• Peer victimization;
  • Social anxiety;
  • Psychophysiology;
  • Associative information network;
  • Vulnerability


  1. Top of page
  2. Abstract
  3. Method
  4. Results
  5. Discussion
  6. References

The study investigated the role of an associative information network as a mechanism underlying the relation of peer victimization and social anxiety disorder (SAD). A sample of N = 80 was divided according to diagnosis (SAD vs. no diagnosis) and amount of peer victimization (low vs. high). Responses to memory of a personally experienced aversive social situation and to imagining a standardized negative social situation were assessed. In terms of skin conductance level, subjects with SAD and peer victimization were more reactive to the memory script than the other three groups while responses to the standardized script did not vary. As to heart rate, there were no differences between the groups. Emotional responses presented with an inconsistent pattern. The results provide a first indication that associative memory structures resulting from aversive social experiences might play a role in the development and maintenance of SAD, but further research is needed.

Social anxiety disorder (SAD) is characterized by persistent fear and avoidance of social situations. It is one of the most frequent psychological disorders with a typical onset at late childhood or adolescence that is accompanied by severe functional impairment (Stein & Kean, 2000; Wittchen, Stein, & Kessler, 1999). SAD results from a combination of genetic vulnerability (Kendler, Neale, Kessler, Heath, & Eaves, 1992) and a detrimental learning history including dysfunctional parenting practices (Harvey, Ehlers, & Clark, 2005; Lieb et al., 2000) as well as stressful social experiences (Kuo, Goldin, Werner, Heimberg, & Gross, 2011; Simon et al., 2009). In particular, SAD has been associated with peer victimization (e.g., Harvey et al., 2005; Kingery, Erdley, Marshall, Whitaker, & Reuter, 2010; Siegel, La Greca, & Harrison, 2009). Peer victimization includes direct or overt aggression (e.g., physical, verbal attacks) and indirect forms such as relational victimization (e.g., ignoring, rejection, exclusion) and reputational victimization (attempting to damage the reputation of a peer, e.g., by spreading rumors). Peer victimization can bring about maladjustment such as loneliness, low self-worth, depression, and anxiety disorders (for a review, see Hawker & Boulton, 2000). Relational peer victimization is associated with subclinical social anxiety (La Greca & Harrison, 2005; Siegel et al., 2009) as well as SAD (Gren-Landell, Aho, Andersson, & Svedin, 2011; Marteinsdottir, Svensson, Svedberg, Anderberg, & von Knorring, 2007). Regarding the causality of this association, longitudinal observations showed that experiences of relational peer victimization are both a predictor and a consequence of social anxiety (La Greca & Harrison, 2005; Siegel et al., 2009).

There is reason to assume that memories of peer victimization remain extraordinarily vivid in subjects with SAD and may thus contribute to the maintenance of the disorder. For example, subjects with SAD report recurrent intrusive images of stressful social situations (Hackmann, Clark, & McManus, 2000; Hackmann, Surawy, & Clark, 1998). These images are not restricted to visual perceptions, but can include body sensations and acoustic perceptions and are most commonly related to memories of adverse events that preceded the onset of SAD. As a consequence of these findings, social traumatic memories have recently been identified as a potential target for treatment (Nilsson, Lundh, & Viborg, 2012; Wild & Clark, 2011). Changing memory representations using rescripting techniques seems to reduce intrusive images as well as symptoms of SAD (Nilsson et al., 2012).

The recent emphasis on memory processes in SAD points towards the usefulness of a trauma perspective, as has been established for the study of posttraumatic stress disorder (PTSD). Current theories of traumatic memories (Brewin, Dalgleish, & Joseph, 1996; Foa & Kozak, 1986) refer to Lang's (1979) bioinformational theory to conceptualize representations of traumatic memories. In this view, a traumatic event is encoded in the form of an associative network that connects stimulus representations about the situation, cognitive appraisals of the stimulus as well as response representations including emotional and physiological reactions. The activation of the network by internal or external cues causes the coactivation of other stimulus and response items, leading to the intrusive recurrence of the original physiological and emotional responses in the form of intrusive images.

Although relational peer victimization does not imply a threat to life and limb and therefore does not qualify as a potentially traumatic event as defined in the current PTSD criteria, relational victimizations may still generate intense associative memory networks as they evoke intense emotional reactions such as feelings of helplessness and fear (Storch & Esposito, 2003). The immediate consequences of peer victimization include intrusive images, similar to PTSD symptoms (Carney, 2008; Mynard, Joseph, & Alexander, 2000). Furthermore, intense physiological alarm responses are not restricted to physical threat but can also be found for threats to social belongingness (MacDonald & Leary, 2005). Relational peer victimization during sensitive periods in childhood and adolescence may be a prototype of such events that elicit social pain, shame, and physiological stress. Repeated intense experiences can then lead to a pathological traumatic memory representation in the form of an associative network. Figure 1 shows a schematic description of an exemplary representation of an associative information network adapted for SAD.


