Compassionate and Uncompassionate Self-Responding

What it is

Two items, roughly. "I try to be understanding towards the parts of myself I do not like" is a compassionate item. "I am disapproving of my own flaws" is an uncompassionate one. Answer the second one as somebody having a bad year, and your total self-compassion score falls, even if you scored well on every kind item.

The question is whether that fall is telling you about your self-compassion, or about your bad year.

In effect

Both sides, each stated as its advocate would state it

Peter Muris's case. Reverse-scoring self-criticism and calling the result self-compassion is a measurement trick. It guarantees the construct will correlate with lower distress, because distress has been built into the instrument. The negative items index vulnerability, not protection. Muris and Nicola Petrocchi separated the six parts and found the three harsh ones tied significantly more strongly to mental health problems than the three kind ones did. Report the parts separately, or at least as two groups, and the apparent power of self-compassion shrinks to something ordinary. A protective quality that dissolves when you stop reverse-scoring neuroticism is not a protective quality.

Kristin Neff's case. Compassionate and uncompassionate responding are two ends of one continuum with a neutral middle, not two independent things. Self-compassion is by definition a reduction in harsh self-judgment as well as an increase in kindness, so stripping the negative items removes half the construct. Her strongest empirical card is an analysis across twenty samples and 11,685 people, testing whether one general factor plus six specific ones fits how people answer. It does, and the general factor carries nearly all the reliable information in their answers. That supports one total score plus six subscales, and it does not support splitting into two separate composites. And the clinching point is one that scoring cannot explain: when people are trained in the practice, the kind subscales go up and the harsh ones go down at the same time, which is what a single dimension does.

Where we land, September 2026

Both are right, about different things.

Muris wins the measurement. A large part of the headline correlation between self-compassion and wellbeing is a questionnaire correlating with a near-copy of itself, and anybody quoting those correlations should say so.

Neff wins the thing itself. The training evidence moves both halves together, and the intervention trials measure depression, stress and anxiety directly, so they do not depend on the scoring argument at all.

The case for being kinder to yourself does not rest on the questionnaire. It rests on what happens when people do it.

Why this matters beyond one instrument

It is a clean worked example of a general problem. Build a measure that includes the absence of a bad thing, then discover that the measure predicts the absence of bad things. The prediction was installed at the design stage. Anybody reading a wellbeing finding is entitled to ask what the questionnaire had already assumed.

What it does not say

It does not say self-compassion is fake. The intervention trials measure outcomes directly, and they survive the scoring argument untouched.

It does not say Neff's scale should be abandoned. The disagreement is about how it is scored and reported, not about whether it measures anything.

It does not resolve the argument. This page is graded contested because the two sides are still publishing at each other, and both of them are making a real point.

It gives no reader a score to interpret. Nothing here is a self-test, and the whole subject of this page is why a total score from this instrument is harder to interpret than it looks.

It is not about self-criticism as a clinical matter. The claim that the harsh items are really measuring neuroticism is a measurement argument, and it is not a statement about anybody's mental health.

---