Contribution
- A theoretical revision of MFT: Fairness is split into Equality and Proportionality, taken from Rai and Fiske’s relationship regulation (pp. 6–8).
- A new 36-item instrument developed from the start in several cultures, with a published item list and scoring key (Appendix).
- Evidence that the higher-order structure of the foundations varies by population (Study 2 networks), and that the MFQ-2 predicts related constructs better than the MFQ-1 (Study 3).
Key insight
The five-foundation model was measured almost entirely in English on WEIRD internet samples, and the MFQ-1 is not invariant across cultures. Building the instrument in many populations at once, with Fairness split, yields scales that compare across nations for five of six foundations, and shows that the familiar two-cluster structure of moral values is a WEIRD pattern, not a universal one.
Assumptions
- Six foundations, defined in Table 2 (pp. 11–12). Care: avoiding emotional and physical damage to another individual. Equality: equal treatment and equal outcome. Proportionality: reward in proportion to merit or contribution. Loyalty: cooperating with ingroups and competing with outgroups. Authority: deference to legitimate authorities and defence of traditions. Purity: avoiding bodily and spiritual contamination and degradation.
- Proportionality = equity. The paper uses “proportionality” for equity because “equity” now means equality of outcomes in some usage (fn. 3).
- Samples. Online panels (CloudResearch, Qualtrics), stratified by age, gender and, where possible, ideology; educated enough to complete an online survey (p. 65).
- WEIRDness is Muthukrishna et al.’s cultural distance from the U.S.
Key results
- Development (Studies 1a–1c, N = 3,360; U.S., India, Iran, Ecuador, China). 116 items cut to 90, 71 and then 50; Iran’s sample stopped at 100 of a planned 200 because sanctions prevented paying participants (fn. 4).
- Structure (Study 2, N = 3,902, 19 nations, about 205 each, May 2021). ESEM on 36 items: CFI .979, RMSEA .024. MGCFA across nations: CFI .896, SRMR .070, taken as configural invariance. ω ranges .73–.95; Purity is lowest (.73 in Japan).
- Invariance (Table 6, p. 30). Alignment non-invariance of intercepts: Care 5.3%, Equality 21.9%, Proportionality 11.4%, Loyalty 24.6%, Authority 16.7%, Purity 39.5%. Only Purity exceeds the 25% threshold; the “temple” item is the most non-invariant.
- Equality and Proportionality. Mean r = .21 (SD .13) across nations, from .47 (UAE) to −.04 (Belgium); more orthogonal in WEIRDer nations.
- Cultural distance (p. 32). Distance from the U.S. correlates with all six; most strongly with Purity (r = .43), then Loyalty (.25).
- Gender. Women higher on Equality and Purity; men higher on Proportionality, Loyalty and Authority; no significant Care difference.
- Religion. Purity’s association with religiosity is the largest (B = .75); cross-religion variation exceeds cross-nation variation.
- Ideology. Care and Equality go with the left; Proportionality, Loyalty, Authority and Purity with the right.
- Networks (pp. 38–41). In the U.S., Care–Fairness and the binding three form separate clusters. Across the 19 nations, exploratory graph analysis finds one dimension in most and two in three (Ireland, New Zealand, Peru). Authority is the most central node in 14 of 19.
- Validity (Study 3, N = 1,410). 17 of 18 preregistered correlations with criterion scales are supported (the miss: Authority with left-wing authoritarianism). Mean adjusted R² over 30 outcomes: MFQ-2 .37, MFQ-1 .30; adding the MFQ-2 to the MFQ-1 adds 13.7% on average.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Equality and Proportionality are distinct foundations | moderate | weak correlation in 19 nations; separate criterion validities (redistribution vs. merit preference) |
| C2 | The MFQ-2’s six-factor structure holds across cultures | moderate | ESEM and MGCFA fit; four items load more strongly on Authority than on their own factor (Corrections) |
| C3 | MFQ-2 scores can be compared across nations | moderate for five foundations, not for Purity | alignment results, Table 6 |
| C4 | Purity and Loyalty are the “least WEIRD” foundations | weak–moderate for Purity | the Purity correlation (.43) uses the subscale that fails invariance; 19 nations, Conley SEs |
| C5 | The individualizing–binding split is a WEIRD phenomenon | moderate | network structure in 19 nations; exploratory |
| C6 | The MFQ-2 predicts criterion variables better than the MFQ-1 | moderate | Study 3, mostly U.S. participants (82%) |
| C7 | Curry et al. 2019 omitted Purity, an important moral foundation | assertion with indirect support | p. 64; supported only by C4 and religiosity results, not by a test of MAC |
Concepts
- MFQ-2 — 36 statements rated 1 (does not describe me at all) to 5 (describes me extremely well); the MFQ-1’s “Relevance” format is dropped (p. 13).