Figure 1. Schematic description of the associative information network adapted for social anxiety disorder, modified according to Neuner, Schauer, & Elbert (2009).

Download figure to PowerPoint

In order to investigate the associative information network, we used the script-driven imagery paradigm (Lang, Levin, Miller, & Kozak, 1983). In this method, an imagery instruction is used to activate emotional associative networks leading to verbal, behavioral, and somatovisceral responses, which can be measured using self-reports and physiological assessments. This method has been used as a symptom provocation test for the study of various anxiety disorders (Cuthbert et al., 2003; Lang & McTeague, 2009; Pitman, Orr, Forgue, de Jong, & Clairborn, 1987). In these experiments, subjects were instructed to imagine personalized as well as standardized scripts of typical feared situations while physiological reactions were assessed. McTeague et al. (2009) found patients with SAD showing higher self-reported fear and physiological arousal during imagery of negative social situations; however, other studies did not find clear effects of social anxiety (e.g., Lang et al., 1983).

The aim of this study was to elucidate the character of traumatic social memory representations in subjects with SAD. In particular, we assumed that traumatic memory processes play a role in a subgroup of subjects with SAD who report high levels of peer victimization. Using the script-driven imagery paradigm, we aimed at determining the specific characteristics of these memories for this subgroup using physiological as well as self-report measures. To isolate the effects of traumatic memories of past events from the effects of imagery of aversive social situations per se, we compared the reactions to the reliving of a personal social negative memory with the imagination of a standardized social negative script. We hypothesized that, in a comparison of four groups (SAD with low and high levels of peer victimization, healthy subjects with low and high levels of peer victimization), we would be able to identify the subgroup of SAD with peer victimization on the basis of their reactions to the script imagery, with a higher contrast between memory script and standardized script than the other groups. In order to examine if heightened physiological and emotional reactivity is actually due to traumatic social memory representation and not merely ascribable to the negative valence of stimuli, personalized and standardized nonsocial negative scripts were applied in addition to the social negative scripts.


  1. Top of page
  2. Abstract
  3. Method
  4. Results
  5. Discussion
  6. References


Eighty adults were recruited from the Bielefeld University and its associated psychotherapy outpatient clinic. The sample was divided into four subsamples according to diagnosis (SAD vs. no diagnosis) and level of peer victimization (low vs. high). Groups of low and high extent of peer victimization were determined based on the median split of an instrument of relational peer victimization (see below). Participants were included in the SAD group if they met DSM-IV criteria for social phobia. They were classified as having no diagnosis if they did not meet criteria of any DSM-IV diagnosis in the last 2 years and did not have a history of affective disorders. Exclusion criteria for the whole sample included substance dependence, actual or prior psychosis, and acute suicidality. Table 1 summarizes demographics, diagnoses, maltreatment exposure, and psychometrics separated by group. In the group of subjects with SAD and a history of peer victimization, 4% had a part-time job, 87% were pupils or students, and 9% were unemployed. In the group of subjects with SAD but without peer victimization, 94% were pupils or students, and 6% were unemployed. Six percent of the healthy subjects with peer victimization had a part-time job, and 94% were students or pupils, whereas in the group of healthy controls without peer victimization, 100% were pupils or students.

Table 1. Demographic Data, Diagnoses, and Psychometric Data, by Groups
VariableSAD and peer victimization (n = 23)SAD only (n = 16)Peer victimization only (n = 17)Control group (n = 24)Analysis Overall F
  1. Note. Values with indices were analyzed for differences between the groups: levels not connected by the same letter were significantly different in a pairwise comparison (Tukey test after analysis of variance, p < .05). SAD = social anxiety disorder; FBS = questionnaire of aversive social experiences in the peer group; CTQ = Childhood Trauma Questionnaire; SPS = Social Phobia Scale; SIAS = Social Interaction Anxiety Scale; BSI GSI = General Severity Index of Brief Symptom Inventory; BDI = Beck Depression Inventory.

  2. a

    χ2 test.

  3. ***p < .001.

