Mapping the Moral Domain

Abstract

The moral domain is broader than the empathy and justice concerns assessed by existing measures of moral competence, and it is not just a subset of the values assessed by value inventories. To fill the need for reliable and theoretically grounded measurement of the full range of moral concerns, we developed the Moral Foundations Questionnaire on the basis of a theoretical model of 5 universally available (but variably developed) sets of moral intuitions: Harm/Care, Fairness/Reciprocity, Ingroup/Loyalty, Authority/Respect, and Purity/Sanctity. We present evidence for the internal and external validity of the scale and the model, and in doing so we present new findings about morality: (a) Comparative model fitting of confirmatory factor analyses provides empirical justification for a 5-factor structure of moral concerns; (b) convergent/discriminant validity evidence suggests that moral concerns predict personality features and social group attitudes not previously considered morally relevant; and (c) we establish pragmatic validity of the measure in providing new knowledge and research opportunities concerning demographic and cultural differences in moral intuitions. These analyses provide evidence for the usefulness of Moral Foundations Theory in simultaneously increasing the scope and sharpening the resolution of psychological views of morality.

Full text

Mapping the Moral Domain 1
Running Head: MAPPING THE MORAL DOMAIN
Mapping the Moral Domain
Jesse Graham, Brian A. Nosek, Jonathan Haidt
University of Virginia
Ravi Iyer
University of Southern California
Spassena Koleva, Peter H. Ditto
University of California, Irvine
March 9, 2010
Author note: This research was supported by Institute for Education Sciences and Jacob Javits
fellowships to the first author and a grant from the National Institute of Mental Health (R01
MH68447). Correspondence concerning this article should be sent to jgraham@virginia.edu.
Supplemental data and analyses can be found at www.moralfoundations.org.

Mapping the Moral Domain 2
Abstract
The moral domain is broader than the empathy and justice concerns assessed by existing
measures of moral competence, and it is not just a subset of the values assessed by value
inventories. To fill the need for reliable and theoretically-grounded measurement of the full range
of moral concerns, we developed the Moral Foundations Questionnaire (MFQ) based on a
theoretical model of five universally available (but variably developed) sets of moral intuitions:
Harm/care, Fairness/reciprocity, Ingroup/loyalty, Authority/respect, and Purity/sanctity. We
present evidence for the internal and external validity of the scale and the model, and in doing so
present new findings about morality: 1. Comparative model fitting of confirmatory factor
analyses provides empirical justification for a five-factor structure of moral concerns. 2.
Convergent/discriminant validity evidence suggests that moral concerns predict personality
features and social group attitudes not previously considered morally relevant. 3. We establish
pragmatic validity of the measure in providing new knowledge and research opportunities
concerning demographic and cultural differences in moral intuitions. These analyses provide
evidence for the usefulness of Moral Foundations Theory in simultaneously increasing the scope
and sharpening the resolution of psychological views of morality.
Keywords: morality, scale validation, moral foundations, culture, values

Mapping the Moral Domain 3
Mapping the Moral Domain
How can we measure moral concerns when people disagree about what “morality”
means? To address this problem we created the Moral Foundations Questionnaire (MFQ,
presented in the appendix), a measure of the degree to which individuals endorse each of five
intuitive systems posited by Moral Foundations Theory: Harm/care, Fairness/reciprocity,
Ingroup/loyalty, Authority/respect, and Purity/sanctity (Haidt & Graham, 2007; Shweder, Much,
Mahapatra & Park, 1997). People vary in the extent to which they endorse, value, and use these
five foundations, providing an opportunity to better understand moral diversity. In this article, we
explain the need for a scale broader than conventional morality scales. We describe the
development of the scale, which involved multiple rounds of item analysis using large
heterogeneous samples. We present the validation of the scale, organized into evidence
confirming internal and external validity. Finally, we describe its pragmatic validity – that is, the
practical usefulness of both the theory and the measure in providing new insights about moral
psychology. A major goal of Moral Foundations Theory is to expand the range of phenomena
studied in moral psychology so that it matches the full range of moral concerns, including those
found in non-Western cultures, in religious practices, and among political conservatives. Here we
present what we have learned about the content and structure of the moral domain using the
Moral Foundations Questionnaire.
What is the Moral Domain?
A great variety of scales are used in moral psychology to measure stages of moral
reasoning (e.g., the Defining Issues Test-2; Rest et al., 1999), moral identity (e.g., the Moral
Identity Scale; Aquino & Reed, 2002), empathy (e.g., the Interpersonal Reactivity Index; Davis,
1983), and moral deficits such as psychopathy (e.g., Levenson’s Self-report Psychopathy Scale;
Levenson et al., 1995). Although these scales measure different aspects of morality, they all share
the assumption (explicit or implicit) that the moral domain is limited to concerns about
individuals harming or unfairly treating other individuals. This is, in part, a reflection of the
dominance of Lawrence Kohlberg’s (1969, 1971) ideas about morality as justice, and his
subsequent debate with Carol Gilligan (1982) about her alternative conception of morality as
care. Both sides agreed that morality was about how well or poorly individuals treated other
individuals. Turiel (1983, p.3) codified this approach into a widely-cited definition of the moral
domain as “prescriptive judgments of justice, rights, and welfare pertaining to how people ought
to relate to each other.” Any values that were not related to “justice, rights, and welfare” (e.g.,
patriotism, authority, or chastity) were considered non-moral, and were relegated to the domain
of social convention or the domain of personal choice (Turiel, Hildebrandt, & Wainryb, 1991).
Definitions of morality in philosophy also frequently stress rules or codes of conduct that reduce
harm to others (e.g., Gert, 2005; Singer, 1979).
Kohlberg certainly noticed that people sometimes justified moral judgments by referring
to group-level moral concerns such as authority, loyalty, and tradition, but he thought that such
thinking was immature and conventional, a part of the “law and order” ethos of stage 4. With
sufficient opportunities for role-taking, adolescents were said to move beyond stage 4 and to
begin using post-conventional reasoning based on an understanding of justice. Kohlberg’s stage
theory has been criticized on many grounds; one criticism relevant for our purposes was that
Kohlberg’s post-conventional morality enshrined politically liberal ideals as developmental
endpoints (Hogan, Johnson, & Emler, 1978; Shweder, 1982; Sullivan, 1977; for related critiques
see Puka, 1994). This critique was backed up by the demonstration that liberals routinely obtain
higher principled reasoning scores on the Defining Issues Test, but that conservative students

Mapping the Moral Domain 4
rose to liberal levels when told to “respond as a left-winger would” (Emler, Renwick, & Malone,
1983). Conservatives could reason at the “higher stage,” but were not doing so presumably
because they had different priorities in their moral reasoning. Despite these critiques, the notion
that “true” moral concerns involve considerations of harm or justice exclusively has persisted in
psychological research.
Moral psychology is currently experiencing a renaissance as social psychologists,
neuroscientists and behavioral economists begin to treat moral judgment and decision-making as
a central topic of inquiry (see Haidt, 2008, for a review). Yet even today, research is still largely
limited to issues of harm and fairness (including rights). Major literatures include decision-
making about runaway trolleys, in which participants must make tradeoffs between harm and
rights (Greene et al., 2009; Hauser, Young, & Cushman, 2008); economic games, which pit
fairness against self-interest (Fehr & Gachter, 2002); and the perennially popular topic of
altruism and prosocial behavior (e.g., Moll, Krueger, Zahn, Pardini, Oliveira-Souza, & Grafman,
2006). Whether carried out in a scanner, on a website, or in a business school lab room using real
money, morality is still usually operationalized as helping (vs. harming) or as playing fair (vs
cheating).
Kohlberg and Turiel based their circumscription of the moral domain on a line of
enlightenment thinking running from Immanuel Kant to John Stuart Mill to John Rawls in which
the autonomy and/or welfare of the individual are the starting point for ethical inquiry. Yet even
Kant (1797/1996, p. 179) had intuitions of a broader moral domain. He wrote that masturbation
was “in the highest degree opposed to morality,” although he granted that “it is not so easy to
produce a rational demonstration” of its wrongness. And before the Enlightenment, philosophers
routinely considered a much broader moral domain. Much of ancient moral philosophy, from
Greece to India to Japan, was virtue-based. These societies valued benevolence and fairness, but
they also emphasized group-level concerns about social order, authority, duty, loyalty to one’s
family or group, and controlling one’s carnal desires to cultivate one’s soul or gain a more
favorable rebirth (Larue, 1991; Shweder et al., 1997). Aristotle’s Nicomachean Ethics (350
BCE/1985) dealt with a wide range of questions related to human flourishing and ideal ways of
living in a community; it treated virtues as habits developed by repeated use. Similarly, the
history of theories and interventions in moral education includes not only instructions in care and
justice, but training in how to be a good group member (Graham, Haidt, & Rimm-Kaufmann,
2008).
Western inquiry into ethics must therefore be seen as having made a sharp turn away from
the full spectrum of moral concerns addressed by thinkers of the past and valued in folk
moralities around the world (See Appiah, 2008, ch. 1, on the causes of this turn). Pincoffs (1986)
documented the gradual narrowing of philosophical inquiry in the West since the 19th century as
philosophers moved away from virtue ethics and adopted what Pincoffs called “quandary ethics”
– the study of how people should resolve quandaries, particularly those that pit the rights or
welfare of one person against those of another. Should psychologists follow philosophers in this
turn to quandary ethics? Or should we keep our eyes focused on what is, rather than on the
Enlightenment conception of what ought to be?
Broadening the Moral Domain
Cross-cultural research on moral judgment has revealed that Turiel’s definition of the
moral domain works well among educated and politically liberal Westerners, for whom harmless
offenses are rarely condemned, even when they are disgusting or disrespectful (Haidt, Koller, &

Mapping the Moral Domain 5
Dias, 1993).1 However, research on people in India (Shweder, Mahapatra, & Miller, 1987),
people of lower social class in Brazil and the United States (Haidt, Koller, & Dias, 1993), and
conservatives in the United States (Graham, Haidt, & Nosek, 2009; Haidt & Hersh, 2001; Jensen,
1998) has revealed moral considerations beyond the individual-based concerns of harm and
fairness, involving concerns about spiritual purity and degradation (even for acts that involve no
harm), concerns about proper hierarchical role fulfillment, and moral expectations of loyalty to
the local or national group. Illustrative of this breadth are open-ended responses participants in
our studies gave when asked to define morality in their own words. Parallel to Turiel’s definition,
many made reference to harm and human welfare (e.g., “Avoiding harm to others”), and fairness
or justice (“Morality is doing the right things to ensure fair treatment for all”). However, others
made reference to moral issues beyond justice, rights, and welfare, and to morally valuable
entities that were not individuals (e.g., “Morality is having a system [that] protects the social
institutions of family, community, and country.”). Others made reference to duty, obedience,
respect, and preserving tradition (e.g., “Matters of duty, irrespective of one’s own personal
desires or ends”). And some made reference to God or religious norms, decency, the soul, and
maintaining purity of mind (e.g., “not having dirty thoughts”). Scales that attempt to measure
morality by assessing attitudes about harm and fairness are thus leaving out much of what
people—even Westerners—explicitly and spontaneously include in their descriptions of the
moral domain.
If morality really does vary by culture, class, politics, and era, then psychologists need a
definition of the moral domain that is not based on a list of specific content areas (e.g., justice,
rights, and welfare). To meet this need, Haidt and Kesebir (2010, p. 800) proposed an alternative
approach that defines moral systems by their function:
Moral systems are interlocking sets of values, virtues, norms, practices, identities,
institutions, technologies, and evolved psychological mechanisms that work together to
suppress or regulate selfishness and make social life possible.
This functionalist approach has at least two advantages over content-based approaches. First, in
its emphasis on moral systems it encourages psychologists to look outward, beyond individual
minds and psychological mechanisms. Moralities emerge as large numbers of people interact
with each other, constrained and enabled by culturally and historically specific sets of institutions
and technologies. A second advantage is that this definition makes it possible to recognize a wide
variety of societies as constituting moral systems, at least descriptively, even if those societies are
structured in ways that many researchers believe to be normatively immoral (e.g., patriarchies
and theocracies).
Values and morals. Virtue ethics, character education, and the ethical codes of most
cultures and religious traditions suggest that the moral domain is broader than the issues of harm
and fairness currently represented in moral psychology scales. One possible existing framework
for broadening the moral domain is research on values, which Schwartz (2007, p. 712) defines as
“trans-situational goals that vary in importance and serve as guiding principles in the life of a
person or a group” (see also Rohan, 2000 and Feldman, 2003 for reviews). Most efforts to
identify the range of values begin with long lists of possible values, covering everything from
cleanliness to hedonism, and then use factor analysis of endorsement ratings to identify a smaller
set of core values (Rokeach, 1973; Schwartz, 1992). Values research has much to offer the
empirical study of morality, and is too often ignored by moral psychologists. Clearly, many
1 Turiel and his students demonstrated their “domain distinction” cross culturally (e.g., Hollos, Leis, & Turiel, 1986),
but their moral violations invariably involved direct physical or emotional harm, not harmless violations implicating
issues of loyalty, respect, or purity.

