The efficacy of cognitive and behavioral interventions upon music performance anxiety: A meta-analysis

Abstract

Cognitive and behavioral interventions are widely used in dealing with anxiety problems. Many musicians face high anxiety levels when they have to perform in front of an audience, thus requiring efficient interventions. Studies investigating interventions for music performance anxiety (MPA) are highly heterogeneous, making it difficult to implement a meta-analytic analysis for all the intervention types investigated. However, there are several studies that investigate the effects of cognitive and behavioral interventions. Thus, the main aim of our study was to investigate the efficacy of cognitive and behavioral interventions in managing MPA symptoms in musicians. We reviewed 14 studies which included musicians (N = 392) who were confronted with high/dysfunctional levels of MPA. Overall, findings indicated a positive medium effect size (Cohen’s d = 0.53). We also tested the influence of several moderators related to intervention characteristics (e.g., type of intervention, type of control group, selection criteria) or participant characteristics (age, percentage of women). Although the differences we found were not statistically significant, the trend of the results was in line with previous research. Theoretical and practical implications for MPA are further discussed.

Keywords

music performance anxiety meta-analysis cognitive and behavioral interventions anxiety musicians

Musicians confront anxiety, music performance anxiety (MPA) being one of the most researched emotional problems in this professional category. It affects musicians of all ages and levels of performance and its severity varies on a continuum (Brugués, 2011). Estimates of musicians who declare MPA has a detrimental effect on their performance range between 16.5% and 60% (Fernholz et al., 2019).

Currently, MPA is considered a particular type of specific phobia in the International Classification of Diseases, Tenth Revision (ICD-10; Dilling & Freyberger, 2010) or a subtype of social phobia in the Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM-5; performance only subtype; American Psychiatric Association [APA], 2013). The association between social phobia and MPA has been observed in several studies. Estimates of musicians with high levels of MPA who also met criteria for social phobia vary between 24% and 33% (Ackermann et al., 2014; Barbar et al., 2014; Osborne & Franklin, 2002; Osborne & Kenny, 2005). Although both conditions have the fear of negative evaluation as a core element, researchers have proposed several differences between the two conditions (for a review, see Kenny, 2011).

MPA has been 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.” Higher levels of MPA are to be expected in performance situations which involve high ego investment (solo performances, recitals, concerts) or evaluative performances (auditions; Kenny, 2009, p. 433).

Steptoe (2001) grouped MPA symptoms into the following categories: (1) affective symptoms: apprehension about being negatively evaluated, and feelings of dread or panic; (2) cognitive symptoms, for example, memory slips, concentration problems, negative self-talk, as well as narrowed attention toward perceived threats, both internal threats (e.g., one’s personal evaluation of the performance) and external threats (e.g., one’s evaluation by the audience); (3) somatic symptoms: hyperventilation, trembling hands and legs, sweaty hands, dry mouth, and increased heart rate; and (4) behavioral symptoms: failures in technique, loss of posture, tremor. More recently, behavioral symptoms were grouped by Juncos et al. (2017) into (1) overt avoidance behaviors (e.g., avoiding auditions, solos, practicing) and (2) covert avoidance behaviors (e.g., avoiding eye contact with jurors and audience members, avoiding expressing one’s self, avoiding more challenging repertoire.

For levels which impair musicians’ well-being or their professional development, several psychological treatment options have been developed, offering a highly heterogeneous intervention literature. Some of them are based on the principles of classic paradigms in psychology (Kenny et al., 2014; Nagel et al., 1989). Other interventions integrate ideas from physical therapy or music therapy (Alexander Technique interventions, improvisation-based interventions, music therapy; Allen, 2013; Montello, 1990; Montello et al., 1990; Valentine et al., 1995). Physical activity has also been integrated in MPA interventions (yoga interventions; Khalsa & Cope, 2006; Stern et al., 2012) as well as hypnosis (Stanton, 1993). Mixed interventions combine techniques from various paradigms (e.g., hypnotherapy, success imagery, and rational emotive therapy; Stanton, 1993).

Several reviews (Fernholz et al., 2019; Kenny, 2005; McGinnis & Milling, 2005) and one meta-analysis (Goren, 2014) have synthesized the results and characteristics of these studies. Kenny (2005) reviewed 22 published and unpublished studies that tested the efficacy of MPA interventions in samples of student musicians and professional musicians. Almost half of the included studies included behavioral, cognitive, and cognitive and behavioral interventions (cognitive behavioral therapy [CBT]; 11 studies). The other half included combined interventions (CBT + placebo, CBT + biofeedback) and other interventions (music therapy, Alexander Technique, Ericksonian resource retrieval, meditation, and hypnotherapy). Overall, Kenny (2005) affirmed that interventions for MPA were highly heterogeneous, some of them seemed promising in reducing MPA levels (cognitive and behavioral interventions and combined interventions) and highlighted several methodological issues (lack of control group, insufficient sample size, various measures of MPA, not certified therapists conducting the intervention).