Age, M (SD)25.35 (4.99)25.00 (4.24)24.65 (4.55)22.83 (4.07)1.42
Sex, % female709471926.68a
Relationship status (%)    8.60a
Education level (%)    1.99a
Less than high school9666 
High school83887679 
College degree961817 
Current comorbid diagnoses (%)     
Affective disorder74a63a0b0b3.39***a
Anxiety disorder80005.08a
Other disorders46002.30a
FBS, M (SD)22.52 (5.30)a9.13 (2.60)b,d18.53 (4.17)c6.75 (3.21)d9.93***
CTQ, M (SD)44.96 (12.69)a46.13 (12.35)a36.18 (9.57)b32.58 (5.26)b8.70***
Emotional maltreatment11.78 (5.27)a12.13 (4.90)a,b8.53 (3.22)a,c6.5 (1.13)c,d9.98***
Emotional neglect13.74 (4.21)a13.06 (4.84)a9.0 (3.43)b8.63 (3.02)b9.87***
Physical maltreatment7.04 (3.20)6.50 (2.34)6.18 (2.38)5.38 (0.82)2.08
Sexual maltreatment5.43 (1.20)6.19 (2.43)5.29 (0.59)5.42 (1.18)1.34
SPS, M (SD)35.78 (14.88)a32.38 (13.93)a16.47 (10.04)b9.29 (5.54)b6.09***
SIAS, M (SD)47.57 (12.38)a45.38 (13.40)a23.59 (8.88)b17.21 (10.57)b7.93***
BDI, M (SD)16.52 (9.95)a11.81 (8.57)a,b11.47 (4.32)b5.21 (3.31)c9.93***
BSI GSI, M (SD)1.19 (0.67)a0.84 (0.47)a0.74 (0.34)a0.35 (0.18)b3.60***


Clinical interview

All subjects underwent structured diagnostic interviews by MSc-level clinicians using the structured Mini International Neuropsychiatric Interview (MINI; Lecrubier et al., 1997), which allows for the determination of diagnoses based on the DSM-IV (APA, 2000).

Self-report measures

Social anxiety was measured using the Social Phobia Scale and Social Interaction Anxiety Scale (SPS and SIAS, Mattick & Clarke, 1998, German version by Stangier, Heidenreich, Berardi, Golbs, & Hoyer, 1999). These are short self-report measures consisting of 20 items each. The SPS assesses scrutiny fears while the SIAS measures fears in general social interactions. Each item is scored on a 5-point scale, indicating how distinctive the statement is for the respondent (0 = not at all to 4 = extremely) with some items reverse scored. Previous research has yielded support for the reliability and adequate convergent validity of the SPS and SIAS German version, while stating insufficient discriminant validity (Heinrichs et al., 2002; Stangier et al., 1999).

The Childhood Trauma Questionnaire (CTQ, Bernstein & Fink, 1998) is a well-validated 28-item self-report questionnaire consisting of five subscales (sexual abuse, physical abuse, emotional abuse, emotional neglect, and physical neglect). The items are rated from 1 (never true) to 5 (very often true) with some items reverse scored. Studies investigating the psychometric properties of the German version indicate a sufficient to excellent internal consistency concerning most of the subscales (Wingenfeld et al., 2010). Because of the deficient internal consistence of the subscale physical neglect and its high intercorrelations with the other subscales (Klinitzke, Romppel, Häuser, Brähler, & Glaesmer, 2012), this subscale was not considered in our analysis.

Peer relational victimization was assessed with the Questionnaire of Aversive Social Experiences in the Peer Group (Fragebogen zu belastenden Sozialerfahrungen in der Peer Group, FBS, Sansen, Iffland, Catani, & Neuner, 2013). As there was a lack of retrospective measures of peer victimization in the German language, this event list was constructed in order to assess stressful social experiences encountered in peer groups. The construction of the questionnaire was based on conceptual considerations. We generated 22 items describing aversive social situations such as rejection, exclusion, being laughed at, insulted, and teased by peers (e.g., “In class nobody wanted to sit next to me,” “Other children or adolescents banned me from their games or activities”). For each situation, subjects should report if the experience happened during childhood (6–12 years), during adolescence (13–18 years), or if they have never experienced it. The items are summed for a Childhood Scale (0–22), an Adolescence Scale (0–22), and a total score (0–44), which was used for this study. The results of a first evaluation study (Sansen et al., 2013) indicate satisfying psychometric properties. The questionnaire presented with good stability over a 20-month period. Construct validity could be confirmed through correlations between the FBS and psychological distress as well as social anxiety. Subjects with high levels of social anxiety had significantly higher FBS scores compared to subjects with low levels of social anxiety, which indicated discriminative validity. The FBS was previously applied in several studies examining the role of peer victimization in terms of psychopathology that argue for a good fitness of the instrument (e.g., Iffland, Sansen, Catani, & Neuner, 2012; Sansen, Iffland, & Neuner, 2014). In the present study, internal consistency of the FBS was high with Cronbach's α = .87 for the Childhood Scale, α = .86 for the Adolescence Scale, and α = .89 for the sum score.

The German version of the Beck Depression Inventory (BDI-II, Hautzinger, Keller, & Kühner, 2006) was used for measuring the severity of self-reported depression during the previous 2 weeks. The questionnaire shows good psychometric properties in clinical and nonclinical samples (Kühner, Bürger, Keller, & Hautzinger, 2007).

