The bifactor structure of the Self-Compassion Scale: Bayesian approaches to overcome exploratory structural equation...

The rapidly expanding self-compassion research is driven mainly by Neff's (2003a, 2003b, 2023) six-factor Self-Compassion Scale (SCS). Despite broad agreement on its six-first-order factor structure, there is much debate on SCS's global structure (one- vs. two-global factors). Neff et al. (2019) argue for an exploratory structural equation model (ESEM) with six specific and one global bifactor (6ESEM + 1GlbBF) rather than two global factors (6ESEM + 2GlbBF). However, ESEM's methodological limitations precluded testing the appropriate 6ESEM + 2GlbBF, relying instead on a model combining ESEM and traditional confirmatory factor analysis (6ESEM + 2CFA). Although intuitively reasonable, this alternative model results in internally inconsistent, illogical interpretations. Instead, we apply recent advances in Bayesian SEM frameworks and Bayes structural equation models fit indices to test a more appropriate bifactor model with two global factors. This model (as does 6CFA + 2GlbBF) fits the data well, and correlations between compassionate self-responding (CS) and reverse-scored uncompassionate self-responding (RUS) factors (∼.6) are much less than the 1.0 correlation implied by a single bipolar factor. We discuss the critical implications for theory, scoring, and clinical application for the SCS that previously were inappropriately based on this now-discredited 6ESEM + 2GlbCFA. In applied practice, we endorse using scores representing the six SCS factors, total SCS, and CS and RUS components rather than relying solely on one global factor. Our approach to these issues (dimensionality, factor structure, first-order and higher order models, positive vs. negatively oriented constructs, item-wording effects, and alternative estimation procedures) has wide applicability to clinical measurement (see our annotated bibliography of 20 instruments that might benefit from our approach). (PsycInfo Database Record (c) 2023 APA, all rights reserved).