McGinnis and Milling (2005) reviewed 10 studies published in scientific journals categorized as (1) behavioral and cognitive therapies, (2) biofeedback, (3) hypnosis, (4) music enhanced therapy, and (5) cognitive behavioral versus medication. The effect size for the included interventions was Cohen’s d = 0.56 at posttreatment and Cohen’s d = 1.36 at follow-up. However, the follow-up indicators were estimated only for four studies. The authors also highlighted the heterogeneity of the treatments and designs and specified many methodological problems (overreliance of self-reported measures, absence of treatment manual). The highest estimated effect size was for music-enhanced therapies (d = 1.57 and posttreatment) and medication had the lowest estimated effect size at posttreatment (d = 0.01).

Fernholz et al. (2019) reviewed 43 studies to establish MPA prevalence, risk factors, and treatment effects. The authors reported that CBT and beta-blockers were most frequently researched and found beneficial results for CBT. Also, several methodological issues previously mentioned were still problematic in MPA research: (1) studies lack a common definition of MPA, (2) no inclusion criteria or different inclusion criteria, (3) no randomization, (4) waiting lists as control groups or lack of a control group, and (5) single blinding of participants.

One meta-analysis merged the results of individual studies to quantify the efficacy of non-pharmacological MPA interventions (Goren, 2014). This meta-analysis included 29 studies that were split in four categories: (1) behavioral, (2) complementary and alternative (e.g., meditation, yoga, Ericksonian, biofeedback, and music therapy), (3) cognitive, and (4) combined (which included combination of two or more types of therapies, with cognitive and behavioral interventions being a part of this category as well). The overall effect size was Cohen’s d = 0.64 and the combination of two or more types of therapies was found to be most effective. However, three aspects should be considered. First, potential moderators were not tested. Second, this meta-analysis combined several interventions in the complementary and alternative category due to the small number of publications on each intervention (e.g., biofeedback, meditation, yoga, music therapy, and Ericksonian). However, complementary and alternative category interventions have different mechanisms of change and their efficacy should be separately investigated. And third, since 2014 new studies of MPA interventions have been published.

To sum up, previous reviews and Goren’s meta-analysis included various intervention studies and several limitations are noted. First, in 2005 some studies did not provide sufficient data and were highly heterogeneous (diverse subject groups, treatments, outcomes), which made it difficult to be synthesized in a meta-analysis (see Kenny, 2005). Second, unpublished papers presented serious methodological issues (see Patston, 1996). Third, often interventions such as hypnotherapy, biofeedback, music therapy, Alexander Technique, included fewer studies, sometimes even just one per treatment category. Fourth, previous meta-analyses did not include moderator analyses.

Thus, it was considered that a more specific synthesis would have clear theoretical and clinical implications. A meta-analysis was preferred due to its advantages (increasing statistical power, opportunity to test several variables that could influence the effect size, etc.; Goren, 2014). As many studies in the MPA intervention literature investigated cognitive and behavioral interventions, the main purpose of this project was to establish the efficacy of cognitive and behavioral interventions in reducing MPA. The specific objectives of this meta-analysis were

- to establish the overall effects of cognitive and behavioral interventions across studies; and

- to test the moderator role of the following variables: type of intervention, type of control group, intervention environment, type of outcome, selection criteria, grouping of participants, type of design, percentage of women, and age.

Method

Literature search

Potentially relevant studies were found following a systematic search of PsycInfo, Google Scholar, Pro-Quest databases, and ResearchGate through May 2019, which was repeated in October 2020. The following keywords were used: “music performance anxiety” paired with “treatment,” “cognitive-behavioral intervention,” “stage fright” paired with “intervention,” “cognitive behavioral treatment,” and “behavioral interventions” paired with “music performance anxiety.” The references within most relevant reviews and meta-analysis were also investigated (Fernholz et al., 2019; Goren, 2014; Kenny, 2005; McGinnis & Milling, 2005; Nagel, 2010).

Selection of studies

Following the search procedure, 60 studies were identified (see Figure 1). Eighteen duplicates were removed and the remaining articles were screened for relevance. Forty-two studies were screened based on their abstracts. Ten papers were further excluded, with nine of them being reviews and one being a meta-analysis. A total of 32 relevant articles were further analyzed for relevance based on their full text. The selection criteria were (1) studies that assessed MPA (i.e., self-reported measures, behavioral measures, or physiological measures); (2) studies which were published in peer-reviewed journals; (3) English-language publications; (4) studies that did not combine CBT with drugs or other forms of therapy (yoga, Alexander Technique, improvisation desensitization, etc.); (5) studies that had sufficient data to estimate effect size; and (6) the intervention that had a specified protocol (authors presented a table which offered a minimal description of the treatment sessions). Eighteen studies were excluded due to the following reasons: (1) positive psychology intervention, guided imagery and improvisation plus systematic desensitization (three studies); (2) CBT + medication (one study); (3) studies not published in peer-reviewed journals (five studies); (4) case studies (two studies); (5) poster presentations (three papers); (6) theoretical articles (three studies); and (7) not measuring MPA (one study). Fourteen studies met the inclusion criteria. Four of them were single-subject design studies, with the rest being randomized clinical trials (see Figure 1).

Figure 1.

PRISMA Flowchart.

Coding for moderators

Categorical moderators (subtype of intervention, type of control group, intervention environment, selection criteria, intervention administration, and type of design) and continuous moderators (age, percentage of women) were analyzed. To provide coding consistency and construct validity, the coding scheme was developed by the first two authors according to the conceptual and operational definitions for all the concepts involved. Further on, the coding procedure was performed by both authors. All instances of disagreement were resolved through consensus. Table 1 presents the moderators and examples for each moderator or a short description.