The Brief Symptom Inventory (BSI, German version by Franke, 2000) is a 53-item self-report questionnaire measuring psychopathology throughout the previous 7 days. The BSI is the short version of the Symptom Checklist 90 (SCL-90, Derogatis, 1977). The items are answered on a 5-point scale from 0 (not at all) to 4 (extremely). Nine symptom dimensions are assessed: somatization, obsession-compulsion, interpersonal sensitivity, depression, anxiety, hostility, phobic anxiety, paranoid ideation, and psychoticism. In addition, different global indices of distress can be calculated, for example, the Global Severity Index (GSI). The mean of the GSI in a German norm sample is M = 0.31 (SD = .02) (Franke, 2000). Reliability and validity of the German version can be judged as satisfactory (Franke, 2000; Geisheim et al., 2002).

For the purpose of measuring subjective responses to the script-driven imagery task, we asked the subjects to specify the intensity of fear and shame and to rate their imagery ability (“How easily could you put yourself in the situation?”) and familiarity of the described situation (“How familiar was the situation?”) each on a 7-point Likert scale.

Physiological assessment

Skin conductance level (SCL) and heart rate (HR) were registered and digitized by a Varioport biosignal recording device (Becker Meditec, Karlsruhe, Germany), that was controlled by a Windows computer with Variograph software (Becker Meditec). SCL was obtained through 9-mm electrodes filled with isotonic paste placed on the nondominant thenar and hypothenar surface. HR signals were obtained from three disposable Ag-AgCl electrodes placed on the manubrium sterni, the lowest part of the sternum, and the lowest left rib. Physiological signals were simultaneously recorded with a sampling rate of 512 Hz.


The ethics committee of the psychology department of Bielefeld University approved the study. Participants provided informed consent at the outset of the study. The investigation consisted of two parts that were conducted on separate days. At first assessment, clinical diagnoses were made by a clinician on the basis of the MINI (Lecrubier et al., 1997). Before completing the self-report questionnaires individually, subjects were asked to portray different personal experiences: a negative social experience in which they felt rejected, denied, humiliated, or embarrassed, a neutral social situation, a negative nonsocial situation (e.g., confrontation with a feared animal or situation, such as an accident) and a neutral nonsocial situation, including meaning (feelings and interpretations) and response information (bodily reactions, behavior). For each of the personal experiences subjects were requested to rate the subjective stress level caused by the event on a 7-point Likert scale (subjective severity rating). According to the procedure illustrated by Pitman et al. (1987), short vignettes were derived from these individual descriptions, which were standardized regarding number of words. The scripts were recorded in a neutral male voice and edited as 30-s sound files. Subsequent analyses regarding external ratings of the social scripts made by N = 62 students (51% male, age M = 27.95, SD = 8.42) did not detect any differences between the four groups in terms of estimated arousal, fear, and shame. Regarding the estimated stress level, we obtained an interaction effect of diagnosis with valence, F(1,72) = 5.396, p < .05, with comparable high stress ratings referring to the negative script and low stress ratings in terms of the neutral script in which the neutral scripts of the healthy groups were rated slightly more stressful than the neutral scripts of the patients with SAD. In addition to the individual scripts, we used 30-s standardized scripts of an aversive social situation, a neutral social situation, an aversive nonsocial situation, and a neutral nonsocial situation.

The laboratory procedure took place 1 week later. Subjects were seated individually in a comfortable chair in the laboratory. Following the attachment of electrodes, instructions for the imagery tasks were given to the subjects. They received brief relaxation instructions and were asked to refrain from moving during data collection. Subjects were told to listen carefully and vividly imagine each of the presented scripts. Personal and standardized scripts were played back via headphones in a randomized order. Each script consisted of four sequential 30-s periods (baseline, reading, imagery, and recovery) that were separated by a tone. Immediately after the recovery period of each script, subjects rated the extent of fear and shame they had experienced as well as imagery ability and familiarity of the script on the associated questionnaire. HR and SCL were continuously recorded during the whole procedure. After the experiment, subjects were debriefed about the actual aims of the study.

Data Reduction and Analysis

Psychophysiological data were preprocessed and analyzed using MATLAB version 7.7 (The MathWorks, Natick, MA) with the toolboxes ANSLAB (Wilhelm & Peyk, 2006) and Ledalab (available at R waves in the ECG data were identified automatically by ANSLAB software and converted into beats per minute (bpm). Additionally, HR data were inspected visually for artifacts. Artifactual data points were replaced manually, nonrecognized R waves were edited, and sections with high proportions of artifacts were not evaluated. Signals of SCL were converted to microsiemens (μS). Raw skin conductance data were also screened for implausible artifacts that were then manually edited.