- nomological network — the pattern of relations among foundations and with other constructs, estimated here as a psychometric network.
- alignment method — multi-group factor analysis that estimates group means while allowing a few non-invariant parameters; 25% non-invariance is the accepted ceiling.
- WEIRD cultural distance — Muthukrishna et al.’s distance of a population from the U.S. on cultural values.
Connections
- MFT. Extends Graham et al. 2013 (LIT-499), reversing its claim that there is no Equality foundation (p. 114 of that chapter).
- Relationship regulation. The Equality–Proportionality split and their definitions are Rai and Fiske’s (LIT-507), cited throughout pp. 6–8 and 60.
- Morality-as-cooperation.
- LIT-047 (NOTE-034) is named on p. 64 as a rival typology that “left out Purity”.
- LIT-047’s footnote 1 said the MFQ has no reciprocity items. The MFQ-2 defines Equality as including “balanced reciprocity” (p. 60), but none of its six Equality items is about reciprocating; the nearest is a Proportionality item about cheaters being punished.
- The MAC-Q, filed in this batch as LIT-496, found MFQ-1 purity items rated as relevant as loyalty and authority items in UK samples. The MFQ-2 adds that Purity endorsement is higher in less-WEIRD populations.
- Monism. Baumard et al. (LIT-508, “mutualistic fairness”) and Gray and Schein (LIT-494, “all morality is harm”) are named as monist theories put forward in WEIRD populations by WEIRD researchers (p. 55).
- Purity’s definition. The paper drops the MFQ-1 item that mentioned God because it proxied religiosity, and reports that Purity still forms one factor across sexual, linguistic and “natural medicine” content (p. 58). LIT-101 (NOTE-132) argues the puritanical part of purity is cooperative; the MFQ-2 does not separate that part from contamination.
Bearing on the record
- NOTE-034, C7. This paper’s six do not contain MAC’s seven, nor the reverse. MAC has bravery and property, absent here; the MFQ-2 has Purity, Care for any sufferer, and Equality as an outcome preference, absent from MAC. The paper’s charge that MAC omitted Purity is the substantive disagreement, and it is the one MAC’s own questionnaire concedes (as reported in LIT-496).
- NOTE-034’s open question. The MFQ-2 measures endorsement of purity, not moralization coded from ethnography. It shows purity is endorsed more outside WEIRD populations, but its Purity means are the ones the authors say not to compare across groups without caution.
- No THEORY document cites it. No ML instruction.
Note of 2026-10-16: THEORY-435, THEORY-539 and THEORY-613 now cite LIT-500 as a source.
Limitations
- Online, literate, mostly urban samples; no small-scale societies (p. 65).
- About 200 per nation; 19 nations for country-level correlations, with spatial dependence handled by Conley standard errors but not formally tested (p. 65).
- Purity fails intercept invariance, so its cross-national comparisons are the least secure, and they carry the paper’s headline about WEIRDness.
- Study 3 is 82% U.S., so the MFQ-2’s predictive advantage is shown mainly in one country.
- Candidate foundations (liberty, honor, honesty, ownership, efficiency) remain unmeasured (p. 64).
Open questions
- Does Purity’s association with cultural distance survive a Purity scale that is invariant?
- Would a reciprocity scale (returning favours, punishing defectors) load with Equality, with Proportionality, or on its own? That would settle whether MFT’s Fairness conflated reciprocity with distribution, as LIT-047 claims.
- Do bravery and property, MAC’s domains with no MFT counterpart, form factors when written into a cross-cultural pool built the same way?
Corrections
- none to a seeded skim (there was no seed)
- Cross-loading items. The paper says “some Purity items” load below .30 with cross-loadings (p. 24). By Table 4, four items load more strongly on Authority than on their own foundation: “the human body should be treated like a temple” (Purity .21, Authority .37); “the strength of a sports team comes from the loyalty of its members” (Loyalty .10, Authority .36); “proud when a person in their community wins in an international competition” (Loyalty .27, Authority .28); “happy when people are recognized on their merits” (Proportionality .27, Authority .40). The problem is not only Purity’s.
- Dimension count. EGA is said to reveal “one dimension in 16 samples and two dimensions in four samples (Ireland, New Zealand, and Peru)” (p. 40). That is 20 for 19 nations, and three are named.
- Direction label. “Cultural distance from the U.S. (higher WEIRDness)” (p. 32) labels distance as WEIRDness; greater distance from the U.S. is lower WEIRDness, which is how the results are then read. These three may be fixed in the version of record.