Emotional–Familiarity processing fluency moderates the Anxiety–Resilience association: evidence from a Chinese Cross-Sectional survey

Abstract

Music is widely used as an emotion regulation resource, yet music preference is often conceptualized as a stable taste rather than as a context-sensitive processing tendency expressed under affective demand. Drawing on processing fluency and experiential engagement frameworks, the present study examined whether an experiential music preference composite—integrating emotional connectedness and familiarity—modulates the association between anxiety and psychological resilience. Across specifications, the anxiety–resilience association increased as MF_EF increased; this conditional pattern was most evident for state anxiety, whereas the incremental moderating contribution in trait-anxiety models was smaller after accounting for MF_04. These results suggest that emotionally meaningful and familiar music engagement may be better characterized as a conditional experiential resource that becomes salient under transient anxiety, rather than a uniform buffer. These findings refine feature-level accounts of music-based emotion regulation by distinguishing experiential engagement from structural preference processes, while also underscoring the correlational nature of the evidence and the need for longitudinal and cross-cultural validation.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-36988-z.

Introduction

Anxiety is a highly prevalent psychological condition affecting individuals across the lifespan and has become a major global public health concern^1,2. Its chronic course and high comorbidity with other emotional disorders not only compromise psychological resources and social functioning but may also undermine individuals’ capacity to cope with stress^3,4. Emerging evidence further suggests that individual differences in psychological resilience may shape both how anxiety is perceived and how it is regulated⁵.

Theoretically, anxiety can be conceptualized as two distinct yet related constructs: state anxiety and trait anxiety⁶. State anxiety refers to a transient emotional reaction to perceived threats or adverse events, typically characterized by heightened physiological arousal and subjective tension. In contrast, trait anxiety reflects a stable personality disposition, indicating a general tendency to experience anxiety across various contexts. Prior work indicates that these two forms of anxiety can differ in their structural correlates and may show differential responsiveness to intervention strategies⁷. Notably, state anxiety may operate less as a stable liability than as a situational signal, such that its psychological consequences depend on the availability and deployment of regulatory resources in the immediate context. Accordingly, its association with adaptive outcomes may be contingent rather than expressed as a uniform main effect. For this reason, to clarify the mechanisms through which music modulates anxiety, it is important to examine the pathways of both state and trait anxiety, thereby improving explanatory specificity and informing more targeted intervention approaches.

Bonanno defines psychological resilience as the capacity to maintain relatively stable and healthy psychological functioning in the face of adversity⁸. Rather than being a fixed trait, resilience is now widely recognized as a dynamic psychological process influenced by both internal and external factors⁹. In recent years, Bonanno and colleagues have proposed the concept of the “ resilience paradox” ⁴⁴,³² highlighting that although most individuals exhibit resilience following potentially traumatic events, accurately predicting who will demonstrate resilience remains extremely challenging. Together with the challenge model of resilience—which emphasizes that adaptive functioning can emerge through context-dependent responses to manageable stressors—these perspectives converge on the view that resilience develops through multiple pathways rather than any single stable predictor^10,11. From this standpoint, dispositional characteristics traditionally framed as risk factors (e.g., trait anxiety) may, under certain conditions, also be implicated in heterogeneous routes to resilience, reflecting the coexistence of heightened sensitivity, accumulated coping experience, or shared individual-difference structures.

Grounded in the multi-mechanism model of music-evoked emotion proposed by Juslin and Västfjäll (2008)¹², together with the Processing Fluency and Aesthetic Pleasure framework, music can elicit emotional responses through multiple, partially overlapping induction pathways^13,14. Some pathways rely primarily on relatively automatic auditory–perceptual processing and operate with high temporal immediacy (e.g., brainstem reflexes and processing fluency), whereas others depend more strongly on accumulated experience and learned associations (e.g., episodic memory and musical expectancy). Consequently, music-evoked emotions are often highly individualized (Supplementary Table S1).

