The Impact of a Self-Distanced Writing Intervention on Social Anxiety Among College Freshmen: The Mediating Roles of Cognitive Reappraisal and Self-Compassion

Abstract

Background

Social anxiety is common during the transition to university, creating a need for brief, accessible psychological interventions. Traditional expressive writing has shown benefits for anxiety, but its emotionally immersive first-person format may be less suitable for some students. Self-distanced writing offers an alternative by encouraging reflection from an observer-like perspective.

Purpose

We examined whether a four-week self-distanced writing intervention reduced social anxiety more than traditional expressive writing or neutral writing, and whether cognitive reappraisal and self-compassion would mediate any observed advantage.

Participants and Methods

In a three-arm, assessor-blind randomized controlled trial, 187 first-year students with Liebowitz Social Anxiety Scale–Self-Report (LSAS-SR) scores ≥38 were randomly assigned to self-distanced writing (SDW; n=63), traditional expressive writing (EW; n=62), or neutral writing (NW; n=62). Each group completed eight supervised 20-minute sessions over four weeks. We assessed social anxiety, cognitive reappraisal, and self-compassion at baseline, one week after the intervention, and at the one-month follow-up.

Results

Social anxiety decreased over time, with different trajectories across groups. SDW did not differ significantly from EW at the one-week post-intervention assessment, but at the one-month follow-up it showed greater reductions in social anxiety than both EW and NW. Increased self-compassion helped explain the difference between SDW and EW, whereas cognitive reappraisal did not. Neither process helped explain the difference between SDW and NW.

Conclusion

A four-week SDW intervention was associated with greater reductions in social anxiety at the one-month follow-up. The findings suggest that self-compassion may be particularly relevant to how SDW differs from traditional expressive writing. Longer follow-up and further evaluation in university settings are needed.

Introduction

The transition to university represents a critical developmental period marked by substantial psychosocial demands, including adaptation to a new environment, the formation of new social relationships, and changes in self-identity.^1,2 Against this broader vulnerability to internalizing difficulties, social anxiety is particularly relevant because university life repeatedly exposes students to socially evaluative situations, such as classroom participation, peer interaction, presentations, and the formation of new interpersonal networks. Social anxiety disorder is characterized by marked fear of negative evaluation and by avoidance of social situations or enduring them with substantial anxiety and distress.^3–5 Meta-analytic evidence indicates that anxiety is a substantial mental health concern among college students in mainland China.⁶ More specifically, social anxiety also appears to be common among Chinese university students. In a cross-sectional study of 2759 college students, Meng et al reported that 33.38% experienced at least one social anxiety symptom.⁷ More recent research involving 1408 Chinese college freshmen found that approximately one-third met the study’s questionnaire-based screening threshold for social anxiety disorder.⁸ Beyond prevalence, longitudinal evidence suggests that higher social anxiety is associated with poorer academic achievement and weaker social ties across the university years.⁹ Social anxiety has also been associated with poorer academic communication, lower engagement and satisfaction, and poorer socio-emotional functioning in higher education.¹⁰ These findings highlight the need for brief and accessible approaches that may help students manage socially evaluative distress during the transition to university.

Expressive writing (EW) may represent one such low-intensity approach because it provides a structured way for individuals to engage with and reflect on distressing experiences. Writing-based interventions have already been examined in college populations, although the findings have been mixed. Among college students with mild social anxiety, Li and Liu found that a single 15-minute expressive-writing exercise improved several indicators of subsequent social performance, including the amount and rate of speech and the use of positive emotion words, relative to a control writing condition.¹¹ In a randomized trial involving first-year college students, Robertson et al found that anxiety decreased over time in both expressive-writing and non-emotional-writing conditions; within the expressive-writing group, linguistic features were associated with the magnitude of anxiety reduction.¹² A meta-analysis of 146 randomized experimental-disclosure studies similarly reported a small positive average effect, together with substantial variation according to study and participant characteristics.¹³ These findings suggest that writing can provide a feasible means of engaging students with stressful experiences, while also indicating that its effects may depend partly on how those experiences are processed rather than on emotional disclosure alone.

Psychological perspective may be one factor shaping this processing. Traditional EW typically emphasizes emotional disclosure and reflection on stressful experiences.^11–13 Research comparing immersed and distanced analysis suggests that remaining immersed in negative experiences may foster repetitive or maladaptive processing, whereas adopting a distanced perspective can facilitate more adaptive reflection.^14,15 Related work has also extended the study of psychological perspective to help-seeking and narrative engagement.^16,17 Self-distancing offers a theoretically distinct approach by asking individuals to consider their experiences from a psychologically removed, observer-like perspective, such as through a third-person perspective.¹⁸ Research on distancing and reinterpretation suggests that distancing can reduce emotional reactivity during regulation and support more analytical forms of emotional processing.^19,20 Linguistic perspective also appears relevant within writing, as changes in pronoun use have been associated with emotional change during expressive-writing tasks.²¹ Experimental and review evidence further indicates that brief manipulations of a self-distanced perspective can promote more adaptive reflection on negative experiences.^22,23 Park et al extended this work by showing across two longitudinal experiments that expressive writing promoted subsequent self-distancing, which was associated with lower emotional reactivity over time.²⁴ Their work, however, examined self-distancing primarily as a process that emerged from expressive writing rather than as a deliberately trained and repeated writing strategy targeting social anxiety. This distinction provides a rationale for directly incorporating self-distancing instructions into a structured writing intervention.

Cognitive reappraisal and self-compassion provide two plausible psychological processes through which self-distanced writing (SDW) may relate to social anxiety. Cognitive reappraisal involves changing the interpretation of an emotional situation in order to alter its emotional impact and is an important component of cognitive-behavioral approaches to social anxiety.²⁵ Individuals with social anxiety may show difficulties in implementing reappraisal effectively. In adults with social anxiety disorder, less successful down-regulation of negative emotion during reappraisal has been associated with greater symptom severity,²⁶ while experimental evidence in children with anxiety disorders indicates that reappraisal can reduce subjective emotional reactivity.²⁷ Decentering, a related process of viewing thoughts and feelings as mental events rather than identifying with them, has also been associated with both cognitive reappraisal and lower social anxiety.²⁸ Together, these findings provide a theoretical basis for proposing that the psychological distance created by SDW may facilitate more balanced interpretations of socially threatening experiences.

