Written and maintained by CASRAI Editorial Board
Published · Last updated
On this page
Definition in One Paragraph
The Self-Compassion Scale (SCS) is a 26-item psychometric instrument developed by Dr. Kristin Neff (2003) to measure how individuals relate to themselves during experiences of failure, inadequacy, and personal suffering. Rather than assessing self-esteem through evaluative comparison, the instrument conceptualizes self-compassion across three core bipolar dimensions consisting of six subscales: Self-Kindness versus Self-Judgment, Common Humanity versus Isolation, and Mindfulness versus Over-Identification. Respondents rate items on a 5-point Likert scale from 1 (Almost Never) to 5 (Almost Always), providing researchers with six specific facet scores and a psychometrically validated overarching total score.
Formula with Every Symbol Explained
Scoring the SCS requires inverting 13 negative items across three subscales before computing arithmetic subscale means and the global composite index.
The 26 items are allocated across six subscales:
- Self-Kindness (SK): Items 5, 12, 19, 23, 26 (positive)
- Self-Judgment (SJ): Items 1, 8, 11, 16, 21 (negative, reverse-scored)
- Common Humanity (CH): Items 3, 7, 10, 15 (positive)
- Isolation (IS): Items 4, 13, 18, 25 (negative, reverse-scored)
- Mindfulness (MI): Items 9, 14, 17, 22 (positive)
- Over-Identification (OI): Items 2, 6, 20, 24 (negative, reverse-scored)
Negative subscale items are linearly inverted:
x_rev = 6 – x
Where x is the raw rating (1 to 5) and x_rev is the recoded score.
Each subscale mean S_k is calculated as:
S_k = (1 / n_k) × Σ x_i
Where n_k is item count (5 for SK and SJ; 4 for CH, IS, MI, and OI), and x_i represents scored item ratings.
The overall total score (SCS_total) is calculated as the grand mean across the six subscales:
SCS_total = (S_SK + S_SJ_rev + S_CH + S_IS_rev + S_MI + S_OI_rev) / 6
Averaging the subscale means assigns equal weight (16.67%) to each theoretical facet, preserving the 1.00 to 5.00 metric.
Worked Example with Real Numbers and Arithmetic Shown
Consider an adult participant completing the 26-item SCS with the following raw ratings:
| Subscale | Items | Raw Ratings | Scored Ratings | Subscale Mean |
|---|---|---|---|---|
| Self-Kindness | 5, 12, 19, 23, 26 | [3, 4, 3, 4, 3] | [3, 4, 3, 4, 3] | 17 / 5 = 3.40 |
| Self-Judgment (Rev) | 1, 8, 11, 16, 21 | [4, 3, 3, 4, 4] | [2, 3, 3, 2, 2] | 12 / 5 = 2.40 |
| Common Humanity | 3, 7, 10, 15 | [3, 4, 4, 3] | [3, 4, 4, 3] | 14 / 4 = 3.50 |
| Isolation (Rev) | 4, 13, 18, 25 | [4, 4, 3, 3] | [2, 2, 3, 3] | 10 / 4 = 2.50 |
| Mindfulness | 9, 14, 17, 22 | [4, 3, 4, 4] | [4, 3, 4, 4] | 15 / 4 = 3.75 |
| Over-Identification (Rev) | 2, 6, 20, 24 | [4, 3, 3, 4] | [2, 3, 3, 2] | 10 / 4 = 2.50 |
Calculate the overall composite by taking the grand mean of the six subscale means:
Sum of Subscale Means = 3.40 + 2.40 + 3.50 + 2.50 + 3.75 + 2.50 = 18.05
SCS_total = 18.05 / 6 = 3.01
By comparison, averaging all 26 scored items directly yields 78 / 26 = 3.00. The subscale mean approach prevents 5-item subscales from mathematically overweighting 4-item subscales.
Interpretation: What the Result Means
Total SCS scores range from 1.00 to 5.00, with higher scores reflecting greater trait self-compassion. Standard empirical research tiers established by Neff classify individual scores as follows:
| Total Score | Classification | Empirical Characteristics |
|---|---|---|
| 1.00 – 2.49 | Low | Marked self-criticism, perceived isolation, and cognitive over-identification with negative affect. |
| 2.50 – 3.50 | Moderate | Typical adult norm; situational self-compassion that fluctuates across life stressors. |
| 3.51 – 5.00 | High | Sustained emotional warmth, recognition of shared human fallibility, and balanced mindfulness. |
The scale exhibits robust psychometric properties, with full-scale Cronbach’s alpha exceeding .90 (original validation α = .92; 3-week test-retest r = .93). Construct evaluation demonstrates verified convergent and discriminant validity: SCS correlates negatively with depression (r = -.51) and anxiety (r = -.65), and positively with life satisfaction. Although correlating moderately with global self-esteem (r = .59), it remains unrelated to narcissism.