Mapping the Moral Domain 6
values are moral values, even if morality is defined only in terms of welfare and fairness
concerns (e.g., benevolence and universalism). However, in seeking to identify a list of the most
important values, there is a risk that some common moral concerns or intuitions will be missed.
For example, reciprocity, loyalty to one’s team or tribe, and concerns about bodily and spiritual
purity are ubiquitous in anthropological accounts of morality, yet they do not appear among
Schwartz’s ten values. This may be because Schwartz began with an atheoretical exploratory
factor-analytic approach, using Western populations. Even if Westerners care quite a bit about
reciprocity (see Cialdini, 2001, ch. 2), they might not list it when asked about their most
important trans-situational goals. Individuals are often unable to access the causes of their moral
judgments (Haidt & Kesebir, 2010; Wilson, 2002). Furthermore, atheoretical descriptive
approaches are limited in their ability to explain why people hold the values they do.
Moral Foundations Theory
In order to fill the need for a systematic theory of morality, explaining its origins,
development, and cultural variations, we created Moral Foundations Theory (MFT). Haidt and
Joseph (2004) began by surveying the literatures in evolutionary psychology and anthropology,
looking for matches – for virtues and areas of moral regulation that were common (though not
necessarily universal) across cultures, and that had some clear counterpart in evolutionary
thinking. For example, virtues related to fairness and practices of reciprocal gift exchange (e.g.,
Mauss, 1924/1990) bore an obvious similarity to the evolutionary literature on reciprocal
altruism (Trivers, 1971); virtues of purity and practices regulating food and sex (e.g., Douglas,
1966) bore an obvious relationship to the evolutionary literature on disgust (Rozin, Haidt, &
McCauley, 2000). The results of this cross-disciplinary review produced five top candidates for
being the psychological “foundations” upon which cultures construct their moralities. These five
foundations are consistent with, but expand upon, several existing taxonomies of moral concern,
including Fiske’s (1992) four models of social relationships, Shweder et al.’s (1997) account of
the “three ethics” of autonomy, community, and divinity that are found widely around the world,
and Hogan et al.’s (1978) evolution-based socioanalytic theory of moral development. MFT can
therefore be seen as an attempt to specify the “evolved psychological mechanisms” that were part
of the definition of moral systems given earlier. Even if all moral systems are social
constructions, they are constructed by people whose minds are not at all like blank slates
(Marcus, 2004). In this way, MFT allows for intuitive or emotional bases of moral judgments, as
well as more deliberate reasoning processes (cf. Greene et al., 2001; Haidt, 2001).
Haidt and Graham (2007) expanded the theory and modified the names of the foundations
to become: Harm/care, Fairness/reciprocity, Ingroup/loyalty, Authority/respect, and
Purity/sanctity. Harm and Fairness generally correspond to Shweder at al.’s (1997) ethics of
autonomy; Ingroup and Authority to the ethics of community; and Purity to the ethics of divinity.
Haidt and Graham also applied the theory to a particular kind of cultural variation within the
United States: the “culture war” between political liberals and conservatives. Drawing on
Shweder and several political theorists (e.g., Burke, 1790/2003; Mill, 1859/2003; Lakoff, 1996;
Muller, 1997; Sowell, 2002), liberalism was hypothesized to indicate a morality in which the
individual is the locus of moral value. In such a moral world, moral regulation revolves around
protecting individuals from harm or unfair treatment by other individuals, or by the social
system. In contrast, conservatives—at least, the social conservatives of the religious right—try to
create more tightly-ordered communities in which (for example) proper relationships between
parent and child, man and woman, and human and God are part of the aim of moral regulation. In
such a moral world, the individual is not the primary locus of moral value; the building block of

Mapping the Moral Domain 7
society is thought to be the family, and a much greater emphasis is placed on virtues and
institutions that bind people into roles, duties, and mutual obligations.
This analysis led to the hypothesis that liberal morality would prioritize Harm and
Fairness over the other three foundations (because the “individualizing foundations” of Harm and
Fairness are all that are needed to support the individual-focused contractual approaches to
society often used in enlightenment ethics), whereas conservative morality would also
incorporate Ingroup, Authority, and Purity to a substantial degree (because these “binding
foundations” are about binding people together into larger groups and institutions).
The first draft of the Moral Foundations Questionnaire was created in part to test this
hypothesis about ideological differences, and was useful (along with other methods, such as
content coding of religious sermons) in supporting the hypothesis (Graham, et al., 2009).
However, our goal in subsequent theory and measure development – reported in this paper – was
much broader. Moral Foundations Theory provides a conceptual organization for measuring and
describing differences in moral concerns across individuals, social groups, and cultures. In
theory, any pattern of “settings” or endorsement levels for the five foundations is possible. Thus,
a reliable and valid scale was needed to measure the degree to which any individual’s moral
beliefs and concerns rely upon each of the five hypothesized foundations. Further, theory-driven
scale validation is a means of testing hypotheses and theoretical claims (cf. Hogan & Nicholson,
1998). We describe the development of the MFQ below, along with what it revealed about the
structure and variation of the moral domain.
Development of the MFQ
Because we wanted to gauge individual differences in the range of concerns that people
consider morally relevant, the first version of the MFQ (reported in Graham et al., 2009, Study 1,
Appendix A) explicitly asked participants to evaluate the moral relevance of several foundation-
related concerns (e.g., “Whether or not some people were treated differently from others,” for
Fairness). For the second version (reported in Graham et al., 2009, Study 2, Appendices A and B)
we added a new section that assessed levels of agreement with more specific and contextualized
moral judgment statements (see below).
These first two versions of the MFQ were tested on heterogeneous populations (using
ProjectImplicit.org, a popular web-based virtual laboratory), with large sample sizes (total
n=3,825). With these data, we compared different confirmatory factor analysis models that
corresponded to distinct, theory-guided conceptualizations of the possible factor structure (see
supplementary materials at MoralFoundations.org). Using the comparative model fitting
techniques described below, we found evidence that a five-factor solution was an improvement
(weighing both fit and parsimony) over models representing a single morality factor, two factors
(an “individualizing” factor underlying Harm and Fairness items, and a “binding” factor
underlying all other items), and three factors (corresponding to Shweder’s three ethics of
autonomy, community, and divinity).
While the overall model fits were reasonably good (see supplements), some of the
internal consistencies of these early versions of the scale were low. In addition, some items had
weak loadings on the latent factors. Correlation matrices of all items in the second version
revealed that some items that were written to represent one foundation actually related more
strongly to another foundation, and some items correlated so highly with each other that they
appeared redundant. Problematic items were replaced with new items in the third version of the
scale.
Pilot testing for the third version of the MFQ was extensive, involving data from over
28,000 participants (surveyed at YourMorals.org) and external validations with eight

Mapping the Moral Domain 8
conceptually related scales. Item and factor analyses on this third version showed improvements
over versions 1 and 2. No subscale had more than one item that loaded poorly on the latent
factor, so we focused our item analyses on determining how reducing from 40 to 30, 20, or even
10 items would impact the validity of the scale.
We developed a novel method for empirically selecting item combinations that would
maximize both internal and external validity. For all 10 subscales (four relevance items or four
judgments items for each of the five foundations) we identified three criterion scales. The first
criterion scale for each subscale was the corresponding foundation subscale using the alternative
format (for instance, Harm-judgments as internal validity criterion for Harm-relevance). The
second criterion subscale was the corresponding foundation subscale from the “taboo-tradeoffs”
questionnaire (see Graham, et al., 2009, Study 3, Appendix C). Modeled on work by Tetlock,
Kristel, Elson, Green, and Lerner (2000), this questionnaire asks participants to indicate how
much money they would require to perform actions that violate the foundations in a variety of
ways (for instance, “kick a dog in the head, hard” for Harm), including options of performing the
violation for free, as well as refusing to perform it for any amount of money. The third criterion
(used for both relevance and judgments subscales) was an external scale we expected to be
related to one particular moral foundation. For Harm the external criterion was the empathy
subscale of the Interpersonal Reactivity Index (Davis, 1983); for Fairness, the Schwartz Equality
value item (Schwartz, 1992); for Ingroup, the Schwartz Loyalty value item (Schwartz, 1992); for
Authority, the Right-Wing Authoritarianism scale (Zakrisson, 2005); and for Purity, the Disgust
Scale-Revised (Haidt, McCauley, & Rozin, 1994, modified by Olatunji et al., 2008).
In order to quantify how the quality of the scale would decline if shortened, and in order
to select the combination of items that would retain the greatest internal and external validity, we
calculated correlations between every combination of items and the three criterion scales, as well
as a combined item-total correlation with all three criterion scales. An example for Harm-
Relevance is shown in Figure 1 (detailed reports of the criterion analyses for all ten subscales can
be found in the supplements). The top left panel of Figure 1 plots the corrected item-total
correlations with the three criterion scales (Harm-judgments, Harm-taboo-tradeoffs
questionnaire, and IRI-empathy, shown in the other three panels) for every single Harm-
relevance item, every 2-item combination, every 3-item combination, and the 4-item aggregate.
One can see in this figure that a single item, H (“whether or not someone was harmed”) was
negatively impacting the internal and external validity of the Harm-relevance aggregates that
included it, and that a 3-item aggregate excluding it had even better internal and external
reliabilities than the 4-item aggregate. All ten of these analyses revealed that using the best three-
item combination yielded internal and external validities as good if not better than using all four
items; moreover, we found that in most cases the optimal 2-item combinations were nearly as
good as the 3-item combinations (although the 3-item combinations were preferable for their
broader conceptual coverage). The best three items from each subscale were retained for the
fourth and final version of the MFQ, shown in the Appendix. The 20 starred items make up the
short-form MFQ.
Description of Final Items
Throughout the several rounds of item generation and selection, we sought to minimize
conceptual and empirical redundancy among items. On this point, we differ from some
approaches to scale development that prioritize high internal consistency. Internal consistency is
important, but so is comprehensive coverage of the various facets of the construct (the scale
development work of Harrison Gough [1979, 1984] is a good example of this balance; see also
John & Soto, 2007). From our point of view, it is better to have dissimilar items that are
moderately correlated but that each capture a different facet of a foundation than it is to have