Importantly, the present study does not conceptualize music preference as a global or trait-like disposition (e.g., stable tastes for genres or styles). Instead, it focuses on music feature preferences, operationalized as feature-specific processing tendencies—that is, individual differences in how readily particular musical attributes are processed as perceptually or affectively fluent under specific situational demands. From this perspective, these tendencies are dispositional yet context-sensitive: they are neither fixed personality traits nor momentary psychological states, but processing propensities that may be differentially expressed depending on emotional demands.

In the present study, two experiential music feature preferences—preference for emotionally connected music (MF_12) and preference for familiar music (MF_13)—were integrated into a composite index (MF_EF), operationalized as the mean of the two components. Emotional connectedness refers to engagement with music perceived as personally meaningful and emotionally resonant, often linked to autobiographical associations and self-relevant memory processes. Familiarity reflects repeated exposure to specific musical material and is closely related to perceptual processing fluency and learned expectancy formed through prior listening experience. Although these dimensions are theoretically distinguishable, they frequently co-occur in everyday listening—particularly when music is used for emotion regulation or stress coping—providing a pragmatic rationale for modeling their shared variance via MF_EF.

Accordingly, MF_EF was conceptualized as an affectively salient music feature preference, reflecting a tendency to engage with music that is both emotionally resonant and predictably structured. This composite index captures feature-level engagement that affords autobiographical relevance alongside cognitive fluency, which may support efficient emotional engagement under psychologically demanding conditions. Importantly, this form of engagement is not assumed to confer uniform protective effects across emotional states. Rather, its functional relevance is expected to be most pronounced under heightened affective demand, when meaning-oriented engagement may help integrate transient arousal into adaptive regulatory processes. In the present framework, we examined MF_EF as an individual-difference moderator of the association between anxiety (state and trait) and psychological resilience. Specifically, we tested whether MF_EF is associated with differential anxiety–resilience coupling, consistent with the view that emotionally meaningful and familiar music may facilitate adaptive affect regulation. To separate this experiential modulation from lower-level auditory–structural preferences, melodic harmony preference (MF_04) was included as a covariate in subsequent analyses.

To account for lower-level auditory–perceptual fluency, melodic harmony preference (MF_04) served as a control variable. MF_04 reflects sensitivity to harmonically regular and structurally coherent musical features, which are typically processed with high perceptual ease and minimal reliance on autobiographical meaning. Such regularities primarily draw on tonal expectancy and perceptual congruence rather than individualized emotional associations. Although harmonic fluency may shape the general affective tone of listening, it is comparatively less personalized and less contingent on situational emotional demands.

Building on this framework, the present study dissociated basic auditory processing tendencies from affectively salient music engagement by modeling MF_EF as an experiential music feature preference factor while controlling for MF_04. We further examined whether MF_EF relates differently to resilience in the context of state versus trait anxiety, thereby offering a more nuanced account of how personalized music engagement may support psychological adaptation (see Fig. 1).

Fig. 1.

Flow chart of the study. MF = Music Feature Preference, EF = emotional-familiarity processing fluency MF_04= preference for melodic harmony, MF_12= preference for emotionally connected music, MF_13 = preference for familiar music, MF_EF = integrated emotional-familiarity processing fluency = (MF_12 + MF_13)/2SA=State Anxiety, TA= Trait Anxiety and Res=Resilience.

Methods

Participants

A total of 407 participants were recruited between March 8 and March 12, 2025, using an online survey administered via Wenjuanxing (Changsha Ranxing IT Ltd.). The study protocol was approved by the Institutional Review Board of Sahmyook University (IRB approval code: SYU 2025-03-018-001). All procedures complied with the Declaration of Helsinki and relevant institutional guidelines and regulations. Electronic informed consent was obtained from all participants prior to survey completion. Using a cross-sectional design, we examined whether facets of music feature preference—melodic harmony, emotional connectedness, and familiarity—were associated with psychological resilience in relation to both state and trait anxiety in a Chinese sample.

