📄 social_science

Investigating the Factor Structure of the Self-Compassion Scale in Autistic Adults

This study utilized exploratory structural equation modelling on a sample of 228 autistic adults to demonstrate that the six-factor and single-bifactor models of the Self-Compassion Scale provide a valid and well-fitting structure for this population, supporting the use of both total and subscale scores in research and practice.

Original authors: Benjamin Kitts Mersereau, Barbara D'Entremont

Published 2026-07-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Benjamin Kitts Mersereau, Barbara D'Entremont

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: Checking the Map

Imagine you have a very popular map called the Self-Compassion Scale (SCS). For years, this map has been used to help people navigate their feelings of self-kindness versus self-criticism. It was drawn based on how non-autistic people experience these feelings.

However, the researchers in this paper asked a crucial question: "Does this same map work for autistic adults?"

Autistic adults often face unique challenges with self-perception, such as feeling isolated or being very hard on themselves. The researchers wanted to make sure that when autistic adults take this test, the questions are actually measuring what they are supposed to measure, rather than getting lost in translation.

The Three Routes Tested

To find the best map, the researchers tested three different ways of organizing the 26 questions on the scale. Think of these as three different theories about how the questions fit together:

The Six-Part Puzzle (Six-Factor Model): This theory says the scale is made of six distinct pieces that talk to each other. Three pieces are positive (being kind to yourself, feeling connected to others, staying mindful), and three are negative (being harsh, feeling alone, getting overwhelmed).
- Analogy: Imagine a six-piece orchestra. Each instrument plays its own part, but they all harmonize to create the music.
The Two-Team Battle (Two-Factor Model): This theory suggests the scale is just a tug-of-war between two teams: the "Good Team" (compassionate self-responding) and the "Bad Team" (uncompassionate self-responding).
- Analogy: A seesaw where one side goes up only if the other goes down.
The Umbrella Model (Single-Bifactor Model): This theory suggests there is one big "umbrella" of self-compassion that covers everything, but underneath that umbrella, the six specific pieces still exist and do their own work.
- Analogy: A large umbrella (the general feeling of self-compassion) protecting you from rain, while six smaller, distinct pockets underneath hold different tools (the specific components).

The Experiment: Putting the Map to the Test

The researchers gathered data from 228 autistic adults in the UK. They used a sophisticated statistical tool called ESEM (Exploratory Structural Equation Modeling).

Analogy: If standard math is like using a ruler to measure a straight line, ESEM is like using a flexible, 3D scanner that can measure how things bend and twist in complex ways. It allows the data to show its true shape without forcing it into a rigid box.

They ran the data through both the "Six-Part Puzzle," the "Two-Team Battle," and the "Umbrella Model" to see which one fit the autistic adults' answers best.

The Results: Which Map Won?

Here is what the data revealed:

The Two-Team Battle (Two-Factor Model) Failed: This model did not fit the data well.
- Why it failed: It tried to force the complex feelings of autistic adults into a simple "Good vs. Bad" fight. The results showed that the feelings are more nuanced than just two opposing teams.
The Six-Part Puzzle (Six-Factor Model) Worked Well: The six distinct pieces (kindness, judgment, humanity, isolation, mindfulness, over-identification) were all there and working correctly.
- Verdict: This is a valid way to look at the data.
The Umbrella Model (Single-Bifactor Model) Was the Best: This model fit the data the best. It showed that while the six specific pieces are important, they all contribute to one strong, unified feeling of self-compassion.
- Verdict: This is the most accurate representation. It confirmed that you can look at the "big picture" score (the umbrella) and the specific details (the pockets) without them contradicting each other.

What This Means for the Score

The study found that for autistic adults:

You can use the Total Score: Because the "Umbrella" model worked so well, adding up all the answers to get one big score is a valid and reliable way to measure self-compassion.
You can also use the Sub-scores: Because the "Six-Part Puzzle" also worked, looking at specific areas (like how much someone struggles with "isolation" vs. "self-kindness") is also useful.
Don't use the Two-Team Score: The study advises against using the simple "Good vs. Bad" split, as it doesn't capture the full picture for this group.

Important Caveats (The Fine Print)

The researchers were careful to note who was not in this study. The participants all had a formal medical diagnosis of autism.

Analogy: The map was drawn for people who have a specific passport. It hasn't been tested on people who know they are autistic but don't have the official paper, or those with higher support needs who couldn't complete the online survey.
Conclusion: The results are great for the group tested, but we can't be 100% sure it applies to every single autistic person yet.

The Bottom Line

This study is like a quality control check. It confirmed that the Self-Compassion Scale is a valid tool for autistic adults. It works best when viewed as a complex system with a strong overall theme (the Umbrella) supported by six specific parts. Researchers and clinicians can now use this tool with confidence to understand how autistic adults treat themselves, knowing the "map" accurately reflects their terrain.

