"صدای ترس": ارزیابی فرکانس صوتی به عنوان یک شاخص فیزیولوژیک اختلال اضطراب اجتماعی

Abstract The relationship between vocal pitch and social anxiety disorder (SAD) has been examined with encouraging initial results, highlighting increased fundamental frequency (F0) as a physiological indicator of SAD. The present series of studies examined the relationship between F0 emitted during social threat and SAD symptoms. Two independent samples of SAD patients, and a sample of demographically-equivalent non-socially anxious controls (NSACs), completed varying social threat tasks which involved speech. Mean F0 emitted throughout the tasks was examined. Male SAD patients emitted greater F0 in comparison to NSACs across studies. For females, this relationship was significant only when examined in patients with SAD of the generalized subtype, and in response to in vivo social exposures. Furthermore, gender-specific thresholds for overall F0 emitted during social threat were identified which demonstrated excellent differentiation between patients with generalized SAD and NSACs. These results provide additional support for increased F0 as a physiological indicator of SAD.

Introduction Social anxiety disorder (SAD; i.e., social phobia) is characterized by excessive fear of social or performance situations, and is the fourth most common mental disorder, with a lifetime prevalence rate of 12.1% (Kessler et al., 2005). The majority of patients seeking treatment for SAD report moderate-to-severe impairment across various life domains such as education, employment, family relationships, marriage/romantic relationships, and friendships (Schneier et al., 1994 and Stein et al., 1999). Despite the high prevalence of SAD, it is noteworthy that less than 1% of a large sample of research participants who were diagnosed with generalized SAD had this diagnosis recorded in their medical records in the year prior to being surveyed (Katzelnick et al., 2001) – a sobering reminder of the need to better refine the ability to objectively identify, and subsequently treat, patients with SAD. To this end, efforts to enhance the assessment of SAD, with regard to accuracy and efficiency, remain pivotal. Vocal acoustic analyses are a novel assessment approach for SAD, and have received support from a small number of studies to date. There are notable potential advantages for the incorporation of vocal acoustic analyses into the broadband assessment of SAD. First, given that vocal characteristics are innately objective, and direct (i.e., physiologically based), vocal acoustic-related assessment of social anxiety symptoms is less subject to response biases than clinician-administered or self-report measures. Second, although the behavioral assessment tests and vocal acoustic analyses which have been utilized to examine vocal characteristics of SAD patients have been highly controlled in the present and previous studies, our end-goal for this approach is to develop assessments which are highly practical to routinely implement, yet which would be highly objective and powerful in classifying SAD patients. For example, patients could read standardized paragraphs to a clinician while standing and being audio recorded to assess public speaking concerns, 1 and vocal analyses could be performed on the first audio recording segments available of patients interacting with a clinician (e.g., during the initial stages of a clinical interview) to assess social interaction concerns. From a basic research standpoint, anxiety has been associated with increased vocal pitch in humans. Speakers produced higher overall vocal pitch during the repetition of short phrases when asked to imagine themselves in an anxious state than in emotional states including boredom and depression (Johnstone & Scherer, 1999). Furthermore, increased vocal pitch holds specific relevance to SAD within the context of submissive behaviors. Consistent with psycho-evolutionary models of social anxiety ( Gilbert, 2001), some social anxiety researchers have noted consistencies between social behaviors reported/exhibited by patients with SAD and submissive behavioral displays highlighted by comparative research focusing primarily on nonhuman primates ( Kaminer & Stein, 2005). The link between social anxiety-related symptoms (e.g., fears of evaluation) and submissive behaviors has been supported by several studies to date (e.g., see Weeks et al., 2011, Weeks et al., 2008, Weeks et al., 2010 and Weeks et al., 2009). Moreover, given that social anxiety may be related to dominance/submissiveness more so for males than females ( Maner, Miller, Schmidt, & Eckel, 2008), socially anxious males in particular may be likely to exhibit submissive behavioral displays when engaging in social situations. From a comparative standpoint, animals display submission to more dominant others by making themselves sound smaller than they actually