Research design

This study employed a cross-sectional survey design to test whether individual differences in music feature preferences moderate the association between anxiety and psychological resilience. A composite index, MF_EF, was computed as the mean of preference for emotionally connected music (MF_12) and preference for familiar music (MF_13) to represent affectively salient, familiarity-based music engagement. Moderation analyses were conducted using SPSS PROCESS (Model 1) to examine whether MF_EF moderates the relationship between state anxiety (SA) or trait anxiety (TA) and psychological resilience (Res). To account for lower-level auditory–structural preference, preference for melodic harmony (MF_04) was included as a covariate. All continuous variables were mean-centered prior to analysis. Interaction effects were tested using bias-corrected bootstrapping with 5,000 resamples. Separate models were estimated for SA and TA to evaluate potential differences between state anxiety and trait anxiety.

Inventory

State-Trait anxiety Inventory, STAI

Anxiety levels were assessed using the Chinese version of the State–Trait Anxiety Inventory (STAI), originally developed by Spielberger et al. (1964) and later translated and revised by Tsoi et al. (1986). In the present study, the state and trait anxiety subscale was employed, consisting of 40 items designed to evaluate individuals’ general tendency to experience persistent anxiety in daily life. Each item was rated on a 4-point Likert scale. The internal consistency of the trait anxiety subscale in this study was acceptable, with a Cronbach’s alpha coefficient of 0.71^15,16.

Brief resilience Scale, BRS

The Brief Resilience Scale (BRS), originally developed by Smith et al. (2008), was used to assess psychological resilience. A simplified Chinese version, adapted from the traditional Chinese translation validated by Tu et al. (2017), was employed in this study. The BRS consists of six items reflecting a unidimensional structure, designed to measure the ability to recover from stress and adversity. Responses were recorded on a 5-point Likert scale. In the present study, the scale demonstrated acceptable internal consistency, with a Cronbach’s alpha coefficient of 0.79^17,18.

The musical feature preference scale

The Musical Feature Preference Scale (MFPS) was developed using a theory-driven, literature-grounded approach to capture auditory attributes that have been linked to emotion regulation and therapeutic listening. Thirteen candidate items were generated from an integrative mapping of 30 peer-reviewed publications spanning music psychology, affective neuroscience, and music therapy. Building on established frameworks and recent meta-analytic evidence, the item pool covered tempo, timbre, intensity/volume, melodic–harmonic structure, rhythmic regularity, spatial/acoustic properties, and experiential resonance (emotional connection and familiarity). Participants rated each item on a 6 - point Likert scale indicating the extent to which each feature characterized the music they would prefer when experiencing stress.

Content validity was evaluated by mapping each item to at least three sources and cross-checking the mappings by two trained coders. The MFPS was administered to 407 participants. Internal consistency was acceptable (Cronbach’s α = 0.78), and corrected item–total correlations (CITC) exceeded 0.30 for most items. Dimensionality was examined using exploratory factor analysis (EFA) with sampling adequacy confirmed (KMO = 0.853; Bartlett’s χ²(78) = 1286.62, p <.001). A three-component solution accounted for 53.13% of the total variance and item communalities ranged from 0.34 to 0.69 (Supplementary Table S2_1- S2_2, Fig. S1 - S2).

MF_12 (“preference for emotionally connected music”) and MF_13 (“preference for familiar music”) showed strong loadings on the same factor (loadings = 0.780 and 0.662) and satisfactory corrected item–total correlations (CITC = 0.344 and 0.361), indicating shared variance consistent with affect-laden, experience-based processing fluency (Supplementary Tables S2_1 and S2_3). Accordingly, we operationalized their aggregation as a composite moderator (MF_EF), reflecting an emotional–familiarity fluency tendency that captures the co-occurrence of autobiographical relevance and predictability/processing ease in everyday listening.