Mapping the Moral Domain 9
similar items that are highly correlated and capture only a small amount of the foundation’s
scope. As such, our aim in item analysis was not to maximize internal consistency via item
redundancy. Instead, we sought a balance between achieving (a) sufficient internal consistency to
believe that there was a common core, and (b) maximal item heterogeneity to increase
confidence that we were representing the foundation in full.
For the moral relevance items, we attempted to cover a wide conceptual area for each
foundation while avoiding obvious culture-war issues (e.g., one item is about the moral relevance
of rights violations in the abstract, without specifying particular content such as gun rights or gay
rights). Items were generated to capture different instances of a foundational moral concern, for
instance asking about group loyalty in reference to different specific groups (nation, family) as
well as to one’s group left in the abstract.
We had two reasons for adding the judgments subscale to the relevance subscale. First,
we wanted multiple response formats to minimize the impact of variance based on response set
(for instance, some people may be more likely to indicate that everything is morally relevant).
Second, we wanted to supplement the abstract relevance assessments—which, as self-theories
about how one makes moral judgments, may be inaccurate with regard to actual moral judgments
(Nisbett & Wilson, 1977)—with contextualized items that could more directly gauge actual
moral judgments. To fill out the judgments subscale, participants need not directly consider or be
aware of the basis for their moral judgments; they just need to decide whether they endorse or
reject the action or event. In this way, the “relevance” subscale may better assess explicit theories
about what is morally relevant, and the “judgments” subscale may better assess actual use of
moral foundations in judgment (see initial evidence for this conclusion in Graham et al., 2009,
Study 2). These judgments took the form of normative declarations (e.g., “It can never be right to
kill a human being,” for Harm), hypotheticals (e.g., “If I were a soldier and disagreed with my
commanding officer’s orders, I would obey anyway because that is my duty,” for Authority),
virtue endorsements (e.g., “Chastity is an important and valuable virtue,” for Purity), and
opinions about government policies (e.g., “When the government makes laws, the number one
principle should be ensuring that everyone is treated fairly,” for Fairness). With this variety of
both item formats and specific item content, the final version of the MFQ gauges sensitivities to
basic kinds of moral concerns, not just opinions on specific moral issues.
Sample sizes, number of items, and alphas from pilot testing on all four versions of the
MFQ are shown in Table 1. Below we describe validity analyses on the fourth and final version
of the scale shown in the Appendix.
Method
Participants and Procedure
Participants were 34,476 adults (37% female; mean age 36.2) who had previously
registered at YourMorals.org and selected to take the Moral Foundations Questionnaire.
Participants come to the website via many different routes, identified using a web tracker. In this
sample, 40.5% entered the URL directly into their browser. The most common referring sites
were edge.org (29.7%), search engines (8.1%), message boards (3.9%), ted.com (2.2%), and
alternet.org (1.9%). (For an analysis demonstrating that response patterns on the MFQ are similar
for participants from diverse referring sources, see Iyer, 2009). The relevance section preceded
the judgments section, and items within each section were given in an order randomized for each
participant. A subset of this sample also chose to take one or several other surveys on the site.
This is the sample and procedure used for all internal and external validity analyses except the
test-retest, pragmatic, and incremental predictive validity analyses, which are detailed below.

Mapping the Moral Domain 10
Results: Internal Validity
Scale means and internal consistencies. Means, standard deviations, and alphas for each
subscale of the MFQ are presented in Table 2. Means are also given separately for liberals,
moderates, conservatives, and libertarians, to serve as points of comparison. Although the alphas
are not as high as many other scales, they indicate a reasonable internal consistency given that
our goal was to gauge an expansive range of moral concerns with a small number of items across
two different item formats.
Relations between relevance and judgments subscales. Relations between the relevance
and judgments subscales are shown in Table 3. The top panel shows zero-order correlations. To
ensure that these relationships were not solely driven by common relations to political ideology,
the bottom panel shows partial correlations controlling for politics. Both panels show convergent
validity for each foundation as measured by the two formats, as well as discriminant validity in
that these relations are stronger than relations between different foundations, despite high
correlations among many of the foundations (average same-foundation r = .48, average different-
foundation r = .14).2
Test-retest reliability. We gave the MFQ to 123 college students (mean age = 20.1, 69.9%
female) from the University of Southern California. After an average interval of 37.4 days (range
28 to 43 days), participants completed the MFQ a second time. In both instances, the MFQ was
administered via a class website which recorded the date and time of completion automatically.
Question order was randomized for each participant each time. Test-retest Pearson correlations
for each foundation score were .71 for Harm, .68 for Fairness, .69 for Ingroup, .71 for Authority,
and .82 for Purity (all p<.001). These test-retest reliabilities are quite similar in magnitude to the
internal consistencies calculated in Table 2. This suggests that item responses are quite stable
over time and that within-occasion variation is more a function of the broad diversity of
measurement rather than instability.
Exploratory factor analysis. Although Moral Foundations Theory predicts a specific
factor structure for moral concerns, we began with exploratory factor analyses to see what factors
emerged from the items in the absence of conceptual constraints. Factor analysis for all 30 items
of the MFQ was performed using direct oblimin rotation with Kaiser normalization (allowing the
factors to be correlated) and maximum likelihood estimation (Fabrigar, Wegener, McCallum, &
Strahan, 1999). Six factors with eigenvalues greater than 1 emerged, but scree plot and factor
loadings indicated that only the first two factors provided meaningful incremental explanatory
power and interpretability (only 2 of the 120 loadings on the last four factors were above .3; see
also Costello & Osborne, 2005, on factor retention from scree plot analysis). The first two factors
were retained, and are shown in Table 4. As the table indicates, the two factors clearly
corresponded to the binding foundations (Ingroup, Authority, and Purity; factor 1) and
individualizing foundations (Harm and Fairness; factor 2), and the strongest loading for all 30
items was as predicted.
Although this analysis supported our distinction between individual-focused and group-
focused moral concerns, it remains an open question whether we are justified in treating the
theoretically-derived moral foundations as five factors, rather than two. To answer this question
we turn to comparisons between different confirmatory factor analysis models.
Confirmatory factor analysis. The large sample sizes we obtained for each version of the
MFQ allowed us to create structural equation models comparing different theoretically-derived
2 Correlations between the latent foundation factors can be found in Figure 2 (five correlated factors model), and
zero-order correlations between the factors are available in a supplement at the first three authors’ websites.

Mapping the Moral Domain 11
factor structures. Table 5 describes the comparative model fitting with the final version of the
MFQ. The first three numerical columns provide fit statistics for the individual models, and the
last two columns show the degree to which each model was an improvement over the model in
the row above it. Figure 2 shows the different confirmatory factor analysis models for the full
scale; as Table 5 reflects, the same models were constructed for the relevance and judgments
subscales separately as well. In the first step, we compared nested first-order models. Our
hypothesis was that model 4 (five correlated factors: Harm, Fairness, Ingroup, Authority, and
Purity) would provide a better overall model fit than a single morality factor model (1), two-
factor model (2: individualizing and binding, corresponding to the results of the exploratory
factor analysis), and three-factor model (3: corresponding to Shweder et al.’s (1997) ethics of
autonomy, community, and divinity). All three tests (relevance subscale, judgments subscale, and
full MFQ) confirmed these predictions; the overall best model (weighing fit and parsimony) was
the five-factor model in every case.3
In the second step, we tested whether the five factors could be more parsimoniously
modeled with two correlated superordinate factors representing our theoretical distinction of
“individualizing” and “binding” foundations (see the hierarchical model in Figure 2). As Table 6
shows, however, the model with five intercorrelated factors was a significant improvement
(again, weighing both fit and parsimony) over the hierarchical models. In general, confirmatory
factor analyses provide robust support for our five-factor conceptualization of the moral
foundations.
Results: External Validity
Relations to other scales. We identified several other scales and scale items also taken by
participants at YourMorals.org that we predicted would relate to the MFQ foundation scores.
Scales were grouped into five external criteria scale sets, one set for each foundation. Harm
criterion scales were the empathic concern subscale of the Interpersonal Reactivity Index (Davis,
1983; n=134), Levenson’s (1995) Psychopathy Scale (reverse-scored; n=116), Schwartz’s (1992;
n=4,228) Benevolence subscale, and three items from the Adapted Good-Self Assessment
(Barriga, Morrison, Liau, & Gibbs, 2001; n=89) on the importance of being kind/caring,
sympathetic/compassionate, and generous/giving. Fairness scales were Social Dominance
Orientation (reverse-scored, as it measures preference for social inequalities; Pratto, Sidanius,
Stallworth, & Malle, 1994; n=1,215), importance of being fair/just on Barriga’s Good-Self scale,
and endorsement of the social justice item on Schwartz’s values scale. Ingroup scales were the
importance of being loyal/faithful on Barriga’s Good-Self scale, and endorsement of loyalty,
national security, and family security items on Schwartz’s values scale. Authority scales were
Right-Wing Authoritarianism (Zakrisson, 2005; n=1,093), the Traditionalism subscale of the
Progressive and Traditional Justice scale (Haidt, Darley, & Gromet, 2009; n=1,384), and
endorsement of the social order, authority, respect for tradition, honoring parents, and obedience
values on Schwartz’s value scale. Purity scales were the Disgust Scale-Revised (Haidt,
McCauley & Rozin, 1994 modified by Olatunji et al., 2008; n=1,681), self-reported religious
attendance (n=32,607), and endorsement of the self-discipline, clean, and devout items on
Schwartz’s values scale. Items from the same scale were averaged together, and correlations
between the foundations and the scales were averaged together for each criterion group.
Correlations between the foundations and the external criterion scales are shown in Table 7. As
3 We also tested whether a six-factor model separating Authority and Tradition (shown in Figure 2) would improve
upon the five-factor model. However, these six-factor models were a worse fit than the five-factor models, 0.60 ≤ εa
≤ 0.65; in addition, as Figure 2 shows, the Authority and Tradition factors were very highly correlated (r = .96),
further supporting a single latent factor for these items.

Mapping the Moral Domain 12
the table shows, each foundation was the strongest predictor for its own conceptually related
group of external scales (average r = .51, vs. average r = .14 for the off-diagonals). This provides
evidence of both convergent and discriminant validity, despite relatively substantial relations
among the foundations.
Predictive validity for social group attitudes. A subset of the participants who took the
MFQ also took a survey in which they reported their gut reactions to various social groups. We
constructed this survey by first identifying groups conceptually related to each foundation
because they represent either virtues or vices of that foundation. For instance, we predicted that
people whose morality rested heavily on the Harm/care foundation would have especially
positive reactions to “caring” groups such as nurses, and especially negative reactions to
“harming” groups like hunters. The foundation relevance of each group was identified a priori by
the authors. Harm-related groups were nurses, environmentalists, pacifists, vegetarians, and
hunters (r). Fairness-related groups were ACLU members, labor unions, rich people (r), and
CEOs (r). Ingroup-related groups were Americans, U.S. Government, flag burners (r), and illegal
immigrants (r). Authority-related groups were soldiers, police officers, U.S. Marines, U.S.
Military, people who spank their children, and anarchists (r). Purity-related groups were virgins,
highly religious people, spiritual people, atheists (r), prostitutes (r), homosexuals (r), people who
have casual sex (r), and people with tattoos or piercings (r). Groups indicated by “(r)” represented
vices of a foundation and were reverse-scored, and for U.S.-specific groups only U.S. citizens
were included in the analyses. Because attitudes toward social groups and moral foundations
scores are both related to political ideology, we used partial correlations controlling for political
ideology to see which groups would be uniquely predicted by one or more foundations. Partial
correlations between foundations and all social groups were averaged for each set of foundation-
related groups; these averages are shown in Table 8. As the table shows, each foundation was the
strongest predictor (above and beyond politics) of attitudes toward conceptually related social
groups, providing further evidence of predictive and discriminant validity. Beyond validation of
the scale, these results also suggest that attitudes about social groups are in part moral judgments
about those groups. Moral Foundations Theory and the MFQ may be useful for researchers who
want to know which moral concerns are related to prejudice toward any particular group.
Incremental predictive validity. The preceding sections establish that the theorized model
of Moral Foundations as five interrelated factors is a better fit than other plausible models, and
that each of the five factors predict foundation-relevant outcomes. An open question is whether
the MFQ has predictive validity beyond existing measures. Because it measures domains that are
not present in other theoretical conceptions of morality—Ingroup, Authority, and Purity—it
surely expands the predictive validity of morality measures. However, even broader measures
exist, such as the Schwartz Values Scale, which measures endorsement of ten broad classes of
values. This scale contains several values and subscale factors that conceptually overlap with the
moral foundations (e.g., Benevolence with Harm/care, Traditionalism with Authority/respect),
and many self-interested values that we consider to be outside the moral domain and not covered
by the MFQ (e.g., Achievement, Hedonism). Even though Moral Foundations Theory has a
different conceptual starting point (an evolutionary account of why humans have the moral
intuitions they do, contra Schwartz’s factor-analytic approach), it is nonetheless worthwhile to
test whether the MFQ has predictive validity beyond Schwartz’s scale in predicting a variety of
scales, opinions, and self-reported behaviors relevant to morality. Because Schwartz’s scale is
larger both in terms of subscales (10 vs. 5) and items (58 vs. 30), this is a particularly tough test
of the predictive validity of the MFQ.
Data for these analyses came from 10,652 visitors to YourMorals.org who took both the
MFQ and the Schwartz Values Scale (SVS), 92% of whom also took other scales or measures.