Self-compassion offers a complementary process involving self-kindness, recognition of common humanity, and mindful awareness of distress. It has been inversely associated with social anxiety,²⁹ while fear of negative evaluation and shame have been proposed as processes linking lower self-compassion with greater social anxiety.³⁰ Other studies have likewise identified associations among self-compassion, emotion regulation, and social anxiety.^31,32 In first-year college students, Ștefan found that self-compassion statistically mediated associations between coping and subsequent social anxiety during the early months of adjustment to college.³³ A self-distanced perspective may therefore be relevant to both a cognitive pathway, by facilitating less threatening interpretations of social experiences, and an affective pathway, by making it easier to respond to one’s own difficulties with the perspective and kindness that might otherwise be extended to another person.

These lines of evidence also reveal several unresolved issues. Previous college-based studies relevant to anxiety have primarily examined traditional expressive writing rather than interventions that explicitly train self-distancing.^11,12 Research on self-distancing has often focused on brief experimental manipulations and emotional-regulation outcomes.^22,23 Although longitudinal expressive-writing research has shown that self-distancing can emerge following repeated writing, it has generally been examined as a psychological process arising from expressive writing rather than as an explicitly instructed intervention component.²⁴ Thus, repeated training in self-distanced writing among first-year university students with elevated social anxiety remains underexplored. Self-distancing has also been examined in relation to social anxiety and self-focused processing among college students,³⁴ yet it remains unclear whether deliberately incorporating self-distancing into a structured writing intervention provides additional benefit beyond traditional expressive writing. Cognitive reappraisal and self-compassion have separate theoretical and empirical links with social anxiety, but, to our knowledge, they have not been examined together as parallel indirect pathways through which a structured self-distanced writing intervention may differ from both an active expressive-writing condition and a neutral-writing condition. Examining these processes within the same model may help distinguish a more cognitive process of reinterpretation from a more affective process of self-relating, while the use of both active and neutral comparison conditions may help separate the perspective-taking component of SDW from nonspecific effects of writing.

The present study therefore examined a four-week self-distanced writing intervention in college freshmen with elevated social anxiety, using traditional expressive writing and neutral writing as comparison conditions. Specifically, we asked whether SDW would produce greater reductions in social anxiety than traditional EW and NW, and whether any such differences would be statistically explained by parallel changes in cognitive reappraisal and self-compassion. The proposed conceptual framework is illustrated in Figure 1. We specified two directional hypotheses. Hypothesis 1: participants receiving SDW would show greater reductions in social anxiety over time than participants receiving traditional EW or NW. Hypothesis 2: the association between SDW and reductions in social anxiety would show parallel indirect effects through increases in (a) cognitive reappraisal and (b) self-compassion. These pathways were conceptualized as complementary rather than sequential mechanisms.

Figure 1.

Proposed parallel mediation model.

Note: H1 denotes the hypothesized association between writing condition and social anxiety reduction; H2a and H2b denote indirect pathways through cognitive reappraisal and self-compassion, respectively.

Abbreviations: SDW, self-distanced writing; EW, expressive writing; NW, neutral writing.

Materials and Methods

Study Design and Trial Overview

The present study was a three-arm, parallel-group, assessor-blind randomized controlled trial conducted from May 2026 to July 2026, with participants allocated to the three study conditions in a 1:1:1 ratio. Random allocation was used to reduce systematic baseline differences between the study conditions and to support a rigorous comparison of changes associated with the assigned writing interventions. In the present trial, this design allowed SDW to be compared with both an active writing condition, traditional EW, and a neutral writing control under the same intervention schedule and assessment procedures.³⁵ The proposed conceptual framework is presented in Figure 1. Participant progression through eligibility assessment, randomization, allocation, follow-up, and analysis is summarized in the CONSORT flow diagram shown in Figure 2.

Figure 2.

CONSORT flow diagram of participant recruitment, allocation, follow-up, and analysis.

Abbreviations: SDW, self-distanced writing; EW, expressive writing; NW, neutral writing.

Participants completed a baseline assessment before randomization (T1), an assessment one week after the final writing session (T2), and a one-month follow-up assessment (T3). The primary objective was to evaluate the efficacy of a self-distanced writing intervention in reducing social anxiety among college freshmen, with cognitive reappraisal and self-compassion examined as parallel potential mediators. The study was reported in accordance with the Consolidated Standards of Reporting Trials 2010 statement.³⁵ The protocol was registered with the Chinese Clinical Trial Registry (ChiCTR2600125746). Ethical approval was obtained from the Institutional Review Board of the School of Art, Minnan Normal University (Approval No. 2026YS-03-002), and the study was conducted in accordance with the Declaration of Helsinki.

Participants, Eligibility, and Randomization

Participants were first-year undergraduate students recruited from Minnan Normal University in China via online advertisements, classroom announcements, and through the university’s mental health service center. Eligibility criteria included: (1) age between 18 and 20 years; (2) a score of 38 or higher on the self-reported Liebowitz Social Anxiety Scale (LSAS-SR);³⁶ and (3) willingness to provide written informed consent. This cutoff has been reported to provide a favorable balance of sensitivity and specificity for screening social anxiety disorder in Chinese samples using the self-report version.³⁷ Exclusion criteria were: (1) current participation in psychotherapy or pharmacotherapy for anxiety or depression; (2) a history of severe mental illness (eg, schizophrenia, bipolar disorder); (3) inability to comprehend the study procedures or questionnaires. Participants who completed all scheduled study procedures received RMB 50 as compensation for their time.

After providing informed consent and completing the baseline (T1) assessment online, eligible participants were randomly assigned to one of three conditions: (1) the Self-Distanced Writing (SDW) group; (2) the traditional Expressive Writing (EW) group; or (3) the Neutral Writing (NW) control group. The randomization sequence with a 1:1:1 allocation was computer-generated by an independent researcher not involved in recruitment. Allocation was concealed using sequentially numbered, opaque, sealed envelopes. A research assistant opened the next envelope in sequence after the T1 assessment to inform the participant of their group assignment. Outcome assessors at T2 and T3 were blinded to group allocation.

