Are good reasoners more incest-friendly? Trait cognitive reflection predicts selective moralization in a sample of American adults

Abstract

Two studies examined the relationship between individual differences in cognitive reflection (CRT) and the tendency to accord genuinely moral (non-conventional) status to a range of counter-normative acts — that is, to treat such acts as wrong regardless of existing social opinion or norms. We contrasted social violations that are intrinsically harmful to others (e.g., fraud, thievery) with those that are not (e.g., wearing pajamas to work and engaging in consensual acts of sexual intimacy with an adult sibling). Our key hypothesis was that more reflective (higher CRT) individuals would tend to moralize selectively — treating only intrinsically harmful acts as genuinely morally wrong — whereas less reflective (lower CRT) individuals would moralize more indiscriminately. We found clear support for this hypothesis in a large and ideologically diverse sample of American adults. The predicted associations were not fully accounted for by the subjects’ political orientation, sensitivity to gut feelings, gender, age, educational attainment, or their placement on a sexual morals-specific measure of social conservatism. Our studies are the first to demonstrate that, in addition to modulating the intensity of moral condemnation, reflection may also play a key role in setting the boundaries of the moral domain as such.

1 Introduction

The notion that thinking well can make a difference to a person’s moral outlook has a long and illustrious history in the annals of Western thought (e.g., Plato [e.g., Protagoras and Meno]; Kant, 1785/1959; Rawls, 1971; Singer, 2005). Ironically, it was David Hume, the reputed uber-sentimentalist, who penned one of the most impassioned testimonials in its defense:

The final sentence, it is probable, which pronounces characters and actions amiable or odious, praise-worthy or blameable…. It is probable, I say, that this final sentence depends on some internal sense or feeling…. But in order to pave the way for such a sentiment, and give a proper discernment of its object, it is often necessary, we find, that much reasoning should precede, that nice distinctions be made, just conclusions drawn, distant comparisons formed, complicated relations examined, and general facts fixed and ascertained …. In many orders of beauty, particularly those of the finer arts, it is requisite to employ much reasoning, in order to feel the proper sentiment; and a false relish may frequently be corrected by argument and reflection. There are just grounds to conclude, that moral beauty partakes much of this latter species …. (Hume, 1751/1983, p. 5, italics added)

More recently, Edvard A. Westermarck — an anthropologist, a philosopher, and Hume’s fellow sentimentalist — offered an intriguing and largely untested conjecture: two parties or peoples who share core moral ideals may nevertheless find themselves in a state of pervasive moral disagreement owing strictly to the degree of cognitive sophistication with which they apply these ideals to an issue at hand.

Most people follow a very simple method in judging of an act. Particular modes of conduct have their traditional labels many of which are learnt with language itself and the moral judgment commonly consists simply in labeling the act according to certain obvious characteristics … But a conscientious and intelligent judge proceeds in a different manner. He carefully examines all the details connected with the act, the external and internal conditions under which it was performed, its consequences, its motive and since the moral estimate in a large measure depends upon the regard paid to these circumstances his judgment may differ greatly from that of the man in the street even though the moral standard which they apply be exactly the same (Westermarck, 1906, pp. 9–10).

Thus, divergent as their philosophical outlooks might have been in other respects (see Westermarck, 1906), Hume and Westermarck appear to be largely in agreement on one key point: a general aptitude for thinking well is likely to exert a profound, even foundational, influence on a person’s moral outlook.

In what follows, we further elaborate and offer an empirical assay of this point by making use of what has long been taken to be one of the most reflexive and cognitively impenetrable moral cognitions ever (see Pennycook, Cheyne, Barr, Koehler, & Fugelsang [2014] for discussion) — the widespread repudiation of consensual sibling incest — and of one of the most demonstrably valid and widely used measures of “thinking well” — the Cognitive Reflection Test (CRT).

1.1 Repudiation of sibling incest and the social intuitionist model

The notion that principled moral opposition to third-party sibling incest is a result of automatic “negative imprinting” enabled by early childhood experiences (whose development both bypasses and is impervious to rational reflection) goes at least as far back as Westermarck’s Histories of Marriage (e.g., Westermarck, 1921) (see Lieberman, Tooby, & Cosmides [2003] for a recent revival of this perspective, but see Royzman, Leeman, & Baron [2009] and Royzman, Goodwin, & Leeman [2011] for a contrarian viewpoint). The putatively non-reflective quality of this opposition is front and center in the lead paragraphs of Jonathan Haidt’s (2001) influential “The emotional dog and its rational tail”. The paper opens with a duo of college-age siblings, Julie and Mark, opting for a night of non-committal sex while vacationing abroad. In Haidt’s (2001) positive analysis, when reading the story, “one feels a quick flash of revulsion … and one knows intuitively that something is wrong” (p. 814), with subsequent reasoning being largely utilized to marshal a range of post-hoc arguments for the validity of the initial impression, while remaining largely impotent to alter the impression itself (see Haidt, Bjorklund, & Murphy, 2000). That is, in Haidt’s politically minded update of Hume’s metaphor, moral reflection, can (at least, in this particular case) aspire to no higher office than that of the press-secretary of the “passions”.¹ (In Haidt’s model, “intuitions”, products of a fast, automatic, hard-to-recover, System-1 processes, take place of the “passions”; nevertheless, many of his candidate “affect-laden” intuitions are clearly what Hume and others would recognize as “passions” par excellence). Since press secretaries are neither paid nor invited to tell their masters what to think, the natural corollary of Haidt’s metaphor is that, while more characteristically reflective individuals may display a superior knack for defending their moral cognitions, whatever they may be (as well as influencing others’ moral cognitions in the process), there should be no unique association between one’s general capacity for private reflection and the content of one’s moral beliefs. This prognosis is strikingly at odds with that of the newer wave of “reflectionist” dual process accounts of moral cognition (see Bucciarelli, Khemlani, & Johnson-Laird [2008], Paxton et al. [2012], and Pennycook et al. [2014], among others; see Saunders [2013] for a thoughtful discussion of the nature of moral reasoning), which, different as they are in some respects, concur that reason is far more than the ever-conniving spin doctor of the passions, but is rather an active and independent contributor to the moral process itself.