Mapping the Moral Domain 13
We used two-step regressions to test whether the five MFQ subscales added incremental
predictive validity beyond the ten SVS subscales for the external criteria (scales and attitudes
toward foundation-related social groups) described above, as well as for positions on a wide
range of political issues. Note that this analysis focuses on the incremental validity of the
aggregate MFQ in comparison to the aggregate SVS, rather than investigating which of the
particular moral concerns predict each criterion variable (for such work, see Koleva, Graham,
Iyer, Ditto, & Haidt, 2010).
In every analysis, the MFQ made a significant improvement to prediction when added to
the SVS (average ∆R² = 8%, all ∆R² significant at p < .001). To provide a point of comparison,
we repeated this analysis by adding the 44-item Big Five personality inventory to the SVS, which
yielded an average ∆R² of only 2%. R² and ∆R² for the scales and foundation-related social group
averages can be found in Table 9. We also investigated R² for the MFQ alone, to further compare
its predictive validity with that of the SVS. As the bolded values in Table 9 show, the MFQ was
actually a more powerful predictor than the SVS for most of the scales and political issue
positions, and all of the social group attitudes. Given that the SVS is a comprehensive, large, and
well-validated measure of values, the MFQ is clearing a high bar in providing unique predictive
validity for outcomes relevant for moral and political psychology, and for the psychology of
prejudice.
Generalizability. Because every step of the scale development used large heterogeneous
samples, we can be more confident about the MFQ’s generalizability than if we had used college
students only (Sears, 1986). However, the samples obtained at ProjectImplicit.org and at
YourMorals.org are not representative of any national or international population – the current
sample is disproportionately from the U.S. (80%), white (87%), male (63%), and educated (mean
education between “completed college” and “some graduate school”) compared to international
or U.S. national averages.
Thanks to the large sample sizes, however, we were able to test whether the five-factor
model of moral concerns was consistent across national and geographic groups. All participants
self-reported their current country of residence, the country in which they grew up (if different),
and the age at which they moved (if they grew up in a different country). For participants who
reported moving to their current country at age 14 or older, the country in which they reported
growing up was used for the cross-cultural analysis. We created 12 location codes, four of which
indicated the four nations from which the largest number of participants had come (U.S., Canada,
U.K., Australia); the other eight location codes indicated multi-nation regions of the world (i.e.
East Asia, Middle East). Model fit information for each location code is shown in Table 10. As
the table shows, the five-factor model of the MFQ is a reasonable or good fit (all εa < .06, all CFI
> .7) for all 11 world regions for which we were able to run the fit analyses, providing evidence
that the measurement and theory of five foundational moral concerns is not specific to U.S. or
Western participants. Notably, although the five-foundation model is a good fit in all these areas,
the data shows much cross-cultural variation in the patterns of moral foundation endorsement.
Even controlling for politics, age, gender, religious attendance, and education, world region is a
significant (ps < .001) predictor of all five foundation scores, indicating that the scale is useful
for measuring and describing cultural differences in moral concerns. We turn to such uses in the
next section.

Mapping the Moral Domain 14
Results: Pragmatic Validity
“Pragmatism asks its usual question. ‘Grant an idea or belief to be true,’ it says, ‘what
concrete difference will its being true make in anyone's actual life? How will the truth be
realized? What experiences will be different from those which would obtain if the belief
were false? What, in short, is the truth's cash-value in experiential terms?’”
— William James (1907/1998), Pragmatism, p. 97
Suppose you were to grant, in a Jamesian spirit, that Moral Foundations Theory is true,
that there really are a few basic intuitive foundations upon which all cultures construct systems of
virtue and vice. What is the “cash-value” of this theory? What new questions does it allow
researchers to ask about morality, and what new understandings can measurement of its
constructs provide? We address this section to researchers who don’t expect to use the Moral
Foundations Questionnaire in their own research, but would like to know what the theory and
measure can tell them about moral psychology. First and foremost, MFT provides a broadened
conception of morality, not only in content domain (including group-focused as well as
individual-focused concerns) but – by showing how moral judgments influence attitudes and
behavior in social situations far removed from quandary ethics – in life domain as well (Rozin,
2006). For instance, the above validation exercises with external scales and social group attitudes
showed that many traits and attitudes that don’t seem to be about morality on the surface
nevertheless show a systematic and theoretically meaningful relationship to moral concerns
measured by the MFQ. We present here three additional findings made possible by MFT’s
broadened definition of morality, and its finer conceptual resolution of morality’s components.
1. Robustness of ideological patterns across cultures. Using a variety of measures and
methods, Graham et al. (2009) showed that liberals value Harm and Fairness concerns more than
conservatives, while conservatives value Ingroup, Authority, and Purity concerns more than
liberals. The vast majority of these participants, however, came from the United States, leaving
open the question of whether these patterns were limited to the particular ideological conflicts of
the United States. Table 11 shows correlations between political ideology4 and the five
foundations for the different world areas described in the Generalizability section. The
correlations indicate that the liberal-conservative patterns found in the U.S. are robust across
national and cultural contexts, both in terms of direction (negative correlations [liberals higher]
for Harm and Fairness, positive correlations [conservatives higher] for Ingroup, Authority, and
Purity) and in terms of magnitude: correlations are consistently strongest for Authority and
Purity, and weakest for Harm. This suggests that across cultures, the most intractable political
debates are likely to involve concerns related to respect for traditions/authorities and
physical/spiritual purity, while the greatest degree of moral commonality may be found in issues
related to harm and care. It also reinforces the claim that political ideology can be self-assessed
and that the unidimensional left-right construct has some degree of common meaning across
societies, despite differences in political party structures and particular national issues (Jost,
2006).
2. East-West differences in foundation endorsement. Moving beyond politics, the
international MFQ data can also be used to describe cross-cultural differences in moral concerns.
We compared participants from Eastern cultures (South Asia, East Asia, and Southeast Asia,
4 Because the terms “liberal” and “conservative” can mean different things in different nations (i.e., the Liberal Party
in Australia is actually center-right ideologically), the political identification item on YourMorals.org clarifies to
participants that the items from “very liberal” to “very conservative” should be read as “very left-wing” to “very
right-wing.”

Mapping the Moral Domain 15
n=2,258) to participants from Western cultures (U.S., U.K., Canada, and Western Europe,
n=104,893). Eastern participants showed stronger concerns about Ingroup (mean difference=.23,
t(107149)=12.42, p < .0001, d = .08) and Purity (mean difference=.25, t(107149)=10.51, p <
.0001, d = .06) compared to Western participants, and were only very slightly more concerned
about Harm, Fairness, and Authority (mean differences < .1, ts < 7, ds < .04). The fact that
differences are concentrated in Ingroup and Purity makes some sense in light of established
cultural differences in collectivism (Triandis, 1995) and the role of purity concerns in daily life
and religious practice, particularly in South Asia (Shweder et al., 1997). But it is noteworthy that
there are not large differences in Authority, given greater sensitivity to social hierarchy (Power
Distance scores) in eastern nations (Hofstede, 2001). The small effect sizes for all the East-West
differences suggest that variation within cultures (e.g., by gender or political ideology) will
exceed the east-West variations given so much attention in cross-cultural research. Here we see
that the increased resolution afforded by MFT allows us to find moral differences we would not
have been able to find otherwise – differences that open up intriguing questions for further
research.
3. Gender differences in foundation endorsement. The MFQ reveals interesting gender
differences as well. In the large international dataset collected at YourMorals.org (49,428
women, 68,812 men), women score higher than men on Harm (mean difference=.47,
t(118238)=99.16, p < .0001, d = .58), Fairness (mean difference=.16, t(118238)=37.75, p <
.0001, d = .22), and Purity (mean difference=.16, t(118238)=25.10, p < .0001, d = .15), with men
just barely higher on Ingroup and Authority (mean differences < .06, ds < .06). Women were
more concerned than men about Harm, Fairness, and Purity, even controlling for political
ideology. As the effect sizes show, these gender differences were much stronger than the
differences between Eastern and Western cultures. The gender patterns make sense in light of
previous research on empathy (Davis, 1983), egalitarianism (Arts & Gelissen, 2001), and disgust
sensitivity (Druschel & Sherman, 1999), but they also show an important divergence from the
political patterns in that Purity is here grouped with Harm and Fairness, rather than Ingroup and
Authority. Here too the finer resolution and broadened scope of MFT allowed us to find and
describe differences in moral personality not possible before.
General Discussion
Moral Foundations Theory (Haidt & Joseph, 2004; Haidt & Graham, 2007) was created
by selecting the closest links between evolutionary accounts of human sociality and
anthropological accounts of the breadth and variability of the moral domain (see especially Fiske,
1992, and Shweder et al., 1997). The findings reported in this article suggest that those
anthropologists were right. From a purely descriptive perspective, the domain of morality
consists of more than just “prescriptive judgments of justice, rights, and welfare” (Turiel, 1983,
p. 3). Furthermore, we found that one does not need to travel to non-Western nations to find this
broader conception of morality. In every country and world region we examined, people on the
political right placed greater emphasis on concerns about ingroup loyalty, respect for authorities
and traditions, and physical/spiritual purity than did people on the political left.
Our data confirm that concerns about Harm and Fairness are so widespread that they
might be said to be universally used foundations of morality (upon which cultures construct
differing ideas as to what counts as harm, or what kinds of distributions are fair). But many
people and cultures are also concerned about the proper relationships of individuals to groups,
authorities, institutions, and roles. And many people and cultures are concerned about how
people treat their own bodies—whether they treat them as playgrounds for their own pleasure, or