Sample Size Justification

An a priori power analysis was conducted for this definitive three-arm RCT using G*Power 3.1 software.³⁸ The analysis was based on a 3 (Group) × 3 (Time) mixed-design ANOVA, assuming a medium effect size (f = 0.25) for the primary interaction effect, an alpha level of.05, and a desired statistical power of.80. This calculation indicated that a total sample of 159 participants would be required. Accounting for an estimated attrition rate of 15%, the target sample size was set at approximately 187 participants.

Intervention Protocol and Control Conditions

All three conditions were conducted in a quiet laboratory setting. Participants attended eight supervised writing sessions over four consecutive weeks, with two sessions each week. Each session lasted 20 minutes, consistent with the duration commonly used in expressive writing research.³⁹

The session duration, laboratory setting, researcher contact, writing materials, and general procedures were held constant across the three conditions so that the principal experimental difference was the writing instruction. At the beginning of each session, participants received an A4 sheet containing the condition-specific instructions and a pen.

A trained research assistant remained present to monitor session timing and procedural adherence, answer procedural questions, and collect the written materials. The research assistant did not provide psychotherapy, interpret participants’ experiences, advise participants about what to write, or otherwise guide the emotional content of the writing.

Self-Distanced Writing (SDW) Group: Participants were instructed to write about a recent anxiety-provoking social situation from a psychologically distanced, observer-like perspective.²⁴ The instructions were: “Please recall a recent social situation that made you feel anxious or troubled. For the next 20 minutes, we ask you to try to analyze and write about this event from the perspective of an observer or a friend. Please use ‘he/she’ or your own name to refer to yourself. Please focus on analyzing the situation, your thoughts, and feelings at that time, as if you were helping a friend understand his or her experience”.

Traditional Expressive Writing (EW) Group: Participants engaged in traditional expressive writing based on established expressive writing procedures.³⁹ The instructions were: “Please recall a recent social situation that made you feel anxious or troubled. For the next 20 minutes, we ask you to write about your very deepest emotions and thoughts regarding this experience. Use ‘I’ to write”.

Neutral Writing (NW) Control Group: Participants completed a factual, nonemotional writing task intended to control for time spent writing, researcher contact, and general task engagement.²⁴ The instructions were: “Please describe, in as much detail and as objectively as possible, all your activities over the past 24 hours. Please avoid describing your feelings or evaluations of the activities”.

The 24-hour window was used only to define the content of the neutral writing task. The SDW and EW prompts asked participants to recall a recent anxiety-provoking social event without specifying a fixed recall window; the 24-hour period in the NW condition was unrelated to the timing of outcome assessments.

Session attendance was recorded. Written materials were identified only by participant ID and were sealed after each session to protect confidentiality.

Intervention Fidelity

Intervention delivery was standardized through printed condition-specific instructions, identical session duration and setting, supervised administration, and attendance recording. These procedures were used to support fidelity to the assigned writing condition. A trained research assistant monitored procedural adherence but did not provide therapeutic guidance or modify the content of participants’ writing.

Assessment Schedule and Rationale

Outcome assessments were conducted at baseline before randomization (T1), one week after completion of the final writing session (T2), and one month after the intervention (T3). The one-week assessment was selected to evaluate relatively early change after participants had completed the full intervention while reducing reliance on short-lived emotional reactions occurring during or immediately after an individual writing session. The one-month assessment was included to examine whether changes were maintained or continued to develop after supervised writing had ended. Longitudinal expressive writing research has shown that self-distancing and subsequent emotional responses can continue to evolve after the writing sessions themselves, providing a rationale for including a follow-up assessment beyond the immediate intervention period.²⁴

Outcome Measures and Language Validation

Outcome assessments were administered online at T1, T2, and T3 by research assistants blinded to group allocation. The primary and secondary outcomes were assessed using previously published and psychometrically evaluated Chinese versions of the following instruments. The present study did not translate the original English instruments de novo. Therefore, no additional forward translation or back translation was conducted specifically for this trial.

Social Anxiety Symptoms. The self-report Liebowitz Social Anxiety Scale (LSAS-SR) served as the primary outcome measure.^36,37 The Chinese version consists of 24 social and performance situations and assesses both fear or anxiety and avoidance. Each fear or anxiety rating and each avoidance rating is scored from 0 to 3. The item ratings are summed to produce a total score ranging from 0 to 144, with higher scores indicating greater social anxiety severity. A score of 38 or higher was used as the eligibility threshold in the present study. The Chinese version has demonstrated good reliability and validity, with a reported Cronbach’s alpha of.90.³⁷

Cognitive Reappraisal. Cognitive reappraisal was assessed using the six-item Cognitive Reappraisal subscale of the Emotion Regulation Questionnaire (ERQ-CR).^40,41 Items, such as “When I want to feel less negative emotion, I change the way I’m thinking about the situation”, are rated on a 7-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree). The ERQ-CR score was calculated as the mean of the six cognitive reappraisal items, with higher scores indicating greater habitual use of cognitive reappraisal. The validated Chinese version has demonstrated good reliability and validity, with a reported Cronbach’s alpha of.83.⁴¹

Self-Compassion. Self-compassion was assessed using the 26-item Self-Compassion Scale (SCS).^42,43 Items are rated on a 5-point Likert scale ranging from 1 (almost never) to 5 (almost always). Negatively worded items were reverse scored, and the overall SCS score was calculated as the mean of all 26 items, with higher scores indicating greater self-compassion. The published Chinese version was used in the present study and has demonstrated good psychometric properties, with a reported Cronbach’s alpha of.84.⁴³

Safety Record

Given the nature of the writing intervention, which involved reflecting on stressful social experiences, a safety monitoring protocol was in place. Participants were informed during consent that writing might bring up temporary discomfort. The supervising research assistant was trained to identify signs of significant distress. All participants were provided with contact information for the university’s psychological counseling center, and a referral protocol was established for any participant expressing acute need for support.