Table 1.

Moderator Variables.

Coding category	Moderator subcategory
Type of intervention	Behavioral (e.g., systematic desensitization, exposure, relaxation groups) Cognitive (cognitive restructuring alone) Cognitive and behavioral (cognitive techniques such as cognitive restructuring and behavioral techniques) ACT intervention (focus on acceptance of one’s distress and other experiential techniques)
Type of control group	Wait-list Alternative treatments (e.g., music analysis) No control group
Type of outcomes	Behavioral (e.g., behavioral anxiety index) Physiological (e.g., heart rate) Subjective (self-reported questionnaires) Music performance (jury evaluation)
Intervention environment	In vivo Virtual reality
Selection criteria	Yes (interviews, cutoff points) No (volunteers)
Intervention administration	Group Individual
Study design	Experimental (randomization and control group) Quasi-experimental (if studies did not use randomization methods and did not include a control group)

Note. Spahn et al. (2015) was excluded from the study design moderator analysis due to the fact that it was not specified if randomization methods were used. ACT: acceptance and commitment therapy.

Procedure

For each of the studies included in the meta-analysis, several variables were retained: study identification data (author and year of publication), type of intervention, type of intervention administration (individual or group intervention), participants category (adults, emerging adults, or adolescents), outcome measures, intervention environment (in vivo or virtual reality), type of control group, intervention duration, percentage of female participants per study, number of participants, mean age of the participants, and selection criteria (yes or no; see Table 2 for the coding of these variables).

Table 2.

Study Characteristics.

Study ID	Type of CBT intervention	Type of intervention administration	Participants	Outcomes	Intervention Environment	Type of control(s)	Intervention duration (no. of sessions)	Women percentage	Age	N	Selection criteria	Type of design
Appel (1976)	Behavioral	Group intervention	Students	PRCP, pulse rates, performance quality	Vivo	Alternative, wait-list	8	50	31	29	Yes	Experimental
Bissonnette et al. (2015)	Behavioral	Individual	Students	PRCP, STAI, PAI, performance quality	VR	Wait-list	6	58.82	21.8	17	Yes	Experimental
Bissonnette et al. (2016)	Behavioral	Individual	Students	Anxiety SUD	VR	No control	6	66	23.3	9	No	Quasi-experimental
Braden et al. (2015)	CBT	Group intervention	Adolescents	MPAI-A, performance quality	Vivo	Wait-list	8	100	13.78	62	No	Quasi-experimental
Clarke et al. (2020)	ACT	Group intervention	Students	K-MPAI, DASS	Vivo	No control	6	66.6	20.33	6	Yes	Quasi-experimental
Hoffman & Hanrahan (2012)	CT	Group intervention	Adults	BAI, PAI, STAI, heart rate, performance quality	Vivo	Wait-list	3	87.88	42.09	33	No	Experimental
Juncos et al. (2017)	ACT	Group intervention	Students	ACQ, K-MPAI, performance quality	Vivo	No control	12	85.71	23.28	7	Yes	Quasi-experimental
Kendrick et al. (1982)	CBT	Group intervention	Students	STAI, SSS, heart rate, performance quality	Vivo	Alternative, wait-list	3	91	18.83	48	Yes	Experimental
Kenny and Halls (2018)	CBT	Group intervention	Adults	K-MPAI, performance quality	Vivo	alternative	4	66.18	44.51	68	No	Experimental
Orman (2003)	Behavioral	Individual	Students	Anxiety SUD, heart rate	VR	No control	12	37.5	20.62	8	No	Quasi-experimental
Osborne et al. (2007)	CBT Behavioral	Group intervention + individual session	Adolescents	MPAI-A, heart rate	Vivo	No control	7	57	13.87	23	Yes	Quasi-experimental
Spahn et al. (2015)	CBT	Group intervention	Students	K-MPAI, performance quality	Vivo	Wait-list	14	67	22.1	21	No	Not specified
Sweeney & Horan (1982)	Behavioral CT CBT	Group intervention	Students	BIA, AATS, pulse rates, performance quality	Vivo	Alternative, wait-list	6	48.97	Not mentioned	49	Yes	Experimental
Van Zyl (2020)	Behavioral	Individual	Students	Self-reported anxiety levels, heart rate	VR	Wait-list	1	41.66	20.6	12	No	Experimental

Note. CBT: cognitive behavioral therapy; CT: cognitive therapy; ACT: acceptance and commitment therapy; RCP: personal report of confidence as a performer; SUD: subjective units of distress; MPAI-A: Music Performance Anxiety Inventory for Adolescents; PAI: Performance Anxiety Inventory; STAI: State and Trait Anxiety Inventory; BAI: Behavior Anxiety Index; ACQ: Anxiety Control Questionnaire; K-MPAI: Kenny Music Performance Anxiety Inventory: SSS: Subjective Stress Scale; BIA: Behavioral Index of Anxiety; AATS: Achievement Anxiety Test Scale; VR: Virtual Reality; DASS: Depression, Anxiety, Stress Scale.