1.2 Cognitive Reflection Test

The Cognitive Reflection Test (CRT) (Frederick, 2005) consists of three mathematical puzzles thought to gauge an individual’s ability and/or willingness to spend more time and mental effort examining and solving a problem characterized by an intuitively compelling but inaccurate response. Despite being only a few years old and three items short, CRT has proved itself as one of the most useful measures of cognitive performance available. To take a few examples, it has been found to predict resistance to common biases in judgments and decisions (Toplak, West, & Stanovich, 2011), including resistance to logical fallacies and overconfidence (Oechssler, Roider, & Schmitz, 2009), more elaborate and thorough heuristic search (Cokely & Kelly, 2009), greater future-mindedness (Frederick, 2005), enhanced forecasting ability (Mellers et al., 2014), a less robust belief in paranormal phenomena (Pennycook et al., 2012), and (under specified conditions) a propensity toward utilitarian moral judgment (Paxton et al., 2012, but see Royzman, Landy, & Leeman, in press). Particularly germane to the present project, Toplak et al. (2011) found CRT to be uniquely related to a variant of the classic outcome bias task (Baron & Hershey, 1988), with high CRT subjects tending to be significantly more resistant to the outcome bias than lower scoring subjects. This finding is of importance given that the outcome bias task’s key cognitive demand lies in being able to set aside one’s privileged knowledge of a decision’s end result while forming a fair assessment of an actor’s a priori decision-making skill (see also Pennycook et al., 2012). This, as we argue below, is also the key cognitive demand involved in one of the most widely used measures of mature moral judgment — the Moral-Conventional Distinction Task.

1.3 The Moral-Conventional Distinction Task (and a tentative typology of harms)

The Moral-Conventional Distinction Task (henceforth, MCDT) was developed as a methodological companion to what has become one of the most generative and influential models of moral cognition to come out of the structural-developmental tradition: the social domain theory (SDT) (see Nucci, 2001; Smetana, 1983; Turiel, 1983; Turiel, Hilderbrandt, & Wainryb, 1991; Turiel, 2002 for overviews, and Blair, 1995; Haidt, Koeller, & Dias, 1993; Nichols, 2002; Shweder, Mahapatra, & Miller, 1987 for further applications and critique). SDT’s fundamental supposition is that socially counter-normative acts may be deemed impermissible in two qualitatively distinct ways. On the one hand, there are prototypically moral transgressions, such as a child engaging in thievery or physical assault; on the other hand, there are prototypically conventional transgressions, e.g., a child coming to school wearing swimwear in lieu of a uniform (Turiel, 2002). Both types of acts are routinely judged inappropriate (or “not OK”) by a majority of children and adults, with the differences encompassing both form and content. With regards to form, conventional transgressions are routinely considered to be less serious and less wrong than their moral counterparts. Most critically, a prototypically immoral act is generally judged to be “not OK” even when its performance is entertained under a set of counterfactual conditions designed to negate its status as a subversion of the existing normative regime.² Thus, an individual may be asked to imagine that the act at hand is no longer “against the rules,” that it has the backing of a legitimate authority, or, perhaps most meaningfully, that it has taken place within a cultural milieu where its performance has been normatively sanctioned by popular support (e.g., Turiel et al., 1991; Turiel, 1983, see Nucci, 2001 and Turiel, 2002 for an overview of the congruence among these measures). While such “what if” propositions tend to generally erode the perceived inappropriateness of socio-conventional misdeeds (e.g., inappropriate dress code), they leave the perceived wrongness of prototypically immoral acts (e.g., stealing) relatively intact, giving them their signature socially transcendent character. This basic differentiation has been observed in a remarkably wide range of subject populations separated by class, religion, geography, cognitive ability, and age (e.g., Hollos, Leis, & Turiel, 1986; Madden, 1992; Nucci, 2001; Smetana & Braeges, 1990; Yau & Smetana, 2003; see Turiel, 2002 for a review).³

One key (and often overlooked) distinction within SDT (e.g., Turiel, 1983) is the relatively nuanced way in which conventional and moral transgressions are thought to differ with respect to their content, with each giving rise to a different sort of harm. Prototypically moral transgressions involve acts that are (perceived to be) intrinsically harmful to others (Turiel, 1983, e.g., p. 44, p.221). In contrast, while prototypically conventional transgressions may cause harm (which is real enough, if for no other reason that they cause disruptions of the established routine and/or hold a potential to undermine the reigning institutional authority), the harms they cause are contingent upon aspects of an existing normative regime that could themselves be deemed arbitrary and modifiable by consensus.