Technical Summary: Investigating the Factor Structure of the Self-Compassion Scale in Autistic Adults

Problem Statement
Self-compassion is a recognized predictor of positive mental health outcomes, yet its measurement in autistic adults remains unvalidated. While the Self-Compassion Scale (SCS) is widely used in general and clinical populations, its factor structure is contested, with competing models including a six-factor correlated model, a two-factor model (distinguishing compassionate vs. uncompassionate self-responding), and a single-bifactor model. Previous research suggests autistic individuals may experience unique challenges in practicing self-compassion, such as difficulties with "common humanity" or heightened self-criticism. Consequently, it is unknown whether established SCS models generalize to the autistic population, necessitating an independent validation to ensure accurate measurement for future interventions.

Methodology
The study utilized a sample of 228 autistic adults (50% female) residing in the United Kingdom, recruited via Prolific and verified through a formal clinical diagnosis screening. Data were sourced from the Open Science Framework (Galvin, 2022). Participants completed the 26-item Self-Compassion Scale (SCS), which assesses six components: self-kindness, self-judgment, common humanity, isolation, mindfulness, and over-identification.

The researchers employed Exploratory Structural Equation Modelling (ESEM) to test three competing factor structures:

A six-factor correlated model.
A two-factor correlated model (Compassionate Self-responding [CS] and Uncompassionate Self-responding [US]).
A single-bifactor model (one general self-compassion factor and six specific component factors).

Analyses were conducted using Mplus (version 8.11) with two estimators: Maximum Likelihood (ML) and Weighted Least Squares Mean and Variance Adjusted (WLSMV), to account for the ordinal nature of the Likert-scale data. Model fit was evaluated using CFI, TLI, RMSEA, and SRMR, with thresholds set at >.95 for CFI/TLI and ≤.05 for RMSEA/SRMR to indicate good fit. Additionally, bifactor-specific indices (Omega, Hierarchical Omega, ECV, and PUC) were calculated in R to assess unidimensionality and reliability.

Key Results

Two-Factor Model: This model demonstrated poor fit, particularly with ML estimation. Even with WLSMV, it failed to meet criteria for RMSEA (.09) and was rejected.
Six-Factor Correlated Model: This model demonstrated acceptable to good fit across both estimators. Factor loadings were generally consistent with prior literature, though self-judgment was weakly defined. Notable cross-loadings were observed (e.g., Item 21 loading more strongly on self-kindness), which is expected in ESEM but suggests item-specific nuances.
Single-Bifactor Model: This model demonstrated the best fit across all indices (with the exception of TLI in the ML estimation, which was still acceptable). The general factor was well-defined (mean loading = .63). While specific component factors (particularly self-judgment and over-identification) showed weaker target loadings compared to the six-factor model, the model yielded fewer significant cross-loadings.
Reliability and Dimensionality: The bifactor model showed high reliability (Coefficient Omega $\omega$ = .97; Hierarchical Omega $\omega_H$ = .91). The general factor explained 94% of the reliable variance. Indices for unidimensionality (ECV = .63; PUC = .86) suggested the scale could be treated as unidimensional for total score calculation, though the values also aligned with the theoretical view of self-compassion as a system of distinct components.

Key Contributions
This study provides the first validation of the SCS factor structure specifically within an autistic adult population. By utilizing ESEM, the authors offer a more nuanced analysis than previous Confirmatory Factor Analysis (CFA) studies, allowing for cross-loadings that better reflect the complex interplay of self-compassion components. The findings replicate the superiority of the single-bifactor model observed in general population studies (Neff et al., 2019; Rakhimov et al., 2023) and extend this support to autistic adults. Furthermore, the study clarifies that while a total score is statistically justifiable, the six subscales also provide distinct, theoretically consistent information.

Significance and Claims
The authors claim that the SCS is a reliable and valid measure for autistic adults, supporting the use of both the total score and the six subscales in research and clinical contexts. The results suggest that the single-bifactor model best captures the construct, aligning with Neff's conceptualization of self-compassion as a unified system of interacting components. However, the authors explicitly note that the two-factor interpretation (separating CS and US) is not supported in this population.

The study concludes that while the bifactor model supports the calculation of a total score, researchers should calculate both general and subscale scores to fully understand self-compassion in autistic adults, given that autistic traits correlate more strongly with negative components. The authors modestly acknowledge limitations, including the exclusion of self-identified autistic individuals without formal diagnoses and those with high support needs or intellectual disabilities, which limits the generalizability of the findings to the broader autistic community. They recommend future research employ Bayesian SEM to test the two-bifactor model and conduct further studies on the scale's convergent and discriminant validity in this population.