Mapping the Moral Domain 16
as temples created by God to house a soul within. These additional moral concerns have not been
included in measures of morality, until now.
The Moral Foundations Questionnaire
The Moral Foundations Questionnaire fills the need for a theoretically-grounded scale
covering the full range of human moral concerns. We found substantial evidence that the scale is
reliable and valid. The scale is internally consistent (both within and between two question
formats) while maintaining conceptual coverage of diverse manifestations of foundation-related
concerns. Test-retest analyses showed stability of foundation subscale scores over time. External
validations of the MFQ using widely-used scales, as well as attitudes toward conceptually related
social groups, showed convergent, discriminant, and predictive validity. Factor analyses
confirmed our theoretical parsing of the moral domain into five sets of concerns: the five-factor
model fit the data better (weighing both fit and parsimony) than competing models, and this five-
factor representation provided a good fit for participants in 11 different world areas. In addition
to the scale itself, we expect that the method introduced in this paper (see Figure 1) for
empirically selecting items to maximize both internal and external validity will also be of use to
researchers.
The best existing instrument for assessing moral concerns beyond harm and fairness is the
Schwartz Values Scale (SVS; Schwartz, 1992), which includes group-level values such as
“tradition” and “conformity.” However, the SVS was created to measure a broad spectrum of
values; it was not designed to cover the moral domain specifically, and it does not cover concerns
about group-loyalty and spiritual purity. In a head-to-head comparison, the MFQ showed
incremental predictive validity beyond the SVS for a diverse array of external scales related to
moral personality, attitudes toward social groups, and opinions about moral and political issues
(see Table 9). Further, in most cases, the MFQ performed even better than the SVS in overall
variance explained of criterion variables, despite its shorter length and narrower conceptual
coverage. This further illustrates the MFQ’s effective measurement properties, balancing
relatively short length and wide coverage of the moral domain.
What MFT and the MFQ can Reveal about Morality
The research reported here indicates that the MFQ is a reliable and valid instrument for
measuring a broad range of moral concerns. But in the process of developing and validating the
MFQ, we also generated a number of substantive discoveries about moral psychology, such as:
1) A map of the moral domain. Because it distinguishes five kinds of moral concerns, and
gives separate evolutionary accounts to explain each of their origins (Haidt & Joseph, 2007),
MFT is not as parsimonious as theories of morality that try to derive the entire moral domain
from one or two principles or processes (usually kin selection plus reciprocal altruism – see
Dawkins, 1976; Hauser, 2006; Joyce, 2006). However, comparisons of different structural
models revealed that the five-factor solution is an improvement over other theoretically-derived
models, even taking into account the relative loss of parsimony. Analyses of international data
showed that this five-factor model was a good fit in every area of the world we were able to
examine. This provides empirical evidence for MFT’s central claim about the structure of human
morality, and points toward the usefulness of this added complexity in revealing new findings
about moral similarities and differences across individuals, groups, and cultures.
2) A guide to where the action is. We found some small and easily interpretable cross-
cultural differences in moral foundation scores: people in Eastern cultures were slightly more
likely to value Ingroup and Purity than people in Western cultures. As with all research that relies
upon educated participants, our cross-national differences would probably have been much larger
if we had found a way to survey rural villagers and the urban poor in Asia. Nonetheless, the
cross-national differences we did find were dwarfed by the within-nation (or within-region)

Mapping the Moral Domain 17
differences we examined, including both ideological differences and sex differences (women
valued Harm, Fairness, and Purity more than men, even controlling for political ideology). With
reliable measures of these different kinds of moral concerns, social and personality psychologists
can now begin to examine many such patterns of similarities and dissimilarities, as well as the
processes behind them.
3) A method for discovering moral prejudices. The finer resolution offered by the MFQ
also revealed the potential role of moral judgment in prejudice. When we examined attitudes
toward various social groups, we found that MFQ subscales indicated varying patterns of moral
concerns that might lead some people to dislike some groups. This suggests that attitudes toward
social groups may often be expressions of moral judgments about those groups – vague intuitions
or explicit convictions that a particular social group upholds or violates one or more foundational
concerns. The moral foundations can thus be used as a kind of decoder ring, allowing us to see
multiple and sometimes unexpected moral threads connecting seemingly unrelated attitudes and
opinions (cf. Koleva et al, 2010). This possibility emphasizes our functional definition of
morality as a description of what motivates people to suppress selfishness, rather than a
prescriptive definition of how one ought to behave. By describing and quantifying a broadened
range of human moral concerns, MFT and the MFQ can aid in our understanding of the dangers
of morality, as well as the benefits.
Limitations and Directions for Further Research
The map of the moral domain that we offer is provisional. We hope that other researchers
will help us improve it. Here are four next steps:
1) Obtain other samples. While the findings reported in this article held across tens of
thousands of participants from around the world, caution is warranted. Most importantly, our
data came from volunteers who are not representative of any definable population. Nonetheless,
the consistency of effects across 11 nations or regions suggests that the observed patterns are, at
least, widespread. It will be useful to compare the present findings with representative samples in
each nation. As a start, the MFQ has been given to nationally representative samples in the USA
(Smith & Vaisey, 2010), which replicated the basic pattern of ideological differences that we
report in Table 11. The problem of non-representativeness is exacerbated in non-English-
speaking nations, because all of our respondents were adult English-speakers with access to the
Internet. Future work will be needed to validate and refine the scale for use with other
populations. This work has already begun, using translations of the MFQ into Chinese, Croatian,
Dutch, Italian, Polish, Russian, and Spanish (all available at MoralFoundations.org), and field
work with populations not reachable via online research (such as the inhabitants of a municipal
dump outside Managua, Nicaragua; Graham, Cox, & Casablanca, 2010).
2) Test developmental hypotheses. Kohlberg traced the development of reasoning about
justice. Gilligan examined the development of reasoning about care. Moral Foundations Theory
indicates that we also need basic descriptive work on the development of judgment and reasoning
about ingroups, authority, and purity. A question of particular interest will be whether moral
maturity involves shifting from one pattern of foundation-usage to another. For example,
Kohlberg’s scoring manual acknowledged that people sometimes justified their judgments by
referring to authority and tradition, but he believed that this was an immature (conventional)
form of morality, which would be transcended with age and additional opportunities for role
taking. As noted in the introduction, this view set off a “stage-ideology” debate, in which
Kohlberg was criticized for treating a liberal set of values as a developmental progression beyond
a conservative set of values (cf. Emler et al, 1983). As a descriptive account of the range of
human moral concerns, MFT does not make normative claims about which kinds of concerns are
better than others (claims about which are more complex, or more mature than others – such as

Mapping the Moral Domain 18
Kohlberg’s [1971] “naturalistic fallacy” – fall into this normativity). However, future work on
how the foundations develop and change in response to environmental changes and life events
could inform this debate by giving more information about how people come to have the moral
concerns they do (and possibly even why people in the same culture and environment hold
different moral values sacred). Clearly, more work is needed on the development of moral
foundations in children, and on the dynamics of moral change throughout adolescence and
adulthood.
3) Go beyond self-report. An additional limitation is that the data reported in this article
consists entirely of self-report measurements of attributes, attitudes, and judgments. Future steps
improving the map of the moral domain should include analyses of morally relevant behaviors,
life events, and implicit attitudes.
4) Look for more foundations. The map we have offered is surely incomplete. Because of
the theory-driven origin of MFT, and the strong empirical support for the theory reported here,
we believe that the five foundations offer a good initial map of the major moral continents.
However, it is quite possible that later research, using different items or different methods, will
reveal that one of these continents is, like Eurasia, really two continents. For example, we
included concerns about liberty in the Fairness foundation, because Fairness gives rise to notions
of rights. Whether such a grouping can be sustained with additional scrutiny is an open question,
as is the question of whether a single foundation underlies intuitions about equality of
opportunities and those about equality of outcomes. It is also possible that two of the five
continents are really riding on a single tectonic plate. For example, Ingroup and Authority tend to
intercorrelate highly across analyses, but our investigation so far suggests that it is reasonable and
useful to keep them separate. Finally, it is possible that some major islands remain to be named.
For example, others have suggested wisdom, waste, authenticity, truth, and self-control as
candidate foundations.
Conclusion
People disagree about the size and content of the moral domain – that is, about what
“morality” means. Researchers therefore need theories that encompass the true breadth of human
morality, and they need measurement tools that can detect a broad array of moral concerns. In
this article we presented Moral Foundations Theory as a way of thinking about morality that goes
beyond harm and fairness (with inspiration from Shweder et al., 1997). And we presented the
Moral Foundations Questionnaire as a reliable, valid, and easy-to-use tool for exploring this
expanded moral domain. The MFQ, and its progeny, will be useful for extending, critiquing, and
otherwise improving psychology’s map of the moral domain.

Mapping the Moral Domain 19
References
Aquino, K., & Reed, A. (2002). The self importance of moral identity. Journal of Personality
and Social Psychology, 83, 1423-1440.
Appiah, K. A. (2008). Experiments in ethics. Cambridge, MA: Harvard.
Aristotle. (350 BCE/1985). Nicomachean Ethics. T. Irwin, trans. Indianapolis: Hackett
Publishing Company.
Arts, W., & Gelissen, J. (2001). Welfare states, solidarity and justice principles: Does the type
really matter? Acta Sociologica, 44(4), 283-299.
Barriga, A. Q., Morrison, E. M., Liau, A. K., & Gibbs, J. C. (2001). Moral cognition: Explaining
the gender difference in antisocial behavior. Merrill-Palmer Quarterly, 47, 532–562.
Burke, E. (1790/2003). Reflections on the revolution in France. New Haven: Yale.
Cialdini, R. B. (2001). Influence: Science and practice (4th Ed.). Boston: Allyn and Bacon.
Costello, A. B., & Osborne, J. W. (2005). Best practices in Exploratory Factor Analysis: Four
recommendations for getting the most from your analysis. Practical Assessment,
Research, and Evaluation, 10, 1-9.
Davis, M. H. (1983). Measuring individual differences in empathy: Evidence for a
multidimensional approach. Journal of Personality and Social Psychology, 44, 113-126.
Dawkins, R. (1976). The Selfish Gene. London: Oxford University Press.
Douglas, M. (1966). Purity and danger. London: Routledge
Druschel, B. A., & Sherman, M. F. (1999). Disgust sensitivity as a function of the Big Five and
gender. Personality and Individual Differences, 26(4), 739-748.
Emler, N., Renwick, S., & Malone, B. (1983). The relationship between moral reasoning and
political orientation. Journal of Personality and Social Psychology, 45, 1073-1080.
Fabrigar, L. R., Wegener, D. T., MacCallum, R. C., & Strahan, E. J. (1999). Evaluating the use
of exploratory factor analysis in psychological research. Psychological Methods, 4, 272-
299.
Fehr, E., & Gachter, S. (2002). Altruistic punishment in humans. Nature, 415, 137-140.
Feldman, S. (2003). Values, ideology, and the structure of political attitudes, in Sears, Huddy, &
Jervis (Eds.), Oxford Handbook of Political Psychology. Oxford, UK: Oxford University
Press, pp. 477-508.
Fiske, A. P. (1992). Four elementary forms of sociality: Framework for a unified theory of social
relations. Psychological Review, 99, 689-723.
Gert, B. (2010). The definition of morality. In E. N. Zalta (Ed.), The Stanford Encyclopedia of
Philosophy. http://plato.stanford.edu/entries/morality-definition/.
Gilligan, C. (1982). In a different voice: Psychological theory and women’s development.
Cambridge, MA: Harvard University Press.
Gough, H.G. (1979). A creative personality scale for the adjective check list. Journal of
Personality and Social Psychology, 37, 1398-1405.
Gough, H.G. (1984). A managerial potential scale for the California psychological inventory.
Journal of Applied Psychology, 69(2), 233-240.
Graham, J., Cox, K., & Casablanca, S. (2009). Moral concerns among sex workers, rural poor,
urban poor, university students, and municipal dump inhabitants in Nicaragua.
Manuscript in preparation.
Graham, J., Haidt, J. & Nosek, B.A. (in press). Liberals and conservatives rely on different sets
of moral foundations. Journal of Personality and Social Psychology.
Graham, J., Haidt, J. & Rimm-Kaufman, S. E. (2008). Ideology and intuition in moral education.
European Journal of Developmental Science, 2, 269-286.