are. Natural selection has resulted in the structural convergence of many animal sounds used in distinct contexts ( Morton, 1977, p. 855), such that birds and mammals (including humans) use higher-pitched sounds (characteristic of smaller animals) when frightened or appeasing. Evidence across a wide range of species demonstrates that vertebrates tend to universally emit higher pitch when avoiding attack ( Collias, 1960). Fundamental frequency (F0) is an objective index of the rate at which the vocal folds open and close across the glottis during phonation, and is the primary determinant of the auditory impression of vocal pitch. The present study was designed to focus exclusively on F0-related analyses. The relationship between vocal pitch (F0) and social anxiety has been examined, with encouraging initial results. Post-pharmacotherapy decreases in state anxiety in SAD patients were accompanied by corresponding decreases in mean and maximum F0 ( Laukka et al., 2008). While these findings are encouraging, it is worth noting that these analyses were restricted to the first 10 s of speeches which were delivered while lying down in a PET scanner. Moreover, this study did not include non-socially anxious control (NSAC) participants. Therefore, this design did not allow for an examination of F0 over time in response to continuous social threat, an assessment of the diagnostic utility of vocal characteristics with regard to SAD, nor measurement of F0 in an ecologically valid context (e.g., while standing [rather than lying down], and able to make eye contact with others). In a more recent study, males who were either high or low in social anxiety took part in a social competition task for the positive attention of a female peer. Specifically, all participants engaged in a semi-structured roleplay involving an 8-min social interaction with two experimental confederates: a female who was trained to relate positively to the participant, and a male who was trained to relate positively to the female confederate, but rudely to the participant (in order to induce implicit social competition for the female peer's attention). Higher social anxiety levels were associated with increased F0 peaks in response to the competitive male confederate, providing evidence for a vocal form of social anxiety-related submissive gesturing within males (Weeks et al., 2011). In addition, the relationship between F0 and social anxiety symptoms has begun to be examined in children. Scharfstein, Beidel, Sims, and Rendon-Finell (2011) found that children diagnosed with SAD emitted significantly greater mean F0 throughout a structured role play in comparison to children diagnosed with Asperger's disorder, although both SAD and Asperger's disorder patients emitted equivalent mean F0 in comparison to a control sample of typically developing peers (i.e., the typically developing peers emitted mean F0 levels between those of the two patient groups). Taken together with the findings of Laukka et al. (2008) and Weeks et al. (2011), the findings of Scharfstein et al. indicate that there may exist important developmental considerations pertaining to the relationship between vocal pitch and SAD, in that SAD-related increases in F0 relative to NSACs may not emerge until adolescence/adulthood – this is an important area for future research. Previous support for the relationship between F0 and social anxiety symptoms in adults notwithstanding, this area of research is still in its infancy, with only two reported studies (Laukka et al., 2008 and Weeks et al., 2011) directly focusing on the subject. The present series of studies was conducted to extend findings on F0 as a physiological indicator of social anxiety. In the first study, individuals diagnosed with SAD and demographically-equivalent NSACs were asked to complete a 4-min impromptu speech task. This study is novel in that it is the first to: examine differences in mean F0 in adults as a function of social anxiety over time during a public speaking task, and to attempt to identify mean F0 thresholds for potentially classifying SAD patients with respect to diagnostic status. The second study extended the examination of F0 emitted during social threat tasks by assessing audio recordings of individuals with a principal diagnosis of SAD (generalized subtype) completing in vivo social exposures using a covert audio recorder. The second study is the first to: examine the relation between F0 and social anxiety symptoms outside of a laboratory environment (thereby enhancing ecological validity); and to control for the effects of general state arousal experienced by speakers. Mean F0 emitted throughout both types of social threat tasks was examined in both studies utilizing specialized vocal analysis equipment. All analyses were performed separately for males and females, given that males and females have distinct F0 ranges ( Behrman, 2007).

