The Normative Insignificance of Neuroscience Selim Berker Harvard University sberker@fas.harvard.edu [penultimate draft of a paper that was eventualy published in Philosophy & Public Afairs 37 (209): 293-329; please cite that version] I. Introduction No doubt when historians of science lok back on the first decade of the twenty-first century, they wil dub it ?The Age of the fMRI.? Functional magnetic resonance imaging has revolutionized the empirical study of the human mind, leading to far-reaching changes in the research paradigms of psychology, economics, and (especialy) the burgeoning field of neuroscience; by one estimate, an average of eight per-reviewed articles employing fMRI were published per day in 207. 1 So perhaps it was inevitable that empiricaly minded philosophers would take some of these fMRI studies to have profound implications for philosophy. Inded, it has recently ben argued that the ground-breaking research by psychologist Joshua D. Grene and coleagues into the neural bases for our moral intuitions should lead us to change our opinions about the trustworthiness of those intuitions. Crudely put, Grene and his coleagues think there are two waring subsystems underlying our moral intuitions: the first makes use of emotional neural processes and generates the sorts of judgments typicaly associated with deontological positions in ethics; the second makes use of more cognitive neural processes and generates the sorts of judgments typicaly associated with utilitarian/consequentialist positions in ethics; and the two subsystems duke it out for one?s overal moral verdict about a given case. 2 By itself, this claim is merely an empirical hypothesis about what, as a mater of fact, causes us to make the moral judgments that we do make. However, Peter Singer and Grene himself have argued that this empirical hypothesis, if true, would also yield conclusions about the sorts of moral judgments that we should make. In particular, Singer and Grene think that the truth of Grene?s empirical hypothesis would give us god grounds to discount our deontological intuitions about 1 Jonah Lehrer, ?Picture Our Thoughts: We?re Loking for To Much in Brain Scans,? The Boston Globe (August 17, 208). 2 Joshua D. Grene, R. Brian Somervile, Leigh E. Nystrom, John M. Darley, and Jonathan D. Cohen, ?An fMRI Investigation of Emotional Engagement in Moral Judgment,? Science 293 (201): 2105-08; Joshua D. Grene, Leigh E. Nystrom, Andrew D. Engel, John M. Darley, and Jonathan D. Cohen, ?The Neural Bases of Cognitive Conflict and Control in Moral Judgment,? Neuron 4 (204): 389-40; and Joshua D. Grene, Sylvia A. Moreli, Kely Lowenberg, Leigh E. Nystrom, and Jonathan D. Cohen, ?Cognitive Load Selectively Interferes with Utilitarian oral Judgment,? Cognition 107 (208): 14-54. 2 cases, but not to discount our utilitarian/consequentialist intuitions about cases. 3 In this article I wish to scrutinize this last claim. More specificaly, I wil argue that once we separate the bad arguments for why Grene et al.?s empirical research has normative implications from the beter arguments for that conclusion, we can see that the neuroscientific results are actualy doing no work in those beter arguments. Or to put my central contention most provocatively: either atempts to derive normative implications from these neuroscientific results rely on a shody inference, or they apeal to substantive normative intuitions (usualy about what sorts of features are or are not moraly relevant) that render the neuroscientific results irelevant to the overal argument. However, my conclusions here are not entirely negative: although I am skeptical about the prospects for deriving normative implications from neural facts about how we hapen to reach moral verdicts, in the article?s final section I sketch a way in which neuroscience could play a more indirect role in sculpting our normative conclusions. It should be clear that much is at stake in this debate. Obviously if Grene?s and Singer?s arguments for why we should privilege our consequentialist intuitions over our deontological ones were sound, there would be far-reaching implications for contemporary debates in first-order ethics. But the implications are even wider than that. So far the only sorts of philosophical intuitions that have ben systematicaly studied using brain-imaging technology have ben moral intuitions about cases, and even then moral intuitions about only a smal class of cases. However, it is only a mater of time before fMRI- based studies of other varieties of philosophical intuitions are conducted?only a mater of time before someone, somewhere, studies what parts of the brain light up when the typical person has intuitions about general phenomena such as knowledge or fre wil, or about specific puzles such as Newcomb?s problem or the sorites paradox. One can almost see how this research wil go. First, no doubt, someone wil hypothesize that there are two separate systems vying for one?s overal verdict about whether a given hypothetical scenario counts as an instance of knowledge, or causation, or fre action. Then, no doubt, someone (possibly the same person) wil conclude that the empirical evidence for this hypothesis gives us 3 Peter Singer, ?Ethics and Intuitions,? The Journal of Ethics 9 (205): 31-52; Joshua D. Grene, ?From Neural ?Is? to Moral ?Ought?: What Are the Moral Implications of Neuroscientific Moral Psychology?? Nature Reviews Neuroscience 4 (203): 847-50; and Joshua D. Grene, ?The Secret Joke of Kant?s Soul,? in Moral Psychology, Vol. 3: The Neuroscience of Morality: Emotion, Brain Disorders, and Development, ed. Walter Sinot-Armstrong (Cambridge, as.: MIT Pres, 208), p. 35-79. 3 god reason to discount the verdicts of one of those systems but not the other. So if the sorts of arguments ofered by Grene and Singer are sucessful, they have the potential to radicaly alter how we go about adjudicating whether philosophical intuitions of a given sort are reliable?and, by extension, to radicaly alter the methodology with which we go about arguing for first-order philosophical claims. Enough speculation, though, about what sorts of empirical research may or may not be conducted in the future, and about what sorts of philosophical arguments may or may not be ofered on the basis of that research. My task here wil be to focus on the neuroscientific research that has ben conducted into the physiological basis for our moral intuitions about hypothetical cases, and on what the normative implications of that research might be. Thus the first order of business wil be to sumarize the essential details of Grene et al.?s research. I. Neuroscientific Results Grene and his coleagues chose to focus their empirical studies on our moral intuitions about a certain class of cases made famous by Philipa Fot and Judith Jarvis Thomson. 4 Consider the folowing scenario (here I use the exact wording employed by Grene et al. in their studies): troley driver dilema: ?You are at the whel of a runaway troley quickly aproaching a fork in the tracks. On the tracks extending to the left is a group of five railway workmen. On the tracks extending to the right is a single railway workman. If you do nothing the troley wil proced to the left, causing the deaths of the five workmen. The only way to avoid the deaths of these workmen is to hit a switch on your dashboard that wil cause the troley to proced to the right, causing the death of the single workman. Is it apropriate for you to hit the switch in order to avoid the deaths of the five workmen?? 5 Assuming that our topic is moral apropriateness, most people judge that it is apropriate to hit the switch. However, contrast that case with the folowing (again, the wording is Grene et al.?s): 4 Se Philipa Fot, ?The Problem of Abortion and the Doctrine of Double Efect,? Oxford Review 5 (1967): 5-15; Judith Jarvis Thomson, ?Kiling, Leting Die, and the Troley Problem,? The Monist 59 (1976): 204-17; and Judith Jarvis Thomson, ?The Troley Problem,? Yale Law Journal 94 (1985): 1395-415. Al thre articles are reprinted in Ethics: Problems and Principles, ed. John Martin Fischer and Mark Raviza (Fort Worth, Tex.: Harcourt Brace Jovanovich, 192). 5 Grene et al., ?An fMRI Investigation,? suplementary material (available at ). 4 fotbridge dilema: ?A runaway troley is heading down the tracks toward five workmen who wil be kiled if the troley proceds on its present course. You are on a fotbridge over the tracks, in betwen the aproaching troley and the five workmen. Next to you on this fotbridge is a stranger who hapens to be very large. The only way to save the lives of the five workmen is to push this stranger of the bridge and onto the tracks below where his large body wil stop the troley. The stranger wil die if you do this, but the five workmen wil be saved. Is it apropriate for you to push the stranger on to the tracks in order to save the five workmen?? 6 Most people judge that it is not apropriate to push the large stranger. What explains this diference in our moral judgments? On the one hand, it might seem puzling that a majority of people judge diferently about these two cases, since in each what is at stake is the life of five people versus the life of one. But on the other hand, there are myriad diferences betwen these two scenarios that could (it might be thought) explain why most of us make a moral distinction betwen them. The task of trying to fix on a moraly relevant feature of these two scenarios that explains why we are justified in giving difering verdicts about them has come to be known as the troley problem. 7 Usualy the presuposition of this literature is that our moral intuitions about these cases (and others of their ilk) are largely acurate, the goal being to find a plausible moral principle that both agres with and explains our intuitive verdicts about the cases in question. But what makes the troley problem so hard?inded, what has led some to despair of our ever finding a solution to it?is that for nearly every principle that has ben proposed to explain our intuitions about troley cases, some ingenious person has devised a variant of the 6 Ibid. 7 Actualy, this isn?t entirely corect. The troley problem is usualy taken to be, not the problem of explaining our difering verdicts about the fotbridge and troley driver dilemas, but rather the problem of explaining our difering verdicts about the fotbridge dilema and a variant of the troley driver dilema in which you are a bystander who ses the runaway troley and can hit a switch that wil divert the troley onto the sidetrack containing the one person. Inded, Thomson (who introduced the term ?troley problem? into the philosophical lexicon) thinks there is no problem explaining the diference in our intuitive reactions to the troley driver and fotbridge dilemas; for Thomson (and for others folowing her), the real problem is explaining what grounds our diferent judgments about the bystander and fotbridge dilemas. And though Singer?s sumary of Grene et al.?s research sugests that it was the bystander dilema that was tested (se Singer, ?Ethics and Intuitions,? p. 39), and though Grene himself, when describing his research, almost always sumarizes the troley driver dilema in a way that is ambiguous betwen the driver and bystander variants (se Grene et al., ?An fMRI Investigation,? p. 2105; Grene et al., ?Neural Bases,? p. 389; and Grene, ?Secret Joke,? p. 41-42), it is worth pointing out that in al of the published studies I discus in this article, it was only the driver, not the bystander, version of the standard troley dilema that was studied. Perhaps it is being asumed that our judgments about the driver and bystander cases (and their neural corelates) wil be the same; however, many philosophers mark a distinction betwen these two cases, and in her most recent discusion of the troley problem (?Turning the Troley,? Philosophy & Public Afairs 36 [208]: 359-74), Thomson argues that although it is permisible to divert the troley if one is the driver, it is impermisible to divert the troley if one is a bystander. (Thus, on Thomson?s curent way of seing things, there actualy is no troley problem, since the very formulation of that problem contains a false presuposition that there is a moraly relevant diference betwen the bystander and fotbridge cases, but no moraly relevant diference betwen the bystander and driver cases.) 