The Self-Compassion Scale: A Re-evaluation

I am currently working in a clinical psychology lab as part of my honours project. One of the traits we’re measuring in our participants is self-compassion, which is defined as the “recognition of distress or suffering and the desire to alleviate it” [1]. We’ve been using the Self-Compassion Scale (SCS) to measure it, as it was developed by Dr. Kristin Neff in 2003, and remains the only measure of self-compassion to this date [2]. The scale itself has three components, each one with a positive and a negative axis. There is common humanity, which means seeing one’s failures as a part of the human condition, and not feeling isolated. Mindfulness which includes having a balanced awareness of the present experience, and self-kindness which means treating oneself with understanding. The 26 items in the scale fall into each of the positive and negative axes of these measures.

interpretation of Dr. Neff’s Proposed Model (graphic by author).

When the scale first came out, Dr. Neff proposed that the negative items should be reverse-coded and added on to the positive items for one whole measure of “self-compassion”. This is how our lab has been operating. However, this past summer a paper came out challenging this total sum analysis. Lopez et al., ran their own large community sample of participants (n = 1736) and performed a confirmatory factor analysis where they found that the six-factor structure with the all-encompassing higher order of self-compassion could not be replicated [3]. They performed an exploratory factor analysis with the data and found that a two-factor model explained approximately 45.4% of the variance [3]. This two-factor model divided the positive and negative axes of these measures, as though they were separate concepts. It appeared that having positive self-compassion was only weakly negatively correlated with having negative self-compassion (an accumulation of traits like over-identification, self-judgement and isolation) [3]. In addition, both of these measures had different relations with measures of wellness like depressive symptoms, perceived stress, positive affect, and rumination [3].

Interpretation of Lopez et al’s proposed two-factor model (graphic by author).

One of the reasons why the first model might not have been replicable is because Dr. Neff developed and validated it with a college student sample [2]. When tested in a wider demographic of a community sample of adults, the model did not hold, and a new two-factor model appeared to account for more of the variance. These two wildly different results are a good example of why study design, and especially participant selection, are important factors in psychological testing. Researchers should carefully assess how their chosen sample is likely to affect their results, especially when they are developing a brand new scale.

References:

1. Jazaieri H, Jinpa GT, McGonigal K, Rosenberg EL, Finkelstein J, Simon-Thomas E, et al. Enhacing compassion: a randomized controlled trial of a compassion cultivation training program. J Happiness Stud. 2013; 14: 1113–1126.
2. Neff, KD. Development and validation of a scale to measure self-compassion. Self Identity. 2003; 2:85–101.
3. Lopez A, Sanderman R, Smink A, Zhang Y, van Sonderen E, Ranchor A, Schroevers MJ. (2015) A Reconsideration of the Self-Compassion Scale’s Total Score: Self-Compassion versus Self-Criticism.