1.1. Two Interpretive Problems of the Third Tradition

When faced with evidence of the ways people process ethical quandaries, NMN theorists are often confronted with two interpretive choices. First, they must decide whether moral processingFootnote ⁴ tends to function effectively (reflecting core human competence) or is prone to dysfunction (leading to error in judgment). Competence theorists contend that ordinary moral judgment tends to point toward normative moral judgment. In contrast, error theorists contend that ordinary moral judgment tends either to point away from normative moral judgment or to have no reliable relationship with normative moral judgment at all.

Second, NMN theorists must decide whether the study of ordinary moral judgment can provide positive evidence in favor of some normative ethical theory or only negative evidence against some theory. Positive projects assume that competent moral judgment tends to align with, and therefore support, certain views in the canon of philosophical ethics. Thus, positive projects in NMN are about substantiating normative ethical claims on the basis of empirical discoveries, an endeavor that, as noted, runs the risks associated with the is-ought problem.

Negative projects, on the other hand, assume just the opposite (Knobe et al. Reference Knobe, Buckwalter, Nichols, Robbins, Sarkissian and Sommers2012). Negative projects in NMN are essentially about holding normative claims accountable to the descriptive realm of possibilities. Implicitly working from the premise that ought implies can, such projects typically seek to invalidate claims that we should behave (act, judge) in certain ways by revealing that we cannot behave (act, judge) in those ways (e.g., Doris Reference Doris1998). Perhaps the most notable advantage of negative projects over positive ones is their intuitive plausibility.

For instance, to many people it seems patently unfair to hold adolescents, the cognitively impaired, animals, or indeed anyone at all to standards of behavior that they are neurophysiologically incapable of upholding. This sole fact, it could be argued, might be sufficient to defeat any appearance of moral obligation among those individuals. By contrast, it seems patently untrue that we should hold anyone to a given standard of behavior solely because they are capable of—or perhaps even inclined toward—upholding it.

Of course, the principle that ought implies can is not without issue. Some philosophers deny that moral obligation is bounded in this way by physical or practical possibility (Kant [1785] Reference Kant and Paton1956). Even without relying on exotic cases and thought experiments, one can see problems with this principle. For instance, if an individual is physiologically incapable of inhibiting his violent racist tendencies, this would not seem to excuse his violent racist behavior. One might argue that this person, like all people, ought to be egalitarian even though he cannot be egalitarian. Likewise, it could be argued that normative moralization of children is centrally concerned with the enforcement of standards that children are not yet equipped to uphold.

Despite the reasons to believe that descriptive facts about individuals’ capacities cannot always be used to identify the boundaries of their normative obligations, negative projects seem at least to present narrower targets for objection, on average, than positive projects do. Whereas the distinction between competence and error theories concerns the (mis)alignment of ordinary moral judgment and normative moral judgment, the distinction between positive and negative projects concerns the (mis)alignment of normative moral judgment and specific views in philosophical ethics. However, researchers’ evaluations of theoretic (mis)alignment are, in both cases, subject to the influence of a priori views held firmly by the researchers making those evaluations before any evidence has even been collected. Thus, views within NMN typically express two qualitative evaluations that cannot be based entirely on empirical evidence: an evaluation of ordinary moral judgment (competent or errant) and a stance toward some canonical or novel moral theory (positive or negative).

It is in large part against these two kinds of qualitative evaluations that NMN researchers receive pushback. In my concluding remarks (section 5), I will argue that NMN researchers can and should develop at least some research programs that avoid making any such a priori evaluative commitments. But for now, I wish to explore the two major types of pushback these evaluations tend to engender. This exploration may provide some insight into how NMN researchers might minimize the impact of such objections in the future.

1.2. Two Kinds of Objections to the Third Tradition

Objections to NMN seek to establish that the neuroscience of moral judgment is incapable of contributing meaningfully to normative understanding of philosophical ethics. Establishing this in the abstract would be rather difficult, so most critics have focused their attacks on specific projects within NMN. Objections to specific projects within NMN may be thought of as consisting of two main kinds: inferential and methodological.