Statistical Analyses

All statistical analyses were performed using the Statistical Package for Social Sciences SPSS 20.0. Comparisons between the groups in terms of demographic and psychometric data were made using analyses of variance (ANOVA) with Tukey's HSD post hoc test while gender distribution, education level, and relationship status were compared using χ2 test. Spearman correlations were used to examine the relationship between peer victimization and social anxiety, depression, and symptom distress. For the main analyses, both HR und SCL were averaged for the baseline period and the imagery period of the different scripts. Following the procedure of Pitman et al. (1987), difference scores for each script were computed (imagery period minus baseline period), which were considered in the further analyses. A full factorial repeated measures ANOVA was conducted for psychophysiological responses and subjective ratings of emotions. The following factors were included in the ANOVA: between-group factor peer victimization (under vs. over FBS median = 13), between-group factor diagnosis (SAD vs. no diagnosis), and within-subject factors script type (memory of personal script vs. imagination of standardized script), script context (social situation vs. nonsocial situation), and script valence (negative vs. neutral). In case of a significant main or interaction effect that included diagnosis or peer victimization, two groups were subsequently compared using post hoc t tests while in case of an interaction effect involving diagnosis and peer victimization, the four groups were compared using post hoc comparisons and planned contrasts (subgroups with SAD and peer victimization vs. the other three groups). To rule out the possibility that differences between the groups and scripts were just due to differences in imagery ability, familiarity of script, or depressive symptomatology, these variables were entered as covariates in all planned contrasts and post hoc analyses. As the subjective severity rating was not available for the standardized scripts, it was only used as covariate in group comparisons referring to the personal scripts to ensure that group differences cannot only be ascribed to severity of the memorized situations.


  1. Top of page
  2. Abstract
  3. Method
  4. Results
  5. Discussion
  6. References

Descriptive Analysis

Means and standard deviations for age, peer victimization (FBS), maltreatment in the family (CTQ), social anxiety (SPS and SIAS), depression (BDI), psychological symptom distress (BSI), and contribution of relationship status, education level, and current diagnoses are presented in Table 1. As shown, the four groups were comparable in terms of their mean age and gender distribution but differed concerning the extent of emotional maltreatment, emotional neglect, social anxiety, depression, and global symptom distress. With the exception of FBS scores, the SAD subgroups with low and high levels of peer victimization were comparable in terms of psychometric measures. Referring to the healthy groups, not only the FBS sum score differed but also the BDI sum score and the BSI GSI, with significantly higher scores in the group of healthy subjects with high levels of peer victimization in contrast to the control group of healthy subjects with low levels of peer victimization.

Preliminary Analysis

We found positive correlations across the four groups between peer victimization (FBS total score) and social anxiety (SPS: r = .43, p < .001; SIAS: r = .39, p < .001), depression (BDI: r = .42, p < .001) and symptom distress (BSI GSI: r = .49, p < .001).

Results of the Full Factorial Repeated Measures ANOVAs

The ANOVAs revealed numerous significant main and interaction effects of the between-group factors diagnosis and peer victimization and the within-subject factors script type, script context, and script valence. As the question of this work applied to characteristics of memory representations in subjects with SAD and experiences of peer victimization, only the main effects and interaction effects referring to these groups will be presented below.

Psychophysiological responses

Due to an equipment malfunction, no SCL data could be obtained in the case of three subjects, and no HR data could be obtained in the case of four subjects.

Skin conductance level

Referring to SCL, we found a main effect of script type, F(1,72) = 9.254, p < .01, with higher SCL scores during memory of the personal situations than during imagination of the standardized scripts and a main effect of valence, F(1,72) = 10.172, p < .01, with a higher reactivity during imagery of the negative scripts than during imagery of the neutral scripts. Moreover, we found a three-way interaction of peer victimization, valence, and script context, F(1,72) = 6.040, p < .05, and a five-way interaction of diagnosis, peer victimization, script type, valence, and script context, F(1,72) = 5.931, p < .05. Based on the finding of the significant five-way interaction, SCL responses to the scripts were examined in more detail as a function of diagnosis and peer victimization. To reduce the number of factors and thus simplify the model, difference scores between the imagery period of the negative and the imagery period of the neutral script were calculated for the standardized and personalized scripts of social and nonsocial situations. Afterwards, a planned contrast was performed to test the hypothesis that subjects with SAD and a history of peer victimization show greater physiological activation during imagery of the personal experience than the other three groups. As Figure 2 depicts, the planned contrast analysis revealed a higher increase in SCL during memory of the aversive social experience in subjects with SAD and a history of peer victimization in contrast to the other three groups, F(1,73) = 5.15, p < .05. The control of the covariates subjective severity rating, F(1,72) = 2.413, p < .05, imagery ability, F(1,72) = 4.97, p < .05, and familiarity, F(1,72) = 4.86, p < .05, did not change this result. However, after inclusion of the covariate BDI sum score, the difference failed to reach significance, F(1,72) = 5.15, p = .09. Within the standardized script, planned contrasts did not show a significant difference between the groups, t(72) = 1.18, p = .24. Besides the between-subject comparisons, we also conducted within-subject comparisons to examine whether SCL responses to the personal memory script differ from SCL responses to the standardized script in each of the four groups. Repeated measures analyses revealed that none of the groups showed significantly different SCL reactions to the personal versus the standardized script.