Mapping the Moral Domain 20
Greene, J.D., Sommerville, R.B., Nystrom, L.E., Darley, J.M., & Cohen, J.D. (2001). An fMRI
investigation of emotional engagement in moral judgment. Science, 293, 2105-2108.
Greene, J.D., Cushman, F.A., Stewart. L.E., Lowenberg, K., Nystrom, L.E., and Cohen, J.D.
(2009). Pushing moral buttons: The interaction between personal force and intention in
moral judgment. Cognition, 111, 364-371.
Haidt, J. (2001). The emotional dog and its rational tail: A social intuitionist approach to moral
judgment. Psychological Review, 108, 814-834.
Haidt, J. (2008). Morality. Perspectives on Psychological Science, 3, 65-72.
Haidt, J., Darley, J., & Gromet, D. (2009). The Progressive and Traditional Justice Scale.
Manuscript in preparation.
Haidt, J. & Graham, J. (2007). When morality opposes justice: Conservatives have moral
intuitions that liberals may not recognize. Social Justice Research, 20, 98-116.
Haidt, J., & Hersh, M. (2001). Sexual morality: The cultures and emotions of conservatives and
liberals. Journal of Applied Social Psychology, 31, 191-221.
Haidt, J., & Joseph, C. (2004). Intuitive Ethics: How Innately Prepared Intuitions Generate
Culturally Variable Virtues. Daedalus, pp. 55-66, Special issue on human nature.
Haidt, J., & Joseph, C. (2007). The moral mind: How 5 sets of innate intuitions guide the
development of many culture-specific virtues, and perhaps even modules. In P.
Carruthers, S. Laurence & S. Stich (Eds.), The Innate Mind, Vol. 3. (pp. 367-391). New
York: Oxford.
Haidt, J., & Kesebir, S. (2010). Morality. In S. T. Fiske, D. Gilbert & G. Lindzey (Eds.),
Handbook of Social Psychology, 5th Edition (pp. 797-832). Hoboken, NJ: Wiley.
Haidt, J., Koller, S., & Dias, M. (1993). Affect, culture, and morality, or is it wrong to eat your
dog? Journal of Personality and Social Psychology, 65, 613-628.
Haidt, J., McCauley, C., & Rozin, P. (1994) . Individual differences in sensitivity to disgust: A
scale sampling seven domains of disgust elicitors. Personality and Individual
Differences, 16, 701-713.
Hauser, M. (2006). Moral minds: How nature designed our universal sense of right and wrong.
New York: HarperCollins.
Hauser, M., Young, L., & Cushman, F. (2008). Reviving Rawls' linguistic analogy: Operative
principles and the causal structure of moral action. In W. Sinnott-Armstrong (Ed.), Moral
psychology, Vol. 2: The cognitive science of morality (pp. 107-143). Cambridge, MA:
MIT Press.
Hofstede, G. (2001). Culture's consequences: Comparing values, behaviors, institutions, and
organizations across nations. New York: Sage Publications.
Hogan, R., Johnson, J. A., & Emler, N. P. (1978). A socioanalytic theory of moral development.
In W. Damon (Ed.), New directions for child development: Vol. 2. Moral development
(pp. 1-18).San Francisco: Jossey-Bass.
Hogan, R., & Nicholson, R. A. (1988). The meaning of personality test scores. American
Psychologist, 43, 621–626.
Hollos, M., Leis, P., & Turiel, E. (1986). Social reasoning in Ijo children and adolescents in
Nigerian communities. Journal of Cross-Cultural Psychology, 17, 352-374.
Iyer, R. (2009). Robustness of Liberal-Conservative Moral Foundations Questionnaire
Differences. Polipsych.com. Retrieved from
http://www.polipsych.com/2009/09/18/robustness-of-liberal-conservative-moral-
foundations-questionnaire-differences/ on January 5, 2010.
James, W. (1907/1998). Pragmatism: A new name for some old ways of thinking and The
Meaning of Truth. Cambridge, MA: Harvard University Press.

Mapping the Moral Domain 21
Jensen, L. A. (1998). Moral divisions within countries between orthodoxy and progressivism:
India and the United States. Journal for the Scientific Study of Religion, 37, 90-107.
John, O. P., & Soto, C. J. (2007). The importance of being valid: Reliability and the process of
construct validation. In R. W. Robins, R. C. Fraley, & R. F. Krueger, Handbook of
Research Methods in Personality Psychology (pp. 461-494). New York: Guilford.
Jost, J. T. (2006). The end of the end of ideology. American Psychologist, 61, 651–670.
Joyce, R. (2006). The evolution of morality. Cambridge, MA: MIT.
Kant, I. (1797/1996). The metaphysics of morals (M. Gregor, Trans.). Cambridge, UK:
Cambridge.
Kohlberg, L. (1969). Stage and sequence: The cognitive-developmental approach to
socialization. In D. A. Goslin (Ed.), Handbook of socialization theory and research.
Chicago: Rand McNally.
Kohlberg, L. (1971). From is to ought: How to commit the naturalistic fallacy and get away with
it in the study of moral development. In T. Mischel (Ed.), Cognitive Development and
Epistemology (pp.151-235). New York: Academic Press.
Koleva, S., Graham, J., Iyer, R., Ditto, P. H., & Haidt, J. (2010). Beyond red and blue: The moral
foundations of the culture war. Manuscript under revision.
Lakoff, G. (1996). Moral politics : How liberals and conservatives think. Chicago: University of
Chicago Press.
Larue, G. A. (1991). Ancient ethics. In P. Singer (Ed.), A companion to ethics (pp. 29-40).
Malden, MA: Blackwell.
Levenson, M. R., Kiehl, K. A., & Fitzpatrick, C. M. (1995). Assessing psychopathic attributes in
a noninstitutionalized population. Journal of Personality and Social Psychology, 68, 151-
158.
Marcus, G. (2004). The birth of the mind. New York: Basic.
Mauss, M. (1924/1990). The Gift: The form and reason for exchange in archaic societies.
London: Routledge.
Mill, J. S. (1859/2003). On liberty. New Haven, CT: Yale University Press.
Moll, J., Krueger, F., Zahn, R., Pardini, M., de Oliveira-Souza, R., & Grafman, J. (2006). Human
fronto-mesolimbic networks guide decisions about charitable donation. Proceedings Of
The National Academy Of Sciences Of The United States Of America, 103, 15623-15628.
Muller, J. Z. (1997). What is conservative social and political thought?. In Muller, J. Z. (ed.),
Conservatism: An Anthology of Social and Political Thought from David Hume to the
present. Princeton, NJ: Princeton University Press, pp. 3–31.
Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know: Verbal reports on mental
processes. Psychological Review, 84, 231-259.
Olatunji, B. O., Haidt, J., McKay, D., David, B., (2008). Core, animal reminder, and
contamination disgust: Three kinds of disgust with distinct personality, behavioral,
physiological, and clinical correlates. Journal of Research in Personality, 42. 1243-1259.
Pincoffs, E. L. (1986). Quandaries and virtues: Against reductivism in ethics. Lawrence, Kansas:
University of Kansas.
Pratto, F., Sidanius, J., Stallworth, L. M., & Malle, B. F. (1994). Social Dominance Orientation:
A personality variable predicting social and political attitudes. Journal of Personality and
Social Psychology, 67, 741-763.
Puka, B. (Ed.). (1994). The Great Justice Debate: Kohlberg Criticism. Moral Development: A
Compendium, Vol. 4. New York: Routledge.
Rest, J., Narvaez, D., Thoma, S. J., Bebeau, M. J. (1999). DIT2: Devising and testing a new
instrument of moral judgment. Journal of Educational Psychology, 91, 644-659.

Mapping the Moral Domain 22
Rohan, M. J. (2000). A rose by any name? The values construct. Personality and Social
Psychology Review, 4, 255-277.
Rokeach, M. (1973). The nature of human values. New York: Free Press.
Rozin, P. (2006). Domain denigration and process preference in academic psychology.
Perspectives on Psychological Science, 1, 365-376.
Rozin, P., Haidt, J., & McCauley, C. R. (2000). Disgust. In M. Lewis & J. Haviland (Eds.)
Handbook of emotions, 2nd edition, (pp.637-653). New York: Guilford Press.
Schwartz, S. H. (1992). Universals in the content and structure of values. In M. P. Zanna (Ed.),
Advances in experimental social psychology, Vol. 25 (pp. 1-65). New York: Academic
Press.
Schwartz, S. H. (2007). Universalism values and the inclusiveness of our moral universe.
Journal of Cross-Cultural Psychology, 38, 711-728.
Sears, D. O. (1986). College sophomores in the laboratory: Influences of a narrow database on
social psychology’s view of human nature. Journal of Personality and Social Psychology,
51, 515-530.
Shweder, R. A. (1982). Liberalism as destiny. Contemporary Psychology, 27, 421-424.
Shweder, R. A., Mahapatra, M., & Miller, J. (1987). Culture and moral development. In J. Kagan
& S. Lamb (Eds.), The emergence of morality in young children (pp. 1-83). Chicago:
University of Chicago Press.
Shweder, R. A., Much, N. C., Mahapatra, M., & Park, L. (1997). The "big three" of morality
(autonomy, community, and divinity), and the "big three" explanations of suffering. In A.
Brandt, & P. Rozin (Eds.), Morality and health (pp. 119-169). New York: Routledge.
Singer, P. (1979). Practical ethics. Cambridge: Cambridge University Press.
Smith, C. & Vaisey, S. (2010). Charitable giving and moral foundations in a nationally-
representative sample. Manuscript in preparation, University of North Carolina.
Sowell, T. (2002). A conflict of visions: The ideological origins of political struggles. New York:
Basic.
Sullivan, E. V. (1977). A study of Kohlberg’s structural theory of moral development: A critique
of liberal social science ideology. Human Development, 20, 352-376.
Tetlock, P.E., Kristel, O., Elson, B., Green, M., & Lerner, J. (2000). The psychology of the
unthinkable: Taboo trade-offs, forbidden base rates, and heretical counterfactuals. Journal of
Personality and Social Psychology, 78, 853-870.
Triandis, H. C. (1995). Individualism and collectivism. Boulder, CO: Westview.
Trivers, R. L. (1971). The evolution of reciprocal altruism. Quarterly Review of Biology, 46, 35-
57.
Turiel, E. (1983). The development of social knowledge: Morality and convention. Cambridge,
England: Cambridge University Press.
Turiel, E., Hildebrandt, C., & Wainryb, C. (1991). Judging social issues: Difficulties,
inconsistencies, and consistencies. Monographs of the Society for Research in Child
Development, 56, 1-103.
Wilson, T. D. (2002). Strangers to ourselves: Discovering the adaptive unconscious.
Cambridge, MA: Belknap Press.
Zakrisson, I. (2005). Construction of a short version of the Right-Wing Authoritarianism (RWA)
scale. Personality and Individual Differences, 39, 863-872.