Data Management and Preanalysis Screening

Before hypothesis testing, data completeness and missing observations were summarized for each study condition and assessment point. Descriptive statistics were calculated for all outcome measures by group and time point. Baseline comparability across the three randomized conditions was evaluated using one-way analysis of variance for continuous variables and chi-square tests for categorical variables. For repeated measures analyses, Mauchly’s test was used to evaluate the sphericity assumption. Greenhouse-Geisser adjusted degrees of freedom were reported when sphericity was violated. Box’s M test was used to evaluate equality of covariance matrices, and multivariate tests based on Pillai’s trace were examined as robustness checks when covariance assumptions were of concern.

Data Analysis

All statistical analyses were conducted using IBM SPSS Statistics version 26.0. All tests were two-tailed, with statistical significance set at p <0.05. Effect sizes for analyses of variance were reported as partial eta squared. Mediation results were reported as unstandardized regression coefficients, with bias-corrected bootstrap confidence intervals for indirect effects.

Primary Analysis

The primary analysis used a 3 Group × 3 Time mixed-design analysis of variance conducted with the SPSS GLM Repeated Measures procedure. Writing condition, SDW, EW, or NW, was the between-participant factor, and assessment time, T1, T2, or T3, was the within-participant factor. Because the GLM Repeated Measures procedure requires complete observations for the repeated outcome variables entered into the model, the primary repeated measures analysis was based on participants with complete T1, T2, and T3 data. This resulted in 181 participants for the primary longitudinal analysis, comprising 61 participants in SDW, 61 in EW, and 59 in NW.

Mauchly’s test was used to assess sphericity, with Greenhouse-Geisser correction applied when necessary. A significant Group × Time interaction was followed by Bonferroni-adjusted between-group comparisons. Within-group changes from T1 to T3 were additionally described using paired-samples t tests, and between-group differences in T1 to T3 change scores were examined using one-way analysis of variance followed by Bonferroni-adjusted comparisons. The same longitudinal analytic framework was applied to cognitive reappraisal and self-compassion.

Mediation Analysis

Parallel mediation was tested using the PROCESS macro (Model 4)⁴⁴ with 5000 bias-corrected bootstrap resamples. To decompose the three-group effect into theoretically meaningful contrasts, two separate parallel mediation models were computed. Contrast D1 (primary analysis) compared SDW against the active control (EW), with NW omitted. Contrast D2 (supplementary analysis) compared SDW against the neutral control (NW), with EW omitted. This approach prevents collinearity and allows for the testing of distinct psychological mechanisms underlying the intervention effects. The outcome was change in social anxiety (Δ LSAS-SR = T3 − T1), and parallel mediators were changes in cognitive reappraisal (Δ ERQ-CR = T2 − T1) and self-compassion (Δ SCS = T2 − T1). Indirect effects were considered significant when the 95% bias-corrected bootstrap CI excluded zero.

Missing Data

Participants with incomplete outcome data across T1, T2, and T3 were excluded from the corresponding repeated-measures ANOVA. For PROCESS analyses, cases with missing data on variables required for the relevant mediation model were excluded from that analysis.

Sensitivity Analyses

Several sensitivity and robustness analyses were performed to assess the robustness of the findings: (a) ANCOVAs on T2 and T3 LSAS-SR scores with baseline LSAS-SR as covariate; (b) multivariate tests based on Pillai’s trace for the mixed-design ANOVA findings; and (c) Box’s M tests to evaluate equality of covariance matrices.

Results

Participant Flow and Baseline Characteristics

The flow of participants through the trial is depicted in Figure 2. A total of 187 first-year students met inclusion criteria (LSAS-SR ≥ 38, age 18–20 years) and were randomly assigned to the SDW group (n = 63), EW group (n = 62), or NW group (n = 62). All 187 participants completed the baseline (T1) assessment. At T2, 186 participants (SDW: n = 63; EW: n = 61; NW: n = 62) provided outcome data; at the one-month follow-up (T3), 181 participants (SDW: n = 61; EW: n = 61; NW: n = 59) completed the assessment. The overall attrition rate was 3.2% (6 of 187 participants). The primary repeated-measures analyses included 181 participants with complete outcome data across T1, T2, and T3 (SDW: n = 61; EW: n = 61; NW: n = 59).

Attrition analyses indicated that the six participants lost to follow-up did not differ significantly from completers on baseline LSAS-SR (p =0.783), ERQ-CR (p =0.075), age (p =0.185), or gender (p =0.688). Baseline SCS scores were higher among dropouts (M = 4.12, SD = 0.45) than among completers (M = 3.48, SD = 0.61), p =0.012.

Baseline demographic and clinical characteristics of the three groups are presented in Table 1. The groups did not differ significantly on the primary outcome, LSAS-SR, F(2, 184) = 0.15, p =0.864, or on the secondary outcomes: ERQ-CR, F(2, 184) = 2.25, p =0.109, and SCS, F(2, 184) = 0.57, p =0.569. Gender distribution was comparable across groups, χ²(2) = 2.45, p =0.294. A small but statistically significant difference in age was observed, F(2, 184) = 4.07, p =0.019; however, the absolute difference (< 0.3 years) was not considered clinically meaningful.

Table 1.

Baseline Demographic and Clinical Characteristics

Variable	SDW (n=63)	EW (n=62)	NW (n=62)
	M (SD)	M (SD)	M (SD)
Age (years)	18.65 (0.68)	18.39 (0.55)	18.40 (0.49)
Gender, n (%)
Female	35 (56%)	32 (52%)	26 (42%)
Male	28 (44%)	30 (48%)	36 (58%)
LSAS-SR	62.68 (17.87)	63.15 (15.43)	61.60 (15.72)
ERQ-CR	4.60 (1.20)	4.30 (1.09)	4.71 (1.03)
SCS	3.55 (0.64)	3.44 (0.59)	3.52 (0.61)

Abbreviations: SDW, self-distanced writing; EW, expressive writing; NW, neutral writing; LSAS-SR, Liebowitz Social Anxiety Scale–Self-Report; ERQ-CR, Emotion Regulation Questionnaire–Cognitive Reappraisal; SCS, Self-Compassion Scale.

