Workplace PERMA-Profiler (and PERMA-Profiler)

Structural validity Lowcontestedevidence form: canonical

direct occupational samples; also a general sample, adults, student veterans, haemodialysis, autistic adults and adolescents (flag basis: occupational samples; large Chinese general sample; Australian adult study; US student-veteran sample; Chinese haemodialysis sample; autistic adults; stronger in adolescent adaptations)

state: assessed · rubric v1.6 · literature as of 2026-09-01 · grade confirmed 2026-09-03

The claimed five-factor structure is only partly and inconsistently supported, and the factor structure is the most contested aspect of the instrument. In occupational samples the correlated five-factor model has typically reached only marginal fit: the Japanese Workplace PERMA-Profiler gave CFI 0.892, TLI 0.858, RMSEA 0.105 and SRMR 0.051 (Watanabe et al. 2018), the Korean version CFI 0.909, TLI 0.881, RMSEA 0.110 and SRMR 0.054 (Choi et al. 2019), and the Chinese workplace version CFI 0.887, TLI 0.855, RMSEA 0.114 and SRMR 0.060 (Yang et al. 2024); all three show RMSEA above conventional thresholds. The Polish Workplace version is the clearest positive case, with the five-factor model reaching desired fit and outperforming a single-factor model in a large sample (N=1070) (Fortuna et al. 2025). Evidence from the general (non-workplace) PERMA-Profiler diverges further: an Australian adult study could replicate neither the theorised five-factor model nor a single-factor alternative and concluded the psychometric properties were insufficient (Ryan et al. 2019); a US student-veteran sample yielded a two-factor solution (emotional versus performance character strengths) (Umucu et al. 2019); and a Chinese haemodialysis sample extracted two dimensions (Qian et al. 2025). Where the five-factor model does hold, it is often as five highly correlated first-order factors: in a large Chinese general sample the first-order five-factor solution fitted better than a higher-order well-being factor (CFI 0.943 after modification) (Nie et al. 2024), and a French study confirmed both a correlated five-factor and a second-order model while its exploratory analysis suggested a more parsimonious three factors (Jourdan et al. 2026). In autistic adults a bifactor model with a dominant general factor fitted better than the five-factor model (Grosvenor et al. 2023). Five-factor support is stronger in adolescent adaptations (Fernandes et al. 2024; Waigel & Lemos 2023). Across studies the Engagement domain is the recurrent weak point.

Confidence note (legacy, first pass; not the basis of the grade): Low. Many studies but findings are inconsistent across populations and forms; five-factor fit is frequently marginal, competing two-factor, three-factor, single-factor and bifactor solutions recur, and the workplace form specifically has few confirmatory studies.

Convergent and discriminant validity Moderatecontestedevidence form: canonical

direct Workplace-form samples; also an autistic-adult study (flag basis: Japanese and Korean workers; autistic-adult study)

state: assessed · rubric v1.6 · literature as of 2026-09-01 · grade confirmed 2026-09-03

Convergent and discriminant patterns are generally in the expected direction, though discriminant separation from distress measures is sometimes weak. For the Workplace form in Japanese and Korean workers, well-being domains correlated moderately to strongly and positively with job satisfaction and work-related psychosocial factors and inversely with psychological distress (Watanabe et al. 2018), and positively with work engagement and life mental well-being while correlating negatively with burnout, occupational stressors and stress responses (Choi et al. 2019). For the general form, a large Chinese study reported convergent correlations of r=0.53 to 0.85 and discriminant correlations of r=-0.19 to -0.38 (Nie et al. 2024). A French study found strong positive associations with flourishing and perceived happiness (convergent) and weaker associations with anxiety, depression, loneliness and negative emotion (discriminant), while cautioning that some correlations with anxiety and depression nonetheless reached moderate to strong magnitude (Jourdan et al. 2026). An Australian study found moderate correlations with depression, anxiety and stress (r=-0.374 to -0.645) but no meaningful association with objectively measured physical activity or sleep (Ryan et al. 2019), and an autistic-adult study confirmed expected correlations with mental-health conditions and life satisfaction (Grosvenor et al. 2023).

Confidence note (legacy, first pass; not the basis of the grade): Moderate. Multiple studies across the general and workplace forms show consistent, theory-congruent convergent correlations; discriminant validity is adequate but repeatedly weaker against depression and anxiety, and no association with objective (non-self-report) markers.