7. Results Means and SDs for patients’ and NSA controls’ responses to all study measures are displayed in Table 2. 7.1. Preliminary analyses As an integrity check of the classification of GSAD patients and NSAC participants, an analysis of variance (ANOVA) was performed examining LSAS scores obtained across the two groups. As expected, the SAD patient sample obtained significantly greater LSAS scores than the NSAC sample, F (1, 36) = 121.73, p < .001, and this was a very large effect, Cohen's d = 3.38 ( Cohen, 1988). Furthermore, it is worth noting that the obtained LSAS scores were consistent with cutoff scores for classifying patients with GSAD (≥60) and non-socially anxious individuals (<30), respectively (see Mennin et al., 2002), providing further support for the classification of our patient and NSAC samples. Of the 16 patients who were included in the Study 2 analyses, 50% (n = 8) were randomly selected for the purposes of examining inter-rater reliability for the vocal acoustic analyses. Specifically, research assistants uninformed about the study hypotheses reviewed the audio recordings and performed independent CSL analyses. Bivariate correlations were computed for each of these sets of ratings. Mean F0 values demonstrated excellent inter-rater reliability (all rs > .996, all ps < .001). It is worth noting that the NSAC participants completed only one social threat task, whereas the patients completed two social exposures. In order to assess whether mean F0 or state anxiety levels changed significantly over the course of the two exposures (e.g., due to habituation of anxiety across exposures), paired-samples t-tests were performed within the full patient sample (to maximize statistical power). No significant changes in either mean F0, t (12) = −.73, p = .48, or state anxiety ratings, both ts (14) < |−.93|, both ps > .37, were obtained across the 2 days of exposure completion, providing support for comparing mean F0 values obtained from the single NSAC speech sample to the mean F0 values obtained from both patient exposures. Thus, in order to prevent range restriction, mean F0 values and state anxiety ratings obtained from the NSACs were pooled with the respective values obtained from the patients for both exposures (see Table 2). Importantly, no significant group differences were obtained between the patient and NSAC samples with regard to state anxiety either (a) in response to (i.e., peak during exposure [patients] vs. pre-speech [NSAC]), or (b) following (i.e., post-exposure), the exposures, all Fs (1, 35) < 3.84, all ps > .06. Thus, although the SAD patients rated their anxiety as higher than the NSAC participants, the patients and NSACs were indicated to experience statistically equivalent levels of state anxiety across the social exposures – this aspect of our control condition allowed us to examine whether GSAD patients emitted higher vocal pitch (F0) relative to non-socially anxious individuals while controlling for overall state anxiety level (i.e., is elevated vocal pitch a feature specific to SAD patients within social situations, or is it simply a function of elevated arousal/state anxiety?). 7.2. Analyses within the full sample 7.2.1. Hypothesis 1: Relation between social anxiety symptoms and mean F0 Bivariate correlations were calculated separately by gender (but within the full sample, to prevent range restriction) to test the hypothesis that mean F0 emitted throughout the exposures would relate positively to both SAD symptom severity (i.e., ADIS-IV CSRs) and social anxiety symptoms experienced over the previous week (i.e., LSAS scores). In addition to examining mean F0 emitted during both types of exposure separately and the baseline period, average overall mean F0 across both types of exposures for the patients was calculated. Of note, mean F0 emitted by the NSACs during the impromptu speech was imputed for both exposure types, and therefore also served as overall mean F0 during exposures for the NSAC subsample (see Table 3). A Bonferroni correction (.05/5 = .01) was applied in order to control for the number of comparisons. Table 3. Zero-order correlations in overall Study 2 sample between social anxiety symptoms experienced in the previous week and indices of vocal fundamental frequency (F0) in response to social threat tasks, reported separately by gender. Measure/index Males Females ADIS-IV-L Clinician's Severity Rating Mean F0 – impromptu exposure .88* .56* Mean F0 – pre-planned exposure .78* .54* F0 – grand mean across exposure type .90* .58* Mean F0 – baseline assessment .59 .17 Liebowitz Social Anxiety Scale Mean F0 – impromptu exposure .83* .65* Mean F0 – pre-planned exposure .65* .61* F0 – grand mean across exposure type .86* .66* Mean F0 – baseline assessment .13 .30 Notes: Sample size varies for males from 13 to 15, and for females from 21 to 22, as a result of some in vivo exposures which were assigned to patients in Study 2 which did not involve speaking (see Section 6.3 for details). ADIS-IV-L, Anxiety Disorders Interview Schedule for DSM-IV-Lifetime version. Values obtained from non-socially anxious control participants during impromptu speech task in Study 1 were included in all analyses, to prevent range restriction (see Section 7.1 for details). Bonferroni correction of: *p < .01 was applied. Table options For both males and females, as in Study 1, SAD diagnostic severity related significantly and positively to: mean F0 emitted during both exposures (with the NSAC mean F0 emitted during the speech task utilized for both correlations), and the overall mean F0 across exposure type. Furthermore, extending upon the results from Study 1, social anxiety symptoms experienced over the previous week related significantly and positively to all exposure-related mean F0 values. Moreover, as expected, mean F0 emitted during the baseline did not relate significantly to social anxiety for either gender. 