Source
Curry, Mullins & Whitehouse (2019), LIT-047 (Current Anthropology 60(1), 47–69, with four invited comments and a Reply), read in full in NOTE-034, with the online supplement.
What was actually shown
The corpus. eHRAF searches on the 60 societies of the HRAF Probability Sample Files gave 3,460 paragraphs from 603 sources dated 1704–2009 (median 1967). The two non-blind authors screened them to 1,426 germane paragraphs; 761 of those carry valence, giving 962 observations.
Valence (Table 2). Positive against negative: kin 214/0, group 127/0, reciprocity 151/0, bravery 101/0, respect 133/0, fairness 17/0, property 218/1. The one negative is the Chuuk admiration of open theft as a display of dominance.
Presence (Table 3). Societies with at least one positive observation, of 60: property 54, kin 47, group 44, reciprocity 43, respect 35, bravery 32, fairness 9. A society shows 4.4 of the seven on average. Fairness is rarer than every other domain, and property commoner than bravery, respect and fairness (repeated-measures GLM, Bonferroni, P ≤ .001).
Coverage. The number of morals observed per society correlates with its total eHRAF paragraph count (r = .43, P = .001), so much of the variation in presence is variation in how much was written.
What could have come out otherwise. A society could have been described as despising one of the seven, and the presence counts could have been uniform. The first almost never happened; the second did not.
What this does not say
- That the seven are universal moral rules. The behaviours are coded in virtue terms (“helping”, “respecting”, “being fair”), so a negative is close to self-contradictory. Cases where one of the seven was condemned for costing a larger cooperation (fn. 3), and 18 value conflicts (fn. 11), were not coded as negative. The 99.9% mostly measures what the codebook could see.
- That morality is about cooperation and nothing else. No non-cooperative domain (purity, sexual conduct, treatment of the dead, harm to strangers) was coded for comparison. Curry 2016 (LIT-503) names that comparison as the theory’s own prediction, and this study does not run it.
- That the seven occur equally in every region. The regional ANOVAs are non-significant at about ten societies per region; that is low power, not equality.
- That the codes are reliable. The screening from 3,460 to 1,426 paragraphs was done by the non-blind authors, with no reliability reported. The blind rater’s agreement on the remainder was κ = .58 overall and .14 for fairness; the reported κ = 1 is the consensus after discussion.
- That the mentions record what the society judges. The judgements are those of mostly Western, mostly mid-twentieth-century ethnographers.
Connections
- Alfano, Cheong & Curry (2024), LIT-512 (NOTE-406), run a dictionary over 9,653 Ethics paragraphs from 256 societies. They find most domains present in most societies, but their method cannot detect valence, and it agrees with this study’s hand codes at κ .08–.25. It roughly replicates presence, not the valence result, and a spot-check found admired theft in four societies.
- The account this evidence is offered for is THEORY-468. The questionnaire evidence for its structure is THEORY-505.