Inferential objections claim that the gap between philosophical ethics (including metaethics) and the neuroscience of moral judgment is not (or cannot be) truly bridged by NMN. These objections rail against the relevance of neuroscience—which objectors view as a fundamentally descriptive discipline—to questions of normative ethics and metaethics. For instance, one inferential objection to NMN is that even if we knew for certain that all judgments of guilt were based in part on the perception of intent, this would not prove the positive thesis that intent is a necessary condition for actual guilt (more on this in section 2.2). Inferential objections concern issues such as the a priori adoption of normative ethical stances in the interpretation of empirical data, presumptions about the normative relevance of ordinary moral judgment, and prejudices about the relative worth of different kinds of neurological (and psychological) processes.

Methodological objections seek to discredit the neuroscience of moral judgment itself, at least as it is currently practiced. Such objections typically do not raise concerns about the relevance of neuroscientific descriptions to normative ethics—instead, they deny that those descriptions are accurate in the first place. For instance, Berker (Reference Berker2009) points out that a frequently used set of moral dilemmas meant to distinguish deontological from utilitarian judgment (Greene et al. Reference Greene, Sommerville, Nystrom, Darley and Cohen2001) fails to do so because those dilemmas are poorly designed (as I discuss in section 3.2). Methodological objections may be specific to the neuroscience of moral judgment, or they may critique neuroscience more generally. Among other things, such objections may impugn the soundness of neuroimaging techniques like fMRI, criticize the construction of experimental probes and materials, or question the ability to infer general principles from demographically homogenous samples (see Holtzman Reference Holtzman2017).

While the distinction between methodological and inferential objections may be useful, it is not as hard-and-fast as the distinction between the two types of interpretive problems (section 1.1). This is because inferential and methodological objections sometimes overlap. A poorly designed questionnaire always presents a methodological problem, but if the poverty of its design lies in its conflation of, say, moral and conventional norm violations, then it is also inferentially problematic. Thus, the appropriate response to certain inferential problems may actually involve changes in methodology (section 4.2).

Furthermore, it is worth noting that while both kinds of objection may be especially challenging for NMN theorists, such challenges are not unique to NMN. For instance, philosophers wishing to leverage evolutionary theories for or against certain views in moral metaphysics (Street Reference Street2006) and epistemology (Kahane Reference Kahane2011) face similar inferential objections (Vavova Reference Vavova and Shafer-Landau2014; Holtzman, Reference Holtzmanforthcoming). And every science, not just neuroscience, faces some degree of scrutiny in the form of methodological objections. Still, since this essay aims to attenuate resistance to NMN, I will be returning to these themes throughout the essay.

2.1. In Favor of Positive and Negative Competence Theories

Gerd Gigerenzer touts the competence of ordinary moral judgment and believes that discoveries about the ways people make moral judgments weigh negatively against some canonical views in philosophical ethics. Thus, he is what I call a negative competence theorist. Gigerenzer (Reference Gigerenzer2008) argues that the human mind functions largely on the basis of heuristic processes that seek to satisfice, that is, to find good-enough solutions to problems. And these heuristic processes, he argues, are in many cases more effective than optimization procedures, which seek to find the best solutions. Unlike many of his contemporaries (e.g., Sunstein Reference Sunstein2013), Gigerenzer does not just think that heuristics are more efficient than other approaches to decision making; he believes that in many cases, they are simply better (Gigerenzer and Selten Reference Gigerenzer and Selten2002). For this reason, he thinks that we should view the brain as a paragon of moral decision making in at least some ways.

On these grounds, Gigerenzer mounts an empirical challenge to idealized utilitarian calculus as it is traditionally conceived (Posner Reference Posner1979). Although Gigerenzer's (Reference Gigerenzer2010) ‘aim is not to provide a normative theory’ (530), he argues that findings in the cognitive sciences ‘challenge the normative ideal that maximizing can generally define how people ought to behave’ (529). Therefore, while his philosophical arguments are of a negative nature, Gigerenzer tends to view the processes that typically underwrite moral decision making as competent. Gigerenzer merely suggests that if utilitarianism is correct, then in many cases it may be morally better to satisfice (Simon Reference Simon1956) than to maximize, but how best to satisfice and why we should believe utilitarianism to be correct, Gigerenzer does not say.

