In the target article, Stanford argues that moral demands inhabit a distinct conceptual category, where they are experienced as externally imposed obligations; and that evolutionarily, this externalization protected prosocial individuals from exploitation, ensuring that any felt obligation to conform with a social norm was paired with a conviction that others should conform as well. Thus, externalizing moral demands (i.e., experiencing moral demands as objective) allowed individuals to reap the benefits of prosociality while also policing defectors.

One critical claim for this argument is that morality comprises “a distinctive conceptual category” (target article, sect. 5, para. 15, sect 6, para. 1). Work was reviewed, showing that children categorically distinguish morals from social conventions (Smetana Reference Smetana, Killen and Smetana2006; Turiel Reference Turiel1983); moral properties are distinguished from response-dependent properties (e.g., “yucky”; Nichols & Folds-Bennett Reference Nichols and Folds-Bennett2003); and, critically, morals are rated as categorically more objective than preferences and social conventions (Goodwin & Darley Reference Goodwin and Darley2008; Reference Goodwin and Darley2012), licensing the conclusion that “[moral beliefs are] treated almost as objectively as scientific or factual beliefs … [and] as categorically different from social conventions (Goodwin & Darley Reference Goodwin and Darley2008, p. 1359). However, we doubt that morals comprise a distinct category – at least on the basis of their objectivity, universality, and authority-independence, as has traditionally been argued. First, prior work has not licensed statistical generalizations about a moral domain, as its authors had assumed. Second, our recent work suggests that objectivity is not an essential feature of morality; rather, behaviorally and neurally, moral claims are more akin to preferences.

To preface our statistical criticism: in most cases, morality must be studied using specific stimuli. For instance, stimuli might include asking children whether hitting (e.g., Wainryb et al. Reference Wainryb, Shaw, Langley, Cottam and Lewis2004) or stealing (Tisak & Turiel Reference Tisak and Turiel1988) is acceptable. Goodwin and Darley (Reference Goodwin and Darley2008) asked about discrimination, robbery, and firing into a crowd, among others. The statistical problem is that one must move, by inference, from the specific stimuli to a sampled population (e.g., a moral domain). To make this inference, stimuli must be treated as a random effect (i.e., as a random sample from a population). If stimuli are averaged, then statistical conclusions (e.g., a t-test across subjects) apply only to those stimuli; in this case, stimuli are a fixed effect (and this leaves aside issues with randomly sampling moral stimuli, a problem beyond the scope of this commentary). This “stimuli-as-fixed-effects” fallacy has been identified in other fields (Clark Reference Clark1973) and is easily solved using mixed effects analyses (Baayen et al. Reference Baayen, Davidson and Bates2008), but the problem has been largely ignored within psychology (Judd et al. Reference Judd, Westfall and Kenny2012; Westfall et al. Reference Westfall, Kenny and Judd2014). The moral/conventional distinction has been criticized on the basis of stimulus content (e.g., stimuli typically describe “schoolyard” violations; Kelly et al. Reference Kelly, Stich, Haley, Eng and Fessler2007, p. 121), and these criticisms may be justified; but ultimately, such content-based criticisms are unnecessary, as the original findings never licensed conclusions about a moral domain at all. They licensed conclusions about the exact stimuli that were used.

Prior work has argued that morality is essentially objective, but how to measure meta-ethical judgment has also been a long-standing concern (for an excellent discussion, see Goodwin & Darley Reference Goodwin and Darley2010). Stanford rightly calls attention to the “hybrid character” of morality, where moral claims fall somewhere between “[objective] representations of how things stand in the world itself … and our subjective reactions to those states of the world” (sect. 6, para. 11); however, prior work has rarely allowed participants to express this hybrid nature. For example, if participants are forced to classify moral claims as true, false, or an opinion/attitude (Goodwin & Darley Reference Goodwin and Darley2008), then distinctions between morals, facts, and preferences may appear to be more discrete than they actually are. We attempted to address this issue in a recent study (Theriault et al. Reference Theriault, Waytz, Heiphetz and Young2017), where participants read moral claims (presented alongside facts and preferences) and simultaneously rated the extent that each was “about facts,” “about preferences,” and “about morality” (1–7; “not at all” to “completely”). Moral claims should be more moral-like than fact-like or preference-like; however, the question of interest was which secondary feature would dominate: Are morals largely fact-like? Or are they largely preference-like?

Although prior work has emphasized that moral claims are essentially objective, our work suggested the opposite: that moral claims were perceived as largely preference-like. Among a set of 24 moral claims that we had generated, and also among 22 claims adapted from the moral foundations questionnaire (Graham et al. Reference Graham, Nosek, Haidt, Iyer, Koleva and Ditto2011; Iyer et al. Reference Iyer, Koleva, Graham, Ditto and Haidt2012), participants rated moral claims as significantly more preference-like than fact-like. Furthermore, we scanned subjects as they read the same claims, and found that moral claims elicited widespread activity in brain regions for social cognition and theory of mind (Schurz et al. Reference Schurz, Radua, Aichhorn, Richlan and Perner2014; Van Overwalle Reference Van Overwalle2009), overlapping with activity for preferences, but not facts. Stanford argues that humans have “[gone] in for cognitively complex forms of representation,” and that moral norms have likely been shoehorned into an evolved framework where “the most fundamental division … [is] between how things stand in the world … and our subjective reactions to those states” (sect. 6, para. 11). If this fundamental representational division exists, and if moral demands were (in part) externalized by co-opting cognitive processes that evolved to represent the world, as Stanford seems to argue, then we should see at least some significant overlap between processing for morals and facts. Instead, morals were behaviorally perceived, and neurally represented, as akin to preferences.

Nevertheless, we agree that moral demands are often experienced as external, even if the moral domain is not as unified as prior work has suggested. But this moral externalization may exist along a spectrum: Some moral claims may be experienced as more objective than others. Indeed, we characterized this variability in a recent study, where by-stimuli moral objectivity tracked with activity in social brain regions (Theriault et al., Reference Theriault, Waytz, Heiphetz and Youngunder review). Understanding this variability will be critical for an account of why some moral demands are experienced as obligatory and enforced (e.g., “murder is wrong”) whereas others are not (e.g., “eating meat is wrong”).