MF_04 (loading = 0.672) also showed a substantial CITC (0.53), providing psychometric support for its relevance to melodic–harmonic appraisal (Supplementary Tables S2_1 and S2_3). Conceptually, however, MF_04 primarily indexes sensitivity to structural musical regularity rather than subjective emotional–autobiographical fluency. We therefore treated MF_04 as a theoretically motivated covariate to partial out lower-level melodic–structural preference when estimating the moderating role of MF_EF in the anxiety–resilience association.

Factorial stability across split samples was evaluated using Tucker’s congruence coefficients (φ). Structure-based coefficients indicated high cross-sample similarity for the best-matching factor pairs (e.g., A1–B2 = 0.9214; A3–B1 = 0.9697), consistent with the fact that factor order may permute across subsamples (Supplementary Table S2_4 - S2_5). Taken together, these results support the downstream modeling strategy in which MF_EF is specified as a moderator of the associations between state anxiety (SA)/trait anxiety (TA) and psychological resilience (Res), while MF_04 is included as a covariate to isolate experiential fluency effects from basic melodic–structural preference.

Statistical analyses

Descriptive statistics for categorical demographic variables were summarized as frequencies and percentages (n, %), whereas continuous variables—including state anxiety (SA), trait anxiety (TA), and psychological resilience (Res)—were summarized using means and standard deviations. Group differences across demographic categories (gender, age, economic status, education, marital status, daily music-listening time, and music product purchasing behavior) were examined using independent-samples t tests and one-way analyses of variance (ANOVAs). All analyses were conducted in IBM SPSS Statistics (Version 31.0).

To test the hypothesized moderation framework, regression-based analyses were conducted to examine whether an experiential music preference composite (MF_EF)—computed as the mean of preference for emotionally connected music (MF_12) and preference for familiar music (MF_13)—moderates the association between anxiety and psychological resilience. Separate moderation models were estimated for state anxiety (SA) and trait anxiety (TA), with Res specified as the outcome variable. Preference for melodic harmony (MF_04) was included as a covariate to account for lower-level auditory–structural preference and to isolate the experiential component captured by MF_EF. All continuous variables were mean-centered prior to analysis, and interaction effects were evaluated using bias-corrected bootstrapping with 5,000 resamples (PROCESS Model 1).

Results

Descriptive statistics

A total of 407 participants were included in the final analyses, and demographic characteristics are summarized in Table 1. The sample comprised 168 males (41.3%) and 239 females (58.7%). Most participants were aged 18–35 years (72.2%), and the majority reported a bachelor’s degree or higher. Most participants reported a middle economic status (78.6%), and more than half were married (56.3%). Daily music listening was common, with 54.1% reporting 30–60 min per day; additionally, 78.1% reported having a music application subscription.

Table 1.

Demographic characteristics of the study sample (N = 407). Values are presented as number of participants (n) and percentage (%). Percentages May not sum to 100 due to rounding. “Other” marital status reflects self-reported categories not classified as married or single.

Variable	Category	n	%
Gender	Male	168	41.3
	Female	239	58.7
Age Group	18 ~ 25	154	37.8
	26 ~ 35	140	34.4
	36 ~ 45	96	23.6
	≥ 46	17	4.2
Education Level	High school or below	11	2.7
	Associate degree	34	8.4
	Bachelor’s degree	336	82.6
	Master’s degree or above	26	6.4
Economic Status	Low	62	15.2
	Middle	320	78.6
	High	25	6.1
Marital Status	Married	229	56.3
	Single	175	43
	Other	3	0.7
Daily Music Listening Time	≤ 30 min	67	16.5
	30–60 min	220	54.1
	60–120 min	88	21.6
	≥ 120 min	32	7.9
Music Application Subscription	Subscriber	318	78.1
	Non-subscriber	89	21.9

Because demographic variables were not central to the primary hypotheses, inferential subgroup comparisons were not emphasized. Subsequent analyses therefore focused on regression-based models examining the associations among anxiety, experiential music preference, and psychological resilience.