Outcome MPA measures were classified into one of three categories: (1) MPA self-report measures (e.g., K-MPAI); (2) behavioral measures (e.g., error counting or behavioral indexes, including behaviors as knee trembling, hands trembling, and stiff arms); and (3) physiological measures (e.g., heart rate indicators). Most of the studies were interested in finding whether music performance was also enhanced as a consequence of the interventions. Thus, performance quality measures were also included (see Table 2).

For estimating effect size, Cohen’s d coefficient was used (Cohen, 1988). One effect size was computed for each comparison per outcome at postintervention. Nine studies did not include follow-up measures; thus, analyses based on follow-up data were not conducted. To estimate effect sizes, we used (1) means and standard deviations, when available; (2) measures of effect size reported in the study (Cohen’s d); and (3) between-group t values and sample size.

A random-effects model was used to guide data analysis and interpretation. This model assumes that there is a distribution of true effect size. The combined effect obtained in one meta-analysis represents the mean of the population of true effect (Borenstein et al., 2011). Heterogeneity of effect sizes was assessed by using the Q statistic (Borenstein et al., 2011). This index compares true heterogeneity with random error. A statistically significant Q suggests a true heterogeneity in effect sizes beyond random error. This model was preferred although the heterogeneity index was not statistically significant, due to the fact that studies were methodologically diverse as we will highlight in the “Results” section.

To determine the publication bias, we used the classic Fail-Safe N Test, which was first developed by Rosenthal (1979). It assesses the number of studies needed to nullify the estimated effect size in our sample, by averaging a z-value of zero, which needs to be added for the combined effect size to be statistically insignificant. The smaller this indicator’s value, the higher reason for concern about the sample size. There are two arguments that sustain the cautious interpretation of this indicator: (1) it is an indicator of statistical significance rather that clinical significance, and (2) by assuming that the effect in the other studies is null, it can underestimate the necessary number of studies that would nullify the estimated effect size.

Results

Overall analysis of cognitive and behavioral interventions across studies

Figure 2 shows statistics examining the estimated effect size on MPA at the end of the intervention considering data in the 14 studies (N = 392). Overall, studies indicate a significant positive medium effect size, d = 0.535, p < .01, 95% confidence interval [CI] = [0.340, 0.730]. The estimated effect sizes in each study range from small non-significant effect sizes, for example, d = 0.04, p = .89, 95% CI = [−0.64, 0.74] for Kendrick et al. (1982), to large significant effect sizes, for example, d = 1.20, p = .03, 95% CI = [0.11, 2.29] for Juncos et al. (2017).

Figure 2.

Estimated Effect Size.

Figure 3.

Effect of Percentage of Women.

Figure 4.

Effect of Age.

Heterogeneity analysis

Studies’ distribution analysis indicates no statistically significant heterogeneity, Q(11) = 5.90, p = .95. Although the Q index was not statistically significant, studies were methodologically diverse, which is why we used the random effect model and we further decided to code several moderators. Studies used different interventions (behavioral, acceptance and commitment therapy [ACT], and CBT), included different participants (adolescents and adults), and used different controls or no controls as well as different outcome measures.

Publication bias

This meta-analysis incorporates data from 14 studies, which yield a z-value of 5.42 and corresponding two-tailed p value of <.0001. The fail-safe N is 94. This means that we would need to locate and include 94 “null” studies for the combined two-tailed p value to exceed .050. In other words, there would be a need of 6.7 missing studies for every observed study for the effect to be nullified.

Moderator analysis

Table 3 summarizes the results of the moderator analyses for the categorical moderators. We investigated seven potential categorical moderators (type of intervention, type of control, intervention environment, type of outcome, type of participants, selection criteria, and type of design). Differences in the estimated effect sizes were found but they were not statistically significant, as is further presented.

Table 3.

Moderation Analysis.

Moderator	Categories of the moderator	No. of studies	Cohen d	Lower limit	Upper limit	QB	df	p
Type of CBT intervention	ACT	2	0.854	0.293	1.415	1.538	3	.674
	Behavioral	7	0.494	0.180	0.807
	CBT	6	0.481	0.232	0.729
	CT	2	0.612	0.054	1.170
Type of control(s)	Alternative treatment	4	0.402	0.066	0.739	1.186	2	.553
	No control	5	0.666	0.322	1.010
	Waiting list	8	0.499	0.228	0.771
Intervention environment	In vivo	10	0.545	0.325	0.765	0.040	1	.842
Intervention environment	In VR	4	0.497	0.077	0.977
Type of outcome	behavioral	2	0.631	0.186	1.077	1.082	3	.781
	physiological	7	0.399	0.096	0.701
	music performance	9	0.539	0.307	0.772
	subjective	14	0.570	0.376	0.763
Selection criteria	yes	6	0.563	0.236	0.891	0.045	1	.832
Selection criteria	no	8	0.519	0.277	0.761
Type of administration	Group	9	0.558	0.329	0.788	0.062	1	.803
Type of administration	Individual	4	0.497	0.077	0.917
Type of design	Experimental	7	0.428	0.149	0.707	0.753	1	.385
Type of design	Quasi-experimental	6	0.604	0.320	0.889

Note. There were studies, the design and measures of which allowed us to code many entries per category of moderator. Thus, the total number of studies for type of intervention, type of control, and type of outcome exceeds the total number of studies included in this meta-analysis. Information regarding the coded variables is synthesized in Table 2. The study of Osborne et al. (2007) was not included in the type of administration group for it was mixed, including individual and group sessions. CBT: cognitive behavioral therapy; ACT: acceptance and commitment therapy; CT: cognitive therapy; VR: Virtual Reality; QB: statistic moderator indicator.

