The short answer
In the best general-adult sample we found (U.S. community adults aged 18–80, stratified by age, sex, and social class; N = 377), the mean LEAS total is 61.8 out of 100 (SD 10.8). That works out to ≈ 3.1 per scene. So the typical adult mostly writes at Level 3, naming a single emotion, and adds a blend (Level 4) on some scenes. Under a normal approximation, only ≈ 22 % average 3.5 or more per scene, and only ≈ 5 % average Level 4 or more. English-speaking university students run higher (≈ 3.0–3.6 per scene). The large French general-population sample runs lower (≈ 2.9). Clinical groups span roughly 2.3–3.8 per scene. Within the same study, they usually score 0–0.6 per scene below their controls, and some groups (anorexia, schizophrenia) show no deficit at all.
Owen's working belief was "most adults ≈ 3–3.5, clinical/teens ≈ 2–2.5". The data support the lower half of the adult range (≈ 3.1 is typical, and 3.5 is the top fifth). Clinical samples score below their own controls, but most still average above 2.5 (only 3 of 12 are at or below 2.5). No adult-format LEAS norms for teenagers were found. See Gaps.
1 · The bell curve
Assumptions. The curve is a normal approximation built from the published mean and SD. No paper publishes the raw score distribution, and real LEAS totals are bounded at 0–100 and may skew. The "out of N" counts apply that curve to the sample's N, so they are estimates and not counted participants. The U.S. scale-up assumes the U.S. community sample represents all U.S. adults. It was a 1990s sample from Tucson, AZ and Marshall, MN, so treat that number as illustration only. "Per scene" = total ÷ 20. A per-scene average of 4.0 does not mean every answer was Level 4, because totals mix levels across scenes.
2 · How to read this
The LEAS presents 20 short two-person scenes ("How would you feel? How would the other person feel?"). Each written answer is scored with the Lane et al. (1990) manual. Self and Other are each scored 0–4. The scene's total is the higher of the two, or 5 when both are Level 4 and clearly different from each other. Summed over 20 scenes, the total ranges 0–100. Many clinical studies use a 10-scene form (LEAS-10, range 0–50). This page converts every sample to a per-scene mean (total ÷ number of scenes) so the two forms can sit on one axis. That conversion is derived.
Scores are not directly comparable across languages or eras. Translations, scoring-manual versions, and how wordy people are all shift scores. For example, Japanese students scored 50 and same-age U.S. students scored 63.5 on matched scenes (Igarashi 2011). Treat cross-country differences as partly measurement.
Caveat: the LEAS measures how complex your written description of feelings in imagined scenes is. It is a proxy. It does not show what you notice in your body, say out loud, or do in a real conflict. People can score high through vocabulary, and score low through brevity.
The level ladder
3 · Demographic breakouts
Sex: women score higher in every published sample
Barrett et al. (2000) found women higher in all 7 samples they tested (U.S. community, four U.S. undergraduate pools, U.S. and German medical students). The gap held after controlling for verbal intelligence. The Japanese (Igarashi 2011) and French (Nandrino 2013) data replicate it. The usual gap is about 0.3–0.6 SD.
Who is in the sample: students vs community, by country
Per-scene means (bars) ± 1 SD (whiskers). Selective-university students (Yale, Australian university students) score highest. This partly reflects age and partly education.
Age and education: direction is consistent, group means not extractable
| Factor | Finding | Source |
|---|---|---|
| Age | Younger → higher. Lower scores in older age groups (French N = 750, five age bands, 17–84). | Nandrino 2013 |
| Age | Younger age correlated with higher geLEAS total, r = 0.28, p < .001 (German community, N = 208, ages 18–65). | Herpertz 2023 |
| Age | Main effect of age: younger participants higher (U.S. community norms, N = 381, mean age 42.7). | Chuning 2025 |
| Education | More education → higher. Lowest scores in the lowest-education group (French sample). | Nandrino 2013 |
| Education | Higher academic degree correlated with higher scores, r = 0.34, p < .001. | Herpertz 2023 |
| Education / SES | Main effects of education and SES, plus a sex × education interaction (U.S.). | Chuning 2025 |
| Verbal IQ | Vocabulary predicts LEAS (B ≈ .3–.4), but the sex gap survives controlling for it. | Barrett 2000 |
| Children | A separate instrument (LEAS-C) is used for ages 10–11. Girls score higher. Scores rise with age, but the metric is not comparable to adult totals. | Bajgar 2005 |
Age-band and education-band means from Nandrino 2013 (Table 2) and Chuning 2025 are behind paywalls and could not be read. Only their reported directions are shown here, and no values have been invented. The Herpertz 2023 sign convention is reported as "younger age positively associated".
Clinical vs comparison groups
Clinical groups do not fall at one fixed low level. The largest same-study gaps are chronic fatigue (−0.52 per scene), mixed psychiatric inpatients (−0.60, against an unmatched student group), and essential hypertension. Anorexia (Parling 2010) and schizophrenia (Maaßen 2021, conventional scoring) showed no deficit against matched controls. Study-to-study differences in scoring and country are as large as the diagnosis effects, so the meaningful comparison is clinical vs. control within each study.
4 · Comparison table (every number used)
| Sample | Group | N | Mean | SD | Metric | Per scene | Source |
|---|
derived marks values computed by this page: combined-sex French mean/SD (pooled from the published sex-specific means, SDs, and Ns), the pooled general-population curve, and all per-scene conversions.
5 · Gaps and honest limits
- No true population norm. The best U.S. sample (N = 377) was stratified, not random. It came from two 1990s Midwest/Southwest sites, and all participants were native English speakers without psychiatric history. The 2025 U.S. normative paper re-analyzes this same archive (N = 381). Its tables are paywalled.
- Normality is assumed. Only means and SDs are published, so tails (e.g., % above 80) are model estimates. A bounded, possibly left-skewed real distribution could put fewer people in the top tail than shown.
- Teen samples: no adolescent norms on the adult LEAS were found. Children use the LEAS-C (Bajgar 2005), which is scored differently.
- Some requested sources were not usable: Maroti's Swedish data are a chronic-fatigue study with N = 30 controls (included above), not population norms. Subic-Wrana 2005 (psychosomatic inpatients, N = 386) has no extractable total in open sources. Frewen 2008 (PTSD) used the TAS-20, not LEAS totals. The 2024 depression study (Müller et al., BMC Psychiatry) used computer scoring of self/other subscales only, and found depressed patients lower on Other awareness.
- Cross-sample pooling mixes countries and scoring eras. The pooled curve (U.S. + French general population, N = 1,127) is shown only as a sensitivity check.
Sources
Built Nov 13, 2026. Static page, no tracking. Chart.js is bundled locally.