Mapping the Moral Domain 23
Table 1. Statistics on the four versions of the Moral Foundations Questionnaire
MFQ Version n Number of items Average Subscale α (range)
1 1,613 15 .59 (.39-.70)
2 2,212 43 .71 (.58-.83)
3 28,801 40 .76 (.69-.86)
4 (Final) 34,476 30 .73 (.65-.84)

Mapping the Moral Domain 24
Table 2. Means, standard deviations and alphas for moral foundation subscales
Foundation Subscale αTotal
(n=34,476)
Liberals
(n=21,933)
Moderates
(n=3,203)
Conservatives
(n=4,128)
Libertarians
(n=2,999)
Harm: H-Relevance .70 3.77 (0.86) 3.93 (0.76) 3.68 (0.84) 3.48 (0.89) 3.26 (1.03)
H-Judgments .51 3.08 (1.11) 3.32 (1.01) 2.95 (1.09) 2.48 (1.11) 2.34 (1.17)
Harm Total .69 3.42 (0.84) 3.62 (0.74) 3.31 (0.81) 2.98 (0.84) 2.80 (0.94)
Fairness: F-Relevance .65 3.89 (0.78) 4.04 (0.67) 3.77 (0.77) 3.44 (0.87) 3.66 (0.90)
F-Judgments .40 3.21 (0.93) 3.43 (0.86) 3.00 (0.86) 2.59 (0.87) 2.71 (0.95)
Fairness Total .65 3.55 (0.73) 3.74 (0.63) 3.39 (0.68) 3.02 (0.73) 3.19 (0.79)
Ingroup: I-Relevance .71 2.24 (1.03) 2.06 (0.94) 2.56 (1.00) 3.03 (1.02) 2.16 (1.10)
I-Judgments .46 2.28 (0.98) 2.09 (0.91) 2.59 (0.90) 3.13 (0.85) 2.21 (0.97)
Ingroup Total .71 2.26 (0.87) 2.07 (0.77) 2.58 (0.79) 3.08 (0.79) 2.19 (0.89)
Authority: A-Relevance .67 2.03 (0.95) 1.88 (0.86) 2.37 (0.90) 2.81 (0.91) 1.71 (0.95)
A-Judgments .60 2.52 (1.12) 2.23 (1.01) 2.97 (0.94) 3.74 (0.82) 2.55 (1.14)
Authority Total .74 2.27 (0.90) 2.06 (0.79) 2.67 (0.77) 3.28 (0.71) 2.13 (0.90)
Purity: P-Relevance .68 1.68 (1.11) 1.44 (0.94) 2.09 (1.09) 2.88 (1.11) 1.31 (1.03)
P-Judgments .75 1.41 (1.20) 1.09 (0.96) 1.88 (1.16) 2.90 (1.20) 1.16 (1.10)
Purity Total .84 1.54 (1.08) 1.27 (0.86) 1.99 (1.03) 2.89 (1.07) 1.23 (0.98)
Note. Range for all items and subscales is 0-5. Standard deviations are shown in parentheses.

Mapping the Moral Domain 25
Table 3. Internal relations between relevance and judgments subscales.
MFQ relevance subscales:
MFQ judgments subscales: Harm Fairness Ingroup Authority Purity
Harm .47 .36 .03 .04 .10
Fairness .32 .46 -.09 -.11 -.12
Ingroup -.05 -.13 .48 .43 .40
Authority -.12 -.21 .42 .49 .47
Purity .05 -.09 .44 .53 .74
After partialing political ideology:
MFQ relevance subscales:
MFQ judgments subscales: Harm Fairness Ingroup Authority Purity
Harm .38 .25 .07 .09 .19
Fairness .23 .35 .06 .05 .07
Ingroup .04 -.04 .38 .32 .25
Authority -.02 -.08 .31 .39 .31
Purity .11 -.01 .29 .37 .64
Note. Top panel shows zero-order correlations between subscales; bottom panel shows partial correlations after
removing variance shared with political ideology. The highest correlation in each row is shown in bold.

Mapping the Moral Domain 26
Table 4. Factor loadings from exploratory factor analysis of the Moral Foundations Questionnaire
items
Relevance Items: Factor 1 Factor 2
Harm:
Emotionally 0.01 0.59
Weak 0.09 0.65
Cruel 0.07 0.59
Fairness:
Unfairly 0.01 0.56
Treated -0.11 0.59
Rights -0.18 0.47
Ingroup:
Loyalty 0.52 0.19
Betray 0.48 0.17
Lovecountry 0.67 0.02
Authority:
Traditions 0.61 0.00
Respect 0.69 0.05
Chaos 0.40 0.20
Purity:
Disgusting 0.57 0.21
Decency 0.70 0.10
God 0.64 -0.02
Judgment Items:
Harm:
Animal -0.01 0.39
Kill -0.07 0.35
Compassion -0.01 0.63
Fairness:
Justice 0.09 0.27
Fairly -0.14 0.48
Rich -0.22 0.38
Ingroup:
Team 0.45 -0.03
History 0.49 -0.18
Family 0.34 0.04
Authority:
Sexroles 0.47 -0.24
Soldier 0.48 -0.20
Kidrespect 0.64 -0.05
Purity:
Harmlessdg 0.66 0.03
Unnatural 0.66 -0.07
Chastity 0.67 -0.08
Note. Strongest factor loading for each item indicated in bold.

Mapping the Moral Domain 27
Table 5. Goodness-of-fit indices for structural models representing confirmatory factor analyses
of Moral Foundations Questionnaire data collected on yourmorals.org (N=34,476)
Relevance items χ2 df εaΔχ2/Δdf 95%CI εaΔ
[1] Single factor 57093.3 90 .100
[2] Two correlated factors 20180.0 89 .060 36913.3/1 1.024 – 1.045
[3] Three correlated factors 17658.0 87 .057 2522.0/2 0.184 – 0.199
[4] Five correlated factors 11347.5 80 .047 6310.5/7 0.158 - 0.166
Judgment items χ2 df εaΔχ2/Δdf 95%CI εaΔ
[1] Single factor 32485.5 90 .076
[2] Two correlated factors 17270.0 89 .055 15215.5/1 0.654 – 0.675
[3] Three correlated factors 12191.4 87 .047 5078.6/2 0.264 - 0.279
[4] Five correlated factors 11084.6 80 .047 1106.8/7 0.064 - 0.072
Full MFQ (all items) χ2 df εaΔχ2/Δdf 95%CI εaΔ
[1] Single factor 138995.4 405 .073
[2] Two correlated factors 74542.9 404 .054 64452.5/1 1.357 – 1.378
[3] Three correlated factors 63074.0 402 .050 11468.9/2 0.400 – 0.415
[4] Five correlated factors 53894.1 395 .046 9179.9/7 0.191 – 0.199
Note. εa = root-mean-square error of approximation (RMSEA) for the model. Δχ2/Δdf = change in χ² and degrees of
freedom relative to the previous model. 95%CI εaΔ = confidence interval around RMSEA of the change in fit between
models; if .050 falls within the CI, then model fits are not considered significantly different. Model in bold is the best-
fitting model according to these comparisons.

Mapping the Moral Domain 28
Table 6. Goodness-of-fit indices for structural models representing confirmatory factor analyses of the
MFQ comparing the optimal first-order models (five intercorrelated factors) to hierarchical models
containing two superordinate factors
Relevance items χ2 df εaΔχ2/Δdf 95%CI εaΔ
[1] Hierarchical model 12723.8 85 .049
[2] Five correlated factors 11347.5 80 .047 1376.3/5 0.085 – 0.094
Judgment items χ2 df εaΔχ2/Δdf 95%CI εaΔ
[1] Hierarchical model 11822.3 85 .046
[2] Five correlated factors 11084.6 80 .047 737.7/5 0.061 – 0.070
Full MFQ (all items) χ2 df εaΔχ2/Δdf 95%CI εaΔ
[1] Hierarchical model 55322.9 400 .047
[2] Five correlated factors 53894.1 395 .046 1428.8/5 0.086 – 0.096
Note. εa = root-mean-square error of approximation (RMSEA) for the model. Δχ2/Δdf = change in χ² and degrees of
freedom relative to the previous model. 95%CI εaΔ = confidence interval around RMSEA of the change in fit between
models; if .050 falls within the CI, then model fits are not considered significantly different. Model in bold is the optimal
model (weighing both fit and parsimony) according to these comparisons.

Mapping the Moral Domain 29
Table 7. Pearson correlations of MFQ subscales with external scales and scale items.
MFQ subscales:
External scale criteria group: Harm Fairness Ingroup Authority Purity
Good-Self: kind/caring,
sympathetic/ compassionate,
generous/giving .43 .27 .00 .07 .06
IRI – Empathic Concern .59 .42 -.02 -.02 .14
Schwartz value: benevolence .47 .29 .19 .13 .22
Psychopathy (reversed) .50 .42 .03 .07 .06
Harm scales average .50 .35 .05 .06 .12
Good-Self: fair/just .15 .44 -.10 -.03 -.11
Schwartz value: social justice .52 .52 -.07 -.10 -.01
Social Dominance Orientation
(reversed) .57 .56 -.27 -.39 -.25
Fairness scales average .42 .51 -.14 -.18 -.12
Schwartz values: loyalty, national
security, family security .04 -.04 .53 .48 .37
Good-Self: loyal/faithful .11 .17 .46 .40 .33
Ingroup scales average .07 .07 .50 .44 .35
Schwartz values: social order,
authority, respect for tradition,
honoring parents, obedience .07 -.04 .53 .62 .54
Progressive and Traditional Justice
Scale – Traditionalism -.37 -.25 .35 .37 .30
Right-Wing Authoritarianism -.25 -.37 .56 .65 .70
Authority scales average -.18 -.22 .48 .55 .51
Schwartz values: self-discipline,
clean, devout .08 -.06 .37 .44 .61
Disgust Scale – Revised .23 .12 .21 .19 .34
Religious Attendance .02 -.12 .27 .32 .54
Purity scales average .11 -.02 .28 .32 .50
Note. The highest correlation for each set of scales is shown in bold.

Mapping the Moral Domain 30
Table 8. Partial correlations of foundation scores and attitudes toward foundation-related social
groups.
MFQ subscales:
Social groups aggregate: Harm Fairness Ingroup Authority Purity
Harm-related groups 0.28 0.09 0.00 -0.06 0.11
Fairness-related groups 0.26 0.28 -0.10 -0.16 0.06
Ingroup-related groups 0.10 0.03 0.35 0.22 -0.05
Authority-related groups -0.10 -0.01 0.22 0.29 0.04
Purity-related groups 0.14 -0.04 0.12 0.26 0.44
Note. Figures represent partial correlations controlling for political ideology. The highest partial correlation in each
row is shown in bold.

Mapping the Moral Domain 31
Table 9. Incremental predictive validity comparisons between the MFQ and Schwartz Value
Survey (SVS)
External
Scales: R² for SVS
alone
∆R² adding
MFQ to
SVS
R² for MFQ
alone
∆R² adding
SVS to
MFQ
Harm scales:
IRI-empathic concern 0.39 0.10 0.39 0.10
Psychopathy 0.36 0.04 0.23 0.17
GoodSelf (kind/caring,
sympathetic/compassion
ate, generous/giving) 0.33 0.03 0.21 0.15
Fairness scales:
SDO 0.47 0.11 0.51 0.07
GS fair/just 0.19 0.08 0.16 0.11
Ingroup scale:
GS loyal/faithful 0.28 0.13 0.31 0.10
Authority scales:
RWA 0.53 0.15 0.65 0.03
QAJ-Traditional justice 0.25 0.09 0.30 0.03
Purity scales:
Disgust Scale-Revised 0.10 0.10 0.15 0.06
Religious attendance 0.28 0.07 0.28 0.07
Social Groups:
Harm-related groups 0.27 0.08 0.29 0.06
Fairness-related groups 0.27 0.06 0.28 0.05
Ingroup-related groups 0.25 0.05 0.26 0.04
Authority-related groups 0.24 0.08 0.29 0.03
Purity-related groups 0.26 0.07 0.27 0.06
Issue positions:
Global warming 0.30 0.05 0.27 0.08
Gun Control 0.20 0.07 0.23 0.04
Immigration 0.19 0.06 0.22 0.03
Defense spending 0.26 0.06 0.29 0.03
Flag-burning 0.15 0.04 0.18 0.01
Terrorism 0.16 0.06 0.21 0.01
Torture 0.29 0.08 0.33 0.04
Abortion 0.33 0.10 0.39 0.04
Stem-cell research 0.26 0.07 0.28 0.05
Gay Marriage 0.33 0.12 0.43 0.02
Evolution/Creationism in
schools 0.21 0.08 0.27 0.02
Note. Bold indicates which scale, Schwartz Values Scale (SVS) or Moral Foundations Questionnaire (MFQ), is the
stronger predictor alone.

