Articles
A Yoga Intervention for Music Performance Anxiety in Conservatory Students Judith R.S. Stern, PhD, JD, Sat Bir S. Khalsa, PhD, and Stefan G. Hofmann, PhD Music performance anxiety adversely affects a large minority of musicians. There is a need for additional treatment strategies, especially those that might be more acceptable to musicians than existing therapies. This pilot study examined the effectiveness of a 9week yoga practice on reducing music performance anxiety in undergraduate and graduate music conservatory students, including both vocalists and instrumentalists. The intervention consisted of fourteen 60-minute yoga classes approximately twice a week and a brief daily home practice. Of the 24 students enrolled in the study, 17 attended the post-intervention assessment. Participants who completed the measures at both pre- and post-intervention assessments showed large decreases in music performance anxiety as well as in trait anxiety. Improvements were sustained at 7- to 14-month follow-up. No changes were observed in mood or in music performance anxiety as retrospectively perceived in group performance or practice settings. Participants generally provided positive comments about the program and its benefits. This study suggests that yoga is a promising intervention for music performance anxiety in conservatory students and therefore warrants further research. Med Probl Perform Art 2012; 27(3):123–128.
M
usic performance anxiety (MPA) is a serious, widespread problem for both student and professional musicians and, in its more extreme forms, can impair or even end careers.1,2 Various surveys have reported that an estimated 15 to 25% of musicians experience severe and persistent levels of MPA, and more (59 to 70%) experience less severe levels. For reviews of surveys, see Kenny (2008)1 and Steptoe (2001).3 MPA is defined as “the experience of marked and persistent anxious apprehension related to musical performance that has arisen through specific anxiety conditioning experiences and which is manifested through combinations of affective, cognitive, somatic and behavioral symptoms.”1(p5) Only tentative conclusions can be reached regarding the effectiveness of existing treatments for MPA, given the
Dr. Stern is XXXXXXXXX, Department of Psychology, Boston University; Dr. Khalsa is Assistant Professor of Medicine, Division of Sleep Medicine, Brigham and Women’s Hospital, Harvard Medical School; and Dr. Hofmann is Professor of Psychology, Boston University, Boston, Massachusetts. Dr. Hofmann is supported by NIMH grant MH078308 and is a paid consultant for Merck/Schering-Plough for research unrelated to this project. Address correspondence to: Dr. Stefan G. Hofmann, Department of Psychology, Boston University, 648 Beacon Street, 6th Floor, Boston, MA 022152002, USA. Tel (617) 353-9610, fax (617) 353-9609.
[email protected].
methodological limitations of the current research.2,4 Betablockers appear to reduce somatic symptoms of MPA while improving some aspects of performance quality.4 Other methods of treating MPA that show promise include cognitive-behavioral therapy. Some mind-body interventions, including meditation, have been assessed in a few small studies each.5 Most musicians with MPA, however, do not avail themselves of psychotherapy.6 In a 1986 survey of 2,212 orchestral musicians by the International Conference of Symphony and Opera Musicians (ICSOM), only 25% of musicians suffering from severe stage fright reported trying psychological counseling.7 Yoga, a holistic mind-body practice that includes cognitive (meditation) and somatic (physical postures and breathing exercises) elements, has been described as an “attractive therapeutic option” for anxiety and anxiety disorders, especially for people who reject conventional psychological treatments.8,9(p890) Prior studies of yoga in non-music settings suggest that yoga may be an effective intervention for MPA. For example, a study of a nonclinical population found that a yoga intervention was associated with greater decreases in anxiety compared to a walking intervention.10 Khalsa and colleagues have conducted the only published studies on the effectiveness of yoga as a treatment for MPA.8,11 These non-randomized, controlled preliminary studies recruited participants from adult elite professional musicians studying at the Tanglewood Music Center. In the first study, 10 musicians participated in an 8-week comprehensive yoga lifestyle program offered through the Kripalu Center for Yoga and Health.8 Yoga was available daily, and a meditation session was offered 5 days per week. Participants typically attended about three to five classes per week. Other elements of the intervention included a weekly intensive yoga session and discussion group. Yoga participants experienced statistically significant improvements compared to the control group in both self-reported performance anxiety in solo settings and in mood.8 The second Tanglewood study used a larger sample (30 yoga participants and 15 controls) and included a 1-year follow-up.11 Half of the yoga participants received a yoga lifestyle intervention, and half received an intervention of yoga/meditation only. All yoga participants were required to attend three yoga and/or meditation classes per week, and outcome measures were acquired at baseline and 6 weeks September 2012
123
later. The results confirmed the results of the first study. In addition, the reduction in solo performance anxiety was maintained a year later for the treatment group. The current pilot study expanded the Tanglewood studies by examining students attending a music conservatory, rather than elite professional musicians, and by using a less-intensive yoga program more likely to be within the reach of the average student. We hypothesized that student musicians who participated in the yoga intervention would show improvements in self-reported MPA, general anxiety/tension, and mood that would be sustained over a follow-up period of approximately 1 year. Because trait anxiety has been found to be related to MPA,12,13 we further hypothesized that baseline trait anxiety levels would show a significant positive relationship with baseline MPA levels, which would provide evidence of convergent validity of the MPA measures.