Consider: within the continental United States, right-hand vehicular traffic is the norm; those who breach it would rightly be deemed a danger to others and to themselves. Yet, the act of driving on the left does not carry harmful consequences as such — it causes harm only within a particular normative regime — the socio-conventional environment where the right-hand driving is the norm. As a consequence, the act of driving on the left is typically not considered wrong in and of itself, but is regarded instead as a conventional violation (albeit one of a very serious kind). And the right-hand driving norm itself is regarded as being arbitrary and revisable: few of us would be shocked (much less call for a moral crusade) upon first learning of a place where the converse rule is in effect. Similar considerations hold for another prototypically conventional (not intrinsically harmful) violation: wearing pajamas to work (henceforth, Pajamas, Huebner et al., 2010; please see the Appendix for the details of this and other key vignettes).

Contrast this with cruelty to animals. Like breaches of vehicular traffic rules, puppy torture is a normative no-no. However, along with other “prototypically immoral” acts, e.g., killing and fraud, it has a further feature that grants it a unique status within SDT’s normative hierarchy: it is intrinsically hurtful to the puppies. Unlike the harms resulting from a contravention of the right-hand traffic rule or contingently inappropriate dress code, the physical pain entailed in the puppy torture case cannot be voted or legislated out of existence by a sheer feat of social reorganization. It remains constant under a variety of conceivable normative regimes. Similar considerations hold for subtler varieties of intrinsic harm, such as those instantiated in Huebner et al.’s (2010) Tickets vignette, where an individual callously sells counterfeit concert tickets to the public knowing full well that those who purchased them shall arrive at the concert venue filled with anticipation that will come to nothing.

Thus, contingent harms, unlike intrinsic (and, thus, socially transcendent) harms, can be psychologically nullified under an alternate normative regime. If asked whether the act of wearing pajamas to work ceases to be problematic in a wearing-pajamas-to-work-is-splendid universe, those able to effect the requisite normative regime nullification, i.e., those able to override their existing epistemic viewpoint (a capacity that, as we noted earlier, is likely predicted by CRT), should return an unqualified Yes. By contrast, those who are unable or unwilling to effect normative regime nullification should perceive a contingently harmful act (i.e., Pajamas) as intrinsically harmful and thus transcendently wrong. In other words, cognitive reflectivity (as measured by CRT) should correlate with a tendency to treat acts comprised of contingent harms as merely conventional offenses, rather than offences that are universally, unalterably wrong.

In contrast, intrinsically harmful acts such as puppy torture, or theft (e.g., Greene et al.’s [2001] Wallet vignette), would be expected to yield a CRT-insensitive correlational pattern, since the harm or violation of rights that they produce remains intact regardless of any counterfactual conjectures. Thus, in cases of this kind, those who reflect deeply and effectively would be expected to return the very same verdict as those who reflect little or hardly at all.

In sum, to the degree that individuals’ capacity to effect normative regime nullification is reliably indexed by the CRT, and to the degree that normative regime nullification affects contingently harmful offences, but not their intrinsically harmful counterparts, two empirically testable hypotheses result. First, there should be a significant inverse association between individuals’ CRT scores and their tendency to assign socially transcendent/non-conventional status to paradigmatically conventional (contingently harmful) offences (i.e., Pajamas); second, no such association should obtain between CRT and paradigmatically immoral (intrinsically harmful) acts (i.e., Tickets).⁴

The situation, however, is rendered somewhat more complex by the fact that SDT also recognizes a third class of violations — violations of the so-called nonprototypical class (Turiel et al., 1991). These violations tend to be sexual in nature and feature consenting adults performing acts devoid of violence, exploitation, or compulsion behind closed doors. Perhaps the most-studied of nonprototypical transgressions is incest.⁵ Here, we are speaking specifically of reproductively inert, exploitation-free physical intimacy between consenting adults, as typified in the following Kissing vignette adapted from Haidt, Koeller, and Dias (1993):

Dave and Laura are college seniors. They are also brother and sister. They are quite affectionate with each other and like to cuddle and kiss each other on the mouth. When nobody is around, they find a secret hiding place and kiss each other on the mouth passionately.

On the surface, the vignette is a moral violation par excellence. When introduced in the context of several class discussions, it tended to provoke invariably negative reactions, with the majority of students judging the siblings’ actions “wrong” and “very bad.” Indeed, Turiel et al. (1991) found that a vignette containing a depiction of non-reproductive consensual incest was judged more seriously wrong by his subjects than any other “nonprototypical” offence (including abortion, homosexuality, and pornography).

Notwithstanding this, “the majority of subjects [in Turiel et al.’s sample] judged that incest should be legal both in the United States (53%) and in another country (68%)” (Turiel et al., p. 53; see table 15). Moreover, even Turiel et al.’s most conservative subjects (devout high-school Catholic students) tended to treat incest as a socially contingent and nongeneralized offence — that is, they did not regard incest as continuing to be wrong under an alternative normative regime in which it was permitted. Rape, killing, and theft, on the other hand, were judged as non-contingent and generalized by virtually all (and nearly all subjects also believed that these three should be universally illegal). Moreover, in justifying their negative evaluations of incest, subjects commonly appealed to “custom and tradition” as underlying normative standards, the appeals that have been shown to be intimately tied to conventional act justifications (Turiel, 1983; Turiel, et al., 1991, p. 3; see also Southwood, 2011).

Thus, notwithstanding strong social prohibitions against sibling incest, the existing evidence suggests that consensual sibling sex belongs to a set of “nonprototypical” (Turiel et al. 1991) issues that elicit an MCDT response that is more characteristic of contingently harmful, conventional offences (a la Pajamas) than of their prototypically immoral (intrinsically harmful) counterparts, such as Tickets and Wallet, featuring fraud and thievery, respectively. As such, the argument of the preceding section suggests that individuals’ reactions to a vignette such as Kissing will yield a correlational pattern more similar to a (contingently harmful) conventional offence (i.e., Pajamas) than to an intrinsically harmful offense(e.g., Tickets).

