Source
Curry, Jones Chesters & Van Lissa (2019), LIT-496, read in NOTE-412 from the PsyArXiv manuscript. Yilmaz, Harma & Doğruyol (2021), LIT-513, read in NOTE-397: the same instrument, translated, by a team outside the programme.
What was actually shown
The instrument. The MAC-Q copies the MFQ’s two formats. Relevance asks how relevant a consideration is when deciding right and wrong. Judgement asks for agreement with moral statements. The authors wrote 57 Relevance and 97 Judgement items from the theory, then kept three per domain, chosen for theoretical consistency and item-total correlation within the domain.
UK, n = 1,392 (LIT-496, Study 1). Relevance, seven factors: RMSEA .03, CFI .98, SRMR .03. Judgement, seven factors: RMSEA .06, CFI .87, SRMR .07, which the paper’s own CFI criterion (above .90) rates unacceptable. Both formats as seven domains: CFI .73. The best combined model keeps fourteen correlated factors, seven per format (CFI .93).
US, n = 469 (LIT-496, Study 3a). Relevance gives seven factors. Judgement gives five by parallel analysis (Family, Group/Reciprocity, Martial Virtues, Fairness, Property), and seven only when forced.
Turkey, n = 445 (LIT-513, Study 1). Relevance, seven factors: CFI .949, RMSEA .053, SRMR .046. A single factor fits far worse (CFI .682). Judgement: CFI .828. Full form: CFI .752.
What could have come out otherwise. The Relevance items could have collapsed to fewer factors in a new language and population. In Turkey they did not, though a second-order two-factor model also fits (CFI .937).
What this does not say
- That the seven are distinct moral kinds. The domains were written into the items, which were then selected for fit within each domain, so recovering seven factors is partly a check on construction. The latent correlations in Turkey run high: fairness–property .874, reciprocity–fairness .808, family–group .800.
- That the Judgement format measures the same seven. It does not fit in any of the three samples, and the paper’s description of the UK result as “a good fit for both scales” (p. 23) overstates it.
- That morality is only about cooperation. The MAC-Q has no non-cooperative item, so it cannot test that. Where non-cooperative items appear, the MFQ’s purity items, they are moralized (THEORY-632).
- That the MAC-Q outperforms the MFQ. It does on alphas, retest and Relevance fit. On criterion and incremental validity the authors call the two “broadly comparable” (p. 38).
- That relevance tracks cooperation. In Turkey the seven subscales explain 5.1% of self-reported prosocial intentions. Group is the only unique predictor, and it is negative (β = −.17).
Connections
- The account the instrument operationalizes is THEORY-468. The ethnographic test of the same seven is THEORY-653.
- The UK data cannot be shared, because the consent form “inadvertently precluded data sharing” (LIT-496, endnote vii). The Turkish replication is the only independent check.