Type of intervention as a moderator

The three categories of intervention produced significant effect sizes with ACT resulting in the highest effect size, d = 0.84, 95% CI = [0.29, 1.41], followed by cognitive interventions, d = 0.61, 95% = [0.05, 1.17], and behavioral interventions, d = 0.49, 95% CI = [0.18, 0.80]. Cognitive and behavioral interventions resulted in a significant effect size, d = 0.48, 95% CI = [0.23, 0.72], a value close to the one of behavioral interventions. However, the difference between the three categories was not statistically significant, QB(3) = 1.53, p = .677.

Type of control group as a moderator

Studies which compared CBT with alternative treatment had a smaller estimated effect size, d = 0.40, 95% CI = [0.06, 0.73], followed by studies which used waiting lists as control group, d = 0.49, 95% CI = [0.22, 0.77]. The highest effect size estimate was for studies which had no control group, d = 0.66, 95% CI = [0.32, 1.01]. All the estimated effect sizes were statistically significant. However, the differences between the three categories of this moderator were not statistically significant, QB(3) = 1.18, p = .553.

Intervention environment as a moderator

The environment of the intervention presented similar significant estimated effect sizes for in vivo intervention, d = 0.54, 95% CI = [0.32, 0.76], and for Virtual Reality (VR) intervention, d = 0.49, 95% CI = [0.07, 0.97], but we did not find a statistically significant difference between them, QB(1) = 0.04, p = .842.

Type of outcome as a moderator

Regarding the type of outcome, smaller significant estimated effect sizes were found for the physiological measures of MPA, d = 0.39, 95% CI = [0.09, 0.70], and music performance indicators, d = 0.53, 95% CI = [0.30, 0.77], and higher significant effect sizes for behavioral measures of MPA, d = 0.63, 95% CI = [0.18, 1.07], and subjective measures of MPA, d = 0.57, 95% CI = [0.37, 0.76]. The differences between them, however, were not statistically significant, QB(3) = 1.08, p = .781.

Selection criteria as a moderator

There were no significant differences in the estimated effect sizes between studies which contained selection criteria (six studies) and studies that did not include selection criteria (eight studies), QB(1) = 0.04, p = .832.

Type of intervention administration as a moderator

Significant effect sizes were observed for both group, d = 0.55, 95% CI = [0.32, 0.78], and individual, d = 0.49, 95% CI = [0.07, 0.91] interventions. The difference between them was not statistically significant, QB(1) = 0.06, p = .803.

Type of design as a moderator

Significant effect sizes were observed for both experimental, d = 0.48, 95% CI = [0.14, 0.70], and quasi-experimental studies, d = 0.60, 95% CI = [0.32, 0.70]. The difference between them was not statistically significant, QB(1) = 0.73, p = .385.

Percentage of women and age as moderators

The percentage of women was not a significant moderator as indicated by B = −0.002, p = .644. Also, age was not a significant moderator as indicated by B = 0.001, p = .847.

Discussion

The main aim of this meta-analysis was to investigate the efficacy of CBT interventions in reducing MPA. Interventions’ efficacy was measured by evaluating changes in MPA symptoms and music performance. Following our systematic search, 14 studies were included in our analysis with 392 musicians, from which 91 effect sizes were estimated. Due to the fact that studies were methodologically heterogeneous, we also tested potential moderators.

Results showed that CBT interventions were effective overall in reducing MPA symptoms (d = 0.535). The estimated effect size for all the studies had a positive medium value which was a bit smaller than the one obtained by Goren (2014) and other meta-analyses which investigate the impact of CBT in anxiety disorders (Carpenter et al., 2018; Hofmann & Smits, 2008). However, the overall estimated effect size value was closer to the one obtained in meta-analyses which compared CBT with placebo (Carpenter et al., 2018) and almost 2 times smaller than other meta-analyses which compared CBT group (d = 1.26) or individual (d = 1.23) intervention for anxiety symptoms with wait-list controls (Bandelow et al., 2015), which is an interesting finding as most of the included studies in our meta-analysis used wait-list or no control group. One possible explanation could be the fact that MPA intervention studies, having a smaller number of participants, lack sufficient power. Also, our meta-analysis included behavioral and physiological measures of MPA that could have resulted in smaller estimated effect size.

As was previously mentioned, we tested several moderators, some related to characteristics of the intervention as well as participants’ characteristics. Although none of them were significant moderators, we did find some differences in the expected direction.