Kurt Gray is another noteworthy competence theorist in NMN, but unlike Gigerenzer, he takes research in neuroscience and psychology to reflect positively on at least some moral theories. His views center around two components of moral judgment: the perception of other minds and the ascription of mental states to them. In ‘Mind Perception Is the Essence of Morality’ (Reference Gray, Young and Waytz2012), Gray and his colleagues argue that ‘moral judgments should be linked to perceptions of intention and suffering’ (104) and that, at least in what they call prototypical cases, ‘moral transgressions involve perceptions of suffering’ (108). Gray and his colleagues express strong (albeit qualified, see section 2.2) views on what morality is and on the ways that normative ethical evaluations ought to be made. Thus, Gray suggests that we might be able to leverage descriptive neuroscientific findings about brain regions involved in ordinary moral judgment (e.g., Young and Saxe Reference Young and Saxe2008) into dicta concerning normative moral theory.

In this way, Gray's positive competence theory differs importantly from Gigerenzer's negative competence theory. Both authors believe that ordinary moral judgment fundamentally reflects competent application of normative ethical concepts. But in Gigerenzer's negative view, the neuroscience of (competent) moral judgment can tell us only what kinds of moral judgments are not normative: maximizing happiness ≠ good, maximizing ≯ satisficing. In contrast, Gray's positive stance suggests that (competent) moral judgment may be evidence for what kinds of judgments are normative: intentionally causing suffering = bad, unintentionally causing suffering > intentionally causing suffering.

2.2. Against Positive and Negative Competence Theories

Considered together, the cases of Gigerenzer and Gray illustrate a pair of related points. First, positive projects within NMN are arguably less plausible in many cases than negative projects. Unlike Gigerenzer's negative project, Gray's positive project threatens to jump from neuroscientific evidence that certain factors affect the perception of moral worth to the suggestion that those same elements factor into normative moral worth. In that sense, Gray's argument may have far-reaching implications if one accepts it, but one might reject its implications due to inferential objections.Footnote ⁵ In this regard, his point is that whatever the philosophical merit of an emotion-versus-reason framework for normative moral evaluation might be, our actual moral evaluations operate largely on the basis of two kinds of perceptions of other minds. Gray argues that of these two kinds—the perception of intentional agents and the perception of suffering patients—each requires both emotion and reason (or, as he puts it, ‘cooperation between affect and cognition’ [Gray and Schein Reference Gray and Schein2012: 419]). Thus, the common conception of moral judgment as emerging from a battle between emotion and reason cannot be brought to bear on empirical-descriptive studies of moral processing, and should (on Gray's view) be replaced by Gray's own mind-perception model.

The second point illustrated by this pair of cases is that negative projects within NMN are in many cases arguably more parochial—that is, limited in relevance to specific moral schemes—than positive projects within NMN. In contrast to Gray, Gigerenzer makes the somewhat staid claim that attempting felicific calculus (Bentham [1823] Reference Bentham2005) in the ‘real, uncertain world’ (Gigerenzer Reference Gigerenzer2010: 536) is usually less efficient and sometimes even less accurate than undertaking what we might call ‘felicific estimation’ in such a world. In this regard, Gigerenzer's claims remain largely within the descriptive domain. Gigerenzer is right to say that ‘whether optimizing or satisficing leads to better outcomes in the real, uncertain world is an empirical question’ (Gigerenzer Reference Gigerenzer2010: 536). But as Bruni and his colleagues (Reference Bruni, Matteo and Rini2014) point out, the extent to which normative ethics is concerned with better outcomes (whatever that might mean) is an open, nonempirical question. Whereas utilitarians think that normative ethics is entirely concerned with outcomes, deontologists, virtue ethicists, care theorists, and others disagree.

So, too, is it an open question whether normative ethics is concerned with the real, uncertain world (Nagel Reference Nagel1976), or only with idealized, nonstochastic expectations. If philosophers like Kant (Reference Kant and Paton[1785] 1956) are right, then the real-world outcomes of our behaviors may be largely morally irrelevant. For Kant, all that matters are our intentions to act in accordance with the principles of duty. If Kantians—and many others—are right to take issue with utilitarianism, then Gigerenzer's point about the value of heuristics in improving real-world outcomes, even if correct from a practical standpoint, may have no bearing on normative ethics.

