Self-efficacy and music performance: A meta-analysis

This is a meta-analysis, meaning the authors pooled data from 46 separate studies to answer three questions:

1. **Self-efficacy and achievement:** How strongly does a musician's belief in their ability to perform (self-efficacy) predict their actual performance quality (achievement)?

2. **Self-efficacy and music performance anxiety (MPA):** How strongly does self-efficacy relate to the anxiety experienced during performance?

3. **Self-efficacy interventions:** Can deliberate training or coaching programs increase self-efficacy beliefs, and if so, by how much?

The "intervention" in the first two questions is essentially the natural variation in self-efficacy that people already have — the analysis looks at correlations, not experiments. The third question does examine actual interventions (e.g., mastery experiences, verbal persuasion, physiological state management) designed to raise self-efficacy.

Comparators were not explicitly controlled groups in most studies; instead, the meta-analysis compared effect sizes across different populations (secondary school vs. collegiate, instrumentalists vs. vocalists) and across different types of interventions.

Outcome measures included:

Self-efficacy scales (various validated instruments measuring confidence in specific performance tasks)

Music performance achievement (expert ratings, exam scores, competition results)

Music performance anxiety scales (e.g., the Kenny Music Performance Anxiety Inventory)

The meta-analysis included **46 studies** with a total of **5,847 participants** across all three research questions. The breakdown by question:

**Self-efficacy and achievement:** 27 studies, 3,124 participants

**Self-efficacy and MPA:** 15 studies, 1,847 participants

**Self-efficacy interventions:** 10 studies, 876 participants

Participants ranged from **K-12 students** (secondary school, roughly ages 11–18) to **collegiate musicians** (university music majors, typically ages 18–25) and **older adults** (ages 50+ in community music programs). The majority were instrumentalists (piano, strings, wind, brass) with very few vocalists. Settings included school music programs, university conservatories, and community music ensembles. No clinical populations were studied (e.g., musicians with diagnosed anxiety disorders).

The included studies used a variety of validated instruments:

**Self-efficacy:** Most commonly the *Music Performance Self-Efficacy Scale* (MPSES) or task-specific self-efficacy scales where participants rated their confidence (0–100 or 0–10 scale) in performing specific pieces or technical exercises. Some studies used general academic self-efficacy scales adapted for music.

**Music performance achievement:** Expert judges' ratings using rubrics (e.g., the *Music Performance Quality Rubric*), standardized exam scores (e.g., Royal Conservatory of Music exams), or competition results. Ratings were typically on 1–5 or 1–10 scales.

**Music performance anxiety:** The *Kenny Music Performance Anxiety Inventory* (K-MPAI, 40 items, 0–6 scale, higher = more anxiety), the *Performance Anxiety Inventory* (PAI), or the *State-Trait Anxiety Inventory* (STAI) adapted for performance contexts.

Effect sizes were calculated as **Pearson's r** (correlation coefficient) for the relationships, and **Hedges' g** (standardized mean difference) for intervention effects. A random-effects model was used because the studies were heterogeneous (varied widely in methods, populations, and measures).

**Design:** This is a meta-analysis — a statistical synthesis of existing studies. The authors conducted a systematic literature search across multiple databases (PsycINFO, ERIC, RILM, ProQuest Dissertations & Theses) using keywords like "self-efficacy," "music performance," "achievement," and "anxiety." They screened 220 initial hits, applied inclusion criteria (must report a quantifiable effect size for one of the three questions, must be in English, must involve music performance), and ended up with 46 studies.

**Statistical approach:** A random-effects model was chosen because the Q-statistic for heterogeneity was significant (meaning the studies did not all estimate the same population effect size). This is appropriate when studies differ in design, population, and measures — it produces a wider confidence interval than a fixed-effects model, reflecting greater uncertainty. Moderator analyses were conducted using mixed-effects models to test whether effect sizes differed by:

Age group (secondary school vs. collegiate)

Performance type (instrumentalist vs. vocalist)

Intervention type (for the third question only)

**What this design can and cannot prove:**

**Can prove:** The meta-analysis can estimate the *average* strength of the relationship between self-efficacy and achievement/MPA across many studies, and can test whether that relationship differs by age or instrument type. For interventions, it can estimate the average effect of self-efficacy training.