Behavioral interventions and interventions that combined cognitive and behavioral techniques or focused just on cognitive aspects produced similar effect sizes. This result is also in line with previous research. ACT interventions, although represented by only two studies, had the highest estimated effect size (d = 0.85). However, neither of the studies used a control group, so the effect size might have been overestimated. Future studies should include active control groups to have a more accurate estimation of ACT efficacy. Goren (2014) found that interventions that combined two or more types of interventions resulted in higher effect sizes; however, she mentioned that the difference might not be so significant as to have practical implications. Carpenter et al. (2018) found that Social Anxiety Disorder (SAD) treatments that primarily included exposure techniques produced larger effect sizes than those that included both cognitive and behavioral techniques and cognitive techniques alone. However, these differences were not statistically significant. Nevertheless, in this meta-analysis cognitive interventions produced higher estimated effect size as compared with behavioral or cognitive and behavioral interventions. Although the difference was not statistically significant, this is an unexpected result. One possible explanation for this difference could be the fact that in both of the studies which were included in this category there was a wait-list control group as comparison. Behavioral and cognitive and behavioral interventions also had alternative treatments as comparison, which might have resulted in a smaller effect size.

Studies with no control and wait-list as controls had higher effect sizes than studies that included alternative treatment groups. However, the estimated effect size was much smaller for wait-list comparison than in other meta-analyses that made comparisons using this type of control as it was previously mentioned (Bandelow et al., 2015).

The estimated effect sizes observed for the type of intervention environment, although slightly higher in the case of VR environment, were not statistically significant. This result is in line with previous meta-analyses which concluded that VR for anxiety disorders was more effective when compared with wait-list and not significantly more effective when compared with active treatments (Opriş et al., 2012; Parsons & Rizzo, 2008; Powers & Emmelkamp, 2008) or a rather smaller difference in favor of VR intervention as observed by McCann et al. (2014; Hedge’s g = .19; 95% CI = [.03, .35]).

Smaller estimated effect sizes were observed for physiological and music performance outcomes and similar estimated effect sizes were observed in case of behavioral and self-reported outcomes. This result is also in line with previous research that shows that physiological markers do not change significantly after CBT interventions in SAD. However, changes in participants’ perceptions of their physiological symptoms were observed (Aderka et al., 2013). Also, Kenny (2005) has stated that improvements following behavioral interventions were mostly reported for self-reported measures of MPA. However, they do not always result in improved music performance.

No significant differences were found for the type of intervention administration moderator. This result was different from previous research. For SAD and posttraumatic stress disorder (PTSD), Carpenter et al. (2018) found significant differences, with higher effect size for the individual CBT compared with group CBT interventions. However, our result should be cautiously interpreted due to the fact that only two studies had individual treatments.

Limitations

A number of limitations should be considered regarding this meta-analysis. First, there were a limited number of trails that were included; thus, some of the moderator analyses might need caution in interpretation due to the small number of studies per category. More specifically, there were only four studies that implemented individual administration of CBT and four studies which implemented VR interventions. Second, ACT interventions category included only two studies. Thus, more research is needed to be able to account for more robust findings. Third, most of the included studies had participants who were involved in classical music. Future studies should test CBT interventions on more diverse samples of musicians to gather evidence for the generalization of these results to musicians who play different music genres. Fourth, a robust variance estimation procedure would have increased the validity of the results of this research. It was not implemented, though, for pragmatic and logistic reasons. However, the Comprehensive Meta-Analysis software used for the analyses of this study is one of the most widely used software for meta-analysis and has a largely accepted validity in the scientific community.

Future directions

MPA is an important issue for many musicians that affects their professional and personal well-being. This meta-analysis provided evidence for the overall efficacy of cognitive and behavioral interventions for MPA. However, due to the small number of participants, at an individual level, many of the included studies lack sufficient statistical power to detect changes, which might have influenced the estimation of the effect size. Also, the estimated effect size in our meta-analysis was similar to that found in other research when comparing CBT with placebo interventions, being almost half the size of the one estimated by other meta-analyses that compared CBT interventions with wait-list controls, although the majority of the included studies had no control or wait-list controls. Future research should include more participants and compare cognitive and behavioral protocols with placebo controls, for these control groups are considered to account also for non-specific factor of an intervention (e.g., expectancy effects; Carpenter et al., 2018) and would provide a more accurate reflection of CBT interventions on this specific anxiety form. Regarding the selection of participants, many of the included studies had convenience samples and did not include selection criteria, or if selection criteria were included, it was differently implemented (e.g., clinical interviews, cutoff points from different scales). Future research should be made to provide standardized clinical tools. Although there were not significant differences between subgroup of CBT interventions, future studies should implement more standardized protocols to better group and quantify intervention effects. Regarding the multilevel measures of anxiety, only two studies included behavioral, physiological, and self-reported measures of anxiety. This is a problem for many studies as well as the diversity of measures used to evaluate behavioral and self-reported levels of anxiety. Many studies also included performance quality measures. However, interventions did not always result in performance quality change. This could be because intervention techniques usually targeted anxiety symptoms. Other psychological factors that influence performance quality were not targeted. Perhaps inclusion of techniques from science of expertise could be more frequently implemented if performance quality is targeted in these interventions as well (deliberate practice, Ericsson, 2002; Ericsson & Harwell, 2019).

To sum up, cognitive and behavioral interventions are promising interventions for musicians who confront themselves with MPA; however, more standardized selection criteria, protocols, and the inclusion of placebo controls would result in more robust conclusions.

Footnotes

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Ackermann

B. J.

Kenny

D. T.

O’Brien

Driscoll

T. R.

(2014). Sound practice —Improving occupational health and safety for professional orchestral musicians in Australia. Frontiers in Psychology, 5, Article 973.

