Contribution
- A statement of MFT as four claims, any of which, if disproved, would sink the theory (p. 61): nativism, cultural learning, intuitionism, pluralism.
- Table 2.1: five foundations with adaptive challenge, original and current triggers, emotions and virtues.
- A review of MFT findings, framed as “pragmatic validity”: the theory’s usefulness in generating findings (section 3).
- Five criteria for foundationhood, with the evidence for each current foundation (Table 2.4), and a list of candidate foundations (4.1.5).
Key insight
Pluralism is a methodological bet: if morality has several evolved sources, a one-construct theory will be Procrustean, “cutting off phenomena it cannot explain … and stretching its unitary construct to fit everything else” (p. 103). The list of foundations is provisional; the pluralism is not.
Assumptions
- Nativism as preparedness. “Innate means organized in advance of experience”, not hard-wired (p. 61, after Marcus).
- Modularity optional. Foundations are thought of as sets of modules, but “you do not have to embrace modularity … to embrace MFT” (p. 63).
- Intuitionism. The social intuitionist model (Haidt 2001, LIT-511) is “the prequel to MFT” (p. 66).
- Descriptive only. “Our project is descriptive, not normative” (p. 57; restated at p. 105 in reply to Jost).
- Measurement. Almost all the evidence reviewed uses the MFQ, whose internet samples the chapter concedes are likely Westernized (p. 92).
Key results
- Table 2.1 (p. 68), the five foundations.
| foundation | adaptive challenge | original triggers | virtues |
|---|---|---|---|
| Care/harm | protect and care for children | suffering, distress or neediness expressed by one’s child | caring, kindness |
| Fairness/cheating | reap benefits of two-way partnerships | cheating, cooperation, deception | fairness, justice, trustworthiness |
| Loyalty/betrayal | form cohesive coalitions | threat or challenge to group | loyalty, patriotism, self-sacrifice |
| Authority/subversion | forge beneficial relationships within hierarchies | signs of high and low rank | obedience, deference |
| Sanctity/degradation | avoid communicable diseases | waste products, diseased people | temperance, chastity, piety, cleanliness |
- Table 2.4 (p. 108), evolutionary model per foundation. Care: kin selection (Hamilton 1964) and attachment (Bowlby 1969). Fairness: reciprocal altruism (Trivers 1971). Loyalty: multilevel selection and tribalism. Authority: rank and dominance (de Waal; Boehm). Sanctity: disgust and the behavioural immune system (Rozin et al.; Schaller & Park).
- Table 2.4, innate preparedness. Nonhuman primate and infant evidence is listed for Care, Fairness and Loyalty; Authority has primate evidence only; Sanctity is “not yet shown in NHP or infants”. The chapter presumes Sanctity “is the most recently evolved foundation, perhaps coevolving with human religiosity” (p. 112).
- Ideology (3.2). In large internet samples, Care and Fairness are valued across the spectrum, slightly more by liberals; Loyalty, Authority and Sanctity rise linearly toward conservatives (Figure 2.1). Table 2.3 reports the pattern in 12 world regions, with Authority and Sanctity correlating most strongly with ideology.
- Sanctity in attitudes (p. 88). Across 20 hot-button issues (N = 24,739), Sanctity was the strongest predictor for most, after partialling ideology and demographics.
- Pathogens (p. 93). Country-level historical pathogen prevalence predicts Sanctity, and also Loyalty and Authority (Van Leeuwen et al. 2012).
- Candidates (pp. 104–106). Liberty/oppression, Efficiency/waste and Ownership/theft (prize-winning nominations from Jost, Shulman & Mastronarde, and Wiessner) and Honesty/deception are under study. The authors’ own self-criticism: fairness measures “centered on equality rather than equity” (Iyer), and Loyalty and Authority might be one foundation (p. 106).
- Equality (p. 114). “We do not believe that there is an ‘equality’ foundation”: equality is a special case of equity, or of Liberty.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | There are several innate moral foundations, at least the five listed | moderate for pluralism, weak for the specific five | Table 2.4 evidence is uneven: Sanctity has no primate or infant evidence, Authority no infant evidence |
| C2 | Liberals rely mainly on Care and Fairness, conservatives on all five | strong as a description of MFQ responses | many samples and methods (Figure 2.1; Table 2.3), almost all self-report, mostly internet |
| C3 | Sanctity predicts political attitudes beyond ideology | moderate | Koleva et al. 2012, as reported |
| C4 | Monist theories (harm, fairness) cannot explain disgust’s effects or Care/Sanctity processing differences | weak–moderate | three cited findings (Schnall et al. 2008; Young & Saxe 2011; Inbar et al. 2012); the disgust finding was later disputed (NOTE-132) |
| C5 | The five criteria are what it takes to be a foundation | informal argument | modelled on Ekman’s criteria for basic emotions |
| C6 | MFT engaged in no post hoc evolutionary theorizing | assertion | p. 113; the foundations were matched to existing theories, which the chapter cites |
| C7 | Equality is not a separate foundation | assertion, later reversed | p. 114; Atari et al. 2023 (LIT-500) split Fairness into Equality and Proportionality |
Concepts
- foundation — part of the “first draft” of the moral mind: a set of related modules that makes it easy to learn values about one recurrent adaptive challenge (pp. 61–63).
- original / current triggers — Sperber’s proper and actual domains, renamed (p. 67).
- individualizing / binding — Care and Fairness versus Loyalty, Authority and Sanctity (p. 106).