5 classic troley scenario for which that principle yields counterintuitive results. Thus as with the Getier literature in epistemology and the causation and personal identity literatures in metaphysics, increasingly baroque proposals have given way to increasingly complex counterexamples, and though some have continued to strugle with the troley problem, many others have simply given up and moved on to other topics. 8 Rather than deal with the normative task of proposing principles that justify our responses to troley- like cases, Grene and his coleagues decided to pursue the descriptive task of investigating the physiological processes that underlie our responses to these sorts of cases. Their central empirical hypothesis requires making two distinctions: first a distinction betwen two diferent classes of moral judgments, and second a distinction betwen two diferent classes of psychological processes. The distinction betwen classes of moral judgments is as folows. Notice that a judgment that it is moraly permissible to hit the switch in the troley driver dilema is precisely the sort of verdict predicted by a utilitarian or, more generaly, consequentialist moral framework: since one?s hiting the switch presumably results in a state of afairs with greater agregate wel-being than the state of afairs that would result were one not to hit the switch, acording to most forms of consequentialism one is moraly required?and hence permited?to hit the switch. Folowing Grene, let us cal particular-case moral judgments that, like the judgment that it is moraly permissible to hit the switch in the troley driver dilema, are ?easily justified in terms of the most basic consequentialist principles? characteristicaly consequentialist judgments. 9 One judgment that is not characteristicaly consequentialist is the judgment that it is moraly impermissible to push the overweight individual to his demise in the fotbridge dilema: precisely what sets deontological moral theories apart from consequentialist ones is that deontological theories tend to yield the result that it is impermissible to kil another person in this way, even for the sake of ?the greater god.? So, folowing Grene again, let us cal particular-case moral judgments that, like the judgment that it is moraly impermissible to push the 8 For a survey of the early clasics of the troley problem literature, se the papers colected in Fischer and Raviza (eds.), Ethics: Problems and Principles. More recent clasics not included in that anthology are Judith Jarvis Thomson, The Realm of Rights (Cambridge, Mas.: Harvard University Pres, 190), chap. 7; F. M. Kam, Morality, Mortality, Vol. I: Rights, Duties, and Status (Oxford: Oxford University Pres, 196), chaps. 6-7; and F. M. Kam, Intricate Ethics (Oxford: Oxford University Pres, 207), chaps. 1-6. 9 Grene, ?Secret Joke,? p. 39. 6 obese man in the fotbridge dilema, are ?in favor of characteristicaly deontological conclusions? characteristicaly deontological judgments. 10 This gives us a two-fold distinction betwen types of particular-case moral judgments, which is usualy taken to corespond to a two-fold distinction betwen types of particular-case moral intuitions. 1 There are a number of reasons to be woried about this distinction betwen characteristicaly consequentialist and characteristicaly deontological moral judgments, but I want to put them aside for the time being so that we can get Grene et al.?s empirical hypothesis on the table. The second distinction upon which that hypothesis depends is a distinction betwen two kinds of psychological processes: emotional proceses and ?cognitive? proceses. (Folowing Grene et al.?s useful convention, 12 I use scare quotes when I mean ?cognitive? to refer to specificaly non-emotional information processing, in contrast to the also widespread use of ?cognitive? to refer to information processing in general, as it does in the phrase ?cognitive science.?) Exactly how to flesh out the emotional versus ?cognitive? process distinction is a contentious mater, so it is worth noting that Grene et al. use ?emotional processing? to refer to information processing that involves behavioraly valenced representations that triger automatic efects and hence ?have direct motivational force,? 13 and they use ??cognitive? processing? to refer to information 10 Ibid. Note that these two definitions are not paralel. As Grene uses the expresions, ?characteristicaly consequentialist judgment? means ?judgment suported by the sort of moral principle that typicaly distinguishes consequentialist theories from deontological ones,? whereas ?characteristicaly deontological judgment? means ?judgment in favor of the sort of verdict that typicaly distinguishes deontological theories from consequentialist ones.? (The contrast is at the level of suporting principles in the one case, at the level of particular judgments in the other.) Thus even though nearly al deontologists judge that it is permisible to divert the troley in the troley driver dilema, such a judgment counts as characteristicaly consequentialist but does not count as characteristicaly deontological. 1 In philosophical discusions of the metaphysics and epistemology of intuitions, there is an ongoing debate over whether intuitions just are judgments arived at in a particular way (for example, not as a result of explicit reasoning, testimony, and so on), or whether intuitions are a separate clas of mental entities that stand to intuitive judgments as perceptual experiences stand to perceptual judgments. For the former view, se Alvin Goldman and Joel Pust, ?Philosophical Theory and Intuitional Evidence,? in Rethinking Intuition, ed. Michael R. DePaul and Wiliam Ramsey (Lanham, Maryland: Rowman & Litlefield, 198), p. 179-97; for the later, se George Bealer, ?Intuition and the Autonomy of Philosophy,? in the same volume, p. 201-39. We ned not take a stand on this debate here, since even if moral intuitions are separate entities over and above the moral judgments formed on their basis, there wil usualy be an intuitive moral judgment coresponding to each moral intuition. Thus either we can say (if we identify intuitions with intuitive judgments) that the experiments in question directly study our moral intuitions, or we can say (if we distinguish intuitions from intuitive judgments) that the experiments indirectly study our moral intuitions by colecting data on our moral judgments, which are taken to be tightly corelated with our moral intuitions. In what folows I wil generaly be fairly lax in sliding back and forth betwen talk of judgments and talk of intuitions. (That said, I am not using ?intuition? as that term is used in much of the psychology literature, where it refers to any sort of automatic, spontaneous ?gut feling? that one might have. Se most of the studies cited in David G. Myers, Intuition: Its Powers and Perils [New Haven, Con.: Yale University Pres, 202] for this sort of usage, which has litle to do with what philosophers mean when they talk about intuitions.) 12 Se Grene et al., ?Neural Bases,? p. 389; and Grene, ?Secret Joke,? p. 40. 13 Grene et al., ?Neural Bases,? p. 397. 7 processing that involves ?inherently neutral representations . . . that do not automaticaly triger particular behavioral responses or dispositions.? 14 Emotional processes tend to be fast and frugal (providing quick responses on the basis of a limited amount of information), and they tend to be domain-specific (responding to particular subject maters, rather than any subject mater in general). By contrast, ?cognitive? processes tend to be slow but flexible and domain-neutral, and it has ben found that, at least in non-moral cases, ?cognitive? processes are recruited for such things as abstract reasoning, problem solving, working memory, self-control, and higher executive functions more generaly. 15 Regions of the brain that have ben associated with emotional processing include the folowing: the medial prefrontal cortex (Brodman?s Area [BA] 9/10), the posterior cingulate/precuneus (BA 31/7), the posterior superior temporal sulcus/inferior parietal lobe (BA 39), the orbitofrontal/ventromedial prefrontal cortex (BA 10/1), and the amygdala. Regions of the brain that have ben associated with ?cognitive? processing include the dorsolateral prefrontal cortex (BA 9/10/46) and the parietal lobe (BA 7/40). Thus we have a two-fold distinction betwen types of moral judgments (characteristicaly consequentialist ones vs. characteristicaly deontological ones) and a two-fold distinction betwen types of psychological processes (emotional ones vs. ?cognitive? ones). The natural question to ask is: which sort (or sorts) of processes underlie each sort of judgment? The proposal put forward by Grene and his coleagues is as folows: Grene et al.?s dual-proces hypothesis: Characteristicaly deontological judgments are driven by emotional processes, whereas characteristicaly consequentialist judgments are driven by ?cognitive? processes, and these processes compete for one?s overal moral verdict about a given case. 16 In one way, this is an extremely old picture of how the moral mind works. The idea that reason and passion strugle for one?s overal moral stance was already comonplace by the time Hume wrote the Treatise. ?Nothing is more usual in philosophy, and even in comon life, than to talk of the combat of passion and reason,? Hume tels us (Treatise 2.3.3), before going on to argue against this Combat Model of 14 Grene, ?Secret Joke,? p. 40. 15 The qualification ?at least in non-moral cases? is crucial here: to say at the outset that ?cognitive? proceses handle abstract reasoning and problem solving in al domains (including the moral) is question beging. 16 Se Grene et al., ?Neural Bases,? p. 398; Grene et al., ?Cognitive Load,? p. 145; and Grene, ?Secret Joke,? p. 40-41. 8 the soul (as Christine Korsgard usefuly dubs it). 17 So Grene et al.?s embracing of the Combat Model is not new. 18 What is new, though, is the surprising twist that they give to that model. Whereas deontology is usualy associated with reason and consequentialism with the sentiments, Grene and his coleagues claim that in fact the oposite is true. On their picture, when we contemplate hiting the switch in the troley driver dilema, our more ?cognitive? brain processes perform a col and detached cost-benefit analysis, yielding the verdict that it is permissible to hit the switch, but when we contemplate pushing the heavy felow in the fotbridge dilema, our more emotional brain processes kick in and scream at us, ?Don?t do that!? thus overiding the cost-benefit analysis that would have demed it permissible to push the man. Grene et al.?s dual-process hypothesis is an empirical hypothesis, and as such it yields a number of empirical predictions. Here I mention just two. First, the dual-process hypothesis predicts that contemplation of cases like the fotbridge dilema should produce increased neural activity in regions of the brain associated with emotional processes, whereas contemplation of cases like the troley driver dilema should produce increased neural activity in regions of the brain associated with ?cognitive? processes. Second, the dual-process hypothesis predicts that people who reach a non-standard verdict about cases like the fotbridge dilema should take longer to reach their verdict than those who reach a standard verdict (since they are overiding an emotional response in order to come to their final verdict), whereas people who reach a non-standard verdict about cases like the troley driver dilema should take aproximately as long to reach a verdict as those who reach a standard verdict (since no emotional response is being overiden). Because the results from fMRI machines are statisticaly noisy, Grene and his coleagues could not test these predictions simply by comparing people?s reactions to the fotbridge and troley driver dilemas. Instead, they neded to compare the neural-activity and response-time results when people 17 Se Christine Korsgard, ?Self-Constitution in the Ethics of Plato and Kant,? Journal of Ethics 3 (199): 1-29. Korsgard contrasts the Combat Model of the soul with an alternate Constitution Model that she finds in Plato and Kant. 18 Though perhaps it is new to empirical psychology: as Grene et al. tel the story (?Neural Bases,? p. 397-98; ?Cognitive Load,? p. 145), empirical psychology went through a long period where, under the influence of Lawrence Kohlberg, it was widely believed that proceses of reasoning underwrite the moral judgments of mature adults, folowed by a more recent period in which Jonathan Haidt and others have proposed that moral judgments are primarily driven by automatic, emotional proceses. Se Lawrence Kohlberg, ?Stage and Sequence: The Cognitive-Developmental Aproach to Socialization,? in Handbok of Socialization Theory and Research, ed. David A. Goslin (Chicago, Il.: Rand McNaly, 1969), p. 347-480; and Jonathan Haidt, ?The Emotional Dog and Its Rational Tail: A Social Intuitionist Aproach to oral Judgment,? Psychological Review 108 (201): 814-34. 