Figure 2. Means of skin conductance level (difference score negative script minus neutral script) during memory of a personal aversive social situation and during imagination of a standardized aversive social situation, by groups.

Download figure to PowerPoint

Heart rate

With respect to HR, the repeated measures ANOVA did neither reveal a significant main effect nor an interaction effect of diagnosis, peer victimization, script type, valence, or script context.

Emotional responses during script-driven imagery (self-report)

The results of the ANOVAs showed a main effect of script context, with higher fear ratings in terms of the nonsocial script than the social script, F(1,76) = 23.53, p < .001, and a main effect of valence with higher fear ratings after imagery of the negative scripts than the neutral scripts, F(1,76) = 354.12, p < .001. Moreover, we found an interaction effect of script context with diagnosis, F(1,76) = 5.86, p < .05 (see Figure 3). Post hoc t tests revealed that fear ratings of the nonsocial scripts did not differ between patients with SAD and healthy subjects, t(78) = 0.32, p = .75, while subjects with SAD reported more fear during imagery of the social scripts than the healthy groups, t(78) = −2.47, p < .05. The consideration of the covariates imagery ability, F(1,77) = 4.13, p < .05, and familiarity of the script, F(1,77) = 5.01, p < .05, did not change this result while the difference in terms of the social script was no longer significant when taking the BDI sum score into account, F(1,76) = 2.34, p = .13.


Figure 3. Means of self-reported levels of fear during imagery of the social situations and the nonsocial situations (independent of script context and valence), by diagnosis.

Download figure to PowerPoint


The ANOVA revealed a main effect of script context, F(1,76) = 45.42, p < .001, with higher shame ratings referring to the social scripts in contrast to the nonsocial scripts and a main effect of valence, F(1,76) = 96.74, p < .001, with higher ratings in terms of the negative than the neutral scripts. We found an interaction of peer victimization with script valence, F(1,76) = 4.98, p < .05 (see Figure 4). Post hoc t test showed no difference between the groups in terms of the neutral scripts, t(78) = −0.41, p =.68, whereas subjects with high levels of peer victimization reported more shame than subjects with a low level of peer victimization referring to the negative scripts, t(78) = −2.49, p < .05. Taking into account the covariates imagery ability, F(1,77) = 7.25, p < .05, and familiarity of the script, F(1,77) = 4.47, p < .05, did not change the result while the groups with low and high level of peer victimization did not differ any longer under consideration of the BDI sum score, F(1,76) = 1.66, p = .20. Furthermore, we found an interaction of peer victimization with script type, F(1,76) = 7.84, p < .01 (see Figure 5). Post hoc t test showed that subjects with a high level of peer victimization reported more shame than subjects with a low level of peer victimization in terms of personal memory scripts, t(78) = −3.32, p < .01, while responses to the standardized scripts did not differ, t(78) = −.57, p = .57. The consideration of the covariates imagery ability, F(1,77) = 14.72, p < .001, familiarity of the script, F(1,77) = 10.71, p < .01, subjective severity rating, F(1,77) = 9.45, p < .01, and BDI sum score, F(1,76) = 4.21, p < .05, did not change the results.


Figure 4. Means of self-reported levels of shame during imagery of the neutral situations and the negative situations (independent of script type and script context), by extent of peer victimization.

Download figure to PowerPoint


Figure 5. Means of self-reported levels of shame during memory of personal situations and during imagination of standardized situations (independent of script context and valence), by extent of peer victimization.

Download figure to PowerPoint


  1. Top of page
  2. Abstract
  3. Method
  4. Results
  5. Discussion
  6. References

The aim of this study was to investigate psychophysiological and emotional characteristics of memory imagery in subjects with SAD. Based on the assumption of an associative information network underlying the relationship of peer victimization and SAD, we expected subjects with SAD and a history of peer victimization to show a heightened reactivity during memory of a formerly experienced aversive social event. In order to examine this hypothesis, we compared psychophysiological and emotional responses of healthy subjects and subjects with SAD each with low and high levels of peer victimization with regard to imagery of different situations using a script-driven imagery paradigm. For one thing, our findings confirmed the relationship between peer victimization and SAD; for another thing, we found that SCL responses during memory of aversive social events depend on diagnosis of SAD as well as on a history of peer victimization while no differences emerged during imagination of standardized aversive social events and imagery of aversive nonsocial situations. In terms of HR, we did not obtain significant differences as a function of SAD or peer victimization. The findings of self-reported emotions were inconsistent indicating that fear ratings depend on a diagnosis of SAD while shame ratings differ as a function of former peer victimization experiences.

The correlation of peer victimization with social anxiety is consistent with our expectation and findings of prior studies (Gren-Landell, et al., 2011; La Greca & Harrison, 2005; Siegel et al., 2009). Beyond that, we found a significant relationship with depressive symptoms and general symptom distress, which also is in line with several findings (Benjet, Thompson, & Gotlib, 2010; Dempsey & Storch, 2008; Prinstein, Boergers, & Vernberg, 2001) indicating that the association of peer victimization and negative psychological consequences is rather nonspecific.