This expectation of distinctive correlational patterns for different categories of counternormative acts highlights the important theoretical distinction between our theoretical perspective and that recently articulated by Pennycook et al. (2014). Pennycook et al. predicted that higher-scoring CRT performers may ultimately render less severe moral wrongness judgments against consensual sibling incest and sexual intercourse with a dead chicken. However, according to their account, there is “no implied necessary association between rationality and rejection of the wrongness of vignettes. Rather, the hypothesis is that greater analytic ability and style is related to a greater tendency to reflect upon the details of the vignette and hence, given that the intuitive default is of considerable wrongness, lead to … reductions in the judgments of wrongness.” (Pennycook et al., 2014). On this account, low and high CRT individuals alike begin with a rapid condemnation of the act, followed by some downward adjustment for the high CRT performers, but little or none for the low CRT performers. Thus, low CRT performers would be expected to remain tethered to their initial normative default, rendering significantly more severe judgments across the board. Notably, Pennycook et al. offer no particular argument why this process should work any differently for acts of sexual deviance than for acts of thievery or physical violence.

Why, then, focus on incest and sex with dead poultry? In Pennycook et al’s view, the importance of examining such nonprototypical offenses (Turiel et al., 1991) in the context of the study of individual differences in cognitive style resides largely in the rhetorical value of doing so: being able to show that rational reflection matters here, where affect and intuition seem to reign supreme, signals that it should matter even more elsewhere. That is, as Pennycook et al. point out, because these types of stimuli “elicit a strong disgust based response [, they] provide a very strong test of the role of individual differences in moral judgement because such moral vignettes should be particularly sensitive to intuitive processing and resistant to reflective processing” (Pennycook et al., 2014, p. 194). Thus, viewed in the context of the present MCDT-based research design, Pennycook et al.’s (2014) predictions diverge substantially from ours, yielding no apparent expectation of differential correlational patterns for intrinsically and contingently harmful offenses.⁶

2 Study 1

2.1 Method

Participants.

Seven hundred and twenty participants located in the United States were recruited, paid, and debriefed through Amazon’s Mechanical Turk website. Participants were paid $1.00 to complete a survey that was expected to take roughly ten minutes. Twenty-one participants failed to complete their entire survey and were excluded from analysis, leaving a final sample of N = 698. Five hundred seventy-two of these participants were randomly assigned to participate in Study 1, while the remaining 126 were assigned to participate in Study 2, which was run concurrently.

Procedure.

In order to maintain participants’ attention, they were told at the beginning of the study that a special “code word” would appear at one point in the survey, and that they would need to enter this word at the end of the study in order to receive credit. Such a word did, in fact, appear just before the conclusion of the study.

An ancillary goal of Study 1 was to examine whether any effects of CRT would be moderated by the presence of incidental disgust, an emotional state that has been argued to increase the severity of moral condemnation (see, e.g., Wheatley & Haidt, 2005; Schnall, Haidt, Clore, & Jordan, 2008; Horberg, Oveis, Keltner, & Cohen, 2009). To the best of our knowledge, there has not yet been a single study, published or otherwise, that examined how feelings of disgust interact with trait cognitive reflectivity to predict moral judgments of wrong. Participants in Study 1 were randomly assigned to either experience one of two such inductions — one involving images of disgusting stimuli such as feces and vomit, and one involving ingesting a food or drink that the participant found to be very disgusting — or a control induction involving ingesting plain crackers. The participants in Study 2 received no induction and served as the control group for the participants who received the disgusting-picture induction. There were no main or interactive effects of this manipulation either on the scenarios reported here or additional scenarios used solely for the purpose of assaying the boundaries of the disgust induction, a result which we report in more detail and elaborate upon elsewhere (see Baron, Royzman, & Goodwin, 2013). Because this affect induction had no effects whatsoever (notwithstanding successful manipulations checks) and is peripheral to our primary hypothesis here, we do not discuss it further. A brief methodological description of the procedures can be found in Landy and Goodwin (2014), and additional materials/data pertaining to this aspect of the study are available from the first author upon request.⁷

Table 1: a. Descriptive Statistics for Study 1 (50.9% female).

Table 1: b. The percentage of subjects moralizing (treating as a socially transcendent offence) each act in Study 1, in order of increasing magnitude. Moralization refers to a tendency to deem a violation wrong in the context of the stipulated counterfactual reality where the violation is considered to be okay.

After completing the disgust induction procedure, participants were told that they had completed the “first study” and that they would now be moving on to the “second study.” They were told that they would be reading several vignettes and making judgments regarding the behaviors described. The vignettes, all presented in a randomized order, included the three key scenarios mentioned earlier: Pajamas (a non-intrinsically harmful conventional violation), Kissing (sibling incest), and Tickets (an intrinsically harmful, moral violation)⁸ as well as some additional vignettes. One of the vignettes (Muffin) was designed to represent a harm-free “personal” domain act (see Turiel 1983, Turiel et al., 1991, Turiel, 2002) — an individual switching from a banana muffin to a blueberry muffin as his breakfast meal. The other, the aforementioned Wallet, described a poor man taking money from a wealthy man’s wallet). In this vignette, the intrinsic harm caused by taking another’s property is “balanced out” by the overall gain in utility the act produces (18 out of 18 students in a preliminary study judged that overall utility would indeed be maximized if the poor man kept the cash). We therefore included this vignette to explore whether high CRT individuals responded to it in a different way than they responded to the paradigmatically moral Tickets case (Turiel, 1983), in which the intrinsic harm is unmitigated. The text of the four key vignettes is presented in the Appendix.