Criterion validity: reference standard Absent (the property does not apply to this construct)untestedevidence form: canonical

absence type: category-error the property does not apply to this construct

state: assessed_absent · rubric v1.6 · literature as of 2026-09-01 · grade confirmed 2026-09-03 · under rubric v1.1: grade Low → Absent; status thin → untested; evidence state assessed → assessed_absent; indirectness direct → unflagged (re-read before first publication, correction C-0005)

Workplace wellbeing as the PERMA-Profiler defines it has no diagnostic or health reference standard, and no study retrieved in this pass validates the instrument against one.

Criterion validity: organisational Absent (searched; none found in the sweep to date)untestedevidence form: canonical

absence type: population-general searched; none found in the sweep to date

state: assessed_absent · rubric v1.6 · literature as of 2026-09-01 · grade confirmed 2026-09-03

No study retrieved in this pass links the instrument to organisational records such as sickness absence, staff turnover or medically diagnosed conditions. What exists is cross-sectional, self-reported association: workplace well-being domains were associated with job strain components, with job control and supervisor and coworker support correlating with all five domains while job demands related only to Engagement and Meaning (Yang et al. 2021); well-being domains, principally Positive emotion, Engagement and Relationships, were inversely associated with personal and work-related burnout in a Taiwanese multicentre sample (N=242) (Wu et al. 2025); Positive emotion was inversely associated with prolonged fatigue in a Chinese workforce sample (Yang et al. 2024); in a university-workplace study, workplace well-being mediated the relationship between psychological distress and subjective quality of life and was lower among staff with suspected psychological health conditions (Ng et al. 2025); and among rural veterans, well-being scores were positively associated with mental and physical health outcomes and were lower in those with service-connected disabilities (Umucu et al. 2024). All of these are concurrent associations, convergent rather than criterion evidence, and none is against an organisational outcome.

Internal consistency Highwell-establishedevidence form: canonical

direct Japanese, Korean and workers; also adults, haemodialysis, autistic-adult and adolescent samples (flag basis: Japanese workers; Korean workers; Chinese workers; Australian adults; Chinese haemodialysis sample; autistic-adult sample; adolescent adaptation)

state: assessed · rubric v1.6 · literature as of 2026-09-01 · grade confirmed 2026-09-03

Internal consistency is the instrument's strongest property and is generally acceptable to good, with the Engagement subscale the consistent exception. For the Workplace form, Cronbach's alpha ranged from 0.75 to 0.96 in Japanese workers (Watanabe et al. 2018), 0.70 to 0.95 in Korean workers (Choi et al. 2019) and 0.69 to 0.93 in Chinese workers (Yang et al. 2024); the Polish Workplace version met minimum reliability thresholds (Fortuna et al. 2025). For the general form, alphas were 0.79 to 0.88 in a large Chinese sample (Nie et al. 2024), the total scale reached alpha 0.914 in a Chinese haemodialysis sample (Qian et al. 2025), and an autistic-adult sample reported alpha 0.93 (Grosvenor et al. 2023). In Australian adults subscale alphas ranged 0.80 to 0.93 for all domains except Engagement (alpha 0.66) (Ryan et al. 2019); the French validation likewise noted lower reliability specifically for Engagement (Jourdan et al. 2026). Omega and composite reliability were reported as adequate in an adolescent adaptation (Fernandes et al. 2024). The three-item domains, especially Engagement, are the source of most reliability shortfalls. Sample sizes were 310 Japanese workers at baseline (Watanabe et al. 2018), 316 Korean workers (Choi et al. 2019), 312 Chinese workers (Yang et al. 2024), 1070 Polish workers (Fortuna et al. 2025), 1468 Chinese adults (Nie et al. 2024), 376 haemodialysis patients (Qian et al. 2025), 517 autistic adults (Grosvenor et al. 2023), 439 Australian adults (Ryan et al. 2019), 612 French-speaking adults, for whom the abstract gives no coefficient values (Jourdan et al. 2026), and 1197 Brazilian adolescents aged 11 to 19, for whom the abstract gives no omega value (Fernandes et al. 2024).

High precondition (rubric 1.6): the cited studies that meet the High precondition, each with a sample size and a statistic of this property.

Confidence note (legacy, first pass; not the basis of the grade): High. Many studies across multiple populations, forms and languages report consistent alpha values in the acceptable to good range, with a reproducible and openly reported weakness for the Engagement subscale.