Primary Outcome: Social Anxiety

Descriptive statistics for all outcome variables by group and time point are displayed in Table 2. A 3 (Group: SDW, EW, NW) × 3 (Time: T1, T2, T3) mixed-design ANOVA was conducted on LSAS-SR scores among participants with complete outcome data across the three assessment points. Mauchly’s test indicated violation of the sphericity assumption, W =0.864, χ²(2) = 25.95, p <0.001; therefore, Greenhouse-Geisser corrected degrees of freedom are reported (ε =0.880). Box’s test supported the assumption of equal covariance matrices across groups, M = 14.10, p =0.316.

Table 2.

Descriptive Statistics for Outcome Variables

Measure	SDW (n=61)	EW (n=61)	NW (n=59)
	M (SD)	M (SD)	M (SD)
LSAS-SR
T1 (Baseline)	62.92 (18.11)	63.25 (15.54)	61.41 (16.07)
T2 (1-week post-intervention)	58.92 (18.22)	59.48 (15.41)	59.39 (16.28)
T3 (1-month follow-up)	57.49 (18.24)	59.61 (15.63)	59.29 (16.45)
ERQ-CR
T1 (Baseline)	4.57 (1.20)	4.32 (1.09)	4.65 (1.00)
T2 (1-week post-intervention)	5.29 (1.18)	4.70 (1.25)	4.76 (1.04)
T3 (1-month follow-up)	5.32 (1.13)	4.62 (1.19)	4.77 (1.04)
SCS
T1 (Baseline)	3.53 (0.63)	3.44 (0.59)	3.47 (0.59)
T2 (1-week post-intervention)	3.84 (0.64)	3.58 (0.60)	3.45 (0.67)
T3 (1-month follow-up)	3.82 (0.61)	3.62 (0.62)	3.42 (0.65)

Abbreviations: SDW, self-distanced writing; EW, expressive writing; NW, neutral writing; LSAS-SR, Liebowitz Social Anxiety Scale–Self-Report; ERQ-CR, Emotion Regulation Questionnaire–Cognitive Reappraisal; SCS, Self-Compassion Scale.

The main effect of Time was significant, F(1.76, 313.27) = 135.21, p <0.001, η²_p =0.432, confirming that LSAS-SR scores decreased across assessment points collapsing across groups. The main effect of Group was not significant, F(2, 178) = 0.06, p =0.942, η²_p =0.001. Critically, the predicted Group × Time interaction was significant, F(3.52, 313.27) = 8.38, p <0.001, η²_p =0.086, indicating that the trajectories of social anxiety reduction differed across the three conditions (see Figure 3).

Figure 3.

LSAS-SR trajectories by Group.

Note: Error bars represent standard errors.

Abbreviations: LSAS-SR, Liebowitz Social Anxiety Scale–Self-Report; SDW, self-distanced writing; EW, expressive writing; NW, neutral writing.

To decompose the interaction, within-group paired-samples t-tests examined LSAS-SR changes from T1 to T3. All three groups showed statistically significant reductions: SDW, Δ = −5.43 (SD = 4.26), t(60) = 9.95, p <0.001, dz = 1.27; EW, Δ = −3.64 (SD = 2.84), t(60) = 10.01, p <0.001, dz = 1.28; NW, Δ = −2.12 (SD = 4.14), t(58) = 3.94, p <0.001, dz = 0.51. The proportions of participants showing any reduction (T3 < T1) were 90.2% (SDW), 88.5% (EW), and 69.5% (NW).

A one-way ANOVA on the T1-to-T3 change scores (Δ LSAS-SR) revealed a significant group effect, F(2, 178) = 11.42, p <0.001. Bonferroni-corrected pairwise comparisons (see Table 3) showed that the SDW group exhibited significantly greater LSAS-SR reductions than the EW group, ΔΔ = −1.79, 95% CI [−3.45, −0.13], p =0.030, Cohen’s d = −0.49, and the NW group, ΔΔ = −3.31, 95% CI [−4.98, −1.63], p <0.001, d = −0.79. The comparison between EW and NW was not significant after Bonferroni correction, ΔΔ = −1.52, 95% CI [−3.20, +0.15], p =0.089, d = −0.43.

Table 3.

One-Way ANOVA and Pairwise Comparisons for LSAS-SR Change Scores (T1-to-T3)

Source of Variation	SS	df	MS	F	p	η2_p
One-way ANOVA for T1-to-T3 LSAS-SR change scores
Between Groups (Δ)	329.16	2	164.58	11.42	<0.001	0.114
Within Groups (Δ)	2565.15	178	14.41
Total	2894.31	180
Bonferroni-adjusted pairwise comparisons
Comparison	ΔΔ	95% CI	p	Cohen’s d
SDW vs EW	−1.79	[−3.45, −0.13]	0.030	−0.49
SDW vs NW	−3.31	[−4.98, −1.63]	<0.001	−0.79
EW vs NW	−1.52	[−3.20, +0.15]	0.089	−0.43

Note: ΔLSAS-SR = T3 − T1. Negative values indicate reductions in social anxiety.

Abbreviations: SDW, self-distanced writing; EW, expressive writing; NW, neutral writing; LSAS-SR, Liebowitz Social Anxiety Scale–Self-Report; CI, confidence interval.

To examine whether treatment gains were maintained or extended during the follow-up period, paired-samples t-tests compared T2 to T3 within each group. The SDW group showed a significant further reduction in social anxiety during the follow-up period, Δ = −1.43 (SD = 2.73), t(60) = 4.08, p <0.001, dz = 0.52. In contrast, neither the EW group, Δ = +0.13 (SD = 2.68), t(60) = −0.38, p =0.704, nor the NW group, Δ = −0.10 (SD = 2.73), t(58) = 0.29, p =0.776, showed significant change from T2 to T3.