9 responded to a large number of cases that are ?like the fotbridge dilema? (and hence give rise to deontological judgments) and a large number of cases that are ?like the troley driver dilema? (and hence give rise to consequentialist judgments). But this leads to a problem: how do we sort the former class of dilemas from the later? Grene et al. setled on the folowing ?purely descriptive? 19 way of doing so: cases that are like the fotbridge dilema were demed to involve harm that is brought about in an ?up close and personal? way, whereas cases that are like the troley driver dilema were demed to involve harm that is brought about in an impersonal way. What, though, does it mean for harm to be brought about in an ?up close and personal? way? Here is Grene et al.?s ?first cut? proposal: a moral dilema counts as personal if and only if ?the action in question (a) could reasonably be expected to lead to serious bodily harm (b) to a particular person or a member or members of a particular group of people (c) where this harm is not the result of deflecting an existing threat onto a diferent party.? 20 The basic idea can be helpfuly sumarized in the slogan ?ME HURT YOU?: The ?HURT? criterion [= (a)] picks out the most primitive kinds of harmful violations (e.g., assault rather than insider trading) while the ?YOU? criterion [= (b)] ensures that the victim be vividly represented as an individual. Finaly, the ?ME? criterion [= (c)] captures a notion of ?agency,? requiring that the action spring in a direct way from the agent?s wil, that it be ?authored? rather than merely ?edited? by the agent. 21 Moral dilemas that were demed by a set of independent coders to met conditions (a), (b), and (c) were classified as personal moral dilemas, and al other moral dilemas were classified as impersonal moral dilemas. Then, working under the assumption that the personal versus impersonal moral dilema distinction tracks the dilema-giving-rise-to-a-deontological-judgment versus dilema-giving-rise-to-a- consequentialist-judgment distinction, Grene et al. used this way of divying up moral dilemas into two piles to test their neural-activity and response-time predictions. And what Grene and his coleagues found was truly remarkable. They had a number of subjects respond to aproximately 20 personal moral dilemas, aproximately 20 impersonal moral dilemas, 19 Grene, ?Secret Joke,? p. 43. 20 Grene et al., ?An fMRI Investigation,? p. 2107, n. 9. 21 Grene et al., ?Neural Bases,? p. 389. 10 and aproximately 20 non-moral dilemas while inside fMRI machines. 2 fMRI technology tracks the magnetic signature of oxygenated blod, which is widely taken to be a fairly acurate way of measuring the level of neural activity in diferent portions of the brain. Thus Grene et al. were able to test their predictions about neural activity. Moreover, they found that their predictions were largely borne out. When responding to personal moral dilemas, subjects exhibited increased activity in the folowing brain areas associated with emotional processes: the medial prefrontal cortex, the posterior cingulate/precuneus, the posterior superior temporal sulcus/inferior parietal lobe, and the amygdala. 23 And when responding to impersonal moral dilemas, subjects exhibited increased activity in two classicaly ?cognitive? brain regions, the dorsolateral prefrontal cortex and the parietal lobe. 24 Furthermore, during several trials Grene et al. measured their subjects? response time to each question, and they reported that their response-time prediction was also confirmed: although subjects who gave emotionaly incongruent answers to personal moral dilemas tok almost two seconds longer, on average, to respond than those who gave emotionaly congruent responses, there was no comparable efect for impersonal moral dilemas. 25 Al told, Grene et al.?s empirical results present an impressive case for their dual-process hypothesis. 26 2 In experiment 1 in Grene et al., ?An fMRI Investigation,? nine subjects responded to 14 personal moral dilemas, 19 impersonal moral dilemas, and 20 non-moral dilemas, presented in random order, while inside fMRI machines. (For the exact wording of the dilemas that were used, se the suplementary material at .) Each dilema was presented as thre screns of text, after which the subject was required to give a verdict about the apropriatenes of a proposed action by presing one of two butons (?apropriate? or ?inapropriate?). Subjects were alowed to advance to each subsequent scren of text at their own rate, though they were given a maximum of 46 seconds to read through al thre screns and respond. In experiment 2 in that same article, nine diferent subjects responded to 2 personal moral dilemas, 19 impersonal moral dilemas, and 20 non-moral dilemas, using the same protocol. (The set of personal moral dilemas was altered in experiment 2 to remove a posible confound in the experimental design: se p. 2108, n. 24.) In Grene et al., ?Neural Bases,? 32 new subjects responded to 2 personal moral dilemas, 18 impersonal moral dilemas, and 20 non-moral dilemas, using the same protocol. (These were the same dilemas used in experiment 2 of ?An fMRI Investigation,? except one of the impersonal moral dilemas was droped.) The data from these new subjects together with the data from the nine subjects from experiment 2 in ?An fMRI Investigation? were then analyzed together as a whole. 23 Se Grene et al., ?An fMRI Investigation,? p. 2106, fig. 1, and p. 2107; and Grene et al., ?Neural Bases,? p. 391 and p. 392, table 1. The superior temporal sulcus was originaly labeled ?angular gyrus? in the first study. Activity in the amygdala was not detected in the first study but was detected in the larger second study. Due to a ?magnetic susceptibility artifact,? neither study was able to image the orbitofrontal cortex, another brain area that has ben asociated with emotional procesing (se Grene et al., ?Neural Bases,? p. 2108, n. 21). 24 Se Grene et al., ?An fMRI Investigation,? p. 2106, fig. 1, and p. 2107; and Grene et al., ?Neural Bases,? p. 391 and p. 392, table 1. 25 Se Grene et al., ?An fMRI Investigation,? p. 2107, fig. 3. 26 In the body of this article I have focused on the neuroimaging and response-time findings, since these results are particularly vivid and tend to capture the public?s imagination. However, there have ben a number of folow-up studies which 1 II. Methodological Wories In general it is dangerous (and perhaps futile) for philosophers to resist empiricaly based chalenges by caling into question the methodology of the relevant experiments, or the interpretation of their results. Not only are philosophers often not wel trained at evaluating scientific studies, but they ned to be extremely careful that the (aleged) design flaws to which they point are not ones that could easily be overcome in future research, or ones that in the end are irelevant to the main philosophical issues at stake. 27 Nevertheless, I think it is worth bringing up thre empirical issues about Grene et al.?s research have ben taken to lend further suport to Grene et al.?s dual-proces hypothesis, including the folowing: ? In ?Neural Bases,? Grene et al. found that when subjects contemplated ?dificult? personal moral dilemas (where degre of dificulty was measured by response time), the anterior cingulate cortex (a brain region asociated with conflict monitoring) and the dorsolateral prefrontal cortex (a brain region asociated with abstract reasoning) exhibited increased activity, in adition to regions asociated with emotion. They also found that the level of activity in the dorsolateral prefrontal cortex was positively corelated with consequentialist responses to these ?dificult? personal moral dilemas. ? In a folow-up study, Grene and coleagues found that consequentialist responses to ?dificult? personal moral dilemas tok longer when subjects were required to perform a cognitively intensive task at the same time as responding to the dilemas, but deontological responses did not exhibit this efect. Se Grene et al., ?Cognitive Load.? ? Michael Koenigs, Liane Young, and coleagues found that patients with damage to their ventromedial prefrontal cortex (a brain region asociated with the emotions) gave a greater percentage of consequentialist responses to the personal moral dilemas from Grene et al.?s ?An fMRI Investigation? than control subjects did. Se Michael Koenigs, Liane Young, Ralph Adolphs, Daniel Tranel, Fiery Cushman, Marc Hauser, and Antonio Damasio, ?Damage to the Prefrontal Cortex Increases Utilitarian Moral Judgments,? Nature 46 (207): 908-1. ? Piercarlo Valdesolo and David DeSteno found that respondents who had watched a funy clip from Saturday Night Live were more likely to give a consequentialist response to the fotbridge dilema than those who had watched a clip from a dul documentary beforehand, but there was no comparable efect for the troley driver dilema. Se Valdesolo and DeSteno, ?Manipulations of Emotional Context Shape Moral Judgment,? Psychological Science 17 (206): 476-7. 27 One potential design wory about Grene et al.?s research that fals into the former category (i.e., wories that could easily be overcome in future research) is as folows. In order to have a comparison clas for their neural-activity data, Grene and his coleagues had their subjects respond to a number of non-moral dilemas, such as: turnips dilema: ?You are a farm worker driving a turnip-harvesting machine. You are aproaching two diverging paths. By chosing the path on the left you wil harvest ten bushels of turnips. By chosing the path on the right you wil harvest twenty bushels of turnips. If you do nothing your turnip-harvesting machine wil turn to the left. Is it apropriate for you to turn your turnip-picking machine to the right in order to harvest twenty bushels of turnips instead of ten?? (Grene et al., ?An fMRI Investigation,? suplementary material.) The triger question here was formulated in terms of apropriatenes to make it as paralel as posible to the triger questions for the moral dilemas tested, but one might wory that it sounds very od to ask whether it is ?apropriate? to turn a machine one way rather than the other. (Should we interpret this as some sort of prudential apropriatenes? Is there even such a notion?) Moreover, the answer to this so-caled dilema is completely obvious, and al told I estimate that of the 20 non-moral dilemas used in Grene et al.?s studies, 12 have completely obvious answers, 6 are somewhat les obvious, and only 2 are genuine dilemas; thus one might wory that to many of these non-moral ?dilemas? have readily evident answers for them to serve as an acurate comparison clas. However, both of these wories could easily be avoided in future research: the set of non-moral dilemas could be altered to include a greater number of dificult dilemas, and the triger question for the dilemas (both moral and non-moral) could be phrased in a les awkward way. (Inded, in their ?Damage to the Prefrontal Cortex Increases Utilitarian Moral Judgments,? Koenigs et al. used Grene et al.?s dilemas but rephrased the triger questions so as to ask whether the subjects ?would? perform the action in question, rather than asking whether they dem it ?apropriate? to perform the action, perhaps for this very reason. However, this rewording introduces new problems, since one?s judgments about what one would do and one?s judgments about what it is moraly permisible to do might pul apart.) One potential design wory about Grene et al.?s research that fals into the later category (i.e., wories that ultimately are 12 before turning to the more important question of what, in principle, the normative implications of this sort of research could be. These thre issues don?t entirely undermine the empirical findings, but they do cast them in a diferent light. Moreover, at least one of these issues wil be crucial to our discussion of the normative implications of Grene et al.?s research. First empirical issue: neural activity in at least one brain region asociated with emotion was found to be corelated with consequentialist judgment. In particular, Grene and his coleagues found that activity in the posterior cingulate, a portion of the brain known to be recruited for emotional processes, predicts characteristicaly consequentialist responses to personal moral dilemas (for example, a response that it is apropriate to push the portly gentleman in the fotbridge dilema). 