Aderka

I. M.

McLean

C. P.

Huppert

J. D.

Davidson

J. R.

Foa

E. B.

(2013). Fear, avoidance and physiological symptoms during cognitive-behavioral therapy for social anxiety disorder. Behaviour Research and Therapy, 51(7), 352–358.

Allen

(2013). Free improvisation and performance anxiety among piano students. Psychology of Music, 41(1), 75–88.

American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (DSM-5®). American Psychiatric Publishing.

Appel

S. S.

(1976). Modifying solo performance anxiety in adult pianists. Journal of Music Therapy, 13(1), 2–16.

Bandelow

Reitt

Röver

Michaelis

Görlich

Wedekind

(2015). Efficacy of treatments for anxiety disorders: A meta-analysis. International Clinical Psychopharmacology, 30(4), 183–192.

Barbar

A. E. M.

de Souza Crippa

J. A.

de Lima Osório

(2014). Performance anxiety in Brazilian musicians: Prevalence and association with psychopathology indicators. Journal of Affective Disorders, 152, 381–386.

Bissonnette

Dubé

Provencher

M. D.

Moreno Sala

M. T.

(2015). Virtual reality exposure training for musicians: Its effect on performance anxiety and quality. Medical Problems of Performing Artists, 30(3), 169–177.

Bissonnette

Dubé

Provencher

M. D.

Sala

M. T. M.

(2016). Evolution of music performance anxiety and quality of performance during virtual reality exposure training. Virtual Reality, 20(1), 71–81.

10.

Borenstein

Hedges

L. V.

Higgins

J. P.

Rothstein

H. R.

(2011). Introduction to meta-analysis. John Wiley & Sons.

11.

Braden

A. M.

Osborne

M. S.

Wilson

S. J.

(2015). Psychological intervention reduces self-reported performance anxiety in high school music students. Frontiers in Psychology, 6, Article 195.

12.

Brugués

A. O.

(2011). Music performance anxiety—Part 1: A review of its epidemiology. Medical Problems of Performing Artists, 26(2), 102–105.

13.

Carpenter

J. K.

Andrews

L. A.

Witcraft

S. M.

Powers

M. B.

Smits

J. A.

Hofmann

S. G.

(2018). Cognitive behavioral therapy for anxiety and related disorders: A meta-analysis of randomized placebo-controlled trials. Depression and Anxiety, 35(6), 502–514.

14.

Clarke

L. K.

Osborne

M. S.

Baranoff

J. A.

(2020). Examining a group acceptance and commitment therapy intervention for music performance anxiety in student vocalists. Frontiers in Psychology, 1127.

15.

Cohen

(1988). Statistical power analysis for the behavioral sciences (2nd ed.). Lawrence Erlbaum.

16.

Dilling

Freyberger

H. J.

(2010). World Health Organization Pocket Guide to the ICD-10 Classification of Mental Disorders. Sixth completely revised edition. Verlag Hans Huber.

17.

Ericsson

K. A.

(2002). Attaining excellence through deliberate practice: Insights from the study of expert performance. In Ferrari

(Ed.), The pursuit of excellence through education (pp. 21–55). Lawrence Erlbaum Associates Publishers.

18.

Ericsson

K. A.

Harwell

K. W.

(2019). Deliberate practice and proposed limits on the effects of practice on the acquisition of expert performance: Why the original definition matters and recommendations for future research. Frontiers in Psychology, 10, Article 2396.

19.

Fernholz

Mumm

J. L. M.

Plag

Noeres

Rotter

Willich

S. N.

. . .Schmidt

(2019). Performance anxiety in professional musicians: A systematic review on prevalence, risk factors and clinical treatment effects. Psychological Medicine, 49(14), 2287–2306.

20.

Goren

(2014). A meta-analysis of nonpharmacologic psychotherapies for music performance anxiety [Doctoral dissertation]. http://gradworks.umi.com/36/21/3621049.html)

21.

Hoffman

S. L.

Hanrahan

S. J.

(2012). Mental skills for musicians: Managing music performance anxiety and enhancing performance. Sport, Exercise, and Performance Psychology, 1(1), Article 17.

22.

Hofmann

S. G.

Smits

J. A.

(2008). Cognitive-behavioral therapy for adult anxiety disorders: A meta-analysis of randomized placebo-controlled trials. The Journal of Clinical Psychiatry, 69(4), Article 621.

23.

Juncos

D. G.

Heinrichs

G. A.

Towle

Duffy

Grand

S. M.

Morgan

M. C.

. . .Kalkus

(2017). Acceptance and commitment therapy for the treatment of music performance anxiety: A pilot study with student vocalists. Frontiers in Psychology, 8, Article 986.

24.

Kendrick

M. J.

Craig

K. D.

Lawson

D. M.

Davidson

P. O.

(1982). Cognitive and behavioral therapy for musical-performance anxiety. Journal of Consulting and Clinical Psychology, 50(3), 353–362.

25.

Kenny

D. T.

(2005). A systematic review of treatments for music performance anxiety. Anxiety, Stress, and Coping, 18(3), 183–208.

26.

Kenny

D. T.

(2009). Negative emotions in music making: Performance anxiety. In Juslin

P.N.

Sloboda

J.A.

(Eds.), Handbook of music and emotion: Theory, research, applications. Oxford University Press.