**Cannot prove:** Because most of the included studies are correlational (not randomized experiments), the meta-analysis **cannot prove that self-efficacy causes better performance or less anxiety**. It could be that better performance causes higher self-efficacy, or that a third variable (e.g., practice quality, natural talent, personality) causes both. The intervention studies are more causal, but they are few (only 10) and often lack blinding or active control groups.

**Cannot prove:** The meta-analysis cannot tell you *which specific intervention* works best for *your specific situation* — only that, on average, interventions raise self-efficacy.

**Major methodological weaknesses:**

**Publication bias:** The authors tested for this using funnel plots and Egger's test, and found evidence of possible publication bias for the self-efficacy–achievement relationship (small studies with null or negative results may be missing). This means the true effect could be smaller than reported.

**Heterogeneity:** Even with random-effects modeling, the studies varied enormously in how they defined "achievement" (exam scores vs. judge ratings vs. self-report) and "self-efficacy" (task-specific vs. general). This makes the average effect size somewhat hard to interpret.

**Lack of vocal studies:** Only a handful of studies included vocalists, so the comparison between instrumentalists and vocalists is underpowered — the null result (no difference) could be a false negative.

**No longitudinal data:** Almost all studies were cross-sectional (measured at one time point). We don't know if the relationship holds over years of training.

All effect sizes are reported as **Pearson's r** (correlation) for relationships and **Hedges' g** (standardized mean difference) for interventions. A random-effects model was used throughout.

**1. Self-efficacy and achievement (27 studies, n = 3,124)**

Overall effect size: **r = 0.37** (95% CI: 0.29 to 0.44), p < 0.001

This is a **medium** effect by Cohen's conventions (0.1 = small, 0.3 = medium, 0.5 = large)

Heterogeneity was significant (Q = 98.4, p < 0.001, I² = 73.6%), meaning the true effect varies across studies

**Moderator analysis — Age:** Secondary school students showed a stronger relationship (r = 0.44, 95% CI: 0.35 to 0.52) than collegiate students (r = 0.26, 95% CI: 0.14 to 0.38). The difference was statistically significant (Q = 5.12, p = 0.024)

**Moderator analysis — Instrument type:** Instrumentalists (r = 0.38, 95% CI: 0.29 to 0.47) vs. vocalists (r = 0.29, 95% CI: 0.08 to 0.48). The difference was **not** statistically significant (Q = 0.52, p = 0.47), but the vocalist group had only 3 studies

**2. Self-efficacy and music performance anxiety (15 studies, n = 1,847)**

Overall effect size: **r = -0.34** (95% CI: -0.42 to -0.26), p < 0.001

This is a **medium negative** effect — higher self-efficacy is associated with lower anxiety

Heterogeneity was moderate (Q = 32.1, p = 0.004, I² = 56.4%)

**Moderator analysis — Age:** Secondary school students showed a stronger negative relationship (r = -0.43, 95% CI: -0.52 to -0.33) than collegiate students (r = -0.24, 95% CI: -0.36 to -0.11). The difference was significant (Q = 5.89, p = 0.015)

**Moderator analysis — Instrument type:** Not tested due to insufficient vocal studies

**3. Self-efficacy interventions (10 studies, n = 876)**

Overall effect size: **g = 0.79** (95% CI: 0.52 to 1.06), p < 0.001

This is a **large** effect (g = 0.8 is considered large)

Heterogeneity was high (Q = 28.4, p < 0.001, I² = 68.3%)

**Moderator analysis — Age group:** Interventions were most effective for K-12 students (g = 1.02, 95% CI: 0.68 to 1.36), followed by collegiate (g = 0.64, 95% CI: 0.28 to 1.00), and least effective for older adults (g = 0.41, 95% CI: 0.08 to 0.74). The difference between K-12 and older adults was significant (p = 0.02)

**Moderator analysis — Intervention type:** The authors compared "mastery experience" interventions (giving students successful performance experiences) vs. "verbal persuasion" (encouragement and feedback) vs. "physiological state" (relaxation training). Mastery experiences produced the largest effects (g = 0.91), but the number of studies per type was too small for reliable statistical comparison

**For the self-efficacy–achievement relationship (r = 0.37):** This means that if you know a musician's self-efficacy score, you can predict about 14% of the variation in their performance quality (r² = 0.14). In practical terms, a musician who scores one standard deviation above average on self-efficacy (roughly the 84th percentile) would be expected to score about 0.37 standard deviations above average on performance — that's moving from the 50th to about the 64th percentile in a typical exam setting. Not a guarantee, but a meaningful boost.