28 Grene et al. concede that these results cast doubt on the simplest version of the dual-process hypothesis, acording to which consequentialist judgments are wholy tied to ?cognitive? processes and deontological judgments wholy tied to emotional processes. They write, ?Like David Hume . . . we suspect that al action, whether driven by ?cognitive? judgment or not, must have some afective basis.? 29 However, what is at stake here is whether al moral judgment, not al action, has an afective basis. Also, Hume is a dangerous aly to cal on at this point: Hume?s contention was not just that al moral judgments ?have some emotional component,? as Grene at one point contends Hume?s view to be, 30 but moreover that al moral judgments are entirely driven by the passions. On Hume?s picture, the strugle in our souls is not betwen reason and passion, with a few passions along for the ride on reason?s side; rather, the fundamental strugle is betwen diferent passions, with reason the underling carying out each passion?s whims. ?Reason is, and ought only to be slave of the passions, and can never pretend to any other ofice than to serve and obey them? (Treatise 2.3.3) is not a slight amendment of the dual-process hypothesis; it is a complete subverting of it. not relevant to the philosophical isues at stake) is as folows. Because of the limitations of fMRI technology, Grene and his coleagues were only able to study neural activity when subjects contemplated hypothetical scenarios. Thus their dual-proces hypothesis has, at least so far, not ben tested with regards to moral judgments about actual scenarios (whether curently perceived or merely remembered). However, these limited results are enough for the philosophical questions at isue. Even if we could only conclude from Grene et al.?s research that deontological judgments about hypothetical cases are not trustworthy but could not make a paralel claim with regards to deontological judgments about actual cases, that would stil be an extremely significant result, given the ubiquity of apeals to hypothetical cases in first-order moral theorizing. 28 Se Grene et al., ?Neural Bases,? p. 395, table 4, and p. 397. 29 Grene et al., ?Neural Bases,? p. 397. 30 Grene, ?Secret Joke,? p. 41. 13 In his own writings Grene tries to reinstate a contrast betwen the processes underlying deontological judgments and the processes underlying consequentialist judgments by proposing that the emotions that drive deontological judgments are ?alarm-like,? whereas those that are present during consequentialist judgments are ?more like a curency.? 31 However, this aleged contrast apears to have no empirical backing, short of an apeal by Grene to his own phenomenology when he considers cases like the troley driver and fotbridge dilemas. So in the end I think the best option for Grene and his coleagues is not to slide into a ful-blown Humean sentimentalism, nor to posit a speculative (and potentialy question-beging) phenomenological distinction betwen ?alarm-like? and ?curency-like? emotions, but rather to concede that at least one region of the brain traditionaly associated with the emotions apears to be recruited for characteristicaly consequentialist judgments and then to insist that exactly how crucial a role this region plays in such judgments is a topic for future research. Their finding with regards to the posterior cingulate mudies the waters a bit, but it ned not mean that the dual- process hypothesis is completely of target. Second empirical issue: when interpreted corectly, the response-time data colected by Grene et al. do not, in fact, confirm their prediction about comparative response times. Each of Grene et al.?s personal moral dilemas was aranged so that a response in which the subject dems it ?apropriate? to perform the proposed action (for example, apropriate to shove the hefty stranger in the fotbridge dilema) goes against the hypothesized emotional response, and a response in which the subject dems it ?inapropriate? to perform the proposed action (for example, inapropriate to shove the hefty stranger in the fotbridge dilema) is consonant with the hypothesized emotional response. In their original paper in Science, Grene et al. reported that responses of ?apropriate? to their various personal moral dilemas tok on average 6.8 seconds whereas responses of ?inapropriate? tok on average 5.0 seconds, 32 and they tok this to confirm their response-time prediction. 3 However, this is a statisticaly invalid way of presenting these results. A more acurate way of presenting the results would be to calculate the average diference in response time 31 Grene, ?Secret Joke,? p. 41; se also p. 64-65. 32 Grene et al., ?An fMRI Investigation,? p. 2107, fig. 3. 3 Grene et al., ?An fMRI Investigation,? p. 2107. 14 betwen the ?apropriate? responses and the ?inapropriate? responses for each question, and then to calculate the average of these average diferences. Such a way of calculating things prevents diferences from question to question in the average response time and diferences from question to question in the ratio of ?apropriate? to ?inapropriate? responses (two factors that should be irelevant to the truth of the hypothesis being tested) from skewing the final results. 34 As it turns out, this statistical point ends up matering. The group of personal moral dilemas that Grene and his coleagues tested were a motley bunch, including not only the fotbridge dilema, but also cases such as the folowing: architect dilema: ?You are a young architect visiting one of your construction sites with your boss. Your boss is a despicable individual who makes everyone around him miserable including you. It ocurs to you that if you were to push him of the building you are inspecting he would fal to his death and everyone would think it was an acident. Is it apropriate for you to push your boss of of the building in order to get him out of your life?? 35 hired rapist dilema: ?You have ben dissatisfied with your mariage for several years. It is your distinct impression that your wife no longer apreciates you. You remember how she apreciated you years ago when you tok care of her after she was muged. You devise the folowing plan to regain your wife?s afection. You wil hire a man to break into your house while you are away. This man wil tie up your wife and rape her. You, upon hearing the horible news, wil return swiftly to her side, to take care of her and comfort her, and she wil once again apreciate you. Is it apropriate for you to hire a man to rape your wife so that she wil apreciate you as you comfort her?? 36 So there is a wory that the diference betwen the average response times for al answers of ?apropriate? and the average response time for al answers of ?inapropriate? to personal moral dilemas was largely due to the fact that almost al respondents very quickly answered ?inapropriate? to cases like the 34 To se how these factors could skew the results on Grene et al.?s way of calculating things, consider this. Supose we tok the personal moral dilema that produced the greatest percentage of ?apropriate? answers, and then aded a large amount of filer text to the dilema so as to increase by some set amount the response time for any answer to it. Then?asuming that this change does not afect the ratio of ?apropriate? to ?inapropriate? responses for that dilema?our change wil have raised the average of al answers of ?apropriate? to personal moral dilemas (Grene et al.?s proposed metric) quite a bit more than it raises the average of al answers of ?inapropriate? to personal moral dilemas. However, the average of the average diferences in response time betwen ?apropriate? and ?inapropriate? response for each personal moral dilema (my proposed metric) would be unafected by such a change. 35 Grene et al., ?An fMRI Investigation,? suplementary material. 36 Ibid. 15 architect and hired rapist cases, which are hardly deserving of the epithet ?dilema.? And this was inded what hapened: in a later paper, Grene et al. admit that when their response-time data are analyzed with cases like the architect and hired rapist cases thrown out, they reveal ?no reliable diferences in RT [response time]? betwen those who gave a response of ?apropriate? and those who gave a response of ?inapropriate? to personal moral dilemas. 37 I bring up this issue because in presenting the empirical case for the dual-process hypothesis, Singer and Grene lean quite heavily on the response-time data. 38 Perhaps future research with larger sample sizes wil confirm some version of the response-time prediction; however, it is important to kep in mind that at this point in time the response-time prediction has not ben borne out, which in fact is an empirical strike against the dual-process hypothesis. 39 Third empirical isue: Grene et al.?s tentative criteria for sorting personal from impersonal moral dilemas are an inadequate way of tracking the dilema-giving-rise-to-a-deontological-moral-judgment versus dilema-giving-rise-to-a- consequentialist-moral-judgment distinction. To claim that characteristicaly deontological judgments only concern bodily harms is nothing short of preposterous; after al, the stock in trade of deontology is suposed to involve not just prohibitions on murder and mayhem, but also requirements against lying, promise-breaking, coercion, and the like. 40 But even within the realm of bodily harms, there are an 37 Grene et al., ?Cognitive Load,? p. 146, n. 5. Actualy, when they concede this, Grene et al. are woried about a slightly diferent isue: here they are woried that cases such as the architect and hired rapist dilemas should not be included in the analysis, since an answer of ?apropriate? to such dilemas does not corespond (or does not obviously corespond) to a consequentialist judgment about such a case. Though I share this wory (se my third empirical isue, below), my point here is somewhat diferent. Even if we tos out the response-time data from the architect and hired rapist dilemas, it is stil statisticaly invalid to compare the average response time of every answer of ?apropriate? to the average response time of every answer of ?inapropriate,? rather than averaging the average diferences in response time betwen answers of ?apropriate? and ?inapropriate? for each question. It is not clear to me whether Grene et al. now realize this point. On the one hand, in a more recent study in which they compare subjects? response times when responding to moral dilemas while performing a cognitively intensive task (se n. 26), Grene et al. continue to present their response-time data in the statisticaly invalid maner (se Grene et al., ?Cognitive Load,? p. 149, fig. 1, and p. 150, fig. 2). On the other hand, they write in that same study, ?This general patern also held when item, rather than participant, was modeled as a random efect, though the results in this analysis were not as strong? (ibid., p. 149). 38 Se Singer, ?Ethics and Intuitions,? p. 341-42; and Grene, ?Secret Joke,? p. 4. Se also Grene, ?Reply to Mikhail and Timons,? in Moral Psychology, Vol. 3, p. 105-17, at p. 109. 39 Recently, thre psychologists and one philosopher have reanalyzed Grene et al.?s data from ?An fMRI Investigation? and definitively established that Grene et al.?s response-time prediction was, in fact, disconfirmed by that data. Se Jonathan McGuire, Robyn Langdon, Max Coltheart, and Catriona Mackenzie, ?A Reanalysis of the Personal/Impersonal Distinction in oral Psychology Research,? Journal of Experimental Social Psychology 45 (209): 57-80. 40 Of course, exactly how we cash out these prohibitions/requirements wil vary from one deontological theory to the next. (That said, it is important to kep in mind that a Ten-Comandments-style ?Never, ever kil!? ?Never, ever lie!? etc. version of deontology is not the only option; inded, such a picture is rarely, if ever, defended outside of introductory ethics courses.) 