3.1. In Favor of Positive and Negative Error Theories

Whereas Gigerenzer and Gray both believe that ordinary moral processing tends be driven by underlying competencies that reflect normative ideals, the NMN researchers who have received the most attention have usually argued the opposite. They opine that ordinary moral processing is fundamentally flawed and that the moral faculties we have evolved are no more useful in many of today's contexts than, say, our appendices are. According to these researchers, a great many of the moral judgments people make tend to be erroneous. Nonetheless, these error theorists often think that empirical work can help us draw normative conclusions—either positive or negative. They believe that the neuroscience of moral judgment can reveal the poverty of ordinary moral processing. Negative error theorists argue that insofar as the conclusions reached by impoverished neural processes should not be trusted, neither should the normative ethical perspectives they underwrite. Positive error theorists push the point even further, claiming that by discounting errant moral judgments, we may be able to hone in on the correct, last-man-standing moral perspective (although see section 3.2 for objections; see also Kamm Reference Kamm2009; Meyers Reference Meyers2015).

Perhaps no NMN theorist has generated more controversy than Joshua Greene (Reference Greene and Sinnott-Armstrong2008, Reference Greene2014a). At the turn of the century, Greene and his colleagues (Reference Greene, Sommerville, Nystrom, Darley and Cohen2001, Reference Greene, Nystrom, Engell, Darley and Cohen2004) purported to show via fMRI that automatic processes such as emotion tend to underwrite rule-based (‘deontological’) moral judgments, whereas reason and controlled cognition tend to underwrite outcome-based (‘utilitarian’) judgments.Footnote ⁶ On these grounds and on the (rather large) assumption that reason is more trustworthy than emotion (cf. de Sousa Reference De Sousa1990; Pham Reference Pham2007), Greene has reputedly mounted the argument that normative moral judgment—and thus, moral right and wrong—tends (and ought) to be determined by outcomes rather than by rules. I am not convinced that Greene's argument is quite as simplistic as this, but this does seem to be the received view of his work.

Greene (Reference Greene2014a) is thus usually interpreted to be an error theorist, arguing that though our moral faculties are in many respects amazingly fast and accurate, they are far from perfect. Moreover, when engaged in situations that were common during the evolutionary history of humankind, our relatively efficient moral faculties are especially prone to misfire due to adaptations that were once an aid to survival but do not align with morality (Greene Reference Greene2014b; see also Singer Reference Singer2005). Needless to say, what many perceive to be Greene's presuppositions about just what qualifies as moral ‘misfiring’ (Sunstein Reference Sunstein2005) have not gone uncriticized (section 3.2).

Regardless, for Greene the upshot of his error theory is positive. Once we recognize the evolutionarily familiar but morally irrelevant factors that can influence moral processing, we can begin to identify what sorts of judgments do not depend on those distractors. And once we do, Greene argues, we usually find that utilitarianism wins the day. From this point, Greene is often accused of jumping from the finding that a utilitarian framework usually does win the day (under what he calls ‘impersonal’ circumstances) to the conclusion that it should win the day (under any circumstances). It is not clear to me that Greene actually makes this leapFootnote ⁷ although anyone who does certainly opens herself up to certain strong criticisms (section 3.2).

One frequently overlooked but standout aspect of Greene's work is its clearly stated recognition of the perils of jumping ‘from neural “is” to moral “ought”’ (Greene Reference Greene2003: 846). From the start, Greene and his colleagues acknowledge that their results ‘raise but do not answer’ (Reference Greene, Sommerville, Nystrom, Darley and Cohen2001: 2107) questions about the link between philosophical ethics and the neuroscience of moral judgment. Greene (Reference Greene, Nystrom, Engell, Darley and Cohen2004) also highlights at least one major philosophical implication of his findings that does not take on normative aspirations. He argues that for utilitarian inclinations to win psychological battles against deontology, controlled cognition must win out over emotion.

This is a purely descriptive claim. And if it is true, Greene points out, then we may need to rethink radically the metaethics we have developed over the past 350 years. If deontological principles arise not from the triumph of pure reason, as Kant claimed, but from its absence, poverty, or failure, then the (descriptive) nature of deontology's foundations—even if not the (normative) propriety of those foundations—may have been long misconstrued. It is this empirically contentious—but normatively neutral—proposition that Greene (Reference Greene and Sinnott-Armstrong2008) calls ‘The Secret Joke of Kant's Soul’.

The joke, according to Greene, is that deontology and rationality, those two pillars of Kant's work, are in reality diametrically opposed. Rationality begets consequentialist judgment, according to Greene, whereas deontology is the product of emotional decision making. I take this to be Greene's most interesting point and think it unfortunate that it has largely been overshadowed by the normative arguments usually ascribed to him. While many researchers (including Gray) would disagree with this descriptive claim, it is ultimately an empirical question, not a normative one, and therefore immune to many of the objections to which other aspects of Greene's work have been subjected.