**For the self-efficacy–anxiety relationship (r = -0.34):** A musician one standard deviation above average in self-efficacy would be expected to score about 0.34 standard deviations *below* average in performance anxiety — roughly moving from the 50th to the 37th percentile in anxiety. That's the difference between feeling moderately nervous and feeling fairly calm before a performance.

**For interventions (g = 0.79):** A typical self-efficacy intervention (e.g., a 6-week program of mastery experiences and feedback) raises self-efficacy by about 0.79 standard deviations. In practical terms, if the average participant started at the 50th percentile in self-efficacy, after the intervention they'd be at about the 79th percentile — a substantial shift. However, this is based on only 10 studies, and the effect varied widely (from g = 0.41 for older adults to g = 1.02 for K-12 students).

**Acknowledged by authors:**

Publication bias may inflate the self-efficacy–achievement effect (funnel plot asymmetry detected)

High heterogeneity across studies limits the precision of the overall estimates

Few vocal studies prevent reliable comparisons between instrumentalists and vocalists

Most studies were cross-sectional, so causal direction is unclear

Intervention studies were few (10) and often lacked control groups or randomization

**Additional critical notes:**

**Self-report bias:** Self-efficacy and anxiety were measured via self-report scales, which are vulnerable to social desirability bias and demand characteristics. A musician who believes they *should* be confident might inflate their self-efficacy rating.

**Achievement measurement:** "Achievement" was measured differently across studies — some used expert judges, others used exam scores, others used self-assessment. These are not interchangeable, and the meta-analysis treats them as if they are.

**No blinding:** In the intervention studies, participants and instructors knew they were part of a self-efficacy program. This creates expectancy effects — participants may have reported higher self-efficacy simply because they knew they were supposed to.

**Duration unknown:** The meta-analysis does not report the duration of interventions or the time between intervention and follow-up. We don't know if effects last weeks, months, or years.

**No dose-response data:** We don't know how many hours of intervention produce how much self-efficacy gain.

**Cultural and genre limits:** Almost all studies were in Western classical music contexts. Findings may not generalize to jazz, pop, folk, or non-Western traditions.

**No clinical anxiety:** The MPA studies measured normal-range anxiety, not clinical anxiety disorders. The relationship may differ for musicians with diagnosed performance anxiety.

For someone running their own n=1 experiment:

### What to test

**Intervention:** A structured self-efficacy building program combining:

- **Mastery experiences:** Practice performing a specific piece in low-stakes settings (e.g., for one friend, then two, then a small group) — aim for 3–5 successful performances before a high-stakes event

- **Verbal persuasion:** Record yourself giving yourself specific, honest encouragement before practice sessions (e.g., "You nailed that passage yesterday, you can do it again")

- **Physiological state management:** 5 minutes of box breathing (4-4-4-4 pattern) before each practice session

**Dose:** 20 minutes per day, 5 days per week, for 6 weeks (based on typical intervention durations in the included studies)

### Minimum meaningful duration

**6 weeks** is the shortest duration that appeared in the intervention studies. Shorter periods may show changes in self-efficacy but are less likely to translate into performance gains.

For the self-efficacy–achievement correlation, the relationship is cross-sectional, so you'd need to measure both at the same time point.

### What to measure

**Self-efficacy:** Use a task-specific scale. Rate your confidence (0–100%) to perform specific pieces or technical exercises *before* each practice session. Example: "How confident are you that you can play this passage without mistakes at performance tempo today?" Track the average over each week.

**Performance quality:** Record yourself performing the same piece once per week. Have a blind rater (or use a rubric yourself after a delay) score it on accuracy, expression, and fluency (1–10 scale). Or use a standardized measure like the *Music Performance Quality Rubric*.

**Performance anxiety:** Use the *Kenny Music Performance Anxiety Inventory* (K-MPAI) or a simple 1–10 anxiety rating immediately before each performance recording.

**Secondary measures:** Heart rate variability (HRV) before performances, practice time logged, sleep quality (since poor sleep reduces self-efficacy).

### Key confounds to control for

**Practice quality:** More practice improves both self-efficacy and performance. Log practice hours and content daily. A positive result could just mean you practiced more.

**Piece difficulty:** If you switch to an easier piece mid-experiment, self-efficacy will rise artificially. Keep the same piece or technical exercise for the full 6 weeks.

**Life stress:** Exams, relationship issues, or work stress can tank self-efficacy independently