16 abundance of clear counterexamples to Grene et al.?s proposal. To mention just one: Frances Kam?s famous lazy Susan case is a dilema giving rise to a consequentialist moral judgment, but it is demed a personal moral dilema by Grene et al.?s tripartite ?ME HURT YOU? criteria. In this case, a runaway troley is heading toward five inocent people who are seated on a giant lazy Susan. The only way to save the five people is to push the lazy Susan so that it swings the five out of the way; however, doing so wil cause the lazy Susan to ram into an inocent bystander, kiling him. 41 Kam?s intuition about this case is characteristicaly consequentialist: she thinks it is permissible to push the lazy Susan, thereby kiling the one to save the five. However, in doing so one would initiate a new threat (ME) that causes serious bodily harm (HURT) to a person (YOU), so this case counts as a personal moral dilema acording to Grene et al.?s criteria. Thus Grene et al.?s crucial assumption that we can establish a claim about the psychological processes underlying deontological and consequentialist judgments by testing the difering processes utilized to think about personal versus impersonal moral dilemas is seriously caled into question. Actualy, maters are even worse than that. What we realy have are thre distinctions: (i) the distinction betwen moral dilemas that typicaly elicit a characteristicaly deontological reaction and those that typicaly elicit a characteristicaly consequentialist reaction; (i) the distinction betwen moral dilemas that intuitively involve harm brought about in an ?up close and personal? way and those that do not; and (ii) the distinction betwen moral dilemas that satisfy Grene et al.?s ?ME HURT YOU? criteria and those that do not. The problem is that none of these distinctions matches up with the others. We have already seen how Kam?s lazy Susan case shows that distinction (i) is not distinction (ii). Kam?s case also shows why distinction (i) is not distinction (i): kiling someone by raming a giant lazy Susan tray into him presumably counts as harming that person in an ?up close and personal? maner, yet this case is one that gives rise to a characteristicaly consequentialist judgment. Moreover, there are a variety of cases that show that distinction (i) is not distinction (ii). Most famously, a variant of the fotbridge case in which there is a trapdor under the fat man that you can triger from afar intuitively counts as a case in which someone is not harmed in an ?up close and personal? way, yet such a case is demed to be a personal moral dilema by the ?ME HURT YOU? criteria, since trigering the trapdor 41 Kam, Morality, Mortality, Vol. I: Rights, Duties, and Status, p. 154. 17 initiates a new threat (ME) that causes serious bodily harm (HURT) to a specific individual (YOU). 42 So al thre of these distinctions pul apart. 43 Grene et al. were never under any delusions that their initial proposal for sorting cases that are like the fotbridge dilema from cases that are like the standard troley dilema was fuly adequate; after al, they explicitly caled it a ?first cut? proposal. 4 And Grene himself now admits that their proposal ?does not work? and ?is clearly wrong.? 45 Grene and his coleagues consider it a task for future research to determine the proper way of characterizing the distinction that they tried to capture with the ?ME HURT YOU? criteria. 46 So where does this leave us? Even if some emotional processes are tied to consequentialist judgment, and even if Grene et al.?s response-time prediction does not hold up, and even if their way of maping the deontological versus consequentialist judgment distinction onto the personal versus impersonal dilema distinction is in ned of revision, nonetheless Grene et al.?s neural-activity results strongly sugest that something like the dual-process hypothesis may wel be true, though perhaps in a modified form. 47 The question to which I would like to now turn is: what folows from these findings? 42 If you doubt that the ME criterion holds in this case, kep in mind that it is usualy a standard feature of the fotbridge case and its variants that the fal is what kils the fat man, whose body then serves as a weight to slow down the runaway troley. 43 In the body of this article I have apealed to Kam?s lazy Susan case in order to argue that the personal dilema versus impersonal dilema distinction (whether construed intuitively or in terms of the ?ME HURT YOU? criteria) is not the dilema- typicaly-giving-rise-to-a-deontological-judgment versus dilema-typicaly-giving-rise-to-a-consequentialist-judgment distinction. I chose to use Kam?s case since it is familiar from the troley problem literature. However, a slightly cleaner version of the same case that would equaly wel serve my purposes is as folows: instead of being on a lazy Susan, the five inocent people are on a doly (i.e., a wheled platform) that you can push out of the way of the oncoming troley, but doing so wil cause the doly to rol down a hil and smash an inocent bystander to death. (Note that in order for either of these cases to be a counterexample to Grene et al.?s proposal, it is not necesary that everyone makes a characteristicaly consequentialist judgment about such a case; al we ned is for a characteristicaly consequentialist judgment to be the standard reply.) Kam notes that lazy Susan cases raise dificulties for Grene et al.?s personal versus impersonal distinction in her Intricate Ethics, p. 142-43 and p. 180, n. 34; Grene concedes the point in his ?Reply to Mikhail and Timons,? p. 108. Se also p. 43, n. 37, and p. 418 of Intricate Ethics, where Kam discuses a second sort of case that poses problems for that distinction. 4 Grene et al., ?An fMRI Investigation,? p. 2107. 45 Grene, ?Reply to Mikhail and Timons,? p. 107, 14. 46 Further evidence of the inadequacy of Grene et al.?s ?ME HURT YOU? criteria is provided by Guy Kahane and Nicholas Shackel, ?Do Abnormal Responses Show Utilitarian Bias?? Nature 452:7185 (208): E5-E6. Kahane and Shackel had five profesional moral philosophers categorize the moral dilemas from Grene et al., ?An fMRI Investigation,? and they found that only 5 out of the 19 personal moral dilemas used in both experiments (26%) and only 10 out of the 2 impersonal moral dilemas used in experiment 2 (45%) were demed by a majority of these philosophers to involve a choice in which a deontological option is contrasted with a consequentialist option. The data from Kahane and Shackel?s study can be found at . 47 What about the aditional studies that have ben taken to lend further suport to the dual-proces hypothesis (se n. 26)? Here, to, I think the empirical upshot is far from certain. Some wories about those studies: ? In ?Neural Bases,? Grene et al. tok activity in the anterior cingulate cortex during contemplation of ?dificult? personal 18 IV. Normative Implications: Thre Bad Arguments Grene and Singer think that quite a lot folows from Grene et al.?s experimental findings. 48 In particular, they think that these findings give us god reason to conclude that characteristicaly deontological moral intuitions should not be trusted. Moreover, they think that these findings impugn the epistemic credentials of characteristicaly deontological moral intuitions without impugning the epistemic credentials of characteristicaly consequentialist moral intuitions. Both Singer and Grene draw further conclusions from their claim about the comparative epistemic standing of deontological and consequentialist intuitions. Revisiting an old debate with Rawls, 49 Singer argues that the untrustworthiness of deontological moral intuitions also shows that the method of reflective equilibrium is fundamentaly misguided. 50 And Grene argues that, in adition, the untrustworthiness of deontological moral intuitions shows that al deontologists?even those who, like moral dilemas to provide evidence that there was a conflict betwen two subsystems in the brain. However, as they themselves note (p. 395), the exact function of the anterior cingulate cortex is not curently known, and the hypothesis that it is devoted to conflict monitoring is just one among several. ? While it is true that in ?Cognitive Load? Grene et al. found that consequentialist responses to ?dificult? personal moral dilemas tok longer when the subjects were performing a cognitively intensive task at the same time (as the dual-proces hypothesis predicts), they did not find that subjects gave a lower percentage of consequentialist responses when performing the cognitively intensive task (as the dual-proces hypothesis would also predict). Grene et al. try to explain away this troubling piece of counter-evidence by speculating that the subjects were ?trying to push through? the interference caused by the cognitively intensive task (Grene et al., ?Cognitive Load,? p. 151), but as Adina Roskies and Walter Sinot- Armstrong note, ?this story makes sense only if subjects knew in advance that they wanted to reach a utilitarian judgment.? Se Roskies and Sinot-Armstrong, ?Betwen a Rock and a Hard Place: Thinking about Morality,? Scientific American Mind 19 (208), . ? Jorge Mol and Ricardo de Oliveira-Souza raise some doubts as to whether Koenigs et al.?s data in their ?Damage to the Prefrontal Cortex Increases Utilitarian Moral Judgments? on patients with damage to the ventromedial prefrontal cortex realy suport Grene et al.?s dual-proces hypothesis. Among other wories, Mol and de Oliveira-Souza point out that these patients also had damage to the anterior dorsolateral prefrontal cortex, a more ?cognitive? portion of the brain that Grene et al. found to be corelated with consequentialist judgment in ?An fMRI Investigation? and ?Neural Bases,? so these patients are not clean cases in which only emotional procesing is impaired. Se Mol and de Oliveira-Souza, ?Moral Judgments, Emotions, and the Utilitarian Brain,? Trends in Cognitive Science 1 (207): 319-21, and ?Response to Grene: Moral Sentiments and Reason: Friends or Foes?? Trends in Cognitive Science 1 (207): 323-24. For Grene?s reply, se his ?Why Are VMPFC Patients More Utilitarian? A Dual-Proces Theory of Moral Judgment Explains,? Trends in Cognitive Science 1 (207): 32-23. Se also Kahane and Shackel, ?Do Abnormal Responses Show Utilitarian Bias?? More importantly, though, it is dialecticaly problematic first to apeal to patients with damage to emotional brain regions when making an empirical case for the dual-proces hypothesis and then to go on to argue that the verdicts of these brain regions should be neglected (in efect urging us to be more like these patients), since many patients with this sort of brain damage make moral decisions in their personal lives that count as disastrous when evaluated by just about any plausible normative standard (se Antonio Damasio, Descartes? Eror: Emotion, Reason, and the Human Brain [New York: Putnam, 194]). So even if studies of such brain-damaged patients end up suporting the dual-proces hypothesis, they threaten to do so only at the cost of making Grene?s and Singer?s normative claims les tenable. 48 Se Grene, ?From Neural ?Is? to Moral ?Ought??; Grene, ?Secret Joke?; and Singer, ?Ethics and Intuitions.? 49 Se Peter Singer, ?Sidgwick and Reflective Equilibrium,? The Monist 58 (1974): 490-517. 50 Singer, ?Ethics and Intuitions,? p. 343-49. 19 Kant, 51 don?t explicitly rely on particular-case moral intuitions in their moral theorizing?are rationalizers who construct flimsy post hoc justifications for the very verdicts that they would be led by their emotions to make anyway. 52 I won?t be discussing either of these aditional arguments here, since they both obviously depend on the antecedent claim that we have god reason to discount deontological but not consequentialist particular-case moral intuitions. 53 Before turning to Grene?s and Singer?s central argument against the probative force of deontological intuitions, though, I want to briefly discuss thre bad arguments for that conclusion. On a charitable interpretation of Grene and Singer, these are arguments that they don?t actualy make but which it is extremely tempting to see them as making; on an uncharitable interpretation of Grene and Singer, these are bad arguments that they slopily mix in with their main argument. My guess is that the truth lies somewhere in betwen: although Grene?s and Singer?s primary and most promising line of argumentation does not rely on these thre arguments, I think they ocasionaly give their main argument more rhetorical force by invoking versions of these arguments. So it is worth showing just how unconvincing these thre arguments are before we consider Singer?s and Grene?s main reason for thinking that Grene et al.?s neuroscientific research gives us god reason to privilege our characteristicaly consequentialist intuitions over our characteristicaly deontological ones. The crudest possible argument for the conclusion reached by both Grene and Singer would proced as folows: the ?emotions bad, reasoning god? argument: P. Deontological intuitions are driven by emotions, whereas consequentialist intuitions involve abstract reasoning. C. So, deontological intuitions, unlike consequentialist intuitions, do not have any genuine normative force. 