Test-retest reliability Lowthinevidence form: mixed

direct Japanese, Chinese and Polish workers; also Chinese haemodialysis patients (flag basis: Japanese workers; Chinese workers; Polish workers; Chinese haemodialysis patients)

state: assessed · rubric v1.6 · literature as of 2026-09-01 · grade confirmed 2026-09-03

Test-retest reliability is thinly evidenced and this is a genuine gap, most acute for the Workplace form. The founding paper reports cross-time consistency for the general PERMA-Profiler but does not present a conventional retest coefficient in the abstract retrieved (Butler & Kern 2016). The Japanese Workplace validation collected a one-month follow-up (baseline N=310, follow-up N=86) and reported intraclass correlation coefficients of 0.75 to 0.96, but these ICCs are presented alongside internal-consistency alphas and the abstract does not cleanly separate a retest ICC per domain (Watanabe et al. 2018). The Chinese workplace version similarly reported ICCs of 0.70 to 0.92 in a reliability analysis rather than a dedicated retest design (Yang et al. 2024). The clearest temporal-stability evidence comes from non-workplace forms: the Polish Workplace study included a repeated-measure subsample (N=66) and reported strong stability (Fortuna et al. 2025); a French study reported good temporal stability for the total score and most subscales (Jourdan et al. 2026); and a Chinese haemodialysis study reported a retest correlation of r=0.764 (Qian et al. 2025). No dedicated occupational retest study was located, and retest intervals, when reported, are short (around one month). The absence of a robust, well-powered retest study specifically for the Workplace PERMA-Profiler is a flagged evidence gap.

Confidence note (legacy, first pass; not the basis of the grade): Low. A small number of short-interval studies with modest follow-up samples, mostly reporting ICCs bundled with internal consistency rather than purpose-designed retest coefficients; no dedicated, adequately powered workplace retest study located. Flag: near-absent for the Workplace form.

Measurement invariance Lowthinevidence form: canonical

general general-population sample; also adolescent and open-and-distance-learning student samples (flag basis: large Chinese general-population study; Brazilian adolescent adaptation; open-and-distance-learning students in Botswana)

state: assessed · rubric v1.6 · literature as of 2026-09-01 · grade confirmed 2026-09-03 · under rubric v1.1: indirectness indirect → general (re-read before first publication, correction C-0005)

Invariance testing exists but is concentrated in non-workplace populations and rarely reaches full scalar invariance in occupational samples. A large Chinese general-population study established factorial invariance across genders (Nie et al. 2024). A Brazilian adolescent adaptation demonstrated invariance across age groups and gender (Fernandes et al. 2024), and an adaptation among open-and-distance-learning students in Botswana examined invariance within a five-factor framework (Magare et al. 2022). For work-specific instruments in the wider PERMA family, the Spanish PERMA+4 Positive Functioning at Work scale showed measurement invariance across educational levels (Garcia-Selva et al. 2024). For the Workplace PERMA-Profiler itself, no study retrieved in this pass reports a full multigroup invariance analysis across sex, age, occupation or language, and the level reached (configural, metric or scalar) for occupational deployment is therefore not established. Cross-language equivalence is assumed rather than formally tested, since each translation is validated as a standalone instrument.

Confidence note (legacy, first pass; not the basis of the grade): Low. Invariance is established (up to scalar in some cases) for general-population and adolescent forms, but is absent for the Workplace PERMA-Profiler specifically; occupational and cross-occupation invariance was not located.

Responsiveness and MIC Absent (searched; none found in the sweep to date)untestedevidence form: canonical

absence type: population-general searched; none found in the sweep to date

state: assessed_absent · rubric v1.6 · literature as of 2026-09-01 · grade confirmed 2026-09-03

Responsiveness and a minimal important change threshold are essentially unestablished. No study retrieved in this pass reports a minimal important change, minimal detectable change tied to an anchor, or a formal responsiveness analysis for the Workplace PERMA-Profiler. Some validation studies computed measurement error and, in the Japanese case, measurement-error statistics alongside reliability (Watanabe et al. 2018), but these are not responsiveness or MIC estimates. Intervention and longitudinal PERMA studies exist in the wider literature, yet a defensible, citable MIC value for interpreting change scores in a workplace deployment was not located.