Like Greene, Kumar and Campbell (2012) also seem to be error theorists. However, after rejecting Greene's project on interpretive grounds (section 3.2), they argue that negative philosophical projects may benefit from neuropsychological insights into moral processing. Thus, they are negative error theorists. They claim that when sets of ordinary moral judgments violate what they call consistency reasoning—a notion similar to (though substantively different from) that of wide reflective equilibrium (Rawls Reference Rawls1971)—we can be sure that a flawed ethical judgment has been made. However, they say, we cannot be sure which of the coincident judgments is flawed and therefore cannot deduce a positive conclusion about which judgment ought to be embraced.

3.2. Against Positive and Negative Error Theories

Some authors have criticized error theorists on inferential grounds. For example, Meyers (Reference Meyers2015) has pointed out that moral systems based on neither outcomes nor rule-following are entirely ignored by Greene and other dual-process theorists. As noted earlier (section 2.2), the existence of virtue ethics, the ethics of care, and other moral frameworks suggests that any description of moral processing cannot be mapped neatly onto just two moral systems without the a priori exclusion of many other potentially valid moral systems.

In their inferential critiques of Greene (Reference Greene2003, Reference Greene and Sinnott-Armstrong2008) and Singer (Reference Singer2005), Kumar and Campbell (Reference Kumar and Campbell2012) hone in on another major vulnerability of error theories. As they point out, any assumption about which of two conflicting views results from ‘misfiring’ (Sunstein Reference Sunstein2005: 541) can only be established on a priori, theoretical grounds. Therefore, neuroscience cannot identify what counts as an error in moral processing. However, Kumar and Campbell (Reference Kumar and Campbell2012) themselves also make a problematic assumption, albeit a much subtler one. Their negative error theory relies on the presupposition that whenever a person draws conflicting moral conclusions in equivalent cases, at least one of those conclusions must be mistaken. But this presupposition is problematic for at least two reasons.

First, it is not clear how we should define equivalent cases in a moral context. Like Greene, Kumar and Campbell treat moral prospects as roughly equivalent whenever they involve equivalent economic prospects—for example, number of lives lost and saved (Tversky and Kahneman Reference Tversky and Kahneman1981). But as Sunstein (Reference Sunstein2005) has pointed out, the link between economic prospects (net lives lost) and normative ethical behavior (obligatory, permissible, and forbidden acts) is far murkier than the link between economic prospects and normative economic behavior.Footnote ⁸

Second, even if we grant that some pair of moral judgments presents a legitimate ethical conflict, it does not follow that at least one of these judgments must be wrong. Some philosophers, encouraged by evidence from cross-cultural studies and other research, suggest that not all of the potential building blocks of normative moral systems fit together (Prinz Reference Prinz2008). If these philosophers are right, then any coherent system of moral values will, by necessity, have to leave out certain incommensurable values that nonetheless would be appropriate in, and arguably necessary for, other equally valid systems of moral values (Wong Reference Wong2006). Thus, two mutually exclusive moral judgments may both be valid because the appropriate system of moral values to adopt in any given situation may vary. Furthermore, even if two conclusions—moral or otherwise—really do strictly contradict one another, it may not necessarily be wrong to endorse both (Priest Reference Priest1998).

Kamm (Reference Kamm2009) has made a point similar to that made by Meyers (Reference Meyers2015), and has suggested a methodological cure to what she believes ails Greene's project. Poor design of moral dilemmas used as experimental probes, she alleges, is ultimately what hamstrings Greene's attempted inferences. She gestures at a number of ways these probes might be improved in future work, but her suggestions are hampered by a reliance on the unrefined and inaccurate notion that there exist emotional and rational ‘centers of the brain’ (Kamm Reference Kamm2009: 332).