METHODS Participants All participants attended The Boston Conservatory, a school for performing artists that awards both bachelor’s and master’s degrees. Participants were required to be full-time students (instrumentalists and/or vocalists) who expected to be enrolled during the fall 2007 and spring 2008 semesters and were at least 17 years old. Students were excluded from participating if they reported a medical condition that would prevent them from doing yoga safely, had been informed by a health-care professional that yoga was currently contraindicated, had a preexisting mind/body practice totaling 1 hour or more per week for a month or more preceding the screening interview, or anticipated scheduling conflicts with the study’s assessment schedule and/or anticipated missing more than 3 of the 14 assigned yoga classes. Participants were recruited by emails sent to all Conservatory students by the Conservatory Wellness Coordinator, announcements by school personnel, and in-person recruitment at the Conservatory for several weeks at the beginning of the fall 2007 and spring 2008 semesters. A number of students expressed initial interest informally during in-person recruitment but did not pursue participation on realizing that their schedules conflicted with the yoga class times. Ninety-nine students were assessed for eligibility. Sixty-eight of them were excluded from participation (53 did not meet inclusion criteria, primarily due to scheduling conflicts, 7 refused to participate due to busy schedules, and 8 failed to participate without giving a reason). Twenty-four students enrolled in the study. Twelve participated in the fall semester, and 12 in the spring semester. All participants took the baseline assessment and were assigned to the yoga intervention. The 24 participants’ ages ranged from 18 to 29 years (M 21.7, SD 3.1). Most participants (87.5%) were female. Although most participants (66.7%) were Caucasian, a variety of other races were present (Hispanic 8.3%, Asian 4.2%, Asian/Caucasian 4.2%, African American 4.2%, other 12.5%). Most participants (58.3%) were instrumentalists, 124
Medical Problems of Performing Artists
33.3% were vocalists, and 8.3% were both instrumentalists and vocalists. Seven participants did not attend the post-intervention assessment. Two cited busy schedules, one cited both illness and a busy schedule, one cited family illness, and three gave no reason. In November 2008, follow-up questionnaires were mailed to the 17 participants who attended the post-intervention assessment. Nine participants returned the questionnaires within a few weeks to a few months. Eight participants did not return them for unknown reasons. Procedures The Institutional Review Board of Boston University Charles River Campus approved the study prior to recruitment. All participants provided written informed consent. Students were told that the study would examine the impact of yoga on musical performance, performance anxiety, and mood. In exchange for their participation, students were offered free yoga classes (the intervention) and $20.00. Our intended sample size was limited to 40, the upper limit of the yoga class size stipulated as feasible by the yoga instructor. That sample size would have provided 86.94% power to detect medium effect sizes (Cohen’s d = 0.05), using two-tailed dependent t-tests and a 0.05 alpha level. Despite aggressive recruitment efforts, the intended sample size was not achieved. Of the approximately 500 fulltime students (excluding those studying dance or music theory) who attended the Conservatory during the recruitment period, all of whom were invited to be screened for participation in the study, 24 students (5%) participated. For this uncontrolled trial, baseline and post-intervention assessments (which took place on the Conservatory campus) and the yoga intervention were provided twice over the course of the academic year (once in the fall semester, and again the following spring semester). A follow-up assessment was given by mail 1 year after the end of the fall yoga intervention. Yoga Intervention The intervention included 14 one-hour classes (two classes per week over 9 weeks, with no classes during school vacations or holidays) on the Conservatory campus. The intervention also included home practice 4 days per week using a 16-minute compact disc recorded by the instructor. In addition, the intervention included optional, very brief (spontaneous practice “minis”). Participants were instructed that they could perform a yoga technique (breathing techniques and postures) learned during class or from the compact disc very briefly (seconds or minutes) at any time in response to their body’s need for movement or breath. The highly trained Kripalu yoga instructor, an employee of the Kripalu Center, had previously taught yoga to musicians. The Kripalu yoga intervention was in keeping with the classical hatha yoga tradition with classes including postures, breathing techniques, and meditation. Class attendance was tracked. Participants’ yoga class attendance ranged between 1 to 13 classes (M 9.29, SD 3.44) out of 14 classes. Participants who