After reading each scenario, participants rated the wrongness of each act on a 100-point scale anchored by “Not at all wrong” on the one end and “Extremely wrong” on the other. This was followed by the key MCDT probe, which asked participants to “suppose that there were foreign country A where some time ago everyone came together and decided that a behavior such as this was OK. In your view, would it be wrong or not wrong for [the protagonist’s name] to do what he/she did, assuming he/she was raised and lived in country A?” In keeping with the traditional MCDT format, this question was framed as a dichotomous “wrong”/“not wrong” choice, with “wrong” and “not wrong” responses being categorized as “moralizing” and “non-moralizing” responses, respectively.⁹

After responding to all vignettes, participants completed the three-item CRT and a brief demographic questionnaire, including a measure of political orientation (responses were made on a nine-point scale ranging from “Extremely conservative” to “Extremely liberal” with the midpoint labeled “Middle-of-the-road” [Jost, 2009]). Also included with the demographic questionnaire was the five-item Private Body Consciousness scale (PBC; Miller, Murphy, & Buss, 1981), a widely used measure of a tendency to attend to and make use of the feelings emanating from one’s body, which has been found in some cases to interact with moral judgment, e.g., in one study high-scoring PBC participants (but not their low-scoring counterparts) made harsher moral judgments in a filthy work area than in a clean work area (Schnall et al., 2008).

2.2 Results and discussion

The descriptive statistics for the key variables and the percentage of subjects who rated each scenario as a socially transcendent offence can be found in Tables 1a and 1b, respectively. The key descriptive statistics for CRT are presented in Table 2 and are similar to those obtained by Frederick (2005). The zero-order correlations between CRT, Demographics, Politics, PBC, and the degree of moralization (attribution of social transcendence) for each of the key scenarios (Tickets, Pajamas, Kissing, Wallet) are reported in Table 3.

Table 2: Descriptive statistics for CRT in Study 1.

Table 3: Zero-order correlations between CRT, Demographics, Political ideology, PBC, and the degree of moralization (attribution of social transcendence) for each of the four key scenarios in Study 1.

^* p<.05

^** p<.01

^*** p<.001.

Consistent with our hypothesis, there was a significant inverse correlation between CRT and the extent to which subjects treated Pajamas as transcendentally wrong on the MCDT task (r= −0.144, p = 0.001), combined with a (non-significantly) positive association between CRT and MCDT responses on the Tickets vignette(r = 0.002, p = 0.966). The correlations were significantly different by Steiger’s z test (p = 0.004, one-tailed).¹⁰ Thus, consistent with one of our main hypotheses, we found that individual differences in deliberative capacity (CRT) were unrelated to the moralization of a paradigmatically moral, intrinsically harmful act, but significantly predicted a tendency to moralize a paradigmatically conventional, contingently harmful offence, with the less deliberative subjects being more inclined to treat the act as a socially transcendent (moral-like) wrong.

Consistent with our further prediction, we found a significant inverse association between MCDT responses to Kissing and the CRT (r = −0. 173, p < 0.001). This association was significantly different from that between CRT and Tickets (p = 0.0003, one-tailed, by Steiger’s method), but not statistically different from that between CRT and Pajamas. And, as would be expected, no significant association was found (r = 0.013, p = 0.748) between CRT and the level of moralization accorded to the paradigmatic “personal” act — a decision to switch from one breakfast food to another (Muffin).¹¹

Table 4: Regression coefficients, p-values, and odds ratios for the four key social transcendence/moralization variables (range: 0–1) as a function of CRT, with Age, Sex, Politics, and PBC as covariates.

The significant inverse association between CRT and Kissing (and CRT and Pajamas), accompanied by a non-significantly positive association between CRT and Tickets, is clearly in concert with our general hypothesis. However, as argued earlier, it is also consistent with a set of potentially valid third-variable accounts. While a list comprising all such third-variable candidates would be unmanageably (perhaps, infinitely) long, we identified three candidate variables — a person’s political ideology, his/her sex, and his/her sensitivity to gut feelings — as being of particular theoretical interest and concern. To explore these possible alternatives for the conjoint variation in CRT and MCDT, we ran a series of logistic regressions with Kissing and Pajamas social transcendence ratings as our target variables and Age, Sex, Politics (scored [1–9 scale] in the direction of greater liberalism), and PBC as covariates. As can be seen from Table 4, with all four covariates in place, CRT remains a significant inverse predictor of Kissing and Pajamas (but not of Tickets or Wallet).¹² With two exceptions (Politics and Sex), none of the covariates were significant predictors of the target variables. Politics was a significant inverse predictor of all four variables [Odds Ratios ranging between 0.889 and 0.832; ps ranging between < 0.001 and 0.03], while females tended to be significantly more moralizing about Pajamas than males [Odds Ratio = 1.954, p = 0.006]. Indeed, putting the results of the regression to the side, none of the four aforementioned variables could have conceivably mediated the link between moralization and CRT since, as indicated in Table 3, neither Age nor Politics correlated with CRT, while neither PBC nor Sex correlated with any of the target variables.

On the whole, the results are quite straightforward and fall readily in line with the hypothesized pattern. Yet, some lingering questions remain.