51 On one interpretation of Kant, although he apeals to particular-case intuitions about the conditions under which something is god in the opening paragraphs of Groundwork I, he takes himself to have discharged any apeal to particular-case moral intuitions once he has completed his argument for the form and existence of the Categorical Imperative by the end of Groundwork II. 52 Grene, ?Secret Joke,? p. 6-72. 53 For a reply to Singer?s ?Ethics and Intuitions? that focuses on the question of whether Grene et al.?s research poses problems for the method of reflective equilibrium, se Folke Tersman, ?The Reliability of Moral Intuitions: A Chalenge from Neuroscience,? Australasian Journal of Philosophy 86 (208): 389-405. However, to my mind Tersman is to quick to concede to Grene and Singer that Grene et al.?s research might demonstrate that deontological moral intuitions are unreliable; his main point is that even if this is so, the method of wide reflective equilibrium can take this fact into acount. 20 This is a bad argument. We ned a substantive reason for thinking that intuitions based in emotion are less likely to be reliable than those based in ?reasoning? for this argument to be at al convincing. After al, there is a venerable tradition that sees emotions as an important way of discerning normative truths. 54 One might disagre with this tradition, but showing that it rests on a mistake requires more than mere name-caling. Furthermore, even if the above argument were anything less than a howler, Grene et al.?s findings with regards to the posterior cingulate would cause aditional problems for the argument. If consequentialist intuitions also recruit emotional processes, the pernicious influence of the emotions can hardly be used to drive an epistemic wedge betwen deontological and consequentialist intuitions. 5 One natural way of improving on the ?emotions bad, reasoning god? argument would involve arguing as folows: the argument from heuristics: P1. Deontological intuitions are driven by emotions, whereas consequentialist intuitions involve abstract reasoning. P2. In other domains, emotional processes tend to involve fast and frugal (and hence unreliable) heuristics. C1. So, in the moral domain, the emotional processes that drive deontological intuitions involve fast and frugal (and hence unreliable) heuristics. C2. So, deontological intuitions, unlike consequentialist intuitions, are unreliable. This is also not the best argument. Usualy when we dem something to be a heuristic, we have a god handle on what the right and wrong answers in the relevant domain are; this is certainly the case in most of Daniel Kahneman and Amos Tversky?s wel-known examples of heuristics for logical and probabilistic reasoning. However, in the moral case it is very much up for debate what the right and wrong answers are. So it is question beging to assume that the emotional processes underwriting deontological intuitions consist in heuristics. Or more precisely: it is question beging to assume that just because emotional processes in other domains consist in heuristics, therefore emotional processes in the moral domain consist 54 For contemporary expresions of this sentiment, se Robert C. Solomon, The Pasions (Garden City, N.Y.: Anchor Pres/Doubleday, 1976); Ronald de Sousa, The Rationality of Emotion (Cambridge, Mas.: MIT Pres, 1987); Patricia Grenspan, Emotions and Reasons: An Inquiry into Emotional Justification (New York: Routledge, 198); Michael Stocker, with Elizabeth Hegeman, Valuing Emotions (Cambridge: Cambridge University Pres, 196); Benet Helm, Emotional Reason: Deliberation, Motivation, and the Nature of Value (Cambridge: Cambridge University Pres, 201); and Martha Nusbaum, Upheavals of Thought: The Inteligence of Emotions (Cambridge: Cambridge University Pres, 201). 5 And replacing the ?emotions bad, reasoning god? argument with an ?alarm-like emotions bad, curency-like emotions plus reasoning god? argument is clearly no dialectical improvement either, unles something substantive is said about why curency-like emotions are les problematic than alarm-like ones. 21 in heuristics. How can we proclaim these emotional processes to be quick but slopy shortcuts for geting at the moral truth unless we already have a handle on what the moral truth is? 56 I have just identified the inference from P1 and P2 to C1 as the major problem with the argument from heuristics. However, it is worth briefly mentioning two aditional problems with the argument. First, it is a mater of some dispute whether premise P2 is even true. A number of authors have argued that, in the non-moral domain, the fast and frugal heuristics underwriting emotional processes are often more reliable than their slow but flexible counterparts involving deliberate reasoning. 57 Second, the inference from C1 to C2 is also questionable. After al, it is doubtful that our mental machinery computes al of the actual and expected consequences of an action whenever we make a characteristicaly consequentialist judgment about it. So even if the neural processes underlying deontological intuitions rely upon heuristics, it is likely that the neural processes underlying consequentialist intuitions also make use of heuristics. 58 Al told, the argument from heuristics faces serious and, to my mind, fatal problems. 59 The first two arguments I have just considered involve fixing on the emotional nature of the processes that, acording to the dual-process hypothesis, underlie deontological judgments; the third argument I want to consider takes a diferent tack. Grene and Singer motivate their main argument by teling a ?just so? story about the evolution of our faculty for making deontological judgments about personal moral dilemas. Here is Grene?s version of that story: The rationale for distinguishing betwen personal and impersonal forms of harm is largely evolutionary. ?Up close and personal? violence has ben around for a very long time, reaching back into our primate lineage . . . . Given that personal violence is evolutionarily ancient, predating our recently evolved human capacities for complex abstract reasoning, it should come as no surprise if we have inate responses to personal violence that are 56 Cas Sunstein makes a similar point in ?Moral Heuristics,? Behavioral and Brain Sciences 28 (205): 531-73, at p. 53-34. 57 Gerd Gigerenzer has ben arguing this point for several decades now. Se, among other places, his ?Moral Intuition = Fast and Frugal Heuristics?? in Moral Psychology, Vol. 2: The Cognitive Science of Morality: Intuition and Diversity, ed. Walter Sinot- Armstrong (Cambridge, Mas.: IT Pres, 208), p. 1-26. John Alman and James Wodward also provide a number of nice examples from the psychology literature in which the utilization of automatic, emotional proceses sems to result in beter non- moral decisions than the utilization of more cognitive, deliberative proceses. Se Wodward and Alman, ?Moral Intuition: Its Neural Substrates and Normative Significance,? Journal of Physiology-Paris 101 (207): 179-202, at p. 189-91, 195-96; and Alman and Wodward, ?What Are Moral Intuitions and Why Should We Care about Them? A Neurobiological Perspective,? Philosophical Isues 18 (208): 164-85, at p. 170-71, 174. 58 I am grateful to Louis Menand for this point. 59 Another complication: in formulating the argument from heuristics, I have asumed that heuristics are, by definition, unreliable. However, as Frances Kam has reminded me, this asumption is strictly speaking false. For example, the rule ?Ad up al the digits and se if the result is divisible by 3? is a useful heuristic for determining whether a natural number is divisible by 3, but it is also perfectly reliable. 2 powerful but rather primitive. That is, we might expect humans to have negative emotional responses to certain basic forms of interpersonal violence . . . . In contrast, when a harm is impersonal, it should fail to triger this alarm-like emotional response, alowing people to respond in a more ?cognitive? way, perhaps employing a cost-benefit analysis. 60 Similarly, Singer writes, For most of our evolutionary history, human beings have lived in smal groups . . . . In these groups, violence could only be inflicted in an up-close and personal way?by hiting, pushing, strangling, or using a stick or stone as a club. To deal with such situations, we have developed imediate, emotionaly based responses to questions involving close, personal interactions with others. 61 In light of their apeal to such an evolutionary story, it is very tempting to read both Grene and Singer as making something like the folowing argument: the argument from evolutionary history: P. Our emotion-driven deontological intuitions are evolutionary by-products that were adapted to handle an environment we no longer find ourselves in. C. So, deontological intuitions, unlike consequentialist intuitions, do not have any genuine normative force. However, this is another bad argument. Presumably consequentialist intuitions are just as much a product of evolution?whether directly or indirectly?as deontological intuitions are, so an apeal to evolutionary history gives us no reason to privilege consequentialist intuitions over deontological ones. At one point Singer contends that consequentialist intuitions ?[do] not seem to be . . . the outcome of our evolutionary past,? 62 but I find this claim rather hard to believe. And at one point Grene declares that ?it is unlikely that inclinations that evolved as evolutionary by-products corespond to some independent, rationaly discoverable moral truth,? without realizing that such a claim poses as much a problem for the epistemic eficacy of consequentialist inclinations as it does for the epistemic eficacy of deontological inclinations. 63 In fact, a crass evolutionary argument of this sort poses problems for more than that. Anyone drawing normative implications from scientific findings is comited to mathematical and scientific judgments 60 Grene, ?Secret Joke,? p. 43. 61 Singer, ?Ethics and Intuitions,? p. 347-48. (Actualy, it is somewhat controversial whether the emotional underpinings of our moral judgments have ben retained in relatively unchanged form since our early ancestors. For some empirical evidence that this might not be so, se Wodward and Alman, ?Moral Intuition: Its Neural Substrates and Normative Significance,? p. 183, 187-8.) 62 Singer, ?Ethics and Intuitions,? p. 350. 63 Grene, ?Secret Joke,? p. 72. 23 having genuine normative force, yet presumably our faculty for making such judgments also has an evolutionary basis. Sensing this sort of wory, Singer cals for us to engage in ?the ambitious task of separating those moral judgments that we owe to our evolutionary basis and cultural history, from those that have a rational basis.? 64 However, this is clearly a false dichotomy. Richard Joyce and Sharon Stret ofer more careful versions of the argument from evolutionary history, yet their conclusion is that al of our moral judgments are unjustified (Joyce), or that al of our value judgments would be unjustified if certain realist conceptions of value were true (Stret). 65 The crucial premise in both Joyce?s and Stret?s argument is that moral judgments/intuitions don?t ned to be truth-tracking in order to conduce toward reproductive fitness, or at least that on a realist construal of what they amount to they don?t ned to; it is this premise that gives Joyce and Stret some hope of undercuting the epistemic status of moral judgments without also undercuting the epistemic status of mathematical and scientific judgments. So maybe one could argue that although deontological intuitions don?t ned to be truth-tracking in order to conduce toward reproductive fitness, consequentialist intuitions do, and in this way resuscitate the argument from evolutionary history. However: it is far from clear how this aditional piece of argumentation would go. Also: now al the work in the argument is being done by armchair theorizing about the conection betwen being truth-tracking and being evolutionarily beneficial; the neuroscientific results have completely droped out of the picture. 6 V. Normative Implications: A Beter Argument Now that we have set aside thre bad, but tempting, arguments for why Grene et al.?s neuroscientific findings have normative implications, we can consider Grene and Singer?s main argument for that conclusion. The crucial move they make is to insist that if Grene et al.?s research is corect, then our 64 Singer, ?Ethics and Intuitions,? p. 351. 