|Home | About | Journals | Submit | Contact Us | Français|
The psychological and neurobiological processes underlying moral judgement have been the focus of many recent empirical studies1–11. Of central interest is whether emotions play a causal role in moral judgement, and, in parallel, how emotion-related areas of the brain contribute to moral judgement. Here we show that six patients with focal bilateral damage to the ventromedial prefrontal cortex (VMPC), a brain region necessary for the normal generation of emotions and, in particular, social emotions12–14, produce an abnormally ‘utilitarian’ pattern of judgements on moral dilemmas that pit compelling considerations of aggregate welfare against highly emotionally aversive behaviours (for example, having to sacrifice one person’s life to save a number of other lives)7,8. In contrast, the VMPC patients’ judgements were normal in other classes of moral dilemmas. These findings indicate that, for a selective set of moral dilemmas, the VMPC is critical for normal judgements of right and wrong. The findings support a necessary role for emotion in the generation of those judgements.
The basis of our moral judgements has been a long-standing focus of philosophical inquiry and, more recently, active empirical investigation. In a departure from traditional rationalist approaches to moral cognition that emphasize the role of conscious reasoning from explicit principles15, modern accounts have proposed that emotional processes, conscious or unconscious, may also play an important role16,17. Emotion-based accounts draw support from multiple lines of empirical work: studies of clinical populations reveal an association between impaired emotional processing and disturbances in moral behaviour1–4; neuroimaging studies consistently show that tasks involving moral judgement activate brain areas known to process emotions5–9; and behavioural studies demonstrate that manipulation of affective state can alter moral judgements10,11. However, neuroimaging studies do not settle whether putatively ‘emotional’ activations are a cause or consequence of moral judgement; behavioural studies in healthy individuals do not address the neural basis of moral judgement; and no clinical studies have specifically examined the moral judgements (as opposed to moral reasoning or moral behaviour) of patients with focal brain lesions. In brief, none of the existing studies establishes that brain areas integral to emotional processes are necessary for the generation of normal moral judgements. As a result, there remains a critical gap in the evidence relating moral judgement, emotion and the brain.
Investigating moral judgements in individuals with focal damage to the ventromedial prefrontal cortex (VMPC) provides a key test. The VMPC projects to basal forebrain and brainstem regions that execute bodily components of emotional responses18, and neurons within the VMPC encode the emotional value of sensory stimuli19. Patients with VMPC lesions exhibit generally diminished emotional responsivity and markedly reduced social emotions (for example, compassion, shame and guilt) that are closely associated with moral values1,2,12–14,16, and also exhibit poorly regulated anger and frustration tolerance in certain circumstances20,21. Despite these patent defects both in emotional response and emotion regulation, the capacities for general intelligence, logical reasoning, and declarative knowledge of social and moral norms are preserved20–23. We selected a sample of six patients with adult-onset, focal bilateral VMPC lesions (Fig. 1) as well as both neurologically normal (NC) and brain-damaged comparison (BDC) subjects. Importantly, each of the VMPC patients had striking defects in social emotion but generally intact intellect and normal baseline mood (Tables 1 and and2,2, see also Supplementary Table 1). In particular, all six VMPC patients had impaired autonomic activity in response to emotionally charged pictures (Table 2), as well as severely diminished empathy, embarrassment and guilt (Table 2). All comparison subjects (NC and BDC) had intact emotional processing.
Subjects evaluated moral dilemmas designed to pit two competing considerations against one another. A paradigmatic dilemma of this type presents subjects with the choice of whether or not to sacrifice one person’s life to save the lives of others. One consideration is a utilitarian calculation of how to maximize aggregate welfare, whereas the other is a strong emotional aversion to the proposed action. One model holds that endorsement of the proposed action (the utilitarian response) requires the subject to overcome an emotional response against inflicting direct harm to another person (a ‘personal’ harm7,8). If emotional responses mediated by VMPC are indeed a critical influence on moral judgement, individuals with VMPC lesions should exhibit an abnormally high rate of utilitarian judgements on the emotionally salient, or ‘personal’, moral scenarios (for example, pushing one person off a bridge to stop a runaway boxcar from hitting five people), but a normal pattern of judgements on the less emotional, or ‘impersonal’, moral scenarios (for example, turning a runaway boxcar away from five people but towards one person). If, alternatively, emotion does not play a causal role in the generation of moral judgements but instead follows from the judgements24,25, then individuals with emotion defects due to VMPC lesions should show a normal pattern of judgements on all scenarios.