took the post-intervention assessment attended 4 to 13 classes (M 10.94, SD 2.19). Participants who did not take the post-intervention assessment attended an average of 5.29 classes (SD 2.50). Thirteen participants, all of whom attended the post-intervention assessment, submitted a complete set of home practice logs. They reported practicing on an average of 31.38 days (SD 12.82) (87.17% of the 36 days asked of them). They also reported performing optional “minis” on an average of 33.62 days (SD 16.95). Measures Performance Anxiety Questionnaire (PAQ).12 This measure of MPA lists 20 statements associated with cognitive and somatic symptoms of anxiety. Participants rate each statement as to how frequently they experience it (using a 5-point scale ranging from never [1] to always [5]) in each of three situations: music practice, group music performance, and solo music performance. The PAQ has been validated by a number of studies.12-14 Kenny Music Performance Anxiety Inventory (K-MPAI).13 Participants rate 26 statements using a 7-point Likert scale ranging from strongly disagree (-3) to strongly agree (3). Seventyeight points are added to the raw score, resulting in scores of 0 to 156, with higher scores indicating higher levels of MPA.15 The K-MPAI has shown high internal consistency (unstandardized item alpha of 0.944, and item-total correlations ranging from 0.347 to 0.899), and convergent validity based on significant positive correlations with both subscales of the State-Trait Anxiety Inventory and modified versions of the PAQ.12 Profile of Mood States Brief form (POMS Brief).16 The POMS Brief (originally and still sometimes called the POMS short form or POMS-SF), a measure of mood state and general psychological distress, lists 30 words or short phrases that respondents rate as to the degree to which each describes themselves during the prior week, using a 5-point scale (ranging from not at all [0] to extremely [4]). Six subscale scores (which can range from 0 to 20) assess six mood states, including tension-anxiety. The POMS Brief Total Mood Disturbance (Total) score can range from –20 to 120, with higher scores reflecting greater disturbance. The POMS Brief has shown adequate internal consistency (Cronbach’s = 0.750.93).17,18 Its publisher states that product-moment correlations support an acceptable level of test-retest reliability, and factor analytic replications support the validity of the subscales.19 State-Trait Anxiety Inventory, T-Anxiety Scale, Form Y-2 (STAIT).20 The STAI-T consists of a 20-item subscale of the StateTrait Anxiety Inventory (STAI) that measures trait anxiety (relatively stable differences in anxiety-proneness). Responses can range from almost never (1) to almost always (4). Total scores can range from 20 to 80. This well-established instrument shows excellent internal consistency (median alpha coefficient of 0.90 in normative samples) and relatively high test-retest reliability. Substantial evidence supports convergent, divergent, and construct validity.20
Home Practice Log. We asked participants to track time spent practicing yoga (both using the CD and “minis”) in a daily log. Yoga Program Evaluation Questionnaire. This exit questionnaire includes seven questions using a 10-cm (100-mm) visual analog scale (from not at all at the extreme left to very much so at the extreme right at 100 mm) concerning the participant’s reactions to the yoga program. Scores can range from 0 to 100 mm. The questionnaire also asks for comments about the program. Yoga Program Follow-Up Questionnaire. This measure asks about time spent practicing mind/body techniques after the intervention, and about the extent to which the program changed the respondent’s music performance. Statistical Analyses Nine data points (0.19% of the total) were missing at random from baseline and post-intervention responses to the primary questionnaires for the 17 participants who attended both assessments. We conducted both completer and intent-totreat last observation carried forward (ITT, LOCF) analyses. Of the 9 participants who returned the follow-up questionnaires, 1 omitted the STAI-T and 1 omitted PAQ Group and Solo subscales. For completer analyses comparing post-intervention to follow-up scores, we included participants who fully completed the scale at all three assessment points. These analyses included 8 participants for the K-MPAI and STAI-T and 7 for the PAQ Solo subscale. Our hypotheses called for three families of analyses: i.e., two tests of correlation of baseline scores, six comparisons of baseline and post-intervention scores, and, where the latter tests were significant, comparisons of post-intervention and follow-up scores. We used the Holm-Bonferroni method to address family-wise error rate.