First and foremost, it may be asked whether, based on the evidence available from this study, we should not conclude that the more reflective subjects tend to deliver a more “relativist” (“not wrong in another country”) answer with respect to Kissing and Pajamas simply because they see these acts as less serious infractions even under existing conditions. If so, this may cause such individuals to judge these transgressions as “not wrong” when faced with making a dichotomous choice for the key MCDT question. To address this, we re-ran the aforementioned logistic regression models with Kissing wrongness ratings and Pajamas wrongness ratings as additional covariates. As would be expected, the effects were somewhat weaker but remained quite strong (Odds Ratio = 0.766 (vs. 0.742 in the original regression, p = 0.004) for Kissing, and Odds Ratio = 0.777, (vs. .711, p = 0.017) for Pajamas, when wrongness ratings were accounted for statistically. Thus, it appears that, consistent with our general proposal, high CRT individuals’ tendency not to treat Kissing and Pajamas as genuinely moral (socially transcendent) offenses cannot be fully accounted for by how severely wrong individuals judge these offenses to be under existing normative conditions. (And, again, it is possible that the small reduction in the CRT effect can be explained by an effect of perceived social transcendence on the wrongness judgments themselves; an act may be considered less wrong if it would not be wrong at all somewhere else.) This result, combined with the other disambiguating findings, is broadly congruent with our original conjecture that, at least in part, CRT’s relation to moralization results from high CRT individuals’ capacity to distinguish between contingently and intrinsically harmful acts, treating only the latter as genuinely moral (socially transcendent) offenses.

Second, we should note that, though the results displayed in Table 3 bear out our main theoretical predictions, they also signal some other intriguing correlational patterns that we did not anticipate. One concerns the uniform tendency for political liberals to be less likely to view the counter-normative behaviors of interest (Tickets, Pajamas, Kissing, and Wallet) as transcendently wrong. The crucial difference between this correlational pattern and that found for the “hyper-reflective” (high CRT) subjects is that, while the high CRT subjects exhibited their “relativistic” orientation in a “selectively amoral” fashion (going easy on the practitioners of illicit sex and dress-code rebels, but being as tough as the low-CRT subjects on acts of thievery and fraud), liberals tended to be more “uniformly amoral”, exhibiting a more relativistic normative orientation across-the-board. Similarly, the liberally oriented subjects tended to rate all behaviors (including theft and fraud) as significantly less wrong under existing normative conditions.

3 Study 2

3.1 Method

Participants.

As described above, 126 participants located in the United States were randomly assigned to complete the Study 2 procedure rather than the Study 1 procedure, though the two studies were run concurrently.

Procedure.

As in Study 1, participants were told at the beginning of the study that a special “code word” would appear at one point in the survey, and that they would need to enter this word at the end of the study in order to receive credit. Such a word did appear just before the conclusion of the study.

Participants in Study 2 did not receive any of the disgust inductions included in Study 1, and immediately performed the MCDT with respect to the five vignettes described above. They subsequently completed a number of items that Study 2 shared with Study 1. These included the CRT, PBC, and the demographic questionnaire. They next completed a filler task, followed by two novel items. One new instrument was a 6-item two-facet measure of conservative/liberal orientation (Kemmelmeir, 2008), which included three items concerning evaluative attitudes toward homosexuality and abortion; both issues were investigated by Turiel et al. alongside incest for their 1991 monograph and represent defining sexual morality issues of America’s cultural right (Baumgartner, Francia, Morris & Scavo, 2008). For each item, the subjects indicated their agreement/disagreement on a 5-point scale (varying from “Strongly disagree” to “Strongly agree”). The scale items, given below, were presented to the subjects in a randomized order and scored in the direction of greater conservatism (R = reverse scored): 1. The federal government should do more to control the sale of handguns (anti-regulatory; R); 2. Colleges should prohibit racist/sexist speech on campus (anti-regulatory; R); 3. Wealthy people should pay a larger share of taxes than they do now (anti-regulatory; R); 4. It is important to have laws prohibiting homosexual relations (social); 5. Same sex couples should have the right to legal marital status (social; R); 6. Abortion should be legal (social; R). Kemmelmeir’s original principal component analysis uncovered two factors that accounted for 55.6% of the overall variance, with the social conservatism factor being signaled by high factor loadings of sexual morality items (see items 4, 5, 6) and the second factor characterized by high factor loadings of “small government”/“anti-regulatory” items (items 1,2,3) (see Kemmelmeir, 2008). Our follow-up principal axis factoring analysis with oblique rotation (oblimin) and 0.5 as the conservative cut-off for rotated factor loadings largely confirmed the two-factor solution; the two factors accounted for 63.6% of the variance, with the social conservatism factor being signaled by high factor loadings of all three sexual morality items and the second factor characterized by high factor loadings of two of the three “small government”/“anti-regulatory” items. Consistent with the measure’s reported two-factor structure, the social conservatism composite (α = 0.783) and the anti-regulatory conservatism composite (α = 0.451) were not significantly correlated with each other (r = .12, p = 0.18). The second new measure assessed participants’ level of educational attainment, in years of formal schooling. The new measures were included to rule out potential third-variable explanations of any association between CRT and moralization that could not be ruled out in Study 1.

3.2 Results

The analyses reported here were designed to further examine the relative potency of CRT as a predictor of sibling incest moralization (Royzman et al., 2008; Royzman et al., 2009; Royzman, et al., 2011) by pitting the predictive powers of CRT against a more nuanced, sexual morality-centric measure of conservative/liberal orientation and a measure of educational attainment (see Method for details). The three social conservatism items of the conservatism scale (Kemmelmeier, 2008) concerned judgments about abortion and homosexuality, widely believed to be the defining issues of America’s culture wars (with opposition to both said to be strongly rooted in Christian biblical beliefs) (Baumgartner et al., 2008).