An especially damning methodological objection has been laid out by Berker (Reference Berker2009). He points out that many of Greene's probes—including one asking whether it is morally permissible to hire someone to rape your wife so that you can then console her (Greene et al. Reference Greene, Sommerville, Nystrom, Darley and Cohen2001)—do not seem to present moral dilemmas at all.Footnote ⁹ Gray further suggests that even if Greene's moral dilemmas were well-constructed, the resolution of moral dilemmas is so atypical a function of moral judgment that studying it cannot teach us much about ordinary moral processing. Gray insists that ‘moral dilemmas in which wrongness (deontology) and harm (utilitarianism) conflict are unrepresentative of typical moral cognition’ (Gray et al. Reference Gray, Schein and Ward2014: 1600). Importantly, Greene's exact probes and others like them have been used many times by many researchers since Reference Greene, Sommerville, Nystrom, Darley and Cohen2001. Thus, Berker's and Gray's criticisms, if apt, could have a far-reaching implications.

One more methodological concern about error theories—and about NMN theories more generally—is the lack of ecological validity in the experiments by which they are often defended. For instance, many studies in NMN—especially those conducted by error theorists—use ‘trolley problems’ (Thomson Reference Thomson1976) and other contrived dilemmas like them. Much has been written questioning the legitimacy of treating abstruse cases like these as paradigms of moral decision making (e.g., Meyers Reference Meyers2015). And as just noted, Gray has argued that all moral dilemmas are so unusual that not only their philosophical but also their psychological relevance may be doubted.

4.1. Schematic Neutrality

The most obvious step in producing more robust work in NMN may be to strip the work of a priori normative schemes to whatever extent possible. As easy as this may be to recognize, it will be difficult to achieve. Indeed, the sophistication of many arguments against NMN reveals just how hard it can be even to recognize that one is making normative assumptions in this pioneering subdiscipline. As illustrated in my critique of Kumar and Campbell (Reference Kumar and Campbell2012) above (section 3.2), settling on a normative ethical scheme in which to interpret moral neuropsychological findings actually involves two assumptions, not just one. There is the more obvious assumption regarding which scheme should be adopted, and then there is the more insidious assumption that only one scheme can be right.

Alternatively, some work within NMN might successfully proceed with normative schematic assumptions so long as such work is tagged with the caveat that it will, by necessity, be relatively parochial. There is nothing wrong with deriving normative principles from, for example, Gigerenzer's cognitive science of moral heuristics (section 2.1) as long as it is acknowledged that these normative principles are contingent on certain assumptions about the nature of normative ethics—in Gigerenzer's case, on the presupposition of consequentialism (section 2.2).

It should also be noted that normative schematic presuppositions in NMN are not limited to moral presuppositions. Psychological presuppositions have colored NMN just as much as ethical and metaethical ones have. As discussed in my analysis of Greene's work (section 3.1), normative psychological assumptions, such as the assumption that controlled cognition is better than automatic processing, have also at times infected NMN. Of course, until the end of the nineteenth century, psychology fell entirely within the auspices of philosophy, so in some sense the psychological landscape of NMN has long been shaped—for better or for worse—by traditional moral theorists, including Hume and Kant. Thus, the sort of schematic neutrality I suggest is rather sweeping and might admittedly make it more difficult to draw normative conclusions on the basis of neuroscientific research. Nonetheless, I think it is an important principle to stick to: NMN should not presume the authoritativeness of any one ethical, metaethical, or psychological theory any more than is necessary for theoretic scaffolding. Furthermore, the limitations incurred by such schematic presumptions should be acknowledged and discussed in detail in those philosophical works that are not schematically neutral.

4.2. Psychological Minimalism

I have just argued that a priori evaluative stances toward psychological states—such as what many perceive to be Greene's a priori evaluation of controlled cognition as better than emotion—are a major (though perhaps inevitable) vulnerability of any NMN project of which they are a part. Such evaluative assumptions open projects to certain obvious interpretive criticisms. Thus, a very basic tenet to which NMN theorists might wish to adhere is that the normative superiority of certain psychological states over others—such as controlled cognition over emotion—should not be assumed whenever such assumptions can be avoided. Here, however, I want to discuss how even descriptive psychological claims within NMN can sometimes be deeply problematic.

Recent work has cast serious doubt on our ideas about which neural substrates are involved in which psychological processes (see, e.g., Klein Reference Klein2011). In fact, some authors believe that no emotion-cognition dichotomy may be plausible at all (e.g., Dolan Reference Dolan2002), at least when it comes to classifying neural substrates (Holtzman Reference Holtzman2018). And even if such a dichotomy may be plausible in some contexts, Gray and Schein (Reference Gray and Schein2012) provide reasons to doubt its plausibility within—or at least its relevance to—the moral domain (section 3.2). As such, it may be best, at least until neuroscience advances further, for NMN not even to attempt to distinguish emotional from controlled cognitive activity on the basis of the activation of specific neural substrates.