Regression model results

To test whether experiential music preference moderates the associations between anxiety and psychological resilience, we conducted regression-based moderation analyses using PROCESS Model 1. State anxiety (SA) and trait anxiety (TA) were examined in separate models, with psychological resilience (Res) specified as the outcome. Experiential music preference (MF_EF) was entered as the moderator, and preference for melodic harmony (MF_04) was included as a covariate to align with the reported models. All continuous predictors were mean-centered prior to estimating the interaction terms. Given the small number of predictors, collinearity diagnostics were reported for completeness rather than as a primary inferential focus. As shown in Table S3, all variance inflation factors were low (VIFs ≤ 1.264), indicating no evidence of problematic multicollinearity.

Resilience showed a mean of 3.24 (SD = 0.38), while state anxiety (M = 2.32, SD = 0.26) and trait anxiety (M = 2.38, SD = 0.30) were comparable in magnitude. MF_EF was mean-centered (M = 0, SD = 0.84), and melodic harmony preference (MF_04) exhibited greater variability (M = 2.71, SD = 1.23). Zero-order correlations indicated that psychological resilience was positively associated with state anxiety, trait anxiety, MF_EF, and MF_04 (rs = 0.17–0.29, ps < 0.01). State and trait anxiety were strongly correlated (r =.61, p <.01). MF_EF also showed a moderate correlation with MF_04 (r =.44, p <.01), indicating meaningful shared variance between experiential fluency–related engagement and melodic–harmonic preference. Distributional indices suggested mild to moderate positive skewness across variables (skewness = 0.36–0.90), with kurtosis values generally modest, although state anxiety showed comparatively higher kurtosis (2.20). Given the overlap between MF_EF and MF_04, subsequent models adjusted for MF_04 to estimate the incremental contribution of MF_EF beyond shared structural-preference variance (Supplementary Table S4_1, S4_2).

Prior to hypothesis testing, key regression diagnostics were examined. Visual inspection of standardized residuals plotted against standardized predicted values revealed no substantial deviations from homoscedasticity or linearity, and Durbin–Watson statistics were close to 2 (SA = 1.895; TA = 1.871), providing no indication of problematic residual autocorrelation (Supplementary Table S4_3; Figs. S3–S4).

Primary moderation models (without covariates)

Both overall regression models were statistically significant. In the state anxiety model, the predictors explained 6.8% of the variance in psychological resilience (R² = 0.068, F(3, 403) = 9.82, p <.001). In the trait anxiety model, a larger proportion of variance was accounted for (R² = 0.118, F(3, 403) = 18.03, p <.001). Trait anxiety showed a significant positive association with psychological resilience (B = 0.31, SE = 0.06, p <.001), whereas the main effect of state anxiety did not reach conventional significance (B = 0.14, SE = 0.07, p =.063). Importantly, MF_EF moderated the anxiety–resilience association in both models (SA × MF_EF: B = 0.19, SE = 0.08, p =.017; TA × MF_EF: B = 0.13, SE = 0.07, p =.043； Table 2).

Table 2.

Moderation analyses MF_EF as a moderator of the associations between anxiety and psychological resilience.

Predictor	State Anxiety Model			Trait Anxiety Model
	B	SE	p	B	SE	p
Constant	3.23	0.18	< 0.001	3.23	0.18	< 0.001
Anxiety (SA/TA)	0.14	0.07	0.063	0.31	0.06	< 0.001
MF_EF	0.06	0.02	0.006	0.06	0.02	0.006
Anxiety × MF_EF	0.19	0.08	0.017	0.13	0.07	0.043
Model fit
R²	0.068			0.118
F(df)	9.82 (3,403)		< 0.001	18.03 (3,403)		< 0.001

Conditional effects analyses indicated that the association between anxiety and psychological resilience strengthened as MF_EF increased. In the state anxiety model, the conditional effect of SA on resilience was not significant at low MF_EF (− 1 SD; B = − 0.02, p =.846) and remained marginal at the mean level (B = 0.14, p =.063), but became positive and significant at high MF_EF (+ 1 SD; B = 0.30, p <.001). Johnson–Neyman results further showed that the SA–resilience association was statistically significant only when MF_EF exceeded approximately 0.03, corresponding to 34.89% of the sample (Table 3; Figs. 2 and 3).