To further account for any potential baseline variation, analysis of covariance (ANCOVA) was conducted on T2 and T3 LSAS-SR scores with T1 LSAS-SR entered as a covariate. At T2 (n = 186), the group effect was significant, F(2, 182) = 6.97, p =0.001, η²_p =0.071. Adjusted means were: SDW = 58.54 (SE = 0.42), EW = 58.75 (SE = 0.43), and NW = 60.59 (SE = 0.43). Bonferroni-corrected pairwise comparisons indicated that SDW did not differ significantly from EW at T2 (Δadj = −0.21, p = 1.000), but both SDW (Δadj = −2.05, p =0.001) and EW (Δadj = −1.84, p =0.004) were significantly lower than NW.

At T3 (n = 181), the group effect was larger in magnitude, F(2, 177) = 11.25, p <0.001, η²_p =0.113. Adjusted means were: SDW = 57.12 (SE = 0.49), EW = 58.91 (SE = 0.49), and NW = 60.40 (SE = 0.50). Bonferroni-corrected pairwise comparisons revealed that SDW was significantly lower than both EW (Δadj = −1.79, p =0.030) and NW (Δadj = −3.29, p <0.001). EW did not differ significantly from NW after correction (Δadj = −1.50, p =0.098).

Secondary Outcomes

Cognitive Reappraisal (ERQ-CR). Mauchly’s test was significant, W =0.767, χ²(2) = 47.03, p <0.001; Greenhouse-Geisser correction was applied (ε =0.811). Box’s test supported equal covariance matrices, M = 17.58, p =0.144. The Group × Time interaction was significant, F(3.24, 288.65) = 9.29, p <0.001, η²_p =0.095. Within-group analyses indicated that the SDW group showed a substantial increase in cognitive reappraisal from T1 to T3, Δ = +0.75, t(60) = 5.79, p <0.001. The EW group showed a smaller but significant increase, Δ = +0.30, t(60) = 3.79, p <0.001. The NW group showed no significant change, Δ = +0.12, t(58) = 1.34, p =0.183.

Self-Compassion (SCS). Mauchly’s test was significant, W =0.665, χ²(2) = 72.08, p <0.001; ε =0.749. Box’s test indicated unequal covariance matrices across groups, M = 70.28, p <0.001. Given this violation, multivariate test results (Pillai’s trace) are also reported below and yielded consistent findings. The Group × Time interaction was significant, F(3.00, 266.77) = 15.16, p <0.001, η²_p =0.146. The SDW group showed a significant increase in self-compassion, Δ = +0.29, t(60) = 4.83, p <0.001. The EW group showed a smaller increase, Δ = +0.18, t(60) = 4.72, p <0.001. The NW group showed no significant change, Δ = −0.05, t(58) = −1.73, p =0.088.

Mediation Analysis

Two parallel mediation analyses (PROCESS Model 4)⁴⁴ with 5000 bias-corrected bootstrap resamples examined whether the superior efficacy of SDW in reducing social anxiety at T3 was mediated by T1-to-T2 changes in cognitive reappraisal and self-compassion. Two planned contrasts were tested: the primary contrast compared SDW against EW (D1; SDW = 1, EW = 0; n = 122), and a supplementary contrast compared SDW against the neutral writing control (D2; SDW = 1, NW = 0; n = 120). Both analyses were restricted to participants with complete outcome data at all three time points.

For D1 (SDW vs EW), the total effect of group on Δ LSAS-SR was significant, B = −1.79, SE = 0.69, 95% CI [−3.16, −0.42], p =0.011. After accounting for the two mediators, the direct effect was reduced and no longer statistically significant, B = −1.29, SE = 0.67, 95% CI [−2.62, +0.04], p =0.058. The a-paths confirmed that SDW (vs EW) significantly predicted greater increases in both cognitive reappraisal, a₁ = 0.34, SE = 0.13, p =0.011, and self-compassion, a₂ = 0.16, SE = 0.06, p =0.008. In the full model including both mediators, self-compassion change significantly predicted LSAS-SR reduction, b₂ = −2.35, SE = 1.01, p =0.022, whereas cognitive reappraisal change did not, b₁ = −0.35, SE = 0.45, p =0.437. The indirect effect via self-compassion was statistically significant, B = −0.38, Boot SE = 0.18, 95% BC CI [−0.76, −0.06], indicating that SDW reduced social anxiety partly through its enhancement of self-compassion. The indirect effect via cognitive reappraisal was not significant, B = −0.12, Boot SE = 0.19, 95% BC CI [−0.53, +0.24]. The total indirect effect was significant, B = −0.50, Boot SE = 0.24, 95% BC CI [−1.00, −0.06]. The ratio of the indirect effect to the total effect via the self-compassion pathway (PM) was 21.1%.

For D2 (SDW vs NW), the total effect was significant, B = −3.31, SE = 0.77, 95% CI [−4.83, −1.79], p <0.001, as was the direct effect, B = −2.68, SE = 0.89, 95% CI [−4.44, −0.93], p =0.003. SDW (vs NW) predicted larger increases in both cognitive reappraisal, a₁ = 0.61, SE = 0.14, p <0.001, and self-compassion, a₂ = 0.33, SE = 0.06, p <0.001. However, neither mediator significantly predicted LSAS-SR reduction in the full model (self-compassion: b₂ = −1.44, SE = 1.15, p =0.212; cognitive reappraisal: b₁ = −0.25, SE = 0.52, p =0.626), and none of the indirect effects were significant: self-compassion, B = −0.47, Boot SE = 0.39, 95% BC CI [−1.29, +0.19]; cognitive reappraisal, B = −0.15, Boot SE = 0.36, 95% BC CI [−0.96, +0.50]; total, B = −0.62, Boot SE = 0.45, 95% BC CI [−1.50, +0.26]. The proportion mediated via self-compassion (PM) was 14.2%. Full path coefficients for both contrasts are reported in Table 4.

Table 4.