Mapping the Moral Domain 32
Table 10. Fit indices for a five-factor model of moral concerns across multiple nations and world
regions
Nation or world region N χ2 εaCFI
United States 26,014 42825.0 .048 .824
Canada 2,072 3189.0 .048 .826
Western Europe 1,670 2637.8 .046 .811
United Kingdom 1,175 1978.7 .046 .837
Australia 707 1493.0 .053 .807
Eastern Europe 413 859.2 .043 .876
Latin America 411 907.2 .039 .841
South Asia 299 747.3 .042 .838
Middle East 225 778.4 .051 .794
Southeast Asia 203 820.5 .056 .748
East Asia 163 675.0 .049 .729
Africa 56 --- --- ---
Note. For each fit analysis there were 395 degrees of freedom and 100 parameters estimated. Fit indices could not be
computed for Africa due to insufficient sample size.

Mapping the Moral Domain 33
Table 11. Foundation correlations with political ideology by world area.
USA UK Canada Australia
Western
Europe
Eastern
Europe
Latin
America Africa
Middle
East
South
Asia
East
Asia
Southeast
Asia Average
N 80,322 2,579 4,314 1,563 3,766 888 1,345 153 575 884 479 550
Harm -0.35 -0.25 -0.31 -0.28 -0.22 -0.17 -0.16 -0.04 -0.19 -0.14 -0.19 -0.12 -0.20
Fairness -0.44 -0.40 -0.36 -0.38 -0.33 -0.28 -0.33 -0.35 -0.32 -0.21 -0.24 -0.29 -0.32
Ingroup 0.47 0.42 0.34 0.44 0.35 0.35 0.32 0.39 0.42 0.33 0.28 0.33 0.37
Authority 0.56 0.51 0.50 0.53 0.47 0.43 0.46 0.51 0.56 0.42 0.37 0.48 0.48
Purity 0.58 0.46 0.47 0.52 0.46 0.45 0.48 0.51 0.51 0.45 0.42 0.49 0.49

Mapping the Moral Domain 34
Figure 1. Example of item selection procedure for final version of MFQ
Note. Figure shows how much is gained, in terms of correlations with internal and external validity criteria scales, as
the number of items in each subscale is increased from 1-4. The panels show analyses of all possible item
combinations in the Harm Relevance subscale, plotting their correlations with the Harm judgments subscale (lower
right quadrant), Harm sacredness scale (upper right), IRI-empathy subscale (lower left), and corrected item-total
correlations with all three criterion scales. Letters refer to items in the Harm relevance subscale: H=Harmed,
E=Emotionally, C=Cruel, W=Weak (all items can be found in the Appendix except Harmed, which was dropped
because of this analysis).
Harm relevance
HEWC
EWC
HEW
HEC
HWC
WC
EC
EW
HC
HW
C
W
E
H
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0 1 2 3 4
Corrected item-total correlation (3 criteria)
Harm Relevance
HEWC
EWC
HWC
HEC
HEW
WC
EC
EW
HC
HW
HE
C
W
E
H
0.00
0.10
0.20
0.30
0.40
0.50
0.60
01234
# of items
Correlation with IRI-empathy
Harm Relevance
HEWC
EWC
HWC HEC
HEW
WC
EC
EW
HC
HW
HE
C
W
E
H
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0 1 2 3 4
# of items
Correlation with MFQ Harm judgm ents
Harm Relevance
HEWC
EWC
HWC
HEC
HEW
WC
EC
EW
HC
HW
HE
C
W
E
H
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0 1 2 3 4
# of items
Correlation with Harm-Taboo-Tradeoffs

Mapping the Moral Domain 35
Figure 2. Moral Foundations Questionnaire confirmatory factor analysis model comparisons
Single factor:
WEA
e1
EMO
e2
TRE
e4
UNF
e5
RIT
e6
BET
e8
LOY
e9
RES
e10
TRA
e11
DEC
e14
DIS
e15
CRU
e16
LUV
e18
CHA
e19
GOD
e20
ANI
e21
KIL
e22
COM
e23
JUS
e26
FAI
e27
RIC
e28
TEA
e29
HIS
e30
FAM
e31
SEX
e33
SOL
e34
KID
e35
HDG
e37
UNN
e38
CHA
e40
H - F - I - A - P
.49
.46
.45
.50 .34
.67
.62
.48
.69 .40
.48 .64
.54 .69 .62
.64 .65 .65
.00 .07 .06-.18 .00-.11
-.22
-.14
.08
-.02
-.08
-.02
-----------------------------------------------------------------------------------------------------------------------------
Two correlated factors:
WEA
e1
EMO
e2
TRE
e4
UNF
e5
RIT
e6
BET
e8
LOY
e9
RES
e10
TRA
e11
DEC
e14
DIS
e15
CRU
e16
LUV
e18
CHA
e19
GOD
e20
ANI
e21
KIL
e22
COM
e23
JUS
e26
FAI
e27
RIC
e28
TEA
e29
HIS
e30
FAM
e31
SEX
e33
SOL
e34
KID
e35
HDG
e37
UNN
e38
CHA
e40
HARM -
FAIRNESS
INGROUP -
AUTHORITY -
PURITY
-.04
.48 .56 .58
.28 .49 .37
.49
.46
.46
.49 .34
.67
.59 .64
.39 .35
.59
.63
.62
.47
.69 .41
.47 .64
.55 .70 .62
.64 .65 .65
-----------------------------------------------------------------------------------------------------------------------------
Three correlated factors:
WEA
e1
EMO
e2
TRE
e4
UNF
e5
RIT
e6
BET
e8
LOY
e9
RES
e10
TRA
e11
DEC
e14
DIS
e15
CRU
e16
LUV
e18
CHA
e19
GOD
e20
ANI
e21
KIL
e22
COM
e23
JUS
e26
FAI
e27
RIC
e28
TEA
e29
HIS
e30
FAM
e31
SEX
e33
SOL
e34
KID
e35
HDG
e37
UNN
e38
CHA
e40
PURITY
.58
.71
.73
.70 .70
.67
HARM -
FAIRNESS
INGROUP -
AUTHORITY
-.06
.47 .56 .58
.28 .49 .37
.53
.48
.50
.53 .36
.70
-.01
.80
.59 .64
.39 .35
.59
.63
.63
.46
.73 .43
.52 .67

Mapping the Moral Domain 36
-----------------------------------------------------------------------------------------------------------------------------
Five correlated factors:
WEA
e1
EMO
e2
TRE
e4
UNF
e5
RIT
e6
BET
e8
LOY
e9
RES
e10
TRA
e11
DEC
e14
DIS
e15
CRU
e16
LUV
e18
CHA
e19
GOD
e20
ANI
e21
KIL
e22
COM
e23
JUS
e26
FAI
e27
RIC
e28
TEA
e29
HIS
e30
FAM
e31
SEX
e33
SOL
e34
KID
e35
HDG
e37
UNN
e38
CHA
e40
HARM AUTHORITY
.61
.39
.70
.35
.62
.67
PURITY
.58
.70
.74
.70 .70
.67
FAIRNESS INGROUP
-.09 .88
-.16
.55 .60 .67
.32 .57 .37
.59
.47
.56
.53 .39
.73 .64
.46 .52
.74
.69
.43
.06
-.01
.72 .80
-.15
.72
.09
-----------------------------------------------------------------------------------------------------------------------------
Six correlated factors:
WEA
e1
EMO
e2
TRE
e4
UNF
e5
RIT
e6
BET
e8
LOY
e9
RES
e10
TRA
e11
DEC
e14
DIS
e15
CRU
e16
LUV
e18
CHA
e19
GOD
e20
ANI
e21
KIL
e22
COM
e23
JUS
e26
FAI
e27
RIC
e28
TEA
e29
HIS
e30
FAM
e31
SEX
e33
SOL
e34
KID
e35
HDG
e37
UNN
e38
CHA
e40
HARM
AUTHORITY
.61
.39
.70
.35
.62
.67
PURITY
.58
.70
.74
.70 .70
.67
FAIRNESS INGROUP
-.09 .85
-.15
.56 .60 .67
.32 .57 .37
.59
.47
.56
.53 .39
.73
.53
.76
.71
.06
-.01
.72
-.15
.72
.09
TRADITION .88
.91
-.20
-.01
.75
.96
.47
.62
.41
-----------------------------------------------------------------------------------------------------------------------------
Hierarchical model:
WEA
e1
EMO
e2
TRE
e4
UNF
e5
RIT
e6
BET
e8
LOY
e9
RES
e10
TRA
e11
DEC
e14
DIS
e15
CRU
e16
LUV
e18
CHA
e19
GOD
e20
ANI
e21
KIL
e22
COM
e23
JUS
e26
FAI
e27
RIC
e28
TEA
e29
HIS
e30
FAM
e31
SEX
e33
SOL
e34
KID
e35
HDG
e37
UNN
e38
CHA
e40
HARM AUTHORITY
.62
.39
.69
.35
.63
.67
PURITY
.58
.70
.73
.71 .70
.67
FAIRNESS I NGROUP
.53 .61 .68
.33 .57 .36
.58
.47
.55
.54 .39
.73 .64
.46 .51
.74
.69
.44
INDIVIDUALIZING BINDING
-.10
.75 .96 .89 .99 .81
e50
e51 e52 e53 e54
Note. Models shown for full MFQ scale. Fit and comparison statistics for these models, and for models
including relevance and judgments data separately, are shown in Tables 5 and 6.

Mapping the Moral Domain 37
Appendix. Moral Foundations Questionnaire
Part I – Moral Relevance
(responded to using the following response options: not at all relevant , not very relevant, slightly
relevant, somewhat relevant, very relevant, extremely relevant)
Harm:
EMOTIONALLY - Whether or not someone suffered emotionally*
WEAK - Whether or not someone cared for someone weak or vulnerable*
CRUEL - Whether or not someone was cruel
Fairness:
TREATED - Whether or not some people were treated differently from others*
UNFAIRLY - Whether or not someone acted unfairly*
RIGHTS - Whether or not someone was denied his or her rights
Ingroup:
LOVECOUNTRY - Whether or not someone’s action showed love for his or her country*
BETRAY - Whether or not someone did something to betray his or her group*
LOYALTY - Whether or not someone showed a lack of loyalty
Authority:
RESPECT - Whether or not someone showed a lack of respect for authority*
TRADITIONS - Whether or not someone conformed to the traditions of society*
CHAOS - Whether or not an action caused chaos or disorder
Purity:
DECENCY - Whether or not someone violated standards of purity and decency*
DISGUSTING - Whether or not someone did something disgusting*
GOD - Whether or not someone acted in a way that God would approve of
(An asterisk indicates that the item is also included in the 20 item short-form MFQ.)

Mapping the Moral Domain 38
Part II – Moral Judgments
(responded to using the following response options: strongly disagree, moderately disagree,
slightly disagree, slightly agree, moderately agree, strongly agree)
Harm:
COMPASSION - Compassion for those who are suffering is the most crucial virtue.*
ANIMAL - One of the worst things a person could do is hurt a defenseless animal.*
KILL - It can never be right to kill a human being.
Fairness:
FAIRLY - When the government makes laws, the number one principle should be ensuring
that everyone is treated fairly.*
JUSTICE – Justice is the most important requirement for a society.*
RICH - I think it’s morally wrong that rich children inherit a lot of money while poor children
inherit nothing.
Ingroup:
HISTORY - I am proud of my country’s history.*
FAMILY - People should be loyal to their family members, even when they have done
something wrong.*
TEAM - It is more important to be a team player than to express oneself.
Authority:
KIDRESPECT - Respect for authority is something all children need to learn.*
SEXROLES - Men and women each have different roles to play in society.*
SOLDIER - If I were a soldier and disagreed with my commanding officer’s orders, I would
obey anyway because that is my duty.
Purity:
HARMLESSDG - People should not do things that are disgusting, even if no one is harmed.*
UNNATURAL - I would call some acts wrong on the grounds that they are unnatural.*
CHASTITY - Chastity is an important and valuable virtue.
(An asterisk indicates that the item is also included in the 20 item short-form MFQ.)