Table 3.

Conditional effects of anxiety on psychological resilience across levels of MF_EF.

Anxiety type	MF_EF level/Region	Effect (B)	SE	t	p	LLCI	ULCI	Johnson–Neyman information
State anxiety (SA)	−1 SD	−0.02	0.11	−0.19	0.85	−0.24	0.20
	Mean	0.14	0.07	1.86	0.06	−0.01	0.28
	+ 1 SD	0.30	0.08	3.57	< 0.001	0.13	0.46
	JN threshold	—	—	—	—	—	—	MF_EF ≈ 0.03
	% sample above JN	—	—	—	—	—	—	34.89%
Trait anxiety (TA)	−1 SD	0.20	0.09	2.17	0.03	0.02	0.38
	Mean	0.31	0.06	5.05	< 0.001	0.19	0.43
	+ 1 SD	0.43	0.07	5.88	< 0.001	0.28	0.57
	JN threshold	—	—	—	—	—	—	MF_EF ≈ − 0.93
	% sample above JN	—	—	—	—	—	—	91.15%

Fig. 2.

MF_EF differentially moderates the associations between state and trait anxiety and psychological resilience.

Fig. 3.

Conditional effect of state anxiety on psychological resilience as a function of MF_EF (Johnson–Neyman plot).

Panel A shows the interaction between state anxiety (mean-centered) and MF_EF in predicting psychological Res. Predicted Res is plotted across mean-centered SA at low (− 1 SD), mean, and high (+ 1 SD) levels of MF_EF. Shaded bands indicate 95% confidence intervals. The SA–Res association varies by MF_EF, consistent with a positive SA×MF_EF interaction (B = 0.1886, p =.0172). Labeled points show predicted Res at SA = 0 and SA = 0.50 for each MF_EF level.

Panel B depicts the parallel interaction for trait anxiety. Predicted Res is plotted across mean-centered TA at low (− 1 SD), mean, and high (+ 1 SD) levels of MF_EF. Shaded bands indicate 95% confidence intervals. Labeled points show predicted Res at TA = 0 and TA = 0.50 for each MF_EF level.

Together, the side-by-side panels make the state–trait asymmetry visually explicit: MF_EF acts as a gated experiential resource for state anxiety, selectively converting higher state anxiety into higher resilience at high MF_EF, whereas its moderating role in trait anxiety is more diffuse and less contingent.

In the trait anxiety model, TA showed a positive association with resilience across all probed levels of MF_EF (− 1 SD: B = 0.20, p =.03; mean: B = 0.31, p <.001; +1 SD: B = 0.43, p <.001). Consistent with this pattern, the Johnson–Neyman threshold occurred at MF_EF ≈ − 0.93, indicating that the conditional effect was significant for 91.15% of participants (Table 3; Figs. 2 and 4).

Fig. 4.

Conditional effect of trait anxiety on psychological resilience as a function of MF_EF (Johnson–Neyman plot).

Supplementary models controlling for melodic harmony preference (MF_04)

To evaluate whether the moderation effects of MF_EF reflected variance attributable to lower-level melodic–harmonic preference, supplementary moderation models were estimated with MF_04 included as a covariate. Both adjusted models remained statistically significant (state anxiety model: R² = 0.083, F(4, 402) = 9.11, p <.001; trait anxiety model: R² = 0.129, F(4, 402) = 14.90, p <.001). In these models, MF_04 showed a small positive association with resilience (state model: B = 0.04, SE = 0.02, p =.011; trait model: B = 0.04, SE = 0.02, p =.030). Controlling for MF_04 did not eliminate the moderation effect in the state anxiety model: the SA × MF_EF interaction remained statistically significant (B = 0.16, SE = 0.08, p =.045), whereas the corresponding interaction in the trait anxiety model was attenuated to marginal significance (B = 0.12, SE = 0.07, p =.060; Supplementary Table S4_4; Fig. S5).