7.2.2. Hypothesis 2: Between-groups analyses Multivariate analyses of variance (MANOVAs) were employed to test the hypothesis that SAD patients would emit higher mean F0 throughout the entirety of the two social exposures in comparison to mean F0 emitted by NSACs during the speech task, performed separately by gender. For the male subsample, the multivariate effect for diagnostic group (i.e., GSAD versus NSAC) was significant, F (3, 9) = 14.95, p < .001. Follow-up between-subjects comparisons revealed that the GSAD patients emitted significantly higher mean F0 during the impromptu (M = 144.83, SD = 6.47) and pre-planned exposures (M = 140.20, SD = 5.25) in comparison to the NSAC participants during the impromptu speech (M = 106.35, SD = 11.43), both Fs > 21.75, both ps < .001, and both of these effects were very large, both Cohen's ds > 3.81. In contrast, the GSAD patients emitted equivalent mean F0 during the baseline recording (M = 120.44, SD = 5.51) in comparison to the NSAC participants during the impromptu speech, F = 4.03, p = .07 (see Fig. 3). Differences in mean F0 values obtained across clinical and control samples in ... Fig. 3. Differences in mean F0 values obtained across clinical and control samples in Study 2, plotted according to situation and separately for males and females. Notes: F0, vocal fundamental frequency; GSAD, generalized (subtype) social anxiety disorder patients; NSAC, non-socially anxious controls. The order of pre-planned and impromptu exposures was counterbalanced. Figure options For the female subsample, the multivariate effect for diagnostic group (i.e., GSAD versus NSAC) was also significant, F (3, 17) = 9.45, p < .001. Follow-up between-subjects comparisons revealed that the GSAD patients emitted significantly higher mean F0 during both the impromptu (M = 221.15, SD = 11.92) and pre-established (M = 229.93, SD = 29.70) exposures than the NSAC participants emitted during the impromptu speech (M = 191.04, SD = 19.14), both Fs > 11.56, both ps < .003, and both of these effects were large, both Cohen's ds > 1.56. In contrast, the GSAD patients emitted equivalent mean F0 during the baseline recording (M = 199.44, SD = 8.11) in comparison to the NSAC participants during the impromptu speech, F = 1.52, p = .23 (see Fig. 3). 7.3. Patient-specific analyses 7.3.1. Hypothesis 3 Paired samples t-tests were employed to test the hypothesis that mean F0 emitted by GSAD patients throughout the two in vivo social exposures would be significantly greater than mean F0 emitted during the baseline recording. A Bonferroni correction (.05/3 = .0167) was applied in order to control for the number of comparisons. Mean F0 emitted during both the impromptu and pre-established exposures was significantly greater than mean F0 emitted during the baseline, both ts > 3.13, both ps < .008. Mean F0 emitted during the two exposures was equivalent, t = .27, p = .79. 4 7.3.2. Receiver operating characteristic (ROC) analyses 7.3.2.1. Hypothesis 4 Last, we were interested to identify potential thresholds by gender for overall F0 emitted across social exposures for adequately distinguishing between individuals with GSAD and NSACs, via ROC analyses. In order to be useful from a clinical standpoint, the primary goal of these analyses in Study 2 was to provide clinicians with thresholds which are applicable for the broad range of social exposures (i.e., not to test the distinction between impromptu and pre-established exposures) – in other words, successful identification of thresholds for overall F0 emitted across various social exposures (by gender) could provide clinicians with cutoffs for determining the presence of clinically-severe social anxiety on the basis of vocal pitch, regardless of the nature of the exposure. The ROC curves for males and females are displayed in Fig. 4. Overall mean F0 values demonstrated perfect differentiation between the male GSAD and NSAC subsamples (AUC = 1.00). Potential thresholds were evaluated through consideration of both sensitivity and specificity. In the present sample of males, the overall mean F0 value that provided the optimal balance was 128.93 Hz, correctly classifying 7/7 SAD patients and 8/8 NSACs (sensitivity: 100.0%; specificity: 100.0%). It is noteworthy that the next available lowest F0 threshold in the present sample was 120.97 Hz, which also exhibited excellent sensitivity (100.0%) and specificity (87.5%) – given that this latter identified F0 threshold fell below the F0 threshold identified in Study 1 (121.96 Hz), findings from Study 2 effectively replicated this previously established cutoff for males. For female participants, overall mean F0 values demonstrated excellent differentiation between the GSAD and NSAC subsamples (AUC = .90) – The overall mean F0 value that provided the optimal balance was 213.53 Hz, correctly classifying 8/9 SAD patients and 12/14 NSACs (sensitivity: 87.5%; specificity: 84.6%). Receiver operating characteristic (ROC) curves for overall mean F0 emitted ... Fig. 4. Receiver operating characteristic (ROC) curves for overall mean F0 emitted throughout the social exposures in the generalized social anxiety disorder patient/non-socially anxious control samples in Study 2, plotted separately for males and females. Note: F0, vocal fundamental frequency.