a. (2010). Neuroticism and avoidance of ambiguous stimuli: Better safe than sorry? Personality and Individual
Differences, 49(8), 1001–1006.
Mineka, S., & Zinbarg, R. (2006). A contemporary learning theory perspective on the etiology of anxiety disorders: it’s not what you thought it was. The American
Myers, K. M., Davis, M. (2004). AX+/BX- discrimination learning in the fear potentiated startle paradigm: Possible relevance to inhibitory fear learning in extinction. Learning & Memory, 11(4), 464–475.
Orr, S. P., Metzger, L. J., Lasko, N. B., Macklin, M. L., Peri, T., & Pitman, R. K. (2000). De novo conditioning in trauma-exposed individuals with and without posttraumatic stress disorder. Journal of Abnormal Psychology, 109, 290–298.
Otto, M. W., Leyro, T. M., Christian, K., Deveney, C. M., Reese, H., Pollack, M. H., & Orr, S. P. (2007). Prediction of “fear” acquisition in healthy control participants in a de novo fear-conditioning paradigm. Behavior Modification, 31(1), 32–51.
Pineles, S. L., Vogt, D. S., & Orr, S. P. (2009). Personality and fear responses during conditioning: Beyond extraversion. Personality and Individual Differences, 46(1), 48– 53.
Soeter, M., & Kindt, M. (2010). Dissociating response systems: erasing fear from memory.
Neurobiology of Learning and Memory, 94(1), 30–41.
Tabbert, K., Stark, R., Kirsch, P., & Vaitl, D. (2006). Dissociation of neural responses and skin conductance reactions during fear conditioning with and without awareness of stimulus contingencies. Neuroimage, 32(2), 761-770.
Torrents-Rodas, D., Fullana, M. A., Bonillo, A., Caseras, X., Andión, O., & Torrubia, R. (2013). No effect of trait anxiety on differential fear conditioning or fear generalization. Biological Psychology, 92(2), 185–90.
van Meurs, B., Wiggert, N., Wicker, I., & Lissek, S. (in press). Maladaptive Behavioral Consequences of Conditioned Fear-Generalization: A Pronounced, Yet Sparsely Studied, Feature of Anxiety Pathology. Behaviour Research and Therapy.
Vervliet, B., Vansteenwegen, D., Baeyens, F., Hermans, D., & Eelen, P. (2005). Return of fear in a human differential conditioning paradigm caused by a stimulus change after extinction. Behaviour research and therapy, 43(3), 357-371.
Vervliet, B., Vansteenwegen, D., & Eelen, P. (2004). Generalization of extinguished skin conductance responding in human fear conditioning. Learning & Memory, 11(5), 555-558.
Watson, J. B., & Rayner, R. (1920). Conditioned emotional reactions. Journal of
APPENDIX
Re-analysis of AART data (CS+, CS-)main effect of stimulus, F(1, 62) = 4.68, p = .03,
main effect of response, F(1, 62) = 8.92, p =
.004, trend towards a stimulus x response interaction, F(1, 62) = 3.25, p = .08
Gradients
Fig. A2. Mean raw SCR in μS (error bars
represent standard errors of the mean).
Fig. A3. Mean raw startle EMG in μV (error
bars represent standard error of the mean).
Fig. A1. Mean median RT towards CS+ and
CS- in AART task (error bars represent standard errors of the mean); for full sample, N = 64.
Fig. A4. Mean number of actual avoidance
responses (error bars represent standard error of the mean).
Fig. A5. Mean US expectancy ratings (error