65 Richard Joyce, The Myth of Morality (Cambridge: Cambridge University Pres, 201), chap. 6; Richard Joyce, The Evolution of Morality (Cambridge, as.: IT Pres, 206); and Sharon Stret, ?A Darwinian Dilema for Realist Theories of Value,? Philosophical Studies 127 (206): 109-6. 6 Moreover, even if Grene and Singer could somehow adapt Joyce?s and Stret?s arguments for their purposes, there is another reason why I don?t think this strategy would work: I don?t believe that Joyce?s and Stret?s original versions of the evolutionary argument are convincing. I argue for this claim in a companion piece to this article titled ?The Metaethical Irelevance of Evolutionary Theory.? 24 deontological intuitions are responding to factors that are moraly irelevant, and as such should not be trusted. This sugests the folowing argument: the argument from moraly irelevant factors: P1. The emotional processing that gives rise to deontological intuitions responds to factors that make a dilema personal rather than impersonal. P2. The factors that make a dilema personal rather than impersonal are moraly irelevant. C1. So, the emotional processing that gives rise to deontological intuitions responds to factors that are moraly irelevant. C2. So, deontological intuitions, unlike consequentialist intuitions, do not have any genuine normative force. When sumarizing his central argument, Grene writes, ?There are god reasons to think that our distinctively deontological moral intuitions . . . reflect the influence of moraly irelevant factors and are therefore unlikely to track the moral truth.? 67 And when responding to a comentator on that article, Grene ads, ?I have . . . argued that these [deontological] judgments can be explained in terms of paterns of emotional response and that these paterns reflect the influence of moraly irelevant factors.? 68 Similarly, Singer writes, If . . . Grene is right to sugest that our intuitive responses are due to diferences in the emotional pul of situations that involve bringing about someone?s death in a close-up, personal way, and bringing about the same person?s death in a way that is at a distance, and less personal, why should we believe that there is anything that justifies these responses? . . . [W]hat is the moral salience of the fact that I have kiled someone in a way that was possible a milion years ago, rather than in a way that became possible only two hundred years ago? I would answer: none. 69 And elsewhere in that same article, Singer insists that ?there are no moraly relevant diferences? betwen the troley driver and fotbridge dilemas, which is why we should ignore our emotional responses to the later sort of dilema. 70 What are we to make of the argument from moraly irelevant factors? The first thing to note is 67 Grene, ?Secret Joke,? p. 69-70. 68 Grene, ?Reply to Mikhail and Timons,? p. 17. Se also Grene, ?Secret Joke,? p. 70, where he refers to ?contingent, non-moral feature[s]?; Grene, ?Secret Joke,? p. 75, where he again refers to ?moraly irelevant factors?; and Grene, ?Reply to Mikhail and Timons,? p. 16, where he refers to ?arbitrary features? in the course of arguing that deontologists sufer a ?garbage in, garbage out? problem. 69 Singer, ?Ethics and Intuitions,? p. 347, 348. 70 Singer, ?Ethics and Intuitions,? p. 350. 25 that, as Grene and Singer both admit, 71 premise P2 in this argument apeals to a substantive normative intuition, which presumably one must arive at from the armchair, rather than directly read of from any experimental results; this is why the argument does not derive an ?ought? from an ?is.? I believe that this feature is a virtue of the argument; however, it is also its ultimate undoing. In what folows, I mention four wories that I have about the argument from oraly irelevant factors, in order of increasing significance. First wory: since Grene et al.?s initial characterization of the personal vs. impersonal dilema distinction does not track the gives-rise-to-a-deontological-judgment vs. gives-rise-to-a-consequentialist- judgment distinction, it is far from clear that premise P1 is true. Grene thinks that the eventual acount of the features to which the deontological moral faculty is responding wil have something to do with ?personalness,? broadly construed, 72 but I have my doubts. 73 Moreover, any atempt to precisely characterize the features that give rise to distinctively deontological judgments reintroduces many of the intricacies of the original troley problem: formulating a principle that distinguishes what separates cases- eliciting-a-deontological-judgment from cases-eliciting-a-consequentialist-judgment is likely to be as 71 Se Grene, ?Secret Joke,? p. 6-67; and Singer, ?Ethics and Intuitions,? p. 347. 72 Grene, ?Reply to Mikhail and Timons,? p. 12. 73 In a more recent study (Joshua D. Grene, Fiery Cushman, Lisa E. Stewart, Kely Lowenberg, Leigh E. Nystrom, and Jonathan D. Cohen, ?Pushing Moral Butons: The Interaction betwen Personal Force and Intention in Moral Judgment,? Cognition 11 [209]: 364-71), Grene and his coleagues claim to have discovered that what explains people?s moral judgments about fotbridge-like cases is whether the agent?s action intentionaly harms someone through the use of what they cal personal force, which is present when ?the force that directly impacts the victim is generated by the agent?s muscles? (p. 364). However, there are a number of problems with the study. For instance, many of the contrasting cases have a variety of diferences beyond those identified as candidate explanatory factors, and a number of obvious potential counterexamples to their proposal were not tested (for example, do people judge it just as moraly unaceptable to force the man of the fotbridge by menacing him with a knife, or by threatening to harm his family, or by tricking him into taking a step backwards?). But, ironicaly, the bigest problem with the study is that Grene et al. sem to have identified, without realizing it, a competing explanation for their respondents? verdicts. Grene et al. gathered evidence about the degre to which their respondents unconsciously filed in more realistic asumptions when imagining the scenarios in question, and they found a high degre of corelation betwen a tendency to refuse to asume that it was absolutely certain that the five would be saved if the one is kiled and a tendency to judge that such a course of action is moraly unaceptable (p. 367-68). So, by their own lights, not al of their subjects were responding to the same scenario, and the variation in responses can be partialy explained by the variation in asumptions about the likelihod of the proposed action suceding. (I suspect that varying asumptions about the degre to which the man on the fotbridge might resist, thereby endangering the life of the agent trying to harm him, could also go a long way toward explaining people?s difering verdicts.) More importantly, though, it is simply a mistake to think that by merely surveying people?s opinions about the moral permisibility of certain actions, we can empiricaly study what sorts of factors elicit characteristicaly deontological judgments. Al these studies tel us is that these people make certain moral judgments about certain scenarios, and certain other moral judgments about certain other scenarios; which of these judgments count as characteristicaly deontological is not something we can just read of from the empirical results. (Sometimes Grene sugests that we can sidestep this wory by postulating that philosophers are confused about the meaning of the term ?deontology,? and although they think it refers to an abstract moral theory, in fact it refers to a psychological natural kind, namely the verdicts of the emotional subsystem; se his ?Secret Joke,? p. 37-38. However, in making this claim Grene is comiting himself to an incredibly controversial claim in the philosophy of language. It?s one thing to say that although we think ?water? refers to one sort of physical substance, in fact it refers to another sort of physical substance. Grene?s claim, however, is akin to saying that although we think that ?Goldbach?s conjecture? refers to an abstract mathematical theory, in fact it refers to the physical proces of digestion, or to saying that although we think that the name ?Barack Obama? refers to a certain person, in fact it refers to the number 7.) 26 dificult as the old problem of formulating a principle that distinguishes the permissible options in troley- like cases from the impermissible ones. After al, Grene et al.?s initial ?ME HURT YOU? criteria were inspired by Thomson?s proposed solution to the troley problem in her 1985 article ?The Troley Problem,? and it fel victim to Kam?s lazy Susan case, which was originaly ofered as a counterexample to Thomson?s very proposal. So setling on a fuly adequate acount of the sorts of features to which deontological judgments are responding is likely to be an extremely dificult, if not impossible, task, and until that task has ben completed, we canot be sure whether P1 is true. Second wory: even if we were able to find a way of characterizing the factors which deontological judgments are responding to that makes P1 true, it is far from clear that P2 would stil seem plausible. It is one thing to claim that a faculty which responds to how ?up close and personal? a violation is is responding to moraly irelevant features, but quite another thing to claim that a faculty which responds to whatever the sorts of features are that distinguish the fotbridge case from the troley driver case is responding to moraly irelevant features. Once we fix on what those features are, P2 may wel strike us as false. 74 Third wory: even if P2 does strike us as true, the argument?s conclusion does not folow, for C2 does not folow from C1. Supose we dem some of the features trigering deontological intuitions to, intuitively, be moraly irelevant, thus granting P2. This is a strike against deontological intuitions. However, we can only conclude that consequentialist intuitions should be privileged over deontological intuitions if a paralel case canot be made against consequentialist intuitions. Moreover, it is open to the defender of deontology to reply that, intuitively, the faculty eliciting consequentialist reactions is also responding to moraly irelevant factors, or failing to respond to moraly relevant ones. For example, a deontologist could contend that the neural processes giving rise to consequentialist judgments are failing to respond to moraly relevant factors by ignoring the separateness of persons, or by treating people as 74 Inded, there is a sense in which this objection is already apropos even if we asume Grene et al.?s ?ME HURT YOU? proposal to be fuly adequate. Grene performs a sort of shel game here: first he proposes that the dilemas eliciting deontological reactions are dilemas that, intuitively, involve harm comited in an ?up close and personal? maner, and then he gloses dilemas that, intuitively, involve harm comited in an ?up close and personal? maner in terms of the ?ME HURT YOU? criteria. However, when it comes time to decide whether deontological judgments are responding to moraly relevant factors, Grene switches back to evaluating things in terms of the intuitive up-close-and-personal-harm distinction, rather than in terms of the ?ME HURT YOU? criteria. However, it?s one thing to say that whether one has comited a harm in an ?up close and personal? maner is a moraly irelevant factor, and quite another thing to say that whether one has initiated a new threat that brings about serious bodily harm to another individual is a moraly irelevant factor. 