Gray's work (section 2.1) also reflects a more general problem with characterizing neurophysiological data in psychological terms. Like Young and her colleagues (Reference Young, Cushman, Hauser and Saxe2007; Young and Saxe Reference Young and Saxe2008), Gray focuses on the perception of other minds as the core of moral judgment. But in doing so, Gray assumes that the involvement of brain regions that have been associated with certain psychological processes in some experimental paradigms demonstrates the involvement of those same psychological processes in different (moral) experimental paradigms. However, this kind of abductive reasoning, which in neuroscience is known as reverse inference, is deeply problematic (Poldrack Reference Poldrack2006). There is not a one-to-one or even a one-to-many relationship between neural activity and psychological states and processes; that relationship is more likely many-to-many (Longino Reference Longino2013). Even the implications of double dissociationsFootnote ¹⁰ between neural activity and psychological processes are uncertain (Van Orden et al. Reference Van Orden, Pennington and Stone2001), and so it should be clear that single associations between neural activity and psychological processes do not prove the engagement of those processes whenever that activity is observed (Jacoby Reference Jacoby1991).

Thus, not only moral-philosophical but also psychological inferences from neuroscientific findings should be made cautiously and scrutinized carefully to ensure that such inferences are fully warranted. This is not to say that we should (or even meaningfully could) do away with all such inferences and assumptions. Still, we might adopt a somewhat conservative version of the ideas just developed, as follows: Psychological characterizations of neuromoral processing should be kept to a minimum and flagged as such whenever possible.

4.3. Methodological Integration

Whereas schematic neutrality and psychological minimalism are oriented (to different degrees) toward overcoming interpretive objections to NMN, this last recommendation is directed primarily against methodological objections. To begin, we should acknowledge that NMN could benefit greatly from a shift toward more ecologically valid work. Such work is especially challenging to design and interpret because of the number of uncontrolled variables and potential ambiguities it can involve. However, if NMN theorists hope to draw normative philosophical conclusions from the neuroscience of moral judgment, it may be especially incumbent upon them to grapple with these challenges and to work within the limitations that ecological validity may place on the interpretation of their work. Although Kamm (Reference Kamm2009) is correct in noting that many experimental prompts lack philosophical rigor in their design, Berker (Reference Berker2009), Gray and colleagues (Reference Gray and Schein2012), and others are right to point out that any use of moral dilemmas or of contrived moral vignettes may put researchers ‘on the wrong track’ (Kahane Reference Kahane2012: 519). While both ethical and practical considerations may necessitate that the majority of new research in NMN falls short of perfect ecological validity, keeping this ideal in mind may nonetheless prove valuable.

While increased philosophical scrutiny of experimental materials may be necessary, vignette-driven lab research, if not supplemented by other kinds of research, remains at risk to miss the forest for the trees. In addition to figuring out how to refine the sorts of experiments NMN researchers have been conducting, a broader reflection on the kinds of research methods NMN should use is in order. The third tradition could benefit especially from studying cases in which between-group differences (and similarities) are incontrovertible, in which observable behavioral outcomes obviate notoriously troublesome (Friborg et al. Reference Friborg, Martinussen and Rosenvinge2006) Likert-type self-reports,Footnote ¹¹ and in which the decisions participants make are both realistic and typical of those that people encounter in their daily lives. Additionally, increased study of (im)moral behavior in social situations, rather than among isolated individuals, may be especially valuable, given the social nature of moral decision making.

There is already a great deal of literature in fields cognate to NMN whose potential normative-ethical implications has yet to be explored by neuroethicists. Such work covers topics ranging from the apparent potential of opioid antagonists to treat kleptomania (Grant and Kim Reference Grant and Kim2002) to the automatic down-regulation of affective inhibitory responses among surgeons viewing images of patients that the rest of us would find extremely aversive (Decety et al. Reference Decety, Yang and Cheng2010). In addition to suggesting new directions for future NMN studies, extant work may on its own provide meaningful insights and ideas for those working within the third tradition. By way of summary: Insights into the nature and implications of neuromoral processing should be sought using all the tools at our disposal, which means significant diversification of our methods and the inclusion of methods and findings from other fields.