As predicted, subjects with both a history of peer victimization and a diagnosis of SAD could be identified on the basis of their SCL response. During memory of a personal traumatic social event, they presented with the highest SCL reactions in comparison to groups with either SAD or a history of victimization and a healthy control group. This effect did not occur in response to the imagination of a standardized script about a negative social event. This result corresponds with previous findings of heightened skin conductance reactions to imagery of personal fear imagery as observed by McTeague et al. (2009). However, subjects with SAD and a history of peer victimization showed comparably high SCL responses to a personal and to a standardized aversive social script while subjects with either SAD or peer victimization and healthy controls apparently showed a higher SCL reactivity during imagination of a standardized aversive social situation compared to SCL responses during memory of a personal aversive social event. Even though this comparison did not reach significance in any of the groups, it could indicate that in subjects with SAD and peer victimization both memories of aversive social events that are stored in form of an associative information network and present potentially threatening social situations lead to an activation and heightened SCL responses. In contrast, in subjects with either SAD or peer victimization and healthy persons, memories of aversive social events seem to be adequately elaborated and stored in autobiographical memory; therefore, these memories are supposed to be of minor importance for the present experiencing and only elicit slight arousal while the imagination of a present potentially threatening social situation is more meaningful involving higher SCL responses. Our results lend support to the notion that the differential reactivity of the groups during the script-driven imagery task was not determined by better imagery ability or familiarity of the imagined situation or higher severity of the portrayed events but seems to be mediated by the script itself and can be interpreted as a result of processing of the associative information network. However, heightened SCL reactivity of the SAD subgroup with peer victimization could also be due to more severe maltreatment experiences or psychopathological cognitive processes in consequence of depression. A comparison of the SAD subgroups with low and high levels of peer victimization showed neither differences in scores of emotional, physical, and sexual maltreatment nor in frequency of diagnosis of depression and depression score, but an exploratory analysis showed that the SCL result was no longer significant after inclusion of the depression score. Higher reactivity could therefore also attribute to depression. Concerning the high comorbidity rate between SAD and depression (Kessler, Stang, Wittchen, Stein, & Walters, 1999) the separability of the disorders seems problematic, whereas this confounding is insoluble using statistical methods (Miller & Chapman, 2001). As the SAD subgroups were comparable in terms of depression, the ascription of heightened reactivity to the associative information network based on peer victimization seems more likely. Based on our findings, it may be important to consider subgroups of SAD with and without social traumatic experiences. Biologically discriminable subgroups depending on early aversive life events have been suggested for major depressive disorder before (Heim, Newport, Mletzko, Miller, & Nemeroff, 2008; Heim, Plotsky, & Nemeroff, 2004).

While the findings of SCL were in line with our previous assumptions, we did not find differences depending on SAD or peer victimization in terms of HR. This result is in contradiction to other studies that found heightened heart rate reactivity during imagery of anxiety provoking social scenes in subjects with SAD (Cuthbert et al., 2003; McTeague et al., 2009). We assume that HR might be less sensitive than SCL and therefore does not sufficiently discriminate between responses of the four groups to the personal versed standardized scripts, especially because the subsamples were relatively small. However, much additional work is required to reach a better understanding of the diverging psychophysiological results.

The findings referring to self-reported fear and shame showed a heterogeneous pattern. While fear responses to the nonsocial scripts were rated comparably high of the four groups, fear responses to the social scripts differed as a function of diagnosis indicating more fear in subjects with SAD during confrontation with social scenarios. This finding is in good agreement with a core feature of SAD diagnostic criteria, that is to say, “a marked and persistent fear of one or more social or performance situations in which the person is exposed to unfamiliar people or to possible scrutiny by others” (DSM-IV-TR, APA, 2000). However, our results suggest that subjects with SAD not only experience fear in social situations involving performance requirements, unfamiliar people, or others' critical appraisal but when confronted with social situations in general. This could reflect the tendency to interpret neutral or ambiguous social situations as negative, which has often been found in the context of SAD and is supposed to play a role in its maintenance (e.g., Amin, Foa, & Coles, 1998; Stopa & Clark, 2000). Referring to shame, our results suggest a tendency to experience more shame when confronted with negative events or personal memories in subjects who repeatedly became victims of peer victimization in contrast to nonvictimized ones. Taking into account that shame is considered to be an interpersonal emotion emerging in situations where the entire self is negatively evaluated (Tangney, Wagner, & Gramzow, 1992), such as peer victimization, this result could be a sign of an outlasting effect of peer victimization on the occurrence and experiences of shame. The assumption of a long-term impact of early adverse events on later experiencing of shame is in line with the findings of Wright, Crawford, and Del Castillo (2009) who report a relationship between childhood emotional maltreatment with later symptoms of anxiety and depression that they found to be mediated by specific internalized maladaptive schemas including shame. In addition, Matos and Pinto-Gouveia (2009) showed that memory of early shame experiences presented with traumatic memory characteristics and assume these memory representations to play a role in development of psychopathology. These findings not only elucidate that shame plays a meaningful role regarding the relationship between adverse social experiences and psychopathology but also indicate that shame seems to be stored in an associative memory structure. However, our data suggest that this specific shame network seems to exist in subjects with peer victimization experiences independent of diagnosis.