27 vats of wel-being, or by assuming that al value is to-be-promoted, or by making morality incompatible with integrity, or . . . [insert your favorite anti-consequentialist intuition here]. So basicaly we have just recapitulated the same old batle of intuitions over the plausibility of consequentialism versus deontology in our evaluation of which sorts of factors are and are not moraly relevant. Which leads to my most pressing wory: the neuroscientific results seem to be doing no work in this argument. The epistemic eficacy of consequentialist versus deontological intuitions now apears to be purely a function of what sorts of features out there in the world they are each responding to. We have thre distinctions on the table: 75 ? dilemas that engage emotion processing vs. dilemas that engage ?cognitive? processing; ? dilemas that elicit deontological judgments vs. dilemas that elicit consequentialist judgments; ? personal moral dilemas vs. impersonal moral dilemas. Grene et al.?s dual-process hypothesis posits that the first of these distinctions matches up with the second. In order to experimentaly assess this hypothesis, Grene and his coleagues identified the second distinction with the third one, and then directly tested whether the first distinction matches up with the third. But the argument from oraly irelevant factors only depends on Grene et al.?s identification of the second distinction with the third one. Thus the neuroscientific results are beside the point. In particular: ? The ?emotion-based? nature of deontological intuitions has no ultimate bearing on the argument?s cogency. (Delete ?emotional? from P1 and C1, and the argument is just as plausible.) ? Issues about the evolutionary history of our dispositions to have deontological and consequentialist intuitions are also irelevant to the argument?s cogency. ? Even the claim that these two sets of intuitions stem from separate faculties is irelevant to the argument?s cogency. (The argument would be just as plausible if it turned out that only one faculty was responding to two diferent sorts of factors.) So the apeal to neuroscience is a red hering: what?s doing al the work in the argument from moraly 75 More precisely, we have four distinctions on the table, since we ned to distinguish betwen the intuitive way of cashing out the personal vs. impersonal moral dilema distinction, and the more regimented way of cashing out that distinction in terms of the ?ME HURT YOU? criteria. 28 irelevant factors is (a) Grene?s identification, from the armchair, of the distinction betwen dilemas- eliciting-deontological-reactions and dilemas-eliciting-consequentialist-reactions with the distinction betwen personal and impersonal moral dilemas, and (b) his invocation, from the armchair, of a substantive intuition about what sorts of factors out there in the world are and are not moraly relevant. The basic problem is that once we rest our normative weight on an evaluation of the moral salience of the factors to which our deontological and consequentialist judgments are responding, we end up factoring out (no pun intended) any contribution that the psychological processes underlying those judgments might make to our evaluation of the judgments in question. So we are left with a dilema: apeal to a substantive intuition about what sorts of factors are moraly relevant, and the neuroscientific results drop out of the picture; or kep those results in, and it loks like our only recourse is to one of the bad arguments we have already dismissed. Thus I conclude that the argument from oraly irelevant factors does not advance the dialectic on the relative merits of deontology versus consequentialism. No reasonable philosopher is going to deny that it makes no moral diference whether one harms someone with one?s bare hands or from a distance. However, deontologists are most definitely going to deny that so crude a distinction is what realy underlies their distinctively deontological moral judgments. And once we have in hand the true acount of what sorts of factors underwrite deontological judgments about cases, I claim that evaluation of their moral relevance wil depend on a substantive normative judgment as to whether observing those sorts of moral distinctions is more or less plausible than ignoring the sorts of moral distinctions that consequentialists typicaly ignore. The apeal to neuroscience provides no new traction on this old debate. 76 VI. An Indirect Role for Neuroscience? One of the things that seemed so exciting about Grene?s research was that it promised a new way of 76 To se how litle work the neuroscience is now doing in the argument from oraly irelevant factors, consider this: should we perform aditional experiments to se what parts of the brain light up when certain people make a judgment that such-and-such- factors-picked-out-by-deontological-judgments are not moraly relevant, and what parts of the brain light up when other people make a judgment that such-and-such-factors-ignored-by-consequentialist-judgments are in fact moraly relevant? What would that posibly tel us? (Shal we then perform yet more experiments to se what parts of the brain light up when people make a judgment about the relevance of the factors to which those second-order judgments are responding, and so on, ad infinitum?) 29 finaly resolving the troley problem. Rather than having to hit upon a compact, exceptionless principle that delivers the intuitive verdict about every troley-like case, we could use the neuroscientific results and some philosophical theorizing to discount certain of those intuitive verdicts, making it easier to find a principle that fits the leftover data. However, one of the things I have just argued is that in order to use the neuroscientific results and some philosophical theorizing to discount certain intuitive verdicts about troley-like cases, Grene in efect neds to have already solved the troley problem. Since the argument for why we should discount a certain set of intuitive verdicts depends on evaluating the features to which those intuitive verdicts are responding, we ned an exceptionless (but not necessarily compact) principle delineating the sorts of features that make those sorts of intuitive verdicts kick in. So there is one way in which Grene?s task is more dificult than the traditional troley problem: rather than neding to find a single principle that states what it takes for an option in a troley-like dilema to count as permisible or impermisible, he neds to find two principles: one principle stating what it takes for the deontological faculty to count an option in a troley-like dilema as permisible or impermisible, and another principle stating what it takes for the consequentialist faculty to count an option in a troley-like dilema as permisible or impermisible (plus an acount of how conflicts betwen the two faculties are resolved). But there is another way in which Grene?s task is easier than the traditional troley problem: he doesn?t ned his acount of the sorts of features to which either of these faculties is responding to be rationaly defensible. Moreover, it seems that there is a way in which neuroscience could play an indirect role in this task that Grene has set for himself?and, by extension, a way in which neuroscience could play an indirect role in more traditional atempts at solving the troley problem. Supose we have established that a certain region of the brain is activated when we contemplate a certain class of cases that yield characteristicaly deontological verdicts about what it is moraly permissible to do. Supose, also, that we have independent knowledge that in non-moral cases this brain region is recruited to distinguish betwen (say) intentional and non-intentional action. Then we might try seeing whether what distinguishes this class of moral dilemas from others has something do with the intentional versus non-intentional action distinction. Since neuroscience only provides evidence of corelations, it is not certain that when the brain region in question is recruited for moral cases it is responding to the same 30 sorts of features as when it is recruited for non-moral cases. But the neuroscientific results can give us clues for where to lok when trying to characterize what sorts of features out there in the world each moral faculty is responding to. And this is true whether our ultimate aim is to debunk or to vindicate those verdicts. However, note that, even here, the neuroscientific results play no role after we have the principles stating what sort of features each faculty is responding to: at that point, the argument for whether we should or should not discount the verdicts of one of these faculties proceds entirely via armchair theorizing about whether the sorts of features to which that faculty is responding are or are not moraly relevant. Stil, providing clues for where to lok when atempting to characterize the features to which distinctively deontological and distinctively consequentialist judgments respond is no smal mater. 7 Could neuroscience play a more direct role in our theorizing about the evidential status of moral intuitions? It seems to me that the best-case scenario is this: the best-case scenario: We notice that a portion of the brain which lights up whenever we make a certain sort of obvious, egregious eror in mathematical or logical reasoning also lights up whenever we have a certain moral intuition. In this case, should we discount the moral intuition? That depends on how e fil in the details of the case. If, for al we can see, there is no conection betwen the content of the moral intuition and the content of the mistaken bit of mathematical/logical reasoning, then I am inclined to think we should continue to trust the intuition and hold out for later neuroscience to make finer distinctions betwen the portions of the brain activated in the moral and non-moral cases. (Supose the same part of the brain that lights up whenever we afirm the consequent also lights up whenever we have an intuition that infanticide is impermissible; would you be wiling to start kiling babies on those grounds?) If, on the other hand, we come to see that the moral intuition in question rests on the same sort of confusion present in the mistaken bit of mathematical/logical reasoning, then of course we should discount the moral intuition, but in that case the neuroscience isn?t playing a direct justificatory role. Again, we might not have thought to link the moral intuition to that sort of mathematical/logical blunder if we hadn?t known the neuroscientific results; but again, once we do link them, it seems that we do so from the comfort of an armchair, not from the 7 Although Grene et al.?s atempt in ?An fMRI Investigation? at characterizing the features that give rise to deontological judgment did not rely on neuroscience in the way I have just sketched, there is a sense in which evolutionary theory played just that sort of indirect role, since evolutionary considerations are what partialy led them to try out the proposal they put forward. 31 confines of an experimental laboratory. It is as if, while trying to prove whether or not some mathematical claim is true, your mathematician friend had said to you, ?Why don?t you try using the Brouwer fixed point theorem?? If you end up proving the claim to be true using that theorem, your justification for the claim in no way depends on your friend?s testimony. (After al, she didn?t give away whether she thinks the claim is true or false.) Nonetheless, your friend?s testimony gave you a hint for where to lok when trying to prove or disprove the mathematical claim. So to, I speculate, neuroscience can provide hints for where to lok during our normative theorizing, but ultimately it can play no justificatory role in that task. Despite Grene?s and Singer?s claims to the contrary, learning about the neurophysiological bases of our moral intuitions does not give us god reason to privilege certain of those intuitions over others. 78 78 Earlier versions of this article were presented at the Arch? Philosophical Research Centre in St. Andrews, at the Harvard Kenedy Schol?s Safra Center for Ethics, at the Harvard Humanities Center?s Cognitive Theory and the Arts Seminar, and at the 209 Rocky Mountain Ethics Congres in Boulder, Colorado (where my comentator was Daniel Demetriou). Joshua D. Grene was kind enough to atend both sesions at Harvard and to ofer clarifications and replies. For writen coments on earlier drafts, I am indebted to Jacob Beck, Carol Berker, Tyler Doget, Frances Kam, Christine Korsgard, Seana Shifrin, Judith Jarvis Thomson, and Hasko Vonkriegstein (on behalf of the moral psychology reading group at Toronto University). For helpful coments and discusion, many thanks as wel to Arthur Aplbaum, Sharon Bery, Tim Buton, Yuri Cath, Colin Chamberlain, Norman Daniels, Daniel Demetriou, Tom Dougherty, Mati Eklund, Nir Eyal, Michael Frazer, Johan Frick, Micha Glaeser, Ned Hal, Ulrike Heuer, Jonathan Ichikawa, Colin Klein, Ole Koksvik, Arnon Levy, Louis Menand, Oded Na?aman, Dilip Ninan, Fran?ois Recanati, Simon Ripon, T. M. Scanlon, Susana Siegel, Alison Simons, Leo Ungar, Manuel Vargas, Alex Vorhoeve, Glen Weyl, Liane Young, and Elia Zardini. Finaly, I am grateful to the editors of Philosophy & Public Afairs for a number of extremely helpful sugestions.