RESULTS Table 1 presents the baseline and post-intervention data, as well as the results of paired-sample t-tests comparing those scores using both ITT LOCF and completer analyses. Completer analyses comparing post to follow-up scores were based on the following post-intervention means, with standard deviations (SD) in parentheses: PAQ Solo 43.86 (11.41); KMPAI 45.00 (16.16); STAI-T 35.75 (8.80). Completer followup means (and SD) were: PAQ Solo 43.85 (11.11); K-MPAI 38.75 (17.77); STAI-T 36.00 (7.50). Before Holm-Bonferroni corrections, both PAQ Solo and K-MPAI scores decreased significantly from baseline to postintervention, with large effect sizes for completers, and medium-to-large effect sizes for the ITT sample. PAQ Group scores also decreased significantly, with a medium effect size for completers and a small-to-medium effect size using the ITT LOCF analysis. PAQ Practice scores showed a trend towards a significant decrease with small-to-medium effect sizes using both analyses. After adjusting the alpha level, the PAQ Solo decreases were still significant (ps 0.05/4 = 0.013), and did not indicate a trend (ps > 0.10/4 = 0.025). The PAQ Practice scores no longer showed a trend (ps > 0.10/3 = 0.03). Improvements in PAQ Solo scores were sustained at follow-up for completers, with no change from post-intervention: t(6) = 0.0, p = 1.00. Before a Holm-Bonferroni correction, the K-MPAI scores showed a trend towards a significant decrease from post-intervention to follow-up, with a mediumto-large effect size: t(7) = 2.01, p = 0.084, 95% CI [–1.09, 13.59], d = 0.71. After adjusting the alpha level, however, the results no longer indicated a trend: p >0.10/2 (total number of post-intervention, to-follow-up comparisons) = 0.05. Before adjusting for multiple comparisons, both the completer and ITT analyses of the POMS Brief TensionAnxiety subscale revealed a trend towards a significant decrease from baseline to post-intervention, with small-tomedium effect sizes. After adjusting the alpha level, no trend appeared (ps >0.10/2 = 0.05). The POMS Brief Total scores neither decreased significantly nor showed a trend towards significance (ps >0.10). An exploratory analysis revealed a significant reduction in STAI-T scores from baseline to postintervention, with a large effect size for completers and a medium-to-large effect size using the ITT LOCF analysis. 126
Medical Problems of Performing Artists
The improvement was sustained at follow-up, t(7) = –0.22, p = 0.84. Completers’ most positive responses to the Yoga Program Evaluation Questionnaire concerned general aspects of the program. For instance, for a question about whether the intervention was beneficial, scores averaged 91.35 (SD 9.51). In response to items that focused specifically on musicrelated benefits, respondents’ mean scores ranged from 67.94 to 83.82. Thirteen respondents provided comments, all positive. Eight of the nine respondents who returned the Yoga Program Follow-up Questionnaire reported that they had continued to practice yoga, and one reported practicing meditation. Only one comment had a negative component (that yoga “does not continue to be effective for me”), but added that yoga was a helpful learning tool that “has led me to using new approaches to . . . music . . .” The other comments were entirely positive, with references to feeling calmer and improved breathing and focus. The 17 participants who attended the post-intervention assessment did not differ significantly from those who did not in their demographic characteristics. Nor did completers differ significantly from noncompleters in baseline PAQ, KMPAI, STAI-T, or POMS Brief Tension-Anxiety scores (all ps > 0.39). Noncompleters’ baseline POMS Brief Total scores were, however, significantly higher than completers’ scores:
t(22) = –2.24, p = 0.04. Analysis of the other five POMS Brief subscales revealed significantly higher scores for noncompleters in baseline Anger-Hostility (t[22] = –2.14, p = 0.04) and Fatigue-Inertia (t[22] = –2.15, p = 0.04). Pearson correlations of baseline scores used the ITT sample. As predicted, K-MPAI scores correlated significantly and positively with STAT-T scores: r = 0.52, p = 0.009. This correlation remained statistically significant after adjusting the alpha level: p