- pragmatic validity — a theory’s scientific usefulness for answering and posing questions (p. 57).
- criteria for foundationhood — common in third-party judgments; automatic affective evaluations; culturally widespread; evidence of innate preparedness; an evolutionary model of adaptive advantage (4.2).
Connections
- Founding essay. Haidt and Joseph 2004 (LIT-502) had four modules and ingroup as a footnoted fifth. This chapter promotes ingroup to Loyalty and renames the rest.
- Morality-as-cooperation (LIT-047, NOTE-034 C7). LIT-047 (pp. 49–50)
says MFT has only five foundations, omits kin altruism, reciprocal
altruism, hawkish displays and property, includes two foundations (Care
and purity) tied to no type of cooperation, and is “avowedly ad hoc”
(citing Haidt & Joseph 2011, not read here). Item by item against this
chapter:
- Kin altruism. Present: Care’s model is kin selection (Table 2.4). What MFT lacks is a foundation for obligations to family as such; Care generalizes to “baby seals”. LIT-047 calls Care “generic” for this reason, which is fair.
- Reciprocal altruism. Present: Fairness is Trivers (pp. 69, 114). LIT-047’s footnote 1 says MFT conflates reciprocity with fairness and the MFQ has no reciprocity items. The chapter’s own report that its fairness measures track equality (p. 106) supports the measurement half.
- Hawkish displays (bravery). Absent. Loyalty’s virtues include “self-sacrifice”; no foundation or candidate is about courage.
- Property. Named as a good candidate, Ownership/theft (p. 104), not measured.
- Care and Sanctity. Sanctity is grounded in disease avoidance, as LIT-047 says. The chapter reports Sanctity as moralized and predictive, which is what LIT-047’s design never measured.
- Ad hoc. The chapter’s criteria are the reply. They are explicit, but they filter candidates; they do not derive them.
- Mutualism and dyadic harm. Baumard et al. 2013 (LIT-508) is cited as fairness monism and Gray et al. 2012 as harm monism (pp. 102–104); dyadic morality’s 2018 statement is LIT-494. The chapter’s argument against monism rests on disgust findings that NOTE-132 reports as disputed.
- Relationship regulation. Rai and Fiske 2011 (LIT-507) is treated as an alternative pluralism, compatible with MFT (pp. 106, 115).
- Moral disciplining. LIT-101 (NOTE-132) cites MFQ data in which Loyalty and Authority predict the moralization of victimless excess as strongly as Purity does. With this chapter’s pathogen finding (p. 93), that is a second sign that the binding foundations do not separate cleanly.
- Obligation. Criterion 1 (p. 109) rests the third-party character of moral judgment on “shared intentionality” (Tomasello et al. 2005), the same capacity Tomasello’s account of obligation builds on (LIT-125, read in NOTE-185).
- The MAC-Q. Curry, Jones Chesters and Van Lissa’s questionnaire, filed in the same batch as LIT-496, compared itself with the MFQ and found MFQ purity items rated about as morally relevant as its loyalty and authority items.
Bearing on the record
- NOTE-034, C7 (“MAC is more precise and broader than MFT”). On this
text the claim splits.
- More precise holds as a claim about derivation. MAC’s domains come from a model of cooperation; MFT’s came from a reading of five books (LIT-502) and are now filtered by criteria.
- Broader fails. Two of the four “omissions” are present in MFT’s own table (kin selection under Care, reciprocal altruism under Fairness), and a third is a named candidate (Ownership). Only bravery is truly absent. MFT in turn contains Sanctity and care for non-kin, which MAC’s seven leave out. Neither list contains the other.
- NOTE-034’s open question (are purity, sex and harm moralized as often as cooperation?). This chapter’s evidence says purity is moralized in Western samples and predicts attitudes, but it is self-report from people asked about purity, not blind coding of what a society moralizes. It does not answer the question LIT-047’s design left open; it shows the question is live.
- No THEORY document cites it. No ML instruction.
Note of 2026-10-16: THEORY-435 and THEORY-539 now cite LIT-499 as a source.
Limitations
- The evidence is almost all MFQ self-report from English-language internet samples; the chapter calls them likely WEIRD (p. 92).
- Pragmatic validity is not truth: a generative theory can carve the domain wrongly. The authors grant the list may change (p. 57).
- The criteria are necessary conditions on candidates, with a final distinctness hurdle (p. 114) that has no stated test.
- The evolutionary models are borrowed and matched, not derived; the match for Loyalty relies on group-level selection, which the chapter notes is disputed (fn. 5).
Open questions
- Is Sanctity one foundation or several? Jarudi’s sexual-versus-food purity split (p. 106) and LIT-101’s puritanical-versus-contamination split point the same way.
- Do bravery and property meet the five criteria? Neither has been tested against them.
- Do the binding foundations separate outside WEIRD samples? LIT-500 reports that the individualizing–binding split is not universal.
Corrections
- none to a seeded skim (there was no seed)
- Regions. The chapter says Graham et al. 2011 replicated the ideological pattern “from 11 different world regions” (p. 77) and, later, “across the 12 world regions for which data were available” (p. 93). Table 2.3 has 12 columns.
- Table 2.3 signs. As extracted, every coefficient in Table 2.3 is positive, while the text says Care and Fairness correlate negatively with conservatism (pp. 93–94). The table reports magnitudes, or the signs were lost in extraction; the text gives the direction.