Johnson–Neyman analyses indicated that, in the state anxiety model, the conditional SA–resilience effect reached statistical significance only at relatively higher levels of MF_EF (JN threshold: MF_EF ≈ 0.09), with 34.89% of participants falling above this value. In the trait anxiety model, the conditional effect remained statistically significant across most of the observed MF_EF range (JN threshold: MF_EF ≈ − 0.89), corresponding to 91.15% of the sample above the threshold. Across both models, the probed conditional effects increased monotonically with MF_EF, consistent with a stronger positive anxiety–resilience association at higher MF_EF (Supplementary Table S5; Fig. S6–S7).

Discussion

A central contribution of the present analyses is the asymmetric pattern observed for state versus trait anxiety. The most counterintuitive result was not a uniform positive association between anxiety and resilience, but a conditional pattern specific to state anxiety^19,20. State anxiety was not a reliable main predictor of resilience²¹; rather, its association with resilience became more positive only at higher levels of MF_EF. This interaction remained statistically significant even after controlling for MF_04, suggesting that elevated state anxiety is not inherently adaptive, but may co-occur with higher resilience primarily when meaning- and familiarity-based modes of music engagement are more pronounced. In this sense, the pattern is more consistent with a resource-contingent association than with a generalized buffering effect^22,23.

Viewed through a process-oriented emotion regulation lens, this asymmetry is compatible with the notion that transient anxiety functions less as a stable liability than as a situational signal whose implications depend on the availability and deployment of regulatory resources^24,25. MF_EF may index a meaning-oriented form of music engagement that becomes particularly relevant under heightened arousal, whereas trait anxiety reflects a more enduring dispositional background. In the present sample, trait anxiety showed a positive association with resilience, consistent with the possibility of heterogeneous and layered pathways in which sensitivity can coexist with accumulated coping capacity, cumulative stress exposure, or shared underlying individual-difference structures²⁶. Correspondingly, the incremental moderating contribution of MF_EF to the trait-anxiety model appeared smaller, and was attenuated after adjustment for MF_04, suggesting partial overlap between experiential engagement and structural preference processes.

The selectivity of the state anxiety × MF_EF interaction also makes purely artifactual explanations less compelling, insofar as broad response biases would more typically be expected to yield more uniform associations across predictors. Instead, the relatively narrow Johnson–Neyman region of significance is consistent with a gated pattern in which the state anxiety–resilience association is evident primarily when MF_EF is elevated. Consistent with multi-mechanism accounts of music-evoked emotion, one interpretation is that state anxiety may be more tightly coupled to higher-order²⁷, experience-dependent engagement routes, whereas trait anxiety reflects a more generalized affective background that is less contingent on momentary engagement with such mechanisms²⁸. Although this pattern may appear superficially related to the resilience paradox, it is better characterized as a circumscribed state-level conditional association rather than a generalized coexistence of distress and functioning. Given the cross-sectional design, causal interpretations remain unwarranted; it is equally plausible that more resilient individuals preferentially engage with emotionally meaningful and familiar music when they experience anxiety^29,30.

To evaluate whether the moderating effect of MF_EF could be accounted for by lower-level structural preference, supplementary models included MF_04 as a covariate. MF_04 showed a small positive association with resilience, indicating that melodic–harmonic preference explains a modest proportion of variance in resilient functioning. Importantly, controlling for MF_04 did not eliminate the moderation observed in the state-anxiety model: the state anxiety × MF_EF interaction remained statistically significant, although attenuated, suggesting some shared variance between MF_EF and melodic–harmonic preference while retaining incremental explanatory value. In contrast, the trait anxiety × MF_EF interaction was attenuated to marginal significance after controlling for MF_04, consistent with greater overlap between experiential engagement and structural preference processes in the trait-anxiety model. This attenuation does not indicate that the trait anxiety–resilience association is absent; rather, it suggests that the distinct moderating contribution of MF_EF becomes less clearly separable once shared variance with MF_04 is taken into account.

