PERSONALITY PROCESSES AND INDIVIDUAL DIFFERENCES
Liberals and Conservatives Rely on Different Sets of Moral Foundations
Jesse Graham, Jonathan Haidt, and Brian A. Nosek
University of Virginia
How and why do moral judgments vary across the political spectrum? To test moral foundations theory
(J. Haidt & J. Graham, 2007; J. Haidt & C. Joseph, 2004), the authors developed several ways to measure
people’s use of 5 sets of moral intuitions: Harm/care, Fairness/reciprocity, Ingroup/loyalty, Authority/
respect, and Purity/sanctity. Across 4 studies using multiple methods, liberals consistently showed greater
endorsement and use of the Harm/care and Fairness/reciprocity foundations compared to the other 3
foundations, whereas conservatives endorsed and used the 5 foundations more equally. This difference
was observed in abstract assessments of the moral relevance of foundation-related concerns such as
violence or loyalty (Study 1), moral judgments of statements and scenarios (Study 2), “sacredness”
reactions to taboo trade-offs (Study 3), and use of foundation-related words in the moral texts of religious
sermons (Study 4). These findings help to illuminate the nature and intractability of moral disagreements
in the American “culture war.”
Keywords: morality, ideology, liberal, conservative
Political campaigns spend vast sums appealing to the self-
interests of voters, yet rational self-interest often shows a weak and
unstable relationship to voting behavior (Kinder, 1998; Miller,
1999; Sears & Funk, 1991). Voters are also influenced by a wide
variety of social and emotional forces (Marcus, 2002; Westen,
2007). Some of these forces are trivial or peripheral factors whose
influence we lament, such as a candidate’s appearance (Ballew &
Todorov, 2007). In recent years increasing attention has been paid
Voters who seem to vote against their material self-interest are
sometimes said to be voting instead for their values, or for their
vision of a good society (Lakoff, 2004; Westen, 2007). However,
the idea of what makes for a good society is not universally shared.
The “culture war” that has long marked American politics (Hunter,
1991) is a clash of visions about such fundamental moral issues as
the authority of parents, the sanctity of life and marriage, and the
proper response to social inequalities. Ideological commitments
are moral commitments; they are not necessarily strategies for
In this article we examine moral foundations theory, which was
originally developed to describe moral differences across cultures
(Haidt & Joseph, 2004). Building on previous theoretical work
(Haidt & Graham, 2007), we apply the theory to moral differences
across the political spectrum within the United States. We propose
a simple hypothesis: Political liberals construct their moral sys-
tems primarily upon two psychological foundations—Harm/care
and Fairness/reciprocity—whereas political conservatives con-
struct moral systems more evenly upon five psychological foun-
dations—the same ones as liberals, plus Ingroup/loyalty, Author-
ity/respect, and Purity/sanctity. We call this hypothesis the moral
foundations hypothesis, and we present four studies that support it
using four different methods.
Liberals and Conservatives
Political views are multifaceted, but a single liberal–
conservative (or left–right) continuum is a useful approximation
that has predictive validity for voting behavior and opinions on a
wide range of issues (Jost, 2006). In terms of political philosophy,
the essential element of all forms of liberalism is individual liberty
(Gutmann, 2001). Liberals have historically taken an optimistic
view of human nature and of human perfectibility; they hold what
Sowell (2002) calls an “unconstrained vision” in which people
should be left as free as possible to pursue their own courses of
personal development. Conservatism, in contrast, is best under-
stood as a “positional ideology,” a reaction to the challenges to
authority and institutions that are so often mounted by liberals
(Muller, 1997). Conservatives have traditionally taken a more
Jesse Graham, Jonathan Haidt, and Brian A. Nosek, Department of
Psychology, University of Virginia.
We thank Mark Berry for creating the supplemental text analysis pro-
gram used in Study 4 and thank Yoav Bar-Anan, Pete Ditto, Ravi Iyer,
Selin Kesebir, Sena Koleva, Allison Meade, Katarina Nguyen, Eric Oliver,
Shige Oishi, Colin Tucker Smith, and Tim Wilson for helpful comments on
earlier drafts. This research was supported by Institute for Education
Sciences and Jacob Javits fellowships and a grant from the National
Institute of Mental Health (R01 MH68447). Supplemental information and
analyses can be found at www.moralfoundations.org.
Correspondence concerning this article should be addressed to Jesse
Graham, Department of Psychology, University of Virginia, P.O. Box
400400, Charlottesville, VA 22904. E-mail: email@example.com
Journal of Personality and Social Psychology, 2009, Vol. 96, No. 5, 1029–1046
© 2009 American Psychological Association 0022-3514/09/$12.00DOI: 10.1037/a0015141
pessimistic view of human nature, believing that people are inher-
ently selfish and imperfectible. They therefore hold what Sowell
called a “constrained vision” in which people need the constraints
of authority, institutions, and traditions to live civilly with each
In terms of their personalities, liberals and conservatives have long
been said to differ in ways that correspond to their conflicting visions.
Liberals on average are more open to experience, more inclined to
seek out change and novelty both personally and politically (McCrae,
1996). Conservatives, in contrast, have a stronger preference for
things that are familiar, stable, and predictable (Jost, Nosek, & Gos-
ling, 2008; McCrae, 1996). Conservatives—at least, the subset prone
to authoritarianism—also show a stronger emotional sensitivity to
threats to the social order, which motivates them to limit liberties in
defense of that order (Altemeyer, 1996; McCann, 2008; Stenner,
2005). Jost, Glaser, Sulloway, and Kruglanski (2003) concluded from
a meta-analysis of this literature that the two core aspects of conser-
vative ideology are resistance to change and acceptance of inequality.
How can these various but complementary depictions of ideological
and personality differences be translated into specific predictions
about moral differences? First, we must examine and revise the
definition of the moral domain.
Expanding the Moral Domain
The consensus view in moral psychology has been that morality
is first and foremost about protecting individuals. The most cited
definition comes from Turiel (1983, p. 3), who defined the moral
domain as “prescriptive judgments of justice, rights, and welfare
pertaining to how people ought to relate to each other.” Turiel
(2006) explicitly grounded this definition in the tradition of liberal
political theory from Kant through John Stuart Mill to John Rawls.
Nearly all research in moral psychology, whether carried out using
interviews, fMRI, or dilemmas about stolen medicine and runaway
trolleys, has been limited to issues of justice, rights, and welfare.
When morality is equated with the protection of individuals, the
central concerns of conservatives—and of people in many non-
Western cultures—fall outside the moral domain. Research in India,
Brazil, and the United States, for example, has found that people who
are less Westernized treat many issues related to food, sex, clothing,
prayer, and gender roles as moral issues (Shweder, Mahapatra, &
Miller, 1987), even when they involve no harm to any person (Haidt,
Koller, & Dias, 1993). Shweder, Much, Mahapatra, and Park (1997)
proposed that Western elites are unusual in limiting the moral domain
to what they called the “ethic of autonomy.” They proposed that
morality in most cultures also involves an “ethic of community”
the cohesiveness of groups and institutions) and an “ethic of divinity”
humanity’s baser, more carnal instincts).
Haidt (2008) recently suggested an alternative approach to defining
morality that does not exclude conservative and non-Western con-
cerns. Rather than specifying the content of a truly moral judgment he
specified the functions of moral systems: “Moral systems are inter-
locking sets of values, practices, institutions, and evolved psycholog-
ical mechanisms that work together to suppress or regulate selfishness
and make social life possible” (p. 70). Haidt described two common
kinds of moral systems—two ways of suppressing selfishness—that
correspond roughly to Sowell’s (2002) two visions. Some cultures try
to suppress selfishness by protecting individuals directly (often using
the legal system) and by teaching individuals to respect the rights of
other individuals. This individualizing approach focuses on individu-
als as the locus of moral value. Other cultures try to suppress selfish-
ness by strengthening groups and institutions and by binding individ-
uals into roles and duties in order to constrain their imperfect natures.
This binding approach focuses on the group as the locus of moral
The individualizing–binding distinction does not necessarily cor-
respond to a left-wing versus right-wing distinction for all groups and
in all societies. The political left has sometimes been associated with
socialism and communism, positions that privilege the welfare of the
group over the rights of the individual and that have at times severely
limited individual liberty. Conversely, the political right includes
libertarians and “laissez-faire” conservatives who prize individual
liberty as essential to the functioning of the free market (Boaz, 1997).
We therefore do not think of political ideology—or morality—as a
strictly one-dimensional spectrum. In fact, we consider it a strength of
moral foundations theory that it allows people and ideologies to be
characterized along five dimensions. Nonetheless, we expect that the
individualizing–binding distinction can account for substantial varia-
tion in the moral concerns of the political left and right, especially in
the United States, and that it illuminates disagreements underlying
many “culture war” issues.
Moral Foundations Theory
Several theorists have attempted to reduce the panoply of human
values to a manageable set of constructs or dimensions. The two
most prominent values researchers (Rokeach, 1973; Schwartz,
1992) measured a wide variety of possible values and aggregated
them through factor analysis to define a smaller set of core values.
Schwartz (1992) and Rokeach (1973) both justified their lists by
pointing to the fundamental social and biological needs of human
2004) also tries to reduce the panoply of values, but with a different
strategy: We began not by measuring moral values and factor ana-
lyzing them but by searching for the best links between anthropolog-
ical and evolutionary accounts of morality. Our idea was that moral
intuitions derive from innate psychological mechanisms that co-
evolved with cultural institutions and practices (Richerson & Boyd,
2005). These innate but modifiable mechanisms (Marcus, 2004) pro-
vide parents and other socializing agents the moral “foundations” to
build on as they teach children their local virtues, vices, and moral
focus on morality and because it more strongly suggests cultural
learning and construction.)
To find the best candidate foundations, Haidt and Joseph (2004)
surveyed lists of virtues from many cultures and eras, along with
taxonomies of morality from anthropology (Fiske, 1992; Shweder et
al., 1997), psychology (Schwartz & Bilsky, 1990), and evolutionary
theories about human and primate sociality (Brown, 1991; de Waal,
1996). They looked for matches—cases of virtues or other moral
concerns found widely (though not necessarily universally) across
cultures for which there were plausible and published evolutionary
explanations of related psychological mechanisms. Two clear
matches were found that corresponded to Turiel’s (1983) moral do-
main and Shweder et al.’s (1997) ethics of autonomy. The widespread
GRAHAM, HAIDT, AND NOSEK
human obsession with fairness, reciprocity, and justice fits well with
evolutionary writings about reciprocal altruism (Trivers, 1971). And
vulnerable individuals from harm fits well with writings about the
evolution of empathy (de Waal, 2008) and the attachment system
(Bowlby, 1969). These two matches were labeled the Fairness/
reciprocity foundation and the Harm/care foundation, respectively. It
is noteworthy that these two foundations correspond to the “ethic of
justice” studied by Kohlberg (1969) and the “ethic of care” that
Gilligan (1982) said was an independent contributor to moral judg-
ment. We refer to these two foundations as the individualizing foun-
dations because they are (we suggest) the source of the intuitions that
make the liberal philosophical tradition, with its emphasis on the
rights and welfare of individuals, so learnable and so compelling to so
Haidt and Joseph (2004) found, however, that most cultures did not
limit their virtues to those that protect individuals. They identified
three additional clusters of virtues that corresponded closely to Sh-
weder et al.’s (1997) description of the moral domains that lie beyond
the ethics of autonomy. Virtues of loyalty, patriotism, and self-
sacrifice for the group, combined with an extreme vigilance for
traitors, matched recent work on the evolution of “coalitional psy-
chology” (Kurzban, Tooby, & Cosmides, 2001). Virtues of subordi-
nates (e.g., obedience and respect for authority) paired with virtues of
authorities (such as leadership and protection) matched writings on
the evolution of hierarchy in primates (de Waal, 1982) and the ways
that human hierarchy became more dependent on the consent of
subordinates (Boehm, 1999). These two clusters comprise most of
Shweder et al.’s “ethic of community.” And lastly, virtues of purity
and sanctity that play such a large role in religious laws matched
writings on the evolution of disgust and contamination sensitivity
(Rozin, Haidt, & McCauley, 2000). Practices related to purity and
Such practices also serve social functions, including marking off the
group’s cultural boundaries (Soler, 1973/1979) and suppressing the
selfishness often associated with humanity’s carnal nature (e.g., lust,
hunger, material greed) by cultivating a more spiritual mindset (see
to these three foundations (Ingroup/loyalty, Authority/respect, and
Purity/sanctity) as the binding foundations, because they are (we
suggest) the source of the intuitions that make many conservative and
religious moralities, with their emphasis on group-binding loyalty,
duty, and self-control, so learnable and so compelling to so many
If the foundations are innate, then why do people and cultures
vary? Why are there liberals and conservatives? We take our
understanding of innateness from Marcus (2004), who stated that
innate “does not mean unmalleable; it means organized in advance
of experience.” He uses the metaphor that genes create the first
draft of the brain, and experience later edits it. We apply Marcus’s
metaphor to moral development by assuming that human beings
have the five foundations as part of their evolved first draft, but
that, as for nearly all traits, there is heritable variation (Bouchard,
2004; Turkheimer, 2000). Many personality traits related to the
foundations have already been shown to be moderately heritable,
including harm avoidance (Keller, Coventry, Heath, & Martin,
2005) and right-wing authoritarianism (McCourt, Bouchard,
Lykken, Tellegen, & Keyes, 1999). But foundations are not values
or virtues. They are the psychological systems that give children
feelings and intuitions that make local stories, practices, and moral
arguments more or less appealing during the editing process.
Returning to our definition of moral systems as “interlocking
sets of values, practices, institutions, and evolved psychological
mechanisms” that function to suppress selfishness, it should now
be clear that the foundations are the main “evolved psychological
mechanisms” that are part of the “first draft” of the moral mind.
Elsewhere we describe in more detail the role of narrative, social
construction, and personal construction in the creation of adult
moral and ideological identities (Haidt, Graham, & Joseph, in
press; Haidt & Joseph, 2007).
Overview of Studies
In four studies we examined the moralities of liberals and
conservatives using four different methods that varied in the de-
gree to which they relied on consciously accessible beliefs versus
more intuitive responses. In Study 1, a large international sample
of respondents rated the moral relevance of foundation-specific
concerns. In Study 2, we examined liberals’ and conservatives’
moral judgments as a function of both explicit and implicit polit-
ical identity. In Study 3, we elicited stronger visceral responses by
presenting participants with moral trade-offs by asking them how
much money they would require to perform foundation-violating
behaviors. In Study 4, we analyzed moral texts—religious sermons
delivered in liberal and conservative churches—to see if speakers
in the different moral communities spontaneously used
foundation-related words in different ways. In all four studies we
found that liberals showed evidence of a morality based primarily
on the individualizing foundations (Harm/care and Fairness/
reciprocity), whereas conservatives showed a more even distribu-
tion of values, virtues, and concerns, including the two individu-
alizing foundations and the three binding foundations (Ingroup/
loyalty, Authority/respect, and Purity/sanctity).
Study 1: Moral Relevance
For our first test of the moral foundations hypothesis we used the
most direct method possible: We asked participants to rate how
relevant various concerns were to them when making moral judg-
ments. Such a decontextualized method can be appropriate for gaug-
ing moral values, as values are said to be abstract and generalized
of introspection (Nisbett & Wilson, 1977) and the intuitive quality of
many moral judgments (Haidt, 2001), such a method does not nec-
essarily measure how people actually make moral judgments. As
such, reports of moral relevance are best understood as self-theories
about moral judgment, and they are likely to be concordant with
explicit reasoning during moral arguments. We predicted that liberals
would rate concerns related to the individualizing foundations as
being more relevant than would conservatives, whereas that conser-
vatives would endorse concerns related to the binding foundations as
being more relevant than would liberals.
29) who had registered at the Project Implicit website (https://
THE MORAL FOUNDATIONS OF POLITICS
implicit.harvard.edu/) and were randomly assigned to this study. Dur-
ing registration, political self-identification was reported on a 7-point
scale anchored by strongly liberal and strongly conservative, with
moderate at the midpoint. Overall, 902 participants rated their polit-
ical identity as liberal, 366 as moderate, and 264 as conservative.
Eighty-one participants did not answer the question. Gender, age,
household income, and education level were also assessed at regis-
Participants first read “When you decide whether something is
right or wrong, to what extent are the following considerations
relevant to your thinking?” They then rated 15 moral relevance
items (see Appendix A) on 6-point scales anchored by the labels
never relevant and always relevant. The items were written to be
face-valid measures of concerns related to the five foundations,
with the proviso that no item could have an obvious relationship to
partisan politics. For example, a Fairness item stated “Whether or
not someone was denied his or her rights.” By avoiding mention
of specific “culture war” topics such as gun rights or gay rights,
we minimized the extent to which participants would recognize
the items as relevant to political ideology and therefore draw on
knowledge of what liberals and conservatives believe to guide
their own ratings. Cronbach’s alphas for the three-item mea-
sures of each foundation were .62 (Harm), .67 (Fairness), .59
(Ingroup), .39 (Authority), and .70 (Purity). (For information on
the factor structure of the moral foundations, see the General
Discussion below and refer to the supplement available at
A 16th item stated “Whether or not someone believed in astrol-
ogy.” This item served as a check for whether participants paid
attention, understood the scale, and responded meaningfully. We
expected that high ratings of relevance on this item reflected careless
or otherwise uninterpretable performance on the rest of the scale.
Sixty-five participants (4.0%) were excluded because they used
the upper half of the relevance scale in response to this item.1
The study pool at Project Implicit randomly assigns participants to
one of dozens of studies each time they return to the site (see Nosek,
2005, for more information about the Virtual Laboratory). After
random assignment to this study, participants completed the relevance
items in an order randomized for each participant. They also com-
pleted an Implicit Association Test (IAT) that is not relevant for this
report and is not discussed further. The order of implicit and explicit
tasks was randomized and had no effect on the results.
Figure 1 shows foundation scores (the average of the three
relevance items for each foundation) as a function of self-rated
political identity. As Figure 1 shows, the negative slopes for Harm
and Fairness (the individualizing foundations) means that conser-
vatives rated these issues as less relevant to their moral judgments
than did liberals. Conversely, the positive slopes for Ingroup,
Authority, and Purity (the binding foundations) means that con-
servatives rated these issues as more relevant to their moral judg-
ments than did liberals. (These patterns were consistent across
nearly all individual items for this and the other studies.)
We tested whether the effects of political identity persisted after
partialing out variation in moral relevance ratings for other demo-
graphic variables. We created a model representing the five foun-
dations as latent factors measured by three manifest variables each,
simultaneously predicted by political identity and four covariates:
age, gender, education level, and income. This model is shown in
Figure 2; for clarity, we show the standardized regression esti-
mates for politics only. Including the covariates, political identity
still predicted all five foundations in the predicted direction, all
ps ? .001. Political identity was the key explanatory variable: It
was the only consistent significant predictor (average |?| ? .25;
range .16 to .34) for all five foundations.2
To test whether this moral foundations pattern was unique to the
United States, we created a multigroup version of the model shown in
Figure 2. We had enough participants from the United States (n ?
695) and the United Kingdom (n ? 477) to create separate groups,
and we put participants from other nations (n ? 417) into a third
group (the countries most represented in this third group were Argen-
tina, n ? 61, and Canada, n ? 44). We first constrained the individual
item loadings to be the same across the three groups, to test whether
people in different nations interpreted the items or used the scale
differently. This model was not significantly different from the un-
constrained model, ??2(20) ? 42.27, ?εa? .01, ns, suggesting that
the factor loadings were invariant across our three nation groups. We
then added equality constraints across nation groups to the regression
estimates from political identity predicting each of the five founda-
tions, to see if politics had differential effects across nations. This
model was also not different from the fully unconstrained model,
??2(30) ? 47.79, ?εa? .02, ns, indicating that the effects of politics
on foundation relevance scores were not different across nations. This
suggests that the relations between political identity and moral foun-
dations are consistent across our U.S., U.K., and “other nations”
groups. In all three groups the individualizing foundations were en-
dorsed more strongly by liberals than conservatives, and the binding
foundations were endorsed more strongly by conservatives than lib-
erals. Further, in all three nation groups liberals were more likely than
conservatives to consider individualizing concerns more morally rel-
evant than binding concerns.
Because the latent variable model does not provide insight into the
rank ordering of the different sets of foundations, we compared them
using a repeated-measures general linear model including politics as
a covariate. For the sample as a whole, the aggregated moral rele-
vance ratings for individualizing foundations were higher than the
aggregated ratings for the binding foundations, F(1, 1207) ?
1,895.09, p ? .001, ?p
politics, F(1, 1207) ? 224.34, p ? .001, ?p
2? .61, and this effect was moderated by
2? .16, such that the more
1Although this cutoff criterion was set a priori, we also tested to see if
these participants differed systematically from other participants. These 65
participants had significantly higher individual means and lower SDs on all
five subscales, suggesting they were more likely to have been giving
consistently high ratings, perhaps due to carelessness. They did not differ
from the rest of the sample in terms of gender, age, or political identity. The
removal of their data did not significantly change the results.
2Average |?|s from the model shown in Figure 2 were .05 for age, .09
for gender, .08 for income, and .06 for education.
GRAHAM, HAIDT, AND NOSEK
liberal participants showed a greater difference between the individ-
ualizing and binding moral foundations.
Study 1 provides initial empirical support for the moral foundations
hypothesis. We sampled broadly across the universe of potential
moral concerns. Had we limited the sampling to issues related to
Harm/care and Fairness/reciprocity (Turiel, 1983), liberals would
have appeared to have more (or more intense) moral concerns than
conservatives (cf. Emler, Renwick, & Malone, 1983). By asking
about issues related to binding groups together—namely, Ingroup/
loyalty, Authority/respect, and Purity/sanctity—we found a more
complex pattern. The moral thinking of liberals and conservatives
may not be a matter of more versus less but of different opinions
about what considerations are relevant to moral judgment. Further,
the observed differences were primarily a function of political
identity and did not vary substantially or consistently by gender,
age, household income, or education level, suggesting that these
effects could be a general description of moral concerns between
the political left and right.
Importantly, the differences between liberals and conservatives
were neither binary nor absolute. Participants across the political
spectrum agreed that individualizing concerns are very relevant to
moral judgment. Even on the binding foundations, liberals did not
(on average) indicate that these were never relevant to moral
judgment. As is the case with politics in general, the most dramatic
evidence for our hypotheses came from partisans at the extremes.
The moral relevance ratings were self-assessments of what factors
matter to a person when making moral judgments; they were not
actual moral judgments. We expand the investigation to moral judg-
ments in Study 2.
Study 2: Moral Judgments
In Study 2, we retained the abstract moral relevance assessments
from Study 1 and added more contextualized and concrete items that
could more strongly trigger the sorts of moral intuitions that are said
to play an important role in moral judgment (Haidt, 2001). We
statement about government policy (e.g., “The government should
Self-reported political identity
Slightly Liberal Moderately
Relevance to Moral decisions (0=never, 5=always)
Relevance of moral foundations across political identity, Study 1.
THE MORAL FOUNDATIONS OF POLITICS
strive to improve the well-being of people in our nation, even if it
sometimes happens at the expense of people in other nations” for
Ingroup), one hypothetical scenario (e.g., “If I were a soldier and
disagreed with my commanding officer’s orders, I would obey any-
way because that is my duty” for Authority), and one positive virtue
(e.g., “Chastity is still an important virtue for teenagers today, even if
many don’t think it is” for Purity; see Appendix B). This approach
In addition to asking participants to self-report their political
identification on a single-item liberal–conservative scale, we also
gave participants an IAT (see Nosek, Greenwald, & Banaji, 2006,
for a review) measuring implicit political identity (association
strengths between liberal–conservative and self–other) to test
whether automatic political identities are predictive of differences
in assessments of moral relevance and moral judgments above and
beyond explicit assessments of one’s political identity.
Participants were 2,212 volunteers (62% female, 38% male;
median age 32) from the research pool at Project Implicit. The
automated study assignment excluded visitors who had partic-
ipated in Study 1 and only included citizens or residents of the
United States because some measures contained U.S. political
figures. In the sample 1,174 participants were liberal, 538 were
moderate, and 500 were conservative. Data from 77 participants
were excluded because of high ratings on the astrology item; the
removal of their data did not significantly alter any of the
Moral relevance items.
used in Study 1, with one or two additional items for each of the
five foundations (see Appendix A). Cronbach’s alphas for each
foundation were .71 (Harm), .70 (Fairness), .71 (Ingroup), .64
(Authority), and .76 (Purity). There were three different versions
of the relevance items—answered as oneself (same as Study 1), as
a typical liberal, or as a typical conservative.
Moral judgment items.
Moral judgment statements were rated
on a 6-point scale, from strongly disagree to strongly agree.
Cronbach’s alphas of the judgment items for each foundation were
.50 (Harm), .39 (Fairness), .24 (Ingroup), .64 (Authority), and .74
(Purity). As with relevance items, there were three versions of the
moral judgment items—answered as oneself, as a typical liberal, or
as a typical conservative. For both types of items, only results for
the self versions are reported here.
Explicit political identity was measured dur-
ing registration as described in Study 1. Implicit political identity
was measured with the IAT. The IAT assesses associations among
Relevance items were the same ones
1. Numbers to the left of foundations indicate standardized regression estimates of the effects of political
identity; positive numbers indicate higher for conservatives, negative numbers indicate higher for liberals.
Parameters estimated ? 90, ??2(140) ? 2,016.85, εa? .093. The abbreviations at the right stand for the items
in the scale given in Appendix A. VIO ? violence; SUF ? suffering; HAR ? harm; DIF ? differently; UNF ?
unfair; RIT ? rights; FRN ? friend; LOY ? loyalty; BET ? betray; DUT ? duties; RES ? respect; RAN ?
rank; DIS ? disgust; UNN ? unnatural; PUR ? purity.
Latent variable model testing multiple predictors of moral foundation relevance assessments, Study
GRAHAM, HAIDT, AND NOSEK
two concept categories (liberals and conservatives) and two iden-
tity attributes (self and other). Stimuli were pictures of well-known
U.S. political figures and words corresponding to “self” or “other.”
The IAT was scored with the D algorithm recommended by
Greenwald, Nosek, and Banaji (2003).
To keep the study short, and because we had high power with
a large sample, we used a planned incomplete design in which
participants were randomly assigned to receive four of the six
self-report measures (self, typical liberal, and typical conserva-
tive versions of the relevance and judgments measure). As such,
the Ns for any given measure are, by design, approximately 2/3
of the total sample. Participants completed the IAT and the four
self-report measures in an order randomized for each partici-
Moral relevance ratings replicated findings from Study 1: Self-
reported conservatives endorsed the individualizing foundations
less and binding foundations more than self-reported liberals did.
We created a latent variable model (parameters estimated ? 114),
?2(320) ? 3,712.26, εa? .07, identical in structure to the model
shown in Figure 2, with self-reported political identity and four
covariates (age, gender, household income, and education) predict-
ing each of the five foundations. With the covariates included,
explicit political identity still predicted moral relevance assess-
ments in the predicted direction for Harm (? ? –.27), Fairness
(? ? –.36), Ingroup (? ? .11), Authority (? ? .39), and Purity
(? ? .38), all ps ? .001.
Considered by-item and in aggregate, the moral foundations
hypothesis was supported for direct moral judgments as well:
Conservatives on average agreed with individualizing foundation
judgments less than liberals and with binding foundation judg-
ments more. Average moral judgments for each foundation are
plotted by explicit political identity in Figure 3. We created a latent
variable model (parameters estimated ? 105), ?2(245) ? 2,414.62,
εa? .06, with self-reported political identity and the same four
covariates predicting each of the five foundations. With the four
other demographic factors included, explicit political identity still
predicted moral judgments in the predicted direction for Harm
(? ? –.32), Fairness (? ? –.43), Ingroup (? ? .67), Authority
(? ? .62), and Purity (? ? .57), all ps ? .001.
Ratings for individualizing foundations were higher than ratings
for the binding foundations for both relevance items, F(1, 1205) ?
1,215.62, p ? .001, ?p
635.58, p ? .001, ?p
politics for both relevance items, F(1, 1205) ? 450.42, p ? .001,
2? .50, and judgment items, F(1, 1200) ?
2? .35, and this effect was moderated by
2? .27, and judgment items, F(1, 1200) ? 649.40, p ? .001,
2? .35, such that the more liberal participants showed the
Implicit Political Identity
Implicit political identity correlated strongly with self-reported
political identity (r ? .63, p ? .001). Likewise, the relationship
between implicit political identity and the moral foundations—
both relevance and judgments—replicated the patterns for explicit
political identity. Latent variable models containing implicit polit-
ical identity instead of self-reported political identity yielded the
same pattern of results as those reported above. This confirms, at
minimum, that the differences in moral relevance and moral judg-
ment ratings do not depend on how people interpret the single-item
political identity measure.
Despite their strong positive correlation, implicit and explicit
political identity may each provide incremental validity of moral
foundation relevance and judgment ratings. Implicit and explicit
political identities were entered as simultaneous predictors (along
with age, gender, household income, and education) of foundation
scores in latent variable models for moral relevance and moral
judgments. For moral relevance ratings, explicit political identity
remained a significant predictor for all foundations (|?|s ? .26 to
.34, all ps ? .001) except Ingroup (|?| ? .05, ns), and implicit
political identity showed some incremental predictive validity
(above and beyond explicit political identity) for the binding
foundations (|?|s ? .10 to .15, all ps ? .05) but not for the
individualizing foundations (|?|s ? .05, ns).
We expected that implicit political identity would show stronger
incremental validity beyond explicit political identity in predicting
moral judgments, as applications of moral intuitions and reasons
rather than self-theories about moral judgment. Both implicit and
explicit political identity showed incremental predictive validity
over the other for all five foundations (|?|s ? .16 to .52, all ps ?
.001). This is particularly impressive considering that implicit and
explicit political identities were strongly correlated, leaving sub-
stantially reduced independent variance for prediction in the si-
multaneous models. Implicit political identity was a stronger si-
multaneous predictor of judgments than of relevance assessments
for all five foundations.
Study 2 replicated the political differences in moral relevance
ratings observed in Study 1 and extended support of the moral
foundations hypothesis to concrete moral judgments. Liberals were
more concerned than conservatives about issues of Harm and
Fairness, whereas conservatives were more concerned than liberals
about issues related to Ingroup, Authority, and Purity. Also, for
both moral relevance and moral judgments, the effects were ob-
served across both explicit and implicit political identities. As
before, we found a convergence pattern in which liberals made a
big distinction between the individualizing and binding founda-
tions, whereas conservatives—particularly strong conservatives—
weighted the two kinds of concerns more or less equally (see
Figure 3). Interestingly, implicit political identity contributed
unique predictive validity beyond that of self-reported political
identity for the moral judgments measure (all five foundations) but
did so more weakly for the moral relevance measure (and only for
the binding foundations). This suggests that moral judgments are
influenced by more than explicit self-theories of moral relevance.
Asking someone their political identity was largely sufficient for
THE MORAL FOUNDATIONS OF POLITICS
predicting what they said would be relevant to them in making moral
judgments, but measuring their implicit political identity added
uniquely predictive validity for moral judgments themselves.
Study 3: Moral Trade-Offs
In Study 3, we adapted Tetlock’s work on sacred values and
taboo trade-offs (Tetlock, 2003; Tetlock, Kristel, Elson, Green,
& Lerner, 2000) to make moral judgments more personal and
visceral than they had been in Studies 1 and 2. Tetlock et al.
(2000, p. 853) defined sacred values as “any value that a moral
community explicitly or implicitly treats as possessing infinite
or transcendental significance that precludes comparisons,
trade-offs, or indeed any other mingling with bounded or sec-
ular values.” Participants confronted with choices that involved
trading off a sacred value (such as human life) for a profane
value (such as money saved by a hospital) showed resistance to
the task and feelings of pollution afterwards, as if it were
impure even to contemplate the trade-off.
We generated five potential taboo violations for each moral foun-
dation. For example, how much money would someone have to pay
you to: Kick a dog in the head (Harm)? Renounce your citizenship
hypothesized that because everyone’s morality relies heavily on the
individualizing foundations, neither liberals nor conservatives would
be happy to “prostitute” their values by accepting money in exchange
for violating them. We predicted that liberals would be less likely to
see trade-offs related to the binding foundations as violations of
sacred values and, therefore, would be more willing to perform these
actions for some amount of money.
Participants were 8,193 adults (40% female, 60% male; median
age 34) who volunteered at www.yourmorals.org; 6,728 were from
the United States, 513 were from Europe, 281 were from Canada,
Self-reported political identity
Agreement with moral statements
(0=strongly disagree, 5=strongly agree)
indicates division of agreement and disagreement (2 indicates slight disagreement and 3 indicates slight
Agreement with moral statements across political identity, Study 2. The horizontal line at 2.5
GRAHAM, HAIDT, AND NOSEK
183 were from Latin America, and 488 were from other areas.
During registration at the site, participants self-reported their po-
litical identity. The response options were the 7-point liberal–
conservative scale from Study 1 plus options for libertarian, other,
and don’t know/not political. In the sample 4,679 participants
self-identified as liberal, 847 as moderate/neutral, 1,093 as con-
servative, 1,034 as libertarian, 304 as “other,” and 233 chose
“don’t know/not political.” Three participants did not answer the
question. We focus our analyses on the 6,619 participants who
chose an option on the 7-point liberal–conservative scale.
Materials and Procedure
Participants first self-selected to complete a one-page survey
labeled “Sacredness Survey: What Would You Do for a Million
Dollars?” from a list of 15–20 surveys. They then read the follow-
Try to imagine actually doing the following things, and indicate how
much money someone would have to pay you (anonymously and
secretly) to be willing to do each thing. For each action, assume that
nothing bad would happen to you afterwards. Also assume that you
cannot use the money to make up for your action. If you prefer to
think about Euros or any other currency, please do. The exact amounts
are not very important.
Response options given after each action were $0 (I’d do it for
free), $10, $100, $1,000, $10,000, $100,000, a million dollars, and
never for any amount of money. Below these instructions partici-
pants found a list of 26 actions (listed in Appendix C), presented
in an order randomized for each participant. Cronbach’s alphas for
each foundation were .69 (Harm), .69 (Fairness), .69 (Ingroup), .67
(Authority), and .58 (Purity).
Below we report analyses on the full 8-point scale as a continuous
index of unwillingness to perform the actions. All findings also held
using a stricter binary criterion of sacredness as the refusal to do the
action for any amount of money (vs. selecting any dollar value).
Foundation scores (the average response on the five items for
each foundation) are plotted across political identity in Figure 4.
As this figure shows, Fairness violations were considered the most
taboo overall, with people across the political spectrum choosing
responses whose average was closest to 7 (a million dollars).
Liberals required slightly more money on average to violate the
Harm foundation. However, conservatives required substantially
higher amounts to violate the three binding foundations.
Because the taboo trade-off measure gauged the “sacredness” of
each foundation, we were concerned that political effects may have
been driven by the differential religious attendance of liberals and
conservatives. We created a latent variable model (parameters esti-
mated ? 123), ?2(404) ? 25,148.49, εa? .09, simultaneously pre-
dicting the five foundation sacredness scores with politics and four
covariates: age, gender, education, and religious attendance.3With
covariates included, political identity still predicted sacredness scores
in the predicted direction for Ingroup (? ? .42), Authority (? ? .31),
and Purity (? ? .11), all ps ? .001. However, in this model politics
weakly predicted sacredness scores for Harm (? ? .06) and Fairness
(? ? .07) in directions opposite to predictions (ps ? .001). As
predicted, the aggregated moral sacredness ratings for individualizing
foundations were higher than the aggregated ratings for binding
foundations, F(1, 6596) ? 3,689.66, p ? .001, ?p
effect was moderated by politics, F(1, 6596) ? 236.28, p ? .001, ?p
? .18; the more liberal participants showed a greater difference
between the individualizing and binding moral foundations for their
overall degree of unwillingness to violate the foundations.
Because we had a large sample of libertarians, who are usually
ignored in political–psychological research, we compared their
sacredness reactions to those of liberals and conservatives. Over-
all, libertarians showed less refusal to violate the foundations for
money than did liberals or conservatives. Each of the five average
never scores for libertarians was lower than the corresponding
score for conservatives (all ts ? 9.5, ps ? .001, ds ? 0.43), and
each was lower than the corresponding score for liberals (all ts ?
3.0, ps ? .01, ds ? 0.09).
2? .36, and this
The results supported the moral foundations hypothesis: Liber-
als refused to make trade-offs on most of the individualizing items
but were more willing to perform actions that violated the three
binding foundations. Conservatives, in contrast, showed a more
even distribution of concerns and reported more unwillingness
than did liberals to accept money to act in ways that violate
Ingroup, Authority, and Purity concerns.
The results also challenged our previous finding that liberals
care more than conservatives about Harm and Fairness issues. Do
these results show that we were premature in concluding, from
Studies 1 and 2, that liberals care more about Harm and Fairness
issues, on average, than do conservatives? We do not think so.
Rather, we think there is a general across-the-board political dif-
ference on the permissibility of making moral trade-offs. It is no
coincidence that John Stuart Mill (1859/2003) is a founder of both
liberalism and utilitarianism. Liberals generally justify moral rules
in terms of their consequences for individuals; they are quite
accustomed to balancing competing interests and to fine-tuning
social institutions to maximize their social utility. Conservatives,
in contrast, are more likely to respect rules handed down from God
(for religious conservatives) or from earlier generations (see Mul-
ler, 1997). Conservatives are more often drawn to deontological
moral systems in which one should not break moral rules even
when the consequences would, overall, be positive (Graham,
Nosek, Haidt, Hawkins, & Iyer, 2008). This deontological reluc-
tance to make trade-offs was not triggered by the methods used in
Studies 1 and 2, but, we suspect, it elevated conservatives’ scores
on all foundations in the present study.
A further novel finding of the present study was that libertarians
had the lowest sacredness scores on all five foundations. This
finding supports Tetlock’s predictions that free-market libertarians
would be the least outraged and most open to contractualizing
moral violations (Tetlock et al., 2000; Tetlock, Peterson, & Lerner,
1996). The differences were particularly stark between libertarians
and conservatives on the three binding foundations. Libertarians
3Household income, included in all latent variable models for Studies 1
and 2, was not available in this dataset. Religious attendance was not
available in Studies 1 and 2.
THE MORAL FOUNDATIONS OF POLITICS
may support the Republican Party for economic reasons, but in
their moral foundations profile we found that they more closely
resemble liberals than conservatives.
From Studies 1–3 we steadily increased the degree to which
participants had gut feelings—not just conscious self-theories—
upon which to base their responses. But all of these methods relied
on responses to hypothetical questions, free from social context
and real-world costs. To test our hypothesis in a more ecologically
valid setting, in Study 4 we analyzed behavior in public settings:
sermons delivered in liberal and conservative churches.
Study 4: Moral Texts
Words do the work of politics. As Lakoff (2004) has shown,
liberals and conservatives use different words to create overarch-
ing “frames” that make policies seem morally good or bad. Writing
from the other end of the political spectrum, Luntz (2007) argued
that the Republican Party’s success in the 1990s was due in large
part to its ability to find “words that work.”
We sought out speeches delivered to live audiences that we
could analyze to test the moral foundations hypothesis. We first
examined Republican and Democratic candidates’ convention
speeches, but we discovered that those speeches were so full of
policy proposals, and of moral appeals to the political center of
the country, that extracting distinctive moral content was un-
feasible using the simple word-count procedures we describe
below. We turned instead to sermons delivered in liberal and
conservative churches. Sermons typically contain parables or
direct instruction on the morally right way to live (Spilka,
Hood, & Gorsuch, 1985). They are delivered to and within the
moral community of the congregation, and they are generally
written by the speaker, not by speech writers. Sermons thus
have an advantage over overtly political texts in that they are
more likely to address the moral concerns of a specific and
cohesive moral community, rather than the concerns of a broad
and heterogeneous polity. We identified conservative and lib-
eral Christian denominations, obtained transcribed sermons
Self-reported political identity
Average amount required to violate taboos
Study 3. Response scale was as follows: 1 ? $0 (I’d do it for free), 2 ? $10, 3 ? $100, 4 ? $1,000, 5 ? $10,000,
6 ? $100,000, 7 ? a million dollars, 8 ? never for any amount of money.
Average amount required to violate foundation-related taboo trade-offs across political identity,
GRAHAM, HAIDT, AND NOSEK
from those denominations, and then analyzed their word use
related to each of the five moral foundations.
Quantitative content analysis of texts is considered the most
objective approach to linguistic data (Silverman, 1993); it in-
volves creating categories of words and obtaining objective
counts of instances of those categories within each text. In this
study we used the Linguistic Inquiry and Word Count program
(LIWC; see Pennebaker, Francis, & Booth, 2003). One disad-
vantage of LIWC is that it does not identify individual word
frequencies. We created a supplemental program (available at
www.moralfoundations.org) to identify the words that were
driving the liberal–conservative differences. Then, independent
raters read these words in context and rated the degree to which
the speaker was endorsing or rejecting the relevant moral foun-
We found multiple categorization schemes to identify liberal and
conservative Christian denominations (Marnell, 1974; Shortridge,
1976; Spilka et al., 1985; Woodberry & Smith, 1998). Unitarian
Universalist was consistently regarded as the most liberal church, and
Southern Baptist was regarded as the most conservative of the major
Christian denominations; these two denominations were also found to
be the most politically active at the pulpit (Guth, 1996). We then used
the search engine Google to locate databases of sermons from Uni-
tarian and Southern Baptist churches. We chose only databases or
church websites that contained sermons in text form, rather than as
audio files. Sermons were delivered between 1994 and 2006, and a
majority (?80%) were delivered between 2003 and 2006. A total of
69 liberal (Unitarian) sermons and 34 conservative (Southern Baptist)
sermons were analyzed.
We created a LIWC dictionary, beginning with the core con-
cepts for each of the five foundations: harm and care, fairness and
reciprocity, ingroup and loyalty, authority and respect, and purity
and sanctity. Dictionary development had an expansive phase and
a contractive phase, all occurring before reading the sermons. In
the expansive phase Jesse Graham and five research assistants
generated as many associations, synonyms, and antonyms for the
base foundation words as possible, using thesauruses and conver-
sations with colleagues. This included full words and word stems
(for instance, nation?covers national, nationalistic, etc.). The
resulting lists included foundation-supporting words (e.g., kind-
ness, equality, patriot, obey, wholesome), as well as foundation-
violating words (e.g., hurt, prejudice, betray, disrespect, disgust-
ing). In the contractive phase, Jesse Graham and Jonathan Haidt
deleted words that seemed too distantly related to the five foun-
dations and also words whose primary meanings were not moral
(e.g., just more often means only than fair). See Appendix D for
the final dictionary of 295 words and word stems.
After running the basic LIWC analyses with this dictionary, we
read them in context to ensure that the most influential words were
being used in support of the moral foundation. For example, “Don’t
let some self-interested ecclesiastical or government authority tell you
what to believe, but read the Bible with your own eyes and open your
heart directly to Jesus” was a sentence from a Unitarian sermon. It
contributed to the Authority raw word count (by including the word
authority), but the context reveals that the speaker is in fact question-
ing or rejecting the Authority foundation.
With our supplemental program, we counted word frequencies
for all words in our LIWC dictionary and computed the liberal–
conservative difference in percent usage of each word. We selected
the 23 words and word stems4that yielded more than a 0.02%
difference. For each of the 3,281 uses of these words, we copied
the surrounding context of 2–3 sentences into a spreadsheet and
then scrambled its rows.
Four raters scored each passage, blind to its origins, by assigning it
a score of 1 if the passage supported or was consistent with the values
or concerns of the relevant foundation, a score of ?1 if the passage
seemed to negate or reject the foundation in question, or a score of 0
if the contextual usage was unclear or irrelevant to the foundation.
a reliability of ? ? .79. Ratings from multiple raters were averaged to
create a “contextually validated usage” score for each word. These
usage scores are more valid indicators than the raw counts of how
speakers value each of the five foundations.
The raw percentages of words related to each foundation are
given in Table 1. As the first line shows, 0.44% of all the 177,629
words in our liberal sermon corpus were among the 51 words in
the Harm category of our LIWC dictionary, whereas only 0.26% of
all the 136,706 words in our conservative sermon corpus fell into
this category. These raw percentages show the predicted effects for
four of the five foundations: liberals used Harm and Fairness
words more frequently than did conservatives, whereas conserva-
tives used Authority and Purity words more frequently than did
liberals. The only violation of our prediction occurred for words in
the Ingroup category, which were used more frequently by liberal
than conservative speakers.
As Table 2 shows, however, the contextually validated ratings
support the moral foundations hypothesis for all five foundations.
Reading the difference-driving words in context, we found that
liberal speakers expressed concerns more in line with Harm and
Fairness than did conservative speakers, and conservative speakers
expressed concerns more in line with Ingroup, Authority, and
Purity than did liberal speakers.
Notably, the effect size on the Ingroup foundation reverses sign
and conforms to our prediction. This reversal occurred because the
words communit?, group, individual?, and nation?were used more
frequently in liberal sermons, but analysis of these words in
context revealed that liberals were much more likely than conser-
vatives to use these words in order to reject the foundational
concerns of ingroup loyalty and group solidarity. Usage ratings for
liberal uses of group were close to zero, and usage ratings for
individual?were negative, indicating many uses rejecting ingroup
values or asserting opposing values like independence and indi-
vidual autonomy. Of the 23 words rated in context, 22 showed
liberal–conservative differences in the predicted direction, and 18
of these differences were significant (p ? .05).
4violen?and war (for Harm); justice and justifi?(for Fairness); commu-
nit?, group, individual?, and nation?(for Ingroup); authorit?, command,
father, law, leader?, mother, obedien?, obey, rebel?, submi?, and tradition?
(for Authority); holy, sin, sinner?, and sins (for Purity).
THE MORAL FOUNDATIONS OF POLITICS
If words do the work of politics, then liberal and conservative
religious leaders are using different words to do different work.
Unitarians, known for their political liberalism, talk more about
issues of Harm and Fairness than do Baptist preachers. Baptists, in
contrast, talk more about Authority and Purity than do Unitarians.
The one finding that seemed to contradict our moral foundations
hypothesis—frequency of Ingroup words—turned out, on closer
inspection, to be consistent with the findings of Studies 1–3.
Unitarians were more likely to use Ingroup-related words, but
when Baptists used these words, they used them in ways more
consistent with the values of the foundation. Unitarians were
actually more likely to use some binding-foundation words (indi-
vidual?, authorit?, rebel?, and submi?) in ways contrary to the
foundational concerns, such as praising rebellion or individuality,
than in ways supporting the foundations.
Our reading of the relevant words in context also revealed that
liberal sermons were more overtly political, spending a great deal of
time talking about the negative direction President George W. Bush
was taking the nation. For instance, most of the 78 instances in which
Unitarians said “war” referred to the current war in Iraq, whereas the
18 Baptist uses of “war” never once referred to the Iraq war, instead
tending to describe a theological war between heaven and hell.
By supplementing the LIWC analyses with word-by-word fre-
quency analyses and reading the most central words in context, we
were able to better understand the patterns of word use and improve
the validity of the moral foundations dictionary. This dictionary will
be useful for further analyses of the words used by different moral
communities and moral cultures in a variety of textual media (e.g.,
stories, speeches, letters, e-mails). Most importantly, Study 4 repli-
cated the moral foundation patterns using existing moral texts deliv-
ered within the moral communities of Unitarian Universalists and
Southern Baptists. Overall, the text results lend support to the moral
foundations hypothesis and suggest that these moral differences exist
in the real world and are not artifacts of the questionnaire-based
methods we used in Studies 1–3.
Four studies found support for our moral foundations hypothesis
using four different methods: moral relevance assessments, moral
judgments, unwillingness to violate the foundations for money,
and word use in religious sermons. Across all four studies, liberal
morality was primarily concerned with harm and fairness, whereas
conservative moral concerns were distributed more evenly across
all five foundations. These findings help explain why liberals and
conservatives disagree on so many moral issues and often find it
hard to understand how an ethical person could hold the beliefs of
the other side: Liberals and conservatives base their moral values,
judgments, and arguments on different configurations of the five
The psychology of Harm/care and Fairness/reciprocity has been
studied by moral psychologists for decades in work on moral
reasoning (Gilligan, 1982; Kohlberg, 1969; Turiel, 1983), empathy
(Hoffman, 1982), and equity theory (Walster, Berscheid, & Wal-
ster, 1976). Research related to the three binding foundations,
however, has mostly examined them as sources of immorality. The
psychology of ingroups has been closely related to the psychology
of racism (Allport, 1954; Brewer, 2007); the psychology of au-
thority has been studied as the psychology of fascism and blind
obedience (Altemeyer, 1996; Milgram, 1974); and the psychology
of purity and disgust has been related to the psychology of stigma
(Crocker & Major, 1989). The binding foundations can certainly
motivate horrific behavior, but a common theme in recent morality
research is that moral–psychological mechanisms are often two-
edged swords (de Waal, 1996). Reciprocity, for example, underlies
both justice and blood feuds (Frijda, 1994). Religion brings out the
best and the worst in people (Wilson, 2002). A complete moral
psychology should include within its purview the major topics,
virtues, and phenomena that either liberals or conservatives believe
are part of their morality.
Integration With Other Theories
The five moral foundations provide a taxonomy for the bases
of moral judgments, intuitions, and concerns. Taxonomies are
the building blocks of theory, organizing metaphors that pro-
vide a vehicle for theories to exert explanatory power over
human behavior. Many other taxonomies can be offered to
organize the same content, and no best one can be chosen
without specifying the purpose for which it will be used. If
one’s goal is to describe moral discourse across cultures, then
Raw Percentages of Foundation-Related Words in Liberal and
Conservative Sermons, Study 4
nary for that foundation was used, divided by the total number of words in
the entire liberal or conservative corpus, then multiplied by 100. A negative
effect size indicates the effect was opposite to prediction.
??p ? .01.
Percentages represent the number of times any word in the dictio-
???p ? .001.
Contextual Validity Ratings of Foundation-Related Words in
Liberal and Conservative Sermons, Study 4
1 indicates support of foundation values, –1 indicates rejection of founda-
tion values, and 0 indicates unclear or irrelevant to foundation values.
??p ? .01.
Usages are averaged ratings of uses of 23 words in context, where
???p ? .001.
GRAHAM, HAIDT, AND NOSEK
Shweder et al.’s (1997) three ethics work well and are fully
compatible with the five foundations: the “ethic of autonomy”
is generated by the Harm and Fairness foundations; the “ethic of
community” draws on the Ingroup and Authority foundations;
and the “ethic of divinity” is built on the Purity foundation. If
one’s goal is to describe patterns of social relationships, then
Fiske’s (1992) four models work well, and two of them (equal-
ity matching and authority ranking) are direct matches to two of
the five foundations (Fairness and Authority). If one’s goal is to
classify all values and look at their relationships, then
Schwartz’s (1992) list of 10 value-types works well, and one
quadrant of his multidimensional plot of values includes the
main values related to the Harm and Fairness foundations (e.g.,
social justice, protect the environment), whereas an adjacent
quadrant is a good catalogue of the values related to Ingroup,
Authority, and Purity (e.g., national security, obedient, clean).
We drew on all three of these theories in creating moral foun-
dations theory, which we propose as the best taxonomy for
research on moral judgment, particularly for psychologists who
are interested in moral intuition (Haidt, 2001; Sunstein, 2005)
or who want to explore moral issues beyond harm and fairness
(Haidt & Graham, 2007). The theory is also useful for organiz-
ing and explaining political differences in people’s endorse-
ments of long lists of various values (Feather, 1979). We do not
believe that moral foundations theory offers an exhaustive
taxonomy; there are surely many other psychological systems
that contribute to moral judgment. Fiske’s “market pricing”
relational model, or Schwartz’s value-types of achievement and
hedonism, for example, may point to additional psychological
mechanisms that support some moral systems (such as the
libertarianism of Ayn Rand, 1957). Nonetheless, as a first pass
to identify the most important sources of moral intuition across
cultures, we believe that the five foundations we have identified
are the best starting point and offer the best balance between
explanatory power and parsimony.
Factor Structure of the Moral Foundations
It is not our intention in this article to develop or validate a scale
related to the moral foundations (for such work see Graham, Haidt, et
al., 2008). However, because we are presenting and testing a new
theory of morality that asserts a specific factor structure, we used the
measurement items and large sample sizes of Studies 1–3 to create
structural equation models of different confirmatory factor analyses.
These models are provided as a supplement at www.moralfounda-
tions.org. The supplement shows that models with five correlated
factors (Harm, Fairness, Ingroup, Authority, and Purity) were a sig-
nificant improvement (weighing both fit and parsimony) over three
other models: (a) a single-factor Morality model, (b) a two-factor
model distinguishing individualizing (Harm–Fairness) and binding
(Ingroup–Authority–Purity) factors, and (c) a three-factor model cor-
responding to Shweder et al.’s (1997) three ethics of autonomy
(Harm–Fairness), community (Ingroup–Authority), and divinity (Pu-
rity). Model comparisons were calculated using the FITMOD pro-
gram (Browne, 1991; see also Browne & Cudeck, 1993). All datasets
supported the prediction that a five-factor model would be an im-
provement over the other models. This was evident for relevance,
judgment, and taboo trade-off items. We take this as preliminary
evidence for a five-factor structure of moral concerns and as support
for the theory guiding Studies 1–4.
Limitations and Future Directions
Research on political attitudes benefits from diverse sampling
instead of the “typical” focus on undergraduate samples (Sears,
1986). Studies 1, 2, and 3 drew on large samples heterogeneous in
age, education, income, and occupation within the United States
and beyond, and Study 4 examined sermons written by ministers in
Unitarian and Southern Baptist churches. Nevertheless, these sam-
ples are not representative of the national and international popu-
lations from which they are drawn (Nosek et al., 2007). Studies
1–3 were subject to self-selection, as participants opted in to visit
the websites in the first place, and Study 3 had an additional
selection bias in that participants chose to take that particular
survey. Assuming that selection influences are similar within the
sample, we can confidently interpret internal comparisons but
cannot infer that the specific slopes and means will serve as
accurate parameter estimates of the U.S. or worldwide population.
Such inferences require representative sampling. An initial report
from one nationally representative dataset (Smith & Vaisey, 2008)
in the United States suggests that the moral foundations patterns
reported here replicated for both the moral relevance items (Stud-
ies 1 and 2) and for the moral judgments items (Study 2).
Our conclusions about liberal and conservative moralities depend
not just on the samples we studied but on the particular items we
picked for measurement. The consistency across four measures (rel-
evance, judgments, taboo trade-offs, and text analysis) provides some
confidence for the robustness of these conclusions. Nonetheless, more
validation work is needed to ensure that all the items are discriminat-
ing and gauging the foundations as intended. A broader investigation
of these and other potential foundations is also needed to ensure that
liberal and conservative differences are characterized correctly. There
may well be kinds of fairness that conservatives care more about than
do liberals, or kinds of purity that liberals care more about than do
conservatives. Future investigations will expand the range of mea-
sures and behavioral observations used to assess the moral founda-
tions and their factor structure.
Three of our four studies relied upon self-reported ratings, and
the fourth study relied upon sermons that draw upon the conscious
reflections of the speaker. Although we measured political identity
implicitly as well as explicitly, we did not investigate whether
differences in moral foundations exist implicitly as well as explic-
itly. Haidt’s (2001) social–intuitionist model, along with recent
studies of moral issues (Greene, Morelli, Lowenberg, Nystrom, &
Cohen, 2008; Inbar, Pizarro, Knobe, & Bloom, in press), indicate
that moral judgment relies heavily on automatic processes. Studies
using implicit measurement methods will be essential for under-
standing the ways in which liberals and conservatives make moral
An additional future direction is to move beyond the unidimen-
sional political spectrum we employed in Studies 1–4. The anal-
ysis of libertarians in Study 3 showed that patterns of moral
foundation endorsement may be more complex than the single
liberal–conservative continuum can adequately describe. There
are many types of liberals, many types of conservatives, and many
people who refuse to place themselves along that spectrum. As-
sessing people in terms of their scores on all five foundations may
THE MORAL FOUNDATIONS OF POLITICS
offer insights into why people embrace or reject political labels
and alliances in the complex ways that they do. (See Haidt et al.,
in press, for such an application of moral foundations theory.)
Likewise, though we found consistent effects across U.S., U.K.,
and a nondifferentiated “other nations” group in Study 1, it is
likely that the link between politics and moral foundations will
vary to some degree across cultural contexts. Identifying the so-
cial, economic, ecological, and historical factors that create such
variations will enrich our understandings of morality, politics, and
the connections between them. Finally, experimental work is
needed to better understand the causal nature of the relationships
we have found. Do people first identify with the political left or
right and then take on the necessary moral concerns, or do the
moral concerns come first, or is there reciprocal influence or even
an unidentified third variable at the root of both?
Western societies are growing more diverse, and with diversity
comes differing ideals about how best to regulate selfishness and
about how we ought to live together. Participants in political
debates are motivated in part by moral convictions. Moral foun-
dations theory offers a useful way to conceptualize and measure
such convictions. As research on political psychology thrives (Jost,
2006), we hope that it will clarify the role that morality plays in
political thought and behavior.
Allport, G. W. (1954). The nature of prejudice. Reading, MA: Addison-
Altemeyer, R. A. (1996). The authoritarian specter. Cambridge, MA:
Harvard University Press.
Ballew, C. C., & Todorov, A. (2007). Predicting political elections from
rapid and unreflective face judgments. Proceedings of the National
Academy of Sciences, USA, 104, 17948–17953.
Boaz, D. (1997). Libertarianism: A primer. New York: Free Press.
Boehm, C. (1999). Hierarchy in the forest: The evolution of egalitarian
behavior. Cambridge, MA: Harvard University Press.
Bouchard, T. J. J. (2004). Genetic influence on human psychological traits:
A survey. Current Directions in Psychological Science, 13, 148–151.
Bowlby, J. (1969). Attachment and loss: Vol. 1. Attachment. New York:
Brewer, M. B. (2007). The social psychology of intergroup relations:
Social categorization, ingroup bias, and outgroup prejudice. In A. W.
Kruglanski & E. T. Higgins (Eds.), Social psychology: Handbook of
basic principles (2nd ed., pp. 695–715). New York: Guilford Press.
Brown, D. E. (1991). Human universals. Philadelphia: Temple University
Browne, M. W. (1991). FITMOD: A computer program for calculating
point and interval estimates of fit measures. Unpublished manuscript.
Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model
fit. In K. A. Bollen & J. S. Long (Eds.), Testing structural equation
models (pp. 136–162). Newbury Park, CA: Sage.
Crocker, J., & Major, B. (1989). Social stigma and self-esteem: The
self-protective properties of stigma. Psychological Review, 96, 608–
de Waal, F. B. M. (1982). Chimpanzee politics. New York: Harper & Row.
de Waal, F. B. M. (1996). Good natured: The origins of right and wrong
in humans and other animals. Cambridge, MA: Harvard University
de Waal, F. B. M. (2008). Putting the altruism back into altruism: The
evolution of empathy. Annual Review of Psychology, 59, 279–300.
Emler, N., Renwick, S., & Malone, B. (1983). The relationship between
moral reasoning and political orientation. Journal of Personality and
Social Psychology, 45, 1073–1080.
Feather, N. T. (1979). Value correlates of conservatism. Journal of Per-
sonality and Social Psychology, 37, 1617–1630.
Feldman, S. (2003). Values, ideology, and the structure of political atti-
tudes. In D. O. Sears, L. Huddy, & R. Jervis (Eds.), Oxford handbook of
political psychology (pp. 477–508). Oxford, United Kingdom: Oxford
Fiske, A. P. (1992). Four elementary forms of sociality: Framework for a
unified theory of social relations. Psychological Review, 99, 689–723.
Frijda, N. H. (1994). The lex talionis: On vengeance. In S. H. M. van
Goozen, N. E. van der Poll, & J. A. Sargeant (Eds.), Emotions: Essays
on emotion theory (pp. 263–289). Hillsdale, NJ: Erlbaum.
Gilligan, C. (1982). In a different voice: Psychological theory and women’s
development. Cambridge, MA: Harvard University Press.
Graham, J., Haidt, J., Nosek, B. A., Iyer, R., Koleva, S., & Ditto, P. H.
(2008). Broadening and mapping the moral domain: Development and
validation of the Moral Foundations Questionnaire. Manuscript in prep-
aration, University of Virginia.
Graham, J., Nosek, B. A., Haidt, J., Hawkins, C. B., & Iyer, R. (2008). The
persistence of the gut: Deontological carryover and political ideology.
Unpublished manuscript, University of Virginia.
Greene, J., Morelli, S. A., Lowenberg, K., Nystrom, L. E., & Cohen, J. D.
(2008). Cognitive load selectively interferes with utilitarian moral judg-
ment. Cognition, 107, 1144–1154.
Greenwald, A. G., Nosek, B. A., & Banaji, M. R. (2003). Understanding
and using the Implicit Association Test: I. An improved scoring algo-
rithm. Journal of Personality and Social Psychology, 85, 197–216.
Guth, J. L. (1996). The bully pulpit: Southern Baptist clergy and political
activism 1980–1992. In J. C. Green, J. L. Guth, C. E. Smidt, & L. A.
Kellstedt (Eds.), Religion and the culture wars: Dispatches from the
front. Lanham, MD: Rowan & Littlefield.
Gutmann, A. (2001). Liberalism. In N. J. Smelser & P. B. Baltes (Eds.),
International Encyclopedia of the Social and Behavioral Sciences. New
Haidt, J. (2001). The emotional dog and its rational tail: A social intuition-
ist approach to moral judgment. Psychological Review, 108, 814–834.
Haidt, J. (2008). Morality. Perspectives on Psychological Science, 3,
Haidt, J., & Graham, J. (2007). When morality opposes justice: Conser-
vatives have moral intuitions that liberals may not recognize. Social
Justice Research, 20, 98–116.
Haidt, J., Graham, J., & Joseph, C. (in press). Above and below left–right:
Ideological narratives and moral foundations. Psychological Inquiry.
Haidt, J., & Joseph, C. (2004). Intuitive ethics: How innately prepared
intuitions generate culturally variable virtues. Daedalus: Special Issue
on Human Nature, 133(4), 55–66.
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 University
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,
Hoffman, M. L. (1982). Development of prosocial motivation: Empathy
and guilt. In N. Eisenberg (Ed.), The development of prosocial behavior
(pp. 218–231). New York: Academic Press.
Hunter, J. D. (1991). Culture wars: The struggle to define America. New
York: Basic Books.
GRAHAM, HAIDT, AND NOSEK
Inbar, Y., Pizarro, D. A., Knobe, J., & Bloom, P. (in press). Disgust Download full-text
sensitivity predicts intuitive disapproval of gays. Emotion.
Jost, J. T. (2006). The end of the end of ideology. American Psychologist,
Jost, J. T., Glaser, J., Sulloway, F., & Kruglanski, A. W. (2003). Political
conservatism as motivated social cognition. Psychological Bulletin, 129,
Jost, J. T., Nosek, B. A., & Gosling, S. D. (2008). Ideology: Its resurgence
in social, personality, and political psychology. Perspectives on Psycho-
logical Science, 3, 126–136.
Keller, M. C., Coventry, W. L., Heath, A. C., & Martin, N. G. (2005).
Widespread evidence for non-additive genetic variation in Cloninger’s
and Eysenck’s personality dimensions using a twin plus sibling design.
Behavior Genetics, 35, 707–721.
Kinder, D. E. (1998). Opinion and action in the realm of politics. In D.
Gilbert, S. Fiske, & G. Lindzey (Eds.), Handbook of social psychology
(Vol. 2, 4th ed., pp. 778–867). Boston: McGraw-Hill.
Kohlberg, L. (1969). Stage and sequence: The cognitive-developmental
approach to socialization. In D. A. Goslin (Ed.), Handbook of social-
ization theory and research (pp. 347–480). Chicago: Rand McNally.
Kurzban, R., Tooby, J., & Cosmides, L. (2001). Can race be erased?
Coalitional computation and social categorization. Proceedings of the
National Academy of Sciences, USA, 98, 15387–15392.
Lakoff, G. (2004). Don’t think of an elephant! Know your values and frame
the debate: The essential guide for progressives. White River Junction,
VT: Chelsea Green.
Luntz, F. (2007). Words that work. New York: Hyperion.
Marcus, G. E. (2002). The sentimental citizen: Emotion in democratic
politics. State College: Penn State Press.
Marcus, G. (2004). The birth of the mind. New York: Basic Books.
Marnell, G. M. (1974). Response to religion. Lawrence: University of
McCann, S. J. H. (2008). Societal threat, authoritarianism, conservatism,
and U.S. state death penalty sentencing (1977–2004). Journal of Per-
sonality and Social Psychology, 94, 913–923.
McCourt, K., Bouchard, J. B. J., Lykken, D. T., Tellegen, A., & Keyes, M.
(1999). Authoritarianism revisited: Genetic and environmental influ-
ences examined in twins reared apart and together. Personality and
Individual Differences, 27, 985–1014.
McCrae, R. R. (1996). Social consequences of experiential openness.
Psychological Bulletin, 120, 323–337.
Milgram, S. (1974). Obedience to authority. New York: Harper and Row.
Mill, J. S. (2003). On liberty. New Haven, CT: Yale University Press.
(Original work published 1859)
Miller, D. T. (1999). The norm of self-interest. American Psychologist, 54,
Muller, J. Z. (1997). What is conservative social and political thought? In
J. Z. Muller (Ed.), Conservatism: An anthology of social and political
thought from David Hume to the present (pp. 3–31). Princeton, NJ:
Princeton University Press.
Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know:
Verbal reports on mental processes. Psychological Review, 84, 231–259.
Nosek, B. A. (2005). Moderators of the relationship between implicit and
explicit evaluation. Journal of Experimental Psychology: General, 134,
Nosek, B. A., Greenwald, A. G., & Banaji, M. R. (2006). The Implicit
Association Test at age 7: A methodological and conceptual review. In
J. A. Bargh (Ed.), Social psychology and the unconscious: The automa-
ticity of higher mental processes (pp. 265–292). New York: Psychology
Nosek, B. A., Smyth, F. L., Hansen, J. J., Devos, T., Lindner, N. M.,
Ranganath, K. A., et al. (2007). Pervasiveness and correlates of implicit
attitudes and stereotypes. European Review of Social Psychology, 18,
Pennebaker, J. W., Francis, M. E., & Booth, R. J. (2003). Linguistic inquiry
and word count: LIWC2001 manual. Mahwah, NJ: Erlbaum.
Rand, A. (1957). Atlas shrugged. New York: Random House.
Richerson, P. J., & Boyd, R. (2005). Not by genes alone: How culture
transformed human evolution. Chicago: University of Chicago Press.
Rokeach, M. (1973). The nature of human values. New York: Free Press.
Rozin, P., Haidt, J., & McCauley, C. R. (2000). Disgust. In M. Lewis &
J. M. Haviland-Jones (Eds.), Handbook of emotions (2nd ed., 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., & Bilsky, W. (1990). Toward a theory of the universal
content and structure of values: Extensions and cross-cultural replica-
tions. Journal of Personality and Social Psychology, 58, 878–891.
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.
Sears, D. O., & Funk, C. L. (1991). The role of self-interest in social and
political attitudes. In M. P. Zanna (Ed.), Advances in experimental social
psychology (Vol. 24, pp. 2–91). New York: Academic Press.
Shortridge, J. R. (1976). Patterns of religion in the United States. Geo-
graphical Review, 66, 420–434.
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.
Silverman, D. (1993). Interpreting qualitative data: Methods for analyzing
talk, text and interaction. London: Sage.
Smith, C., & Vaisey, S. (2008). Charitable giving and moral foundations
in a nationally-representative sample. Manuscript in preparation, Uni-
versity of North Carolina.
Soler, J. (1979). The semiotics of food in the Bible. In R. Forster & O.
Ranum (Eds.), Food and drink in history (E. Forster & P. M. Ranum,
Trans., pp. 126–138). Baltimore: Johns Hopkins University Press.
(Original work published 1973)
Sowell, T. (2002). A conflict of visions: The ideological origins of political
struggles. New York: Basic Books.
Spilka, B., Hood, R. W., Jr., & Gorsuch, R. L. (1985). The psychology of
religion: An empirical approach. Englewood Cliffs, NJ: Prentice-Hall.
Stenner, K. (2005). The authoritarian dynamic. New York: Cambridge
Sunstein, C. R. (2005). Moral heuristics. Brain and Behavioral Science, 28,
Tetlock, P. E. (2003). Thinking about the unthinkable: Coping with secular
encroachments on sacred values. Trends in Cognitive Science, 7, 320–
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 Psy-
chology, 78, 853–870.
Tetlock, P. E., Peterson, R., & Lerner, J. (1996). Revising the value
pluralism model: Incorporating social content and context postulates.
In C. Seligman, J. Olson, & M. Zanna (Eds.), Ontario symposium on
social and personality psychology: Values (pp. 25–51). Hillsdale, NJ:
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, United Kingdom: Cambridge University
THE MORAL FOUNDATIONS OF POLITICS