Assumptions and Common Mistakes
Researchers should account for several critical measurement assumptions when administering the SCS:
1. Rejecting the Two-Factor Fallacy
A frequent error is bifurcating the instrument into separate “Compassionate” and “Uncompassionate” composite scores. Bifactor Exploratory Structural Equation Modeling across 20 international samples (Neff et al., 2019; N = 11,685) proved that a single general self-compassion factor explains 95% of reliable variance. Two-factor models exhibit poor fit; researchers should use either the total score or the six subscales. To test a proposed factor structure against your own data, fit it as a confirmatory factor analysis model rather than relying on subscale correlations.
2. The Short Form (SCS-SF) Subscale Trap
The 12-item SCS-SF (Raes et al., 2011) correlates strongly with the long form (r ≥ .97, α ≥ .86). However, because it contains only two items per subscale, subscale reliabilities are inadequate. The SCS-SF must only be used to report an overall total score, never subscale profiles.
3. Floor and Ceiling Effects
In community cohorts, scores follow a normal distribution (M ≈ 3.05, SD ≈ 0.60). In acute psychiatric cohorts, researchers must evaluate potential floor and ceiling effects, as severe depression can compress Self-Kindness scores near 1.00.
4. Omission of Reverse Scoring
Failing to invert the 13 negative items (1, 2, 4, 6, 8, 11, 13, 16, 18, 20, 21, 24, 25) invalidates the composite, severely compromising empirical construct validity.
How to Report It (APA Example)
Manuscripts adhering to APA 7th edition guidelines should report the scale version, response options, sample reliability (α or ω), and descriptive statistics.
Method Section
“Self-compassion was assessed using the 26-item Self-Compassion Scale (SCS; Neff, 2003). Responses were recorded on a 5-point Likert scale from 1 (Almost Never) to 5 (Almost Always). The instrument yields six subscales: Self-Kindness (5 items), Self-Judgment (5 items), Common Humanity (4 items), Isolation (4 items), Mindfulness (4 items), and Over-Identification (4 items). After reverse-scoring negative items, the total score was computed as the grand mean of the six subscales. In the present sample, the scale demonstrated excellent internal consistency (Cronbach’s α = .92, McDonald’s ω = .93).”
Results Section
“Participants reported moderate levels of global self-compassion (M = 3.08, SD = 0.61). Subscale means were M = 3.12 (SD = 0.73) for Self-Kindness, M = 2.87 (SD = 0.80) for Self-Judgment (reverse-scored), M = 3.24 (SD = 0.76) for Common Humanity, M = 2.95 (SD = 0.84) for Isolation (reverse-scored), M = 3.36 (SD = 0.70) for Mindfulness, and M = 2.90 (SD = 0.81) for Over-Identification (reverse-scored).”
How to Calculate It in Excel, R, or Python
Researchers can automate scoring across common data analysis environments using these verified implementations.