The lack of convergence between the psychophysiological and affective responses initially seems contradictory. However on the one hand, divergence can be ascribed to the relative independence of emotional and physiological components (Lang et al., 1983; Lang, Greenwald, Bradley, & Hamm, 1993) with the assumption that emotional self-reports are much more affected by social desirability, lack of awareness of emotional states, or automatic control or inhibition processes than psychophysiological data. On the other hand, lack of consistent results might also be attributed to several methodological facts. The first one refers to sample size and the fact that most group differences barely reached level of significance. Although a considerable number of subjects was studied in this experiment, it may still have been too small to compensate for the subdivision into four groups. Secondly, inconsistent results might be based on the specific composition of our subsamples. As a high percentage of the SAD subgroups with low and high level of peer victimization consisted of students, it may be assumed that the present subjects display a relatively high level of functioning and do not represent cases of severely impaired patients. Although the SAD subgroups and the healthy subgroups differed in terms of different measures, the contrast might not have been sufficiently high enough to influence HR and self-reported emotions. A further problem relating to subgroup composition concerns the group of healthy subjects with peer victimization. These subjects, considered to represent resilient persons, indeed did not meet criteria of mental disorders, but comparison of the healthy subgroups revealed that the resilient subgroup displayed higher depression scores and higher psychological symptom distress. Concerning stress and impairment, the resilient subgroup seems to be located between the SAD groups and the healthy control group, taking into account that the insufficient discrimination of the groups might state a further reason for our ambiguous results.

Although the results of the present investigation are somewhat less clear than we expected, the study reveals novel and interesting findings. To our knowledge, this is the first study to investigate psychophysiological and emotional responses to memory of a personal aversive social experience in contrast to a standardized aversive social event as a function of SAD and peer victimization. Particularly, heightened SCL reactivity of subjects with SAD and peer victimization during memory of the personal aversive social experience points to an altered memory representation of aversive social events in these subjects in the form of an associative information network that might play a role in development and maintenance of SAD. But even the finding that experiencing of shame seems to depend on the extent of former peer victimization experiences is in good agreement with the assumption that early adverse experiences can lead to development of associative memory structures determining future experiencing and behavior. Based on our study, one can conclude that, similar to development of PTSD after a traumatic event (Orr, 1994), one subtype of SAD seems to originate from repeated aversive social experiences that progressively lead to an association of several social stimuli with negative affective, physiological, and bodily responses in memory (Cuthbert et al., 2003; Lissek et al., 2008). The findings of the present investigation lend support to the assumption that activation of these memory representations can lead to network processing, cue emotional memory, and elicit corresponding negative responses in subjects with SAD and a history of social trauma. This might be one reason why patients with SAD experience painful reactions in social situations even if the situation does not really imply social threat.

However, some limitations are worth mentioning. Several methodological problems have been presented before. A further limitation applies to the cross-sectional design of the study, meaning that causal conclusions about the relationship of peer victimization and SAD cannot be drawn. Beyond that, the retrospective assessment of peer victimization might be prone to influences of memory deficits or biases caused by psychopathology, mood, or age. Although retrospective recall biases cannot be eliminated completely, they are not thought to invalidate retrospective studies (Hardt & Rutter, 2004).

In spite of these limitations, our results are encouraging, and it would be beneficial to further investigate the role of adverse social experiences for development and maintenance of SAD as well as the underlying mechanisms. Our results should be replicated in a larger sample, which would allow studying the effects of potential correlates and confounders such as type of SAD (generalized vs. circumscribed) or different maltreatment experiences (peer victimization, emotional, physical, or sexual maltreatment, neglect). A further important question for future work is to observe the formation of an associative information network in the context of a developmental longitudinal study. It would be useful to focus on different objective and subjective measures as well as on different subgroups of subjects. As our knowledge about resilient subjects is scarce, further investigation of this subgroup, their symptomatology, potential resilience, protective, and risk factors is especially required to determine factors that could prevent development of associative memory structures and psychopathology.

Our findings provide relevant indications for researchers and clinicians dealing with SAD. Because memory representations of aversive social events seem to play a role in development and maintenance of SAD, their modification should be an important target of treatment. Therefore, it seems important to consider and further investigate memory processes in subjects with SAD in order to generate more appropriate treatments and preventions of SAD.


  1. Top of page
  2. Abstract
  3. Method
  4. Results
  5. Discussion
  6. References