The composite scores for each factor/subscale were formed by adding the three individual items. As reported above, consistent with the measure’s posited two-factor structure, the social conservatism composite and the anti-regulatory conservatism composite were not significantly correlated with each other (r = .12, p = 0.18). Because of its low internal consistency and because it was conceptually unrelated to our main interest, we excluded the anti-regulatory composite from further consideration.

Table 5 lists the results of a logistic regression with Kissing (social transcendence) as the dependent variable and CRT, Age, Gender, PBC, and the two new items (Social Conservatism and Education) as covariates.

Table 5: Social transcendence (range: 0–1) for Kissing in Study 2 as a function of Sex, Age, PBC, Social Conservatism, Years of Education (Education), and CRT.

Note: Politics was excluded due to concerns about multicollinearity with Social Conservatism; entering the two predictors simultaneously rendered both non-significant, with the CRT level of significance remaining unchanged (p = 0.004). Social Conservatism was retained as the more morally pertinent of the two.

We entered all of the available covariates into a stepwise logistic regression (using the method of conditional backward elimination based on Wald’s test). Only two variables were retained in the final model, Social Conservatism (B = 0.175, odds ratio = 1.191, p= 0.023) and CRT (B = −0.564, odds ratio = 0.569, p = 0.001), with CRT being the stronger of the two predictors.¹³ Thus, CRT once again emerges as a predictor of Kissing after the other variables of interest have been accounted for statistically, thus replicating the pattern observed in Study 1. We should further note that, as in Study 1, the association between CRT and Kissing transcendence was not fully accounted for by the Kissing wrongness ratings when the latter was included as a covariate (B = −0.467; odds ratio = 0.627; p = 0.034).

The rest of the pattern was similar to that reported in Study 1, with one exception. Relatively few participants (13.5 percent) in this sample judged that the Pajamas scenario would still be wrong under a different normative regime, which may explain the lack of any significant association between CRT and the Pajamas transcendence rating (B = −0.163; odds ratio = 0.850; p = 0.489, accounting for Sex, Age, and Education). The remainder of the associations were as expected, including a significant inverse association between CRT and Pajamas wrongness ratings (standardized beta = −.198, p =.04) accompanied by no significant association between CRT and the wrongness/transcendence ratings for either of the intrinsic harm scenarios.

4 General discussion

The main thrust of the present research was to join prior work in social domain theory with a recently validated and popular measure of individual differences in deliberative capacity (CRT) to test the hypothesis that an individual’s propensity towards deliberative thought would be significantly associated with a tendency to treat non-intrinsically harmful, but socially inappropriate acts (wearing pajamas to work, engaging in acts of sexual intimacy with a sibling) as matters of convention rather than morality proper, but that no such association would be found between deliberative capacity and a tendency to moralize acts that are intrinsically harmful to others.

To test this hypothesis, we asked a large and ideologically diverse sample of American adults to complete the CRT along with a streamlined version of the MCDT (moral-conventional distinction task) featuring, among others, a prototypical conventional issue (Pajamas), a widely discussed nonprototypical issue (Kissing), and two vignettes involving prototypically harmful acts against others. We found that, while high CRT performance was significantly associated with a tendency to treat contingently harmful transgressions as arbitrary societal constructions, it bore no relation to the tendency to moralize, or not moralize, acts that are intrinsically harmful to others. Moreover, this pattern remained intact when taking into account political orientation, sensitivity to gut feelings, gender, age, and, in Study 2, educational attainment and a more nuanced measure of social conservatism centered on sexual morality. Nor was the CRT-moralization link fully explained by subjects’ antecedent ratings of the acts’ permissibility. This indicates that CRT’s predictive relation to our key measure (one that no previous study of cognitive style, including Pennycook et al.’s [2014] has tapped) — whether a counter-normative act will be represented as a “genuinely moral” (socially transcendent) rather than merely conventional offense — does not merely reflect a normative spillover from subjects’ reactions to the wrongness probe.

Our pattern of findings may offer a new way to illuminate a very old question: what accounts for the apparently vast differences in moral opinion across time and place? (To see just how old the question is, one need only refer to Herodotus’s famed account [the Histories, Book III, section 38] of King Darius’s futile attempts to persuade some of his subjects to forfeit their customs for the disposal of the dead in favor of their neighbors’.)

In much recent research, apparent group differences in the moralization of actions have been taken as evidence for the existence of different underlying moral systems that are not shared or may be differentially activated across groups (see, for instance Shweder et al., 1987, 1997; Jensen, 1997, 1998; Graham, Haidt, & Nosek, 2009). One commonly cited case in point is the priestly town of Bhubaneswar, where it is judged morally impermissible for a widow to eat fish. At first blush, this seems like a case of fundamentally divergent values par excellence, with the “traditional” Indian informants subscribing to a code of ethics (later dubbed the “Divinity Code” [Schweder et al., 1997]) that their liberal Western counterparts either do not possess at all or possess in a radically attenuated form. However, this interpretation of the evidence has been questioned (Turiel, Killen, & Helwig, 1987). For, as Shweder’s own ethnographic work indicates (see Turiel et al., 1987 for an overview), there is a viable alternative account of the phenomenon: according to the local beliefs, fish acts as a sexual stimulant, increasing the likelihood that the widow will engage in sexual congress, which her husband’s ever-watchful spirit could witness (and presumably agonize over) from its celestial abode. On this interpretation, the ban on eating fish appears to be but another example of a societal attempt to regulate preventable harm. Similar “informational assumptions” (Turiel et al., 1991) about the fabric of the cosmos, the mind, and, especially, the reciprocal relationship between material and spiritual realms, may apparently account for a wide range of other seemingly indisputable cases of fundamental “moral disagreement” (Brandt, 1954).