1. Microsoft Excel
With raw items in columns A2:Z2:
# Reverse-score negative items (Items 1, 2, 4, 6, 8, 11, 13, 16, 18, 20, 21, 24, 25):
=6 - A2
# Subscale Means:
Self-Kindness: =AVERAGE(E2, L2, S2, W2, Z2)
Self-Judgment (rev): =AVERAGE(6-A2, 6-H2, 6-K2, 6-P2, 6-U2)
Common Humanity: =AVERAGE(C2, G2, J2, O2)
Isolation (rev): =AVERAGE(6-D2, 6-M2, 6-R2, 6-Y2)
Mindfulness: =AVERAGE(I2, N2, Q2, V2)
Over-Ident (rev): =AVERAGE(6-B2, 6-F2, 6-T2, 6-X2)
# Grand Total Score:
=AVERAGE(SK_Cell, SJ_Cell, CH_Cell, IS_Cell, MI_Cell, OI_Cell)
2. R Snippet
Using base R for item columns Q1:Q26:
rev_items <- c(1, 2, 4, 6, 8, 11, 13, 16, 18, 20, 21, 24, 25)
df_scored <- df[, paste0("Q", 1:26)]
df_scored[, paste0("Q", rev_items)] <- 6 - df_scored[, paste0("Q", rev_items)]
sk <- rowMeans(df_scored[, paste0("Q", c(5, 12, 19, 23, 26))], na.rm = TRUE)
sj <- rowMeans(df_scored[, paste0("Q", c(1, 8, 11, 16, 21))], na.rm = TRUE)
ch <- rowMeans(df_scored[, paste0("Q", c(3, 7, 10, 15))], na.rm = TRUE)
is <- rowMeans(df_scored[, paste0("Q", c(4, 13, 18, 25))], na.rm = TRUE)
mi <- rowMeans(df_scored[, paste0("Q", c(9, 14, 17, 22))], na.rm = TRUE)
oi <- rowMeans(df_scored[, paste0("Q", c(2, 6, 20, 24))], na.rm = TRUE)
df$scs_total <- rowMeans(cbind(sk, sj, ch, is, mi, oi), na.rm = TRUE)
3. Python Snippet
Using pandas to recode and calculate composites:
import pandas as pd
rev_cols = [f"Q{i}" for i in [1, 2, 4, 6, 8, 11, 13, 16, 18, 20, 21, 24, 25]]
df_scored = df[[f"Q{i}" for i in range(1, 27)]].copy()
df_scored[rev_cols] = 6 - df_scored[rev_cols]
subscales = {
"sk": [5, 12, 19, 23, 26],
"sj": [1, 8, 11, 16, 21],
"ch": [3, 7, 10, 15],
"is": [4, 13, 18, 25],
"mi": [9, 14, 17, 22],
"oi": [2, 6, 20, 24]
}
sub_means = pd.DataFrame({
k: df_scored[[f"Q{i}" for i in v]].mean(axis=1)
for k, v in subscales.items()
})
df["scs_total"] = sub_means.mean(axis=1)
Frequently Asked Questions (FAQ)
What is the difference between the 26-item SCS and 12-item SCS-SF?
The 26-item SCS assesses six subscales with 4 to 5 items each, permitting subscale analysis. The 12-item SCS-SF selects two items per facet for brief administration (r ≥ .97 with long form) and is validated solely for global total score reporting.
Is the Self-Compassion Scale free for research?
Yes. Dr. Kristin Neff grants open permission for non-commercial academic research, student theses, and clinical evaluations without licensing fees, provided the 2003 validation paper is formally cited.
How does self-compassion differ from self-esteem?
While sharing moderate correlation (r ≈ .59), self-esteem hinges on competitive evaluation and perceived superiority, correlating with narcissism. Self-compassion is non-evaluative and remains unrelated to ego defense or downward social comparison.
Can researchers compute separate positive and negative composite scores?
No. Bifactor ESEM testing across 20 international samples (Neff et al., 2019) demonstrated that a general self-compassion factor accounts for 95% of reliable variance. Two-factor splits fail empirical fit criteria and lack psychometric justification.
What test-retest reliability has been established for the SCS?
In the original validation study by Neff (2003), the SCS demonstrated strong test-retest reliability of r = .93 across a 3-week interval, demonstrating high temporal stability.
Sources
- Neff, K. D. (2003). The development and validation of a scale to measure self-compassion. Self and Identity, 2(3), 223–250. URL: https://doi.org/10.1080/15298860309027. Read: 2026-11-10.
- Raes, F., Pommier, E., Neff, K. D., & Van Gucht, D. (2011). Construction and factorial validation of a short form of the Self-Compassion Scale. Clinical Psychology & Psychotherapy, 18(3), 250–255. URL: https://doi.org/10.1002/cpp.702. Read: 2026-11-10.
- Neff, K. D., Tóth-Király, I., Yarnell, L. M., Arimitsu, K., Castilho, P., Ghorbani, N., … & Mantzios, M. (2019). Examining the factor structure of the Self-Compassion Scale in 20 diverse samples: Support for use of a total score and six subscale scores. Psychological Assessment, 31(1), 27–45. URL: https://doi.org/10.1037/pas0000629. Read: 2026-11-10.
- Self-Compassion Research Scales Repository. (2026). The Self-Compassion Scale (SCS) and Scoring Protocols. Dr. Kristin Neff Lab. URL: https://self-compassion.org/self-compassion-scales/. Read: 2026-11-10.