To test for between-group differences in the probability of utilitarian responses given for each scenario type (non-moral, impersonal moral, personal moral), we used a logistic regression fitted with the generalized estimating equations method (Fig. 2). There were no significant differences between groups on the non-moral or impersonal moral scenarios (all P values >0.29, corrected for multiple comparisons). In contrast, for personal moral scenarios, the VMPC group was more likely to endorse the proposed action than either the NC group (odds ratio = 2.81; P = 0.04, corrected) or BDC group (odds ratio = 3.30; P = 0.006, corrected). There was no difference between the NC and BDC groups (odds ratio = 0.85; P = 0.68, uncorrected). These data indicate that the VMPC group’s responses differed only for personal moral scenarios, suggesting that VMPC-mediated processes affect only those moral judgements involving emotionally salient actions.
In a more fine-grained analysis, we examined response patterns within the personal moral scenarios. For seven out of the 21 personal moral scenarios, both comparison groups were at 100% agreement in their judgements. An additional eighth scenario elicited 100% agreement from the BDC group, and near-perfect agreement from the NC group (with only one participant deviating from the shared response). These eight scenarios were therefore classified as ‘low-conflict’ (for example, abandoning one’s baby to avoid the burden of caring for it). The remaining 13 scenarios (none of which elicited 100% agreement from either comparison group) were classified as ‘high-conflict’ (for example, smothering one’s baby to save a number of people). Reaction-time data support this distinction: response latencies in the NC group on high-conflict scenarios were significantly longer than on low-conflict scenarios (t-test with 19 degrees of freedom, t(19) = −3.63; P = 0.002).
Like the patients in the comparison groups, the VMPC patients uniformly rejected the proposed action in every one of the low-conflict scenarios (Fig. 3). In contrast, significant differences emerged for the high-conflict scenarios: the VMPC group was more likely to endorse the proposed action than either the NC (odds ratio = 4.70; P = 0.05, corrected) or BDC group (odds ratio = 5.38; P = 0.02, corrected), with no difference between the NC and BDC participants (odds ratio = 0.87; P = 0.77, uncorrected). Every high-conflict personal scenario elicited the same pattern: a greater proportion of the VMPC group endorsed the action than either comparison group.
To recapitulate, VMPC patients’ judgements differed from comparison subjects’ only for the high-conflict personal moral dilemmas, all of which featured competing considerations of aggregate welfare on the one hand, and, on the other hand, harm to others that would normally evoke a strong social emotion. Low-conflict personal moral scenarios lacked this degree of competition. This difference probably accounts for the greater consensus and faster reaction times on low-conflict personal dilemmas in the comparison groups, and it can also account for the VMPC patients’ pattern of judgements. Evidence suggests that knowledge of explicit social and moral norms is intact in individuals with VMPC damage21,22. In the absence of an emotional reaction to harm of others in personal moral dilemmas, VMPC patients may rely on explicit norms endorsing the maximization of aggregate welfare and prohibiting the harming of others. This strategy would lead VMPC patients to a normal pattern of judgements on low-conflict personal dilemmas but an abnormal pattern of judgements on high-conflict personal dilemmas, precisely as was observed. The specificity of this result argues against a general deficit in the capacity for moral judgement following VMPC damage. Rather, VMPC seems to be critical only for moral dilemmas in which social emotions play a pivotal role in resolving moral conflict4,8,16,17.
It is important to note that the effects of VMPC damage on emotion processing depend on context. In this study, the VMPC patients’ abnormally high rate of utilitarian judgements is attributed to diminished social emotion, whereas in a recent study of the Ultimatum Game, theVMPC patients’ abnormally high rate of rejection of unfair monetary offers was attributed to poorly controlled frustration, manifested as exaggerated anger20. These seemingly contradictory findings highlight two distinct aspects of emotion impairment that are due to VMPC damage. In most circumstances, VMPC patients exhibit generally blunted affect and a specific defect of social emotions, but in response to direct personal frustration or provocation, VMPC patients may exhibit short-temper, irritability, and anger. In the moral judgement task we report here, participants respond to hypothetical actions and outcomes that elicit social emotions related to concern for others. In the Ultimatum Game, in contrast, participants respond to unfair take-it-or-leave-it offers that trigger frustration. In brief, the tasks in the two studies are different in that the Ultimatum Game involves self-interest in a real behavioural setting, whereas the task in the present study focuses on the interest of others described in a hypothetical scenario.
