In plain language
The Self-Compassion Scale (SCS) is the most widely used measure of self-compassion. Researchers agree it captures six specific facets (like self-kindness, mindfulness, and self-judgment), but there has been a long-running dispute about the big picture: does the scale measure one overall self-compassion dimension, or two distinct dimensions—how compassionately you treat yourself and how harshly (uncompassionately) you treat yourself?
Earlier work by Neff and colleagues favoured a single global factor, but the statistical method they used (exploratory structural equation modeling, ESEM) could not actually fit the crucial rival model with two global factors, forcing a hybrid approach that this paper shows produces internally inconsistent, illogical results. Reanalyzing the same data with recent advances in Bayesian structural equation modeling, the authors were able to test the proper bifactor model with six specific factors and two global factors.
That two-global-factor model fit the data well, and compassionate self-responding and (reverse-scored) uncompassionate self-responding correlated only about .6—far from the perfect correlation a single bipolar dimension would require. The practical upshot for researchers and clinicians: report the six facet scores, a total score, and separate compassionate and uncompassionate component scores, rather than relying on one global score alone. The methodological lessons extend well beyond this scale to many clinical instruments.
Key findings
- The widely cited model supporting a single global self-compassion factor (6ESEM + 2GlbCFA) produced internally inconsistent, illogical interpretations because ESEM cannot properly estimate a bifactor model with two global factors.
- Using Bayesian structural equation modeling, the appropriate bifactor model with six specific factors and two global factors (compassionate and uncompassionate self-responding) fit the data well.
- Compassionate self-responding and reverse-scored uncompassionate self-responding correlated about .6—substantially below the correlation of 1.0 implied by a single bipolar global factor, supporting two distinguishable dimensions.
- The authors recommend that applied users score the six SCS facets, the total SCS, and the separate compassionate and uncompassionate components, rather than relying solely on one global factor.
- Conclusions about theory, scoring, and clinical application previously based on the discredited hybrid model need re-evaluation.
- The paper provides an annotated bibliography of 20 clinical instruments facing similar dimensionality, item-wording, and positive-versus-negative-construct issues that could benefit from the same Bayesian approach.
How to cite
APA
Marsh, H. W., Fraser, M. I., Rakhimov, A., Ciarrochi, J., & Guo, J. (2023). The bifactor structure of the Self-Compassion Scale: Bayesian approaches to overcome exploratory structural equation modeling (ESEM) limitations. Psychological Assessment, 35(8), 674–691. https://doi.org/10.1037/pas0001247
BibTeX
@article{marsh2023bifactor,
title = {The Bifactor Structure of the Self-Compassion Scale: Bayesian Approaches to Overcome Exploratory Structural Equation Modeling (ESEM) Limitations},
author = {Marsh, Herbert W. and Fraser, Madeleine I. and Rakhimov, Arman and Ciarrochi, Joseph and Guo, Jiesi},
journal = {Psychological Assessment},
volume = {35},
number = {8},
pages = {674--691},
year = {2023},
doi = {10.1037/pas0001247}
}
Related work
- All publications by Joseph Ciarrochi (searchable, with free PDFs)
- Process-Based Therapy & Idionomic Analysis
Author: Joseph Ciarrochi (ORCID 0000-0003-0471-8100). Free copy hosted with permission for scholarly use. Please cite the published version via the DOI above.