Two Versions of the Same Instrument
Kristin Neff's Self-Compassion Scale exists in two validated forms: the original 26-item long form (Neff, 2003) and a 12-item short form developed later by Raes, Pommier, Neff, and Van Gucht (2011). Both measure the same six subscales. The difference is how many items represent each one.
What the Short Form Trades Away
The 12-item version uses two items per subscale instead of the long form's four to five. Raes et al. (2011) found the short form correlates highly with the long form's total score, which is why it's used in research contexts where survey length is a hard constraint, large-scale studies, repeated-measures designs, or surveys bundling many instruments together.
The cost of that brevity shows up at the subscale level. With only two items measuring, say, isolation, a single ambiguous or oddly-worded response has twice the influence on that subscale score compared to the long form. The total score holds up well; the six individual subscale scores get noisier.
Why JobCannon Uses the 26-Item Long Form
The Self-Compassion Test's main value isn't just a single overall number, it's the subscale breakdown showing specifically where self-judgment, isolation, or over-identification are pulling a score down, and where self-kindness, common humanity, or mindfulness are already strong. That breakdown is the part the short form is weaker at delivering reliably.
The 26-item version takes a few minutes longer to complete than the 12-item short form would. For a test that exists to hand someone something actionable rather than just a single score, that tradeoff runs the other way from a research survey squeezed for length.
Both Are Legitimate
Neither version is "more correct." Researchers running large studies reasonably choose the short form. A test meant to give an individual person a detailed, subscale-level picture reasonably uses the long form. If you've seen a 12-item version elsewhere and wondered why this Self-Compassion Test has 26 questions, that's the reason, more items per subscale, more reliable individual breakdown.