Mediation Path Coefficients (PROCESS Model 4; Parallel Mediation)

Path	Predictor → Outcome	D1: SDW vs EW B (SE), p	D1: 95% CI	D2: SDW vs NW B (SE), p	D2: 95% CI
a1	Group (SDW vs EW/NW) → Δ Cognitive Reappraisal (ΔERQ-CR)	+0.34 (SE = 0.13) p =0.011	[+0.08, +0.60]	+0.61 (SE = 0.14) p <0.001	[+0.33, +0.88]
a2	Group (SDW vs EW/NW) → Δ Self-Compassion (ΔSCS)	+0.16 (SE = 0.06) p =0.008	[+0.04, +0.28]	+0.33 (SE = 0.06) p <0.001	[+0.20, +0.45]
b1	Δ Cognitive Reappraisal (ΔERQ-CR) → ΔLSAS-SR	−0.35 (SE = 0.45) p =0.437	[−1.24, +0.54]	−0.25 (SE = 0.52) p =0.626	[−1.27, +0.77]
b2	Δ Self-Compassion (ΔSCS) → ΔLSAS-SR	−2.35 (SE = 1.01) p =0.022	[−4.35, −0.35]	−1.44 (SE = 1.15) p =0.212	[−3.71, +0.83]
c′	Group (SDW vs EW/NW) → ΔLSAS-SR (direct effect, c′)	−1.29 (SE = 0.67) p =0.058	[−2.62, +0.04]	−2.68 (SE = 0.89) p =0.003	[−4.44, −0.93]
c	Group (SDW vs EW/NW) → ΔLSAS-SR (total effect, c)	−1.79 (SE = 0.69) p =0.011	[−3.16, −0.42]	−3.31 (SE = 0.77) p <0.001	[−4.83, −1.79]
Indirect pathway		D1: SDW vs EW B (Boot SE)	95% BC CI	D2: SDW vs NW B (Boot SE)	95% BC CI
Via Cognitive Reappraisal		−0.12 (0.19)	[−0.53, +0.24]	−0.15 (0.36)	[−0.96, +0.50]
Via Self-Compassion		−0.38 (0.18)	[−0.76, −0.06]	−0.47 (0.39)	[−1.29, +0.19]
Total indirect effect		−0.50 (0.24)	[−1.00, −0.06]	−0.62 (0.45)	[−1.50, +0.26]

Note: D1 = SDW vs EW; D2 = SDW vs NW. ΔLSAS-SR = T3 − T1; ΔERQ-CR = T2 − T1; ΔSCS = T2 − T1. Negative ΔLSAS-SR values indicate reductions in social anxiety, whereas positive ΔERQ-CR and ΔSCS values indicate increases. The arrow (→) indicates the hypothesized directional path from predictor to outcome.

Abbreviations: B, unstandardized coefficient; SE, standard error; Boot SE, bootstrap standard error; CI, confidence interval; BC CI, bias-corrected confidence interval.

Sensitivity Analyses

Several sensitivity and robustness analyses were conducted to evaluate the robustness of the findings. First, the ANCOVA on T2 and T3 LSAS-SR scores with T1 LSAS-SR as covariate confirmed that the Group effect remained significant after adjusting for baseline variation, with the SDW advantage over EW emerging at T3 but not at T2. Second, given the violation of sphericity across all mixed ANOVAs, multivariate tests (Pillai’s trace) confirmed the same pattern of results for LSAS-SR, V =0.143, F(4, 356) = 6.85, p <0.001, η²_p =0.071; ERQ-CR, V =0.130, F(4, 356) = 6.21, p <0.001, η²_p =0.065; and SCS, V =0.198, F(4, 356) = 9.79, p <0.001, η²_p =0.099. Third, Box’s M test confirmed the assumption of equal covariance matrices for LSAS-SR (M = 14.10, p =0.316) and ERQ-CR (M = 17.58, p =0.144), but was significant for SCS (M = 70.28, p <0.001); Pillai’s trace multivariate test corroborated the SCS Group × Time interaction.

Discussion

This study set out to determine whether a structured self-distanced writing intervention could outperform traditional expressive writing in reducing social anxiety among college freshmen, and whether cognitive reappraisal and self-compassion functioned as parallel mediators of any observed advantage. The interaction between group and time confirmed that SDW led to greater reductions in social anxiety than both EW and NW, with the advantage over EW being modest in magnitude (d = −0.49) and emerging during the follow-up period rather than at the one-week post-intervention assessment. Self-compassion served as a statistically significant mediator of the SDW advantage over EW, whereas cognitive reappraisal did not, despite being enhanced by SDW to a greater extent than by EW. A supplementary contrast comparing SDW against the neutral writing control yielded a statistically significant total and direct effect but no significant indirect effects through either mediator, indicating that the self-compassion pathway is specific to the comparison with an active writing comparator rather than the neutral-writing comparator. These findings provide partial support for the proposed dual-mechanism model and point toward an affective pathway, rooted in self-kindness rather than cognitive restructuring, as the one possible pathway associated with the SDW advantage over EW.

The finding that SDW produced greater reductions in social anxiety than both EW and NW, a difference that reached statistical significance at the one-month follow-up, extends the self-distancing literature in a theoretically important direction. Prior work has established that brief, experimentally induced self-distancing can regulate negative affect in the moment and that the cognitive mechanism may involve increased sensitivity to corrective feedback during emotional learning.^45,46 The present study moves beyond acute effects to demonstrate that repeated, structured self-distanced writing over four weeks can generate sustained symptomatic improvement that continues to accrue after the intervention ends. This temporal pattern, an effect that was not yet present at post-intervention but emerged at follow-up, is consistent with what has been described as the delayed, durable effect of expressive writing paradigms,⁴⁷ and it underscores the value of including follow-up assessments in writing intervention trials.

The absence of a statistically significant SDW advantage at T2, coupled with the emergence of a statistically significant advantage at T3, invites careful consideration of the underlying mechanism. In the ANCOVA controlling for baseline severity, SDW and EW were nearly indistinguishable at the one-week post-intervention assessment; only after a month of unsupervised consolidation did SDW pull meaningfully ahead. This pattern aligns with the view that self-distancing operates as an acquired cognitive skill that requires practice beyond the training context to exert its full therapeutic impact, rather than as a single-session cathartic tool.⁴⁸ Participants in the SDW condition, having been repeatedly cued to adopt an observer’s perspective on their own social difficulties, may have internalized this stance and continued to apply it spontaneously in daily life during the follow-up period. By contrast, the expressive writing group, whose gains remained stable but did not grow after the intervention, may have benefited more from the immediate emotional processing afforded by disclosure, a benefit that plateaus once the writing sessions conclude.

