Oxford Studies in Experimental Philosophy [1, 1 ed.] 0198718764, 9780198718765

The new field of experimental philosophy has emerged as the methods of psychological science have been brought to bear o

178 29 2MB

English Pages 400 [383] Year 2015

Report DMCA / Copyright

DOWNLOAD PDF FILE

Recommend Papers

Oxford Studies in Experimental Philosophy [1, 1 ed.]
 0198718764, 9780198718765

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Oxford Studies in Experimental Philosophy

Oxford Studies in Experimental Philosophy Volume 1 EDITED BY

Tania Lombrozo Joshua Knobe Shaun Nichols

1

3

Great Clarendon Street, Oxford, OX2 6DP, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries # The several contributors 2014 The moral rights of the author have been asserted First Edition published in 2014 Impression: 1 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2014933005 ISBN 978–0–19–871876–5 (hbk) ISBN 978–0–19–871877–2 (pbk) Printed and bound by CPI Group (UK) Ltd, Croydon, CR0 4YY Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.

Contents Introduction Tania Lombrozo, Joshua Knobe, and Shaun Nichols 1. Huck vs. JoJo: Moral Ignorance and the (A)symmetry of Praise and Blame David Faraci and David Shoemaker 2. The Cognitive Mechanisms of Intolerance: Do our Meta-Ethical Commitments Matter? Jennifer Cole Wright, Cullen B. McWhite, and Piper T. Grandjean 3. Doing, Allowing, and Enabling Harm: An Empirical Investigation Christian Barry, Matthew Lindauer, and Gerhard Øverland 4. Attributing Responsibility: Actual and Counterfactual Worlds Tobias Gerstenberg and David A. Lagnado 5. The Moral Status of an Action Influences its Perceived Intentional Status in Adolescents with Psychopathic Traits Elise M. Cardinale, Elizabeth C. Finger, Julia C. Schechter, Ilana T.N. Jurkowitz, R.J.R. Blair, and Abigail A. Marsh

1

7

28

62

91

131

6. The Concept of Intentional Action in High-Functioning Autism Edouard Machery and Tiziana Zalla

152

7. A Scientific Case for Conceptual Dualism: The Problem of Consciousness and the Opposing Domains Hypothesis Anthony I. Jack

173

8. Carving Up the Social World with Generics Sarah-Jane Leslie

208

vi

CONTENTS

9. Flavors of “Togetherness”: Experimental Philosophy and Theories of Joint Action Deborah Tollefsen, Roger Kreuz, and Rick Dale

232

10. The Good in Happiness Jonathan Phillips, Sven Nyholm, and Shen-yi Liao

253

11. Causal Reasoning: Philosophy and Experiment James Woodward

294

12. Experimental Philosophy: 1935–1965 Taylor Shaw Murphy

325

Index

369

Introduction Tania Lombrozo, University of California, Berkeley Joshua Knobe, Yale University Shaun Nichols, University of Arizona

The aim of Oxford Studies in Experimental Philosophy is to provide a venue for publishing work in this new field. Experimental philosophy has emerged over the last decade as the methods of psychological science have been brought to bear on traditional philosophical problems. The ambition of the field has been to use these empirical methods as a way to illuminate some of the oldest issues in philosophy. Experimental philosophy is an interdisciplinary field at the intersection of philosophy and psychology. Accordingly, just about all of the papers in this field so far have been published either in philosophy journals or in psychology journals. There is certainly value in this approach, especially insofar as it has led to real engagement with work in each of these disciplines, considered separately. However, there is also a downside. When people publish in philosophy journals, they may feel a certain pressure to focus exclusively on the philosophical side of their work, to the exclusion of the more psychological side; there may be a corresponding pressure to focus on the empirical when publishing in psychology. OSEP is intended to serve as a venue in which people can explore experimental philosophy in all of its aspects, addressing both philosophical and psychological issues, as needed. Looking through this first volume, one finds a mix of chapters by philosophers, chapters by psychologists, and chapters co-authored by people in both disciplines. Yet the work done by these different people does not seem to be easily classifiable



T . LOMBROZO ET AL .

into the different disciplines. Instead, one finds the emergence of a truly interdisciplinary field in which people from different disciplines are working together to address a shared set of questions. The volume is roughly structured into four sections. The first three chapters focus on recent developments in moral psychology—a topic that has seen lively debate and a great deal of progress over the last decade. In the second section, we highlight three contributions that bring new methods to moral psychology: formal modeling and special populations. In the third section, we assemble four chapters that adopt an experimental philosophy approach to novel topics, including intuitive dualism, generics, joint action, and happiness. Finally, in the fourth section, we present two chapters that provide critical and historical context to the development of experimental philosophy.

Moral Psychology: New Findings David Faraci and David Shoemaker investigate the prima facie plausible idea that our judgments of blame and praise should be symmetrical. For instance, if the unfortunate circumstances of a person’s upbringing render him less blameworthy for bad actions, it should also render him less praiseworthy for good actions. However, in a series of experiments, they find that people tend to treat such cases asymmetrically. At least in some cases, a bad upbringing makes a person less blameworthy for bad actions but more praiseworthy for good actions. Faraci and Shoemaker explore various ways in which one might square these results with the idea that blame and praise should be treated symmetrically. Jennifer Cole Wright and colleagues examine the relationship between two ethical issues that might at first seem quite distinct: (a) the metaethical question as to whether certain moral claims are objectively true and (b) people’s willingness to show toleration for those who have opposing values. Wright and McWhite find that their experimental participants tended to be “meta-ethical pluralists,” meaning that each participant tended to think that some moral questions have objective answers while others do not. They then showed a surprising relationship between issues (a) and (b). Specifically, there was a significant correlation whereby participants showed less tolerance for people with opposing values about those questions they regarded as being objective.

INTRODUCTION



Christian Barry and colleagues challenge a long-standing distinction between doing and allowing harm. Backed by a series of novel experimental findings, the authors’ suