What is the main difference between liberals and conservatives moral foundations?

Moral Foundations Theory is pluralistic in two ways. First, “foundations are not [just] values” [Graham et al., 2009, p.1031]. Instead, foundations are psychologically pluralistic in that each foundation is made up of attitudes, cognitions, values, emotions, physiological reactions, and linguistic styles [Graham et al., 2013]. The theory is also pluralistic in the sense that there is not just one moral foundation—there are five or possibly more. To illustrate the point, Haidt and Joseph [2012] suggested that moral foundations are analogous to distinct taste receptors [for salt, sourness, sweetness, etc.] insofar as they are irreducibly pluralistic innate modules. Each moral foundation is an “evolved psychological mechanism” that become elaborated and revised through cultural leaning [Graham et al., 2009, p.1031].

According to the theory, foundational differences materialize not only between countries but also within countries, such as in Culture Wars between liberals and conservatives. This is because liberals tend to have an optimistic view of human nature [Sowell, 2002], which posits that only freedom and liberty are needed to bring out the best in people. Liberals thus rely on the “individualizing” foundations of harm/care and fairness/cheating as a basis of their morality. Conservatives’ more pessimistic view of human nature [Sowell, 2002] prescribes the need for social constraints, institutions, social structure, and order to rein in inherently selfish tendencies and thus make civil society possible; in this way, the “binding” foundations of loyalty, authority, and purity also play a prime role in conservative morality [see Graham et al., 2009; p.1029]. During the developmental period, exposure to practices, lessons, and stories elaborate the various psychological components of different foundations within the moral minds of young liberals and conservatives, rendering them markedly different by adulthood. This might explain why liberals and conservatives might communicate using different moral languages.

What is the evidence that adult liberals and conservatives default to different moral languages? In the seminal empirical article on the topic and precedent for the current investigation, Graham et al. [2009] reported four studies, the first three of which relied on self-report methods and are thus tangential to the present question concerning language. In the crucial Study 4, they showed that the sermons of liberal and conservative pastors varied in their moral languages. For foundations of harm, fairness, authority, and purity, the effects were supportive of the MFH: liberals used more harm and fairness, and less authority and purity language than conservatives. However, liberals also used loyalty words more frequently than conservatives, a finding not predicted by the MFH. It is this particular study and this set of linguistic analyses that the present paper aims to replicate and extend.

Given the extensive support for the MFH coming from other methods [e.g., self-report], it remains unclear why liberals might use more loyalty words than conservatives, rather than vice versa. Graham et al. [2009] attempted to somewhat diminish the significance of this disconfirming finding. Their argument was based on the premise that word frequencies cannot assess whether a speaker has a positive or negative attitude toward a foundation. To illustrate the critique, consider the hypothetical situation in which a liberal were to say “group loyalties have become problematic in this country.” The linguistic analysis would score this statement as being high [dense] in loyalty language because of the mention of the words group and loyalties, both of which are words in the loyalty dictionary, even though the speaker’s attitude toward loyalty is clearly critical. To accommodate this concern, they also had people read and code the texts for the speaker’s attitudes toward each foundation—whether statements were supportive or critical of a foundation—and claimed that “these usage scores are more valid indicators than the raw counts of how speakers value each of the five foundations” [p.1039; emphasis added]. They then tested the MFH with respect to whether liberals and conservatives differed as predicted in their verbalized valuation of each foundation. Unequivocal support the MFH on all five foundations, including loyalty, resulted.

Note that Graham et al. [2009] did not dismiss the relevance of word counts vis-à-vis the MFH. Their point was subtler: that word counts are less relevant to measuring how much a person values a foundation than are contextual usage assessments. I make the theoretical case here that the topics that people bring up [operationalized as word frequencies] are also relevant to MFT and the MFH.

Beginning with first principles, “foundations are not [just] values or virtues” [Graham et al., 2009, p.1031]. Instead, foundations are “psychological systems” and thus psychologically pluralistic in that each foundation is made up of not just values, but also cognitions, emotions, physiological reactions, and languages [Graham et al., 2013]. Following from this principle of psychological pluralism, both the moral languages that people use [operationalized by word counts] and individuals’ verbalized attitudes toward those topics [operationalized by contextualized usage ratings] should both be included within the large umbrella of construct plurality that MFT proposes.

This position is consistent with a long psychological research tradition that has accepted the basic premise that what a person talks about [in positive or negative terms] reflects their psychological states and personality traits. Freud [1901] believed that one could draw inferences about a person’s hidden intentions through observation of slips of the tongue. The Rorschach [1921] inkblot test and the Thematic Apperception Test [e.g., McClelland, 1979, Winter, 1998] asked people to make sense of ambiguous images; what they spontaneously said [or projected] was thought to be revealing of personalities and psychopathologies. Gottschalk [e.g., Gottschalk, Gleser, Daniels, & Block, 1958] and McAdams [1995] continued the tradition by coding stream-of-consciousness thoughts and autobiographical life narratives, respectively. With the advent and expanding availability of computers, researchers began developing computerized methods of quantifying word usage [e.g., Rosenberg and Tucker, 1978, Stone et al., 1966, Weintraub, 1981, Weintraub, 1989].

A century after Freud’s writings on slips-of-the tongue, Pennebaker, Francis, and Booth [2001] released the first version of Linguistic Inquiry and Word Count [LIWC], which has since become enormously popular among psychologists for using text to understand individuals’ states, situations, and personalities [see Tausczik & Pennebaker, 2010, for a review; and used in Graham et al., 2009]. LIWC is built on the premise that “the words we use in daily life reflect what we are paying attention to, what we are thinking about, what we are trying to avoid, how we are feeling, and how we are organizing and analyzing our worlds” [Tausczik & Pennebaker, 2010; p.30]. Indeed, a number of studies have found that word usage [assessed via word counts and LIWC] is a stable feature of personality [e.g., Pennebaker and Graybeal, 2001, Pennebaker and King, 1999] and correlates with personality traits [e.g., Holtgraves, 2011, Hirsh and Peterson, 2009, Lee et al., 2010].

If LIWC analyses are revealing of individuals’ characteristic thoughts and stable personalities and if the finding that liberals use more loyalty words turns out to be robust [i.e. it replicates], then either [a] the MFH regarding loyalty is not supported with respect to word usage; or [b] future moral foundation theorists and researchers ought to tackle the question of why liberals spend so much time criticizing loyalty when loyalty is not a foundation upon which their morality is based. More generally, determining whether there are differences in the moral languages that liberals and conservatives use informs the broader question of whether political ideology should be conceived of as a feature of personality. Evidence that liberals and conservatives use different moral languages would support the notion that ideology is personological.

To my knowledge, 10 years after the publication of Graham et al. [2009], there have been no close or conceptual replications of the moral foundation word frequency analysis. However, some studies are somewhat relevant. One study conceptually “replicated” 1the linguistic attitudinal analysis with a politically diverse set of religious and political active adults and using human coding [McAdams et al., 2008]. Other studies have assessed word frequencies within specific contexts and produced mixed evidence. For instance, when expressing opinions about stem cell research, liberals used more harm [MFH support] and fewer purity words [MFH support] than conservatives [Clifford & Jerit, 2013], and when talking about abortion or same-sex marriage, liberals used more fairness [MFH support] and fewer purity words [MFH support] than conservatives [Frimer et al., 2016, Sagi and Dehghani, 2014].

Environmental issues have produced mixed results vis-à-vis the MFH. Some studies found that liberals talked about climate change and pollution primarily in terms of harm [Feinberg & Willer, 2013; MFH support], whereas others found that liberals drew from the fairness [MFH support], authority [MFH challenge], and purity foundations [Frimer, Tell, & Haidt, 2015; MFH challenge]. And Frimer et al. [2016] found that liberals used more purity [MFH challenge] and fewer fairness words [MFH challenge] than conservatives when talking about the Keystone XL oil pipeline, implying a complete, context-specific reversal of the MFH. In sum, this research shows that the MFH is not uniformly supported across various issues and topics. Whether or not some enduring dispositional difference between liberals and conservatives exists, as the MFH posits, remains inconclusive from the scientific record. The goal of the present research is to inform that question.

The impetus for the present replication studies is threefold. First, Moral Foundations Theory has been hugely influential. According to Google Scholar, the seminal empirical paper [Graham et al., 2009] has been cited ~2500 times, and the popular book that followed [Haidt, 2012] received ~4100 citations as of October 2019. Second, the original study confirmed the MFH for four foundations and disconfirmed it for the fifth. Replication studies may shed light on whether the MFH is indeed supported when it comes to moral language. Third, there seem to be no close or conceptual published replications of the word count version of the MFH.

The goal of the present studies is to apply the “the replication recipe” [Brandt et al., 2014] to test whether the finding that liberals and conservatives use different moral languages [Table 1 of Graham et al., 2009] replicates and generalizes to other contexts. The original study reported that liberals use more harm, fairness, and loyalty words whereas conservatives use more authority and purity words.