The mediation analyses offer a more granular view of the psychological processes underlying SDW’s sustained advantage. The indirect effect via self-compassion was statistically significant, suggesting that SDW reduced social anxiety partly by enhancing participants’ capacity to treat themselves with kindness and to view their struggles as part of the shared human experience. This finding resonates with recent experimental work showing that brief self-compassion inductions can increase willingness to engage with feared social situations⁴⁹ and that self-compassion training produces gains in socially anxious individuals that are clinically meaningful even when they do not surpass those of active relaxation comparators on symptom measures.⁵⁰ The present study contributes a novel demonstration that self-compassion can be enhanced not only through explicit compassion-focused exercises but also through a cognitive reframing of the narrative perspective from which emotional events are processed. Writing about a painful social experience from the outside appears to naturally evoke the same compassionate orientation that one would extend to a friend, bypassing the resistance that can accompany direct self-compassion instructions.

The contrast-specificity of the mediation findings warrants emphasis. When SDW was benchmarked against the neutral-writing control (D2) rather than against EW, the total effect and the direct effect were both significant, but neither the self-compassion nor the cognitive-reappraisal indirect effect reached significance. This pattern indicates that the mediational role of self-compassion emerged specifically in the comparison between two active writing conditions and was not observed when SDW was contrasted with a neutral-writing control. Such contrast-dependent mediation may vary according to the comparison condition and the intervention components that differ between groups.^51,52 Accordingly, the self-compassion pathway should be interpreted as the mechanism by which SDW outperforms another writing-based intervention, rather than as the sole mechanism by which it differs from neutral writing. This qualification is theoretically meaningful: it suggests that self-distancing confers its incremental benefit over a comparable task through an affective, self-compassionate reorientation, whereas its advantage over neutral writing is more direct and less fully accounted for by the measured mediators.

The cognitive reappraisal pathway did not reach statistical significance despite greater increases in reappraisal in the SDW condition. This suggests that changes in habitual reappraisal did not reliably account for the observed reduction in social anxiety in the present mediation model. A related ecological momentary intervention study similarly reported improvements in emotion regulation and self-compassion without clear statistical mediation of clinical outcomes.⁵³ One possible explanation is that the ERQ-CR primarily reflects habitual reappraisal tendencies and may be less sensitive to momentary cognitive changes occurring during individual writing sessions. Future studies using repeated state measures or experience-sampling methods may provide a more sensitive assessment of whether short-term changes in appraisal are associated with subsequent changes in social anxiety.

Theoretical and Practical Implications

The findings offer several theoretical and practical implications that should be considered within the scope of the present trial. The significant indirect effect through self-compassion in the comparison between SDW and EW suggests that changes in the way individuals relate to themselves during reflection may be relevant alongside cognitive reinterpretation. At the same time, the absence of significant indirect effects in the SDW versus NW comparison indicates that neither self-compassion nor cognitive reappraisal should be regarded as a complete explanation of the intervention effect.

From a practical perspective, SDW uses brief and standardized writing instructions that can be delivered within a structured university setting. The present findings therefore support further evaluation of SDW as a potentially accessible approach for students with elevated social anxiety. However, the intervention was delivered under supervised research conditions, and its feasibility, acceptability, and effectiveness in routine university mental health services were not directly examined. Future research should also incorporate behavioral assessments, such as observed social performance, behavioral avoidance, or engagement during socially evaluative tasks, to determine whether changes in self-reported social anxiety are accompanied by observable behavioral change.

Limitations

This study has several limitations. The reliance on self-report measures for the primary outcome introduces the possibility that observed improvements partly reflect expectancy or demand characteristics, a concern that has been documented across the expressive writing literature.¹³ The one-month follow-up, although consistent with the typical duration of writing intervention trials,¹³ cannot speak to whether the SDW advantage is maintained over longer periods or continues to widen. Although SDW significantly enhanced cognitive reappraisal, this change did not mediate social anxiety reduction, potentially because the ERQ-CR captures dispositional reappraisal tendencies rather than the in-the-moment cognitive shifts cultivated by SDW; experience-sampling measures may offer a more sensitive test of this pathway in future work. Unmeasured variables such as baseline depression and trait rumination may have influenced treatment response, although randomization mitigates this concern at the group level. The study did not include an independent manipulation check of state self-distancing. Although intervention fidelity was supported by standardized instructions, supervised administration, and attendance monitoring, these procedures confirm adherence to the assigned writing protocol rather than directly verifying the degree of psychological distance induced by SDW. Future studies could incorporate state self-distancing measures or linguistic indicators, such as the use of first-person and non-first-person pronouns, to assess whether the intended psychological manipulation occurred as expected.

Conclusion

The present study found that a structured four-week self-distanced writing intervention was associated with greater reductions in social anxiety among college freshmen than traditional expressive writing and neutral writing at the one-month follow-up. The difference between SDW and EW was not statistically significant at the one-week post-intervention assessment, indicating that the relative advantage of SDW became apparent during the subsequent follow-up period. Parallel mediation analysis showed a significant indirect effect through increased self-compassion when SDW was compared with EW, whereas cognitive reappraisal did not show a significant indirect effect. Neither proposed mediator showed a significant indirect effect in the comparison between SDW and NW. These findings suggest that self-compassion may be relevant to understanding how SDW differs from traditional expressive writing, while also indicating that the processes associated with SDW are unlikely to be explained by a single pathway. The brief and structured format of SDW warrants further evaluation in university settings. Future research should examine whether the observed effects are maintained over longer follow-up periods, incorporate behavioral assessments of social anxiety alongside self-report outcomes, and directly assess whether SDW produces the intended increase in psychological distance.

Data Sharing Statement

The deidentified individual participant-level data used in the analyses reported in this article will be available from the corresponding author, BoKai Yang (ybk1541@mnnu.edu.cn).

Disclosure

The authors report no conflicts of interest in this work.