Our findings may elucidate this issue by revealing a third and hitherto unexplored reason why different individuals or groups might moralize the same behaviors to varying degrees: the difference may be less a matter of what they think than of how they think, i.e., their ability (and/or willingness) to counterfactualize the existing normative regime on demand. Different cognitive styles could produce different patterns of moralization, leading, per Westermarck (1906), to recognizably distinct normative profiles, even while holding constant the key value positions and the underlying factual assumptions. Residents of Bhubaneswar might simply be unable (or unwilling) to posit an alternate reality in which widows can properly enjoy a plate of cod or in which a child may wear a non-traditional item of clothing to a priestly ceremony.¹⁴

Our demonstration that different cognitive styles are associated with different patterns of issue moralization also has some clear methodological implications. For one, it should give pause to any automatic inference from the evidence of an apparent difference in cultural (Shweder et al., 1987), socioeconomic (Haidt et al., 1993), or ideological (e.g., Jensen, 1997) patterns of moral evaluation to a presumption of fundamentally divergent values as the underlying cause (Graham et al. 2009). The differences could simply result from distinctive cognitive styles, which affect how individuals understand and apply their shared values to facts on the ground. In fact, the overall pattern we report is highly reminiscent of that found in one previous study (where no controls for cognitive style or cognitive ability were used [Haidt et al., 1993]) in which members of economically disadvantaged, relatively uneducated groups were shown to be significantly more likely than their better educated high-SES counterparts to blur the distinction between morality and convention, with the lower SES group treating “harmless” nonprototypical offences, including incest, as moral absolutes. Interestingly, the two SES groups were, however, identical in their condemnation of the one and only truly prototypical moral offense included in the stimulus set: a child pushing another child off a swing. Given the increasingly well-documented association between poverty and impaired cognitive performance, as measured both by indices of fluid intelligence and CRT-like tasks requiring individuals to override an intuitively salient response to perform well (see Mani, Mullainathan, Shafir, & Zhao, 2013 for evidence and discussion), the current analysis strongly indicates that the SES-moralization link found in Haidt et al.’s (1993) study would have been at least partly mediated by measures of cognitive style and cognitive ability had such measures been used, suggesting that the moral-conventional distinction gap between Haidt et al.’s high- and low-SES participants is at least as likely to be a product of the low-SES participants’ cognitive limitations as it is of high-SES participants’ purportedly limited access to the full range of foundational values, whether these are framed in terms of Shweder and colleagues’ original “Big Three” (Shweder, Much, Mahapatra, & Park, 1997) or Haidt and colleagues’ more recently proffered “Big five” (Graham et al., 2009).

One further novel finding to emerge from the present research is the striking divergence between the uniform correlational pattern found between liberalism and MCDT response, and the more selective pattern found between the latter and the CRT; while the higher scoring CRT subjects tended to be “less moralistic” in a category-specific way (“going easy” on the practitioners of illicit sex and dress-code violators, but being as “tough” as their low-CRT counterparts on acts of thievery and fraud), liberals tended to be more “uniformly amoral”, engaging in less moralization across-the-board. Similarly, the liberally oriented subjects tended to rate all behaviors (including theft and fraud) as significantly less wrong. While it is clearly beyond the scope of this paper to adduce a convincing explanation for this effect, we hazard that it partly results from the fact that those judging from within a distinctly liberal mindset may view all deviant acts (including mass fraud) as a shared responsibility between the individual transgressor and the “inauspicious” social environment within which he or she is situationally embedded, while more conservative individuals see the responsibility as resting squarely on the shoulders of the deviant actor alone.

A final and, perhaps, most basic implication of our results is that, in conjunction with the pioneering work of Paxton et al. (2012), Pennycook et al. (2014) and others, they underscore the foundational importance of deliberative thought to the process of moral evaluation. Moreover, the present research adds to both of the aforementioned studies by indicating that, beyond affecting the reported amplitude of one’s moral disapproval, proneness to rational deliberation may play a unique role in setting the boundaries of the moral domain as such, informing one’s very construal of where the realm of convention ends and that of morality proper begins. For people lacking the requisite measure of cognitive sophistication (be they the Brahmans of Orissa¹⁵ [Shweder et al., 1987] or lower-SES Brazilian children [see Haidt, 2012, p.21]) or the will to posit cosmological realities different from one’s own (see Madden, 1993), the line between the two normative domains may be blurred to the point of non-existence.¹⁶

What does this all mean? It has been widely argued in recent moral judgment literature — not to mention historically — that, by and large, reason “can aspire to no higher office” than that of the resourceful and steadfast, but ultimately clueless spinmeister of the passions: an inveterate after-the-fact explainer and justifier with little to no involvement in the judgment process itself. The studies reported herein suggest otherwise. They suggest that, aside from likely playing a key role in adjudicating intrapsychic normative disputes (Paxton et al., 2012; Royzman et al., 2011), reason is in fact an active and powerful contributor to the construction of the moral domain itself, the position that David Hume himself held dear some 250 years ago.