To conclude, the present findings are consistent with a model in which a combination of intuitive/affective and conscious/rational mechanisms operate to produce moral judgements8,22,24–27. Though the precise characterization of these potential systems awaits further work, the current results suggest that the VMPC is a critical neural substrate for the intuitive/affective but not for the conscious/rational system.
Six patients with bilateral, adult-onset damage to the VMPC and twelve brain-damaged comparison patients who had lesions that excluded structures thought to be important for emotions (VMPC, amygdala, insula, right somatosensory cortices) were recruited from the Patient Registry of the Division of Cognitive Neuroscience at the University of Iowa. Twelve healthy comparison subjects with no brain damage were recruited from the Iowa community. Groups were age-, gender- and ethnicity-matched. All participants gave written informed consent.
The neuroanatomical analysis of VMPC patients (Fig. 1) was based on magnetic resonance data for two subjects (those with lesions due to the surgical resection of orbital meningiomas) and on computerized tomography data for the other four subjects (with lesions due to rupture of an anterior communicating artery aneurysm). All neuroimaging data were obtained in the chronic epoch. Each patient’s lesion was reconstructed in three dimensions using Brainvox28. Using the MAP-3 technique, the lesion contour for each patient was manually warped into a normal template brain. The overlap of lesions in this volume, calculated by the sum of n lesions overlapping on any single voxel, is colour-coded in Fig. 1.
Participants made judgements on a series of 50 hypothetical scenarios, which were adapted from a previously published set8. See the Supplementary Information for the full text of the actual scenarios used. Each scenario was presented as text through a series of three screens. The first two described the scenario and the third posed a question about a hypothetical action related to the scenario (“Would you … in order to …?”). Participants read and responded at their own pace, pressing an ‘up’ arrow key to advance from one screen to the next, and a ‘yes’ or ‘no’ button to indicate an answer to the question. ‘Yes’ responses always indicated commission of the proposed action. There was no time limit for reading the scenario description (screens 1 and 2). Participants had a maximum of 25 s to read the final question screen and respond.
We used three classes of stimuli: non-moral scenarios (n = 18), and two classes of moral scenarios subdivided according to the emotional reaction elicited by the proposed action: ‘personal’ (n = 21) or ‘impersonal’ (n = 11), as described previously7,8. To validate this subdivision, an independent group of ten neurologically normal subjects rated the emotional salience of the actions proposed in the moral scenarios. The actions described in personal scenarios were rated as significantly more emotionally salient than the actions described in impersonal scenarios (means were 5.9 and 3.0 on a scale from 1 to 7, respectively; t(31) = −8.90, P<0.0001). Within either class of moral scenarios (personal or impersonal), it was not valid to separately analyse judgements based on the emotional salience of the proposed action (that is ‘high-emotion’ versus ‘low-emotion’ scenarios) because emotionality ratings were remarkably similar for scenarios within each class: 9 of the 11 impersonal scenarios received a mean emotion rating between 1.1 and 3.0, while 20 of the 21 personal scenarios received a mean emotion rating between 5.3 and 6.7.
We further subdivided the personal moral scenarios into ‘low-conflict’ and ‘high-conflict’ on the basis of the reaction times and consensus produced on them by normal subjects. Reaction times on high-conflict scenarios were significantly longer than on low-conflict scenarios (t(19) = −3.63, P = 0.002). Importantly, low-conflict and high-conflict scenarios did not differ in their rated emotional salience (t(19) = −0.85, P = 0.41).
We thank H. Damasio for making available neuroanatomical analyses of lesion patients and for preparing Fig. 1. We thank all participants for their participation in the experiments and R. Saxe for comments on the manuscript. This work was supported by grants from the National Institutes of Health, the National Science Foundation, the Gordon and Betty Moore Foundation, and the Guggenheim Foundation.