Taken together, these findings suggest that MF_EF is better conceptualized as a context-sensitive regulatory resource than as a universal protective factor. Its moderating role appears most robust for transient state anxiety, where meaning-oriented engagement with emotionally meaningful and familiar music may be particularly consequential under heightened arousal. By contrast, at the dispositional level, experiential engagement appears to overlap more substantially with broader, stable individual-difference structures once melodic–harmonic preference is controlled³¹.

Limitations of study and recommendation

Although the present study examined associations among feature-specific music processing tendencies, anxiety, and psychological resilience in a relatively large sample using a theoretically motivated moderation framework, several limitations should be noted.

First, all variables were assessed via self-report, which is vulnerable to response biases (e.g., social desirability), limited introspective access, and recall error. This concern is particularly relevant for subjective constructs such as anxiety and resilience, as well as self-reported music-processing tendencies. Future work would benefit from multimethod assessment, including behavioral tasks, physiological indices (e.g., heart-rate variability, skin conductance), and/or neural measures, to more directly characterize regulatory processes and reduce shared-method variance.

Second, the sample consisted exclusively of adults from mainland China. Because music perception, feature salience, and affective associations are shaped by cultural learning and musical socialization, the psychological meaning and regulatory relevance of melodic harmony preference (MF_04), preference for familiar music (MF_13), and preference for emotionally connected music (MF_12)—and their composite experiential index (MF_EF)—may not generalize straightforwardly to other cultural contexts with different tonal conventions, listening practices, or musical repertoires. Although this cultural setting provides an informative context for evaluating feature-level accounts of music-related regulation, cross-cultural replication and formal measurement invariance testing will be necessary to determine whether the observed configuration-dependent associations reflect culture-general processes or culturally specific patterns.

In addition, the sample was relatively homogeneous in educational attainment, with 82.6% of participants holding a bachelor’s degree. Restricted variability in education and related socioeconomic factors may have constrained variance in psychological and experiential measures, potentially attenuating associations and limiting the detection of subgroup-specific patterns. Future work should therefore recruit samples with greater educational and socioeconomic diversity to better characterize individual differences in music-related processing tendencies and their links to anxiety and resilience.

Importantly, the present study did not evaluate a music-based intervention protocol. Accordingly, the findings should not be interpreted as direct evidence for clinical efficacy or as providing prescriptive guidance for applied intervention design. Instead, they are best viewed as generating theoretically informed, testable hypotheses regarding how distinct musical features and experiential engagement may relate to emotion regulation under varying anxiety conditions.

Building on these limitations, future research should employ longitudinal, experimental, and cross-cultural designs to test whether manipulating feature-level musical attributes (e.g., harmonic regularity, familiarity, autobiographical relevance) produces causal effects on anxiety regulation and resilience. Such work may also clarify whether the observed state–trait dissociation reflects differences in regulatory timescale, attributional processes, or learning dynamics. In this way, the present findings offer a conceptual basis for intervention-oriented research while remaining within the interpretive bounds of a correlational design.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Yinghua Jin, and Hongshan Liu, and Huan He designed the study. Data collection was led by Yinghua Jin and Hongshan Liu, analysed by Hongshan Liu and Huan He. The manuscript was written by Hongshan Liu and reviewed/revised by all authors.

Funding

The Authors received NO FUNDING for this work.

Data availability

The datasets generated and analyzed during the current study are available in the Zenodo repository at https://doi.org/10.5281/zenodo.16916403.

Declarations

Competing interests

The authors declare no competing interests.

Data statement

The data presented in this study are available upon request from the corresponding author (Dr. Yinghua Jin, Email: 925babyshan@gmail.com).