27.

Kenny

D. T.

(2011). The psychology of music performance anxiety. Oxford University Press.

28.

Kenny

D. T.

Arthey

Abbass

(2014). Intensive short-term dynamic psychotherapy for severe music performance anxiety: Assessment, process, and outcome of psychotherapy with a professional orchestral musician. Medical Problems of Performing Artists, 29(1), 3–7.

29.

Kenny

D. T.

Halls

(2018). Development and evaluation of two brief group interventions for music performance anxiety in community musicians. Psychology of Music, 46(1), 66–83.

30.

Khalsa

S. B. S.

Cope

(2006). Effects of a yoga lifestyle intervention on performance-related characteristics of musicians: A preliminary study. Medical Science Monitor, 12(8), CR325–CR331.

31.

McCann

R. A.

Armstrong

C. M.

Skopp

N. A.

Edwards-Stewart

Smolenski

D. J.

June

J. D.

Abamukong

Reger

G. M.

(2014). Virtual reality exposure therapy for the treatment of anxiety disorders: An evaluation of research quality. Journal of Anxiety Disorders, 28(6), 625–631.

32.

McGinnis

A. M.

Milling

L. S.

(2005). Psychological treatment of musical performance anxiety: Current status and future directions. Psychotherapy: Theory, Research, Practice, Training, 42(3), 357–373.

33.

Montello

(1990). Utilizing music therapy as a mode of treatment for the performance stress of professional musicians. New York University.

34.

Montello

Coons

E. E.

Kantor

(1990). The use of music therapy as a treatment for musical performance stress. Medical Problems of Performing Artists, 5(1), 49–57.

35.

Nagel

J. J.

Himle

D. P.

Papsdorf

J. D.

(1989). Cognitive-behavioural treatment of musical performance anxiety. Psychology of Music, 17(1), 12–21.

36.

Nagel

J. J.

(2010). Treatment of music performance anxiety via psychological approaches: A review of selected CBT and psychodynamic literature. Medical Problems of Performing Artists, 25(4), 141–148.

37.

Opriş

Pintea

García-Palacios

Botella

Szamosközi

Ş.

David

(2012). Virtual reality exposure therapy in anxiety disorders: A quantitative meta-analysis. Depression and Anxiety, 29(2), 85–93.

38.

Orman

E. K.

(2003). Effect of virtual reality graded exposure on heart rate and self-reported anxiety levels of performing saxophonists. Journal of Research in Music Education, 51(4), 302–315.

39.

Osborne

M. S.

Franklin

(2002). Cognitive processes in music performance anxiety. Australian Journal of Psychology, 54(2), 86–93.

40.

Osborne

M. S.

Kenny

D. T.

(2005). Development and validation of a music performance anxiety inventory for gifted adolescent musicians. Journal of Anxiety Disorders, 19(7), 725–751.

41.

Osborne

M. S.

Kenny

D. T.

Cooksey

(2007). Impact of a cognitive-behavioural treatment program on music performance anxiety in secondary school music students: A pilot study. Musicae Scientiae, 11(2_suppl), 53–84.

42.

Parsons

T. D.

Rizzo

A. A.

(2008). Affective outcomes of virtual reality exposure therapy for anxiety and specific phobias: A meta-analysis. Journal of Behavior Therapy and Experimental Psychiatry, 39(3), 250–261.

43.

Patston

(1996). Performance anxiety in opera singers [Unpublished master’s thesis]. Sydney Conservatorium of Music, University of Sydney.

44.

Powers

M. B.

Emmelkamp

P. M.

(2008). Virtual reality exposure therapy for anxiety disorders: A meta-analysis. Journal of Anxiety Disorders, 22(3), 561–569.

45.

Rosenthal

(1979). The file drawer problem and tolerance for null results. Psychological Bulletin, 86(3), 638–641.

46.

Spahn

Walther

J. C.

Nusseck

(2015). The effectiveness of a multimodal concept of audition training for music students in coping with music performance anxiety. Psychology of Music, 44(4), 893–909.

47.

Stanton

H. E.

(1993). Research note: Alleviation of performance anxiety through hypnotherapy. Psychology of Music, 21(1), 78–82.

48.

Steptoe

(2001). Negative emotions in music making: The problem of performance anxiety. In Juslin

P. N.

Sloboda

J. A.

(Eds.), Music and emotion: Theory and research (pp. 291–307). Oxford University Press.

49.

Stern

J. R.

Khalsa

S. B. S.

Hofmann

S. G.

(2012). A yoga intervention for music performance anxiety in conservatory students. Medical Problems of Performing Artists, 27(3), 123–128.

50.

Sweeney

G. A.

Horan

J. J.

(1982). Separate and combined effects of cue-controlled relaxation and cognitive restructuring in the treatment of musical performance anxiety. Journal of Counseling Psychology, 29(5), 486–497.

51.

Valentine

E. R.

Fitzgerald

D. F.

Gorton

T. L.

Hudson

J. A.

Symonds

E. R.

(1995). The effect of lessons in the Alexander technique on music performance in high and low stress situations. Psychology of Music, 23(2), 129–141.

52.

Zyl

M. V.

(2020). The effects of virtual reality on music performance anxiety among university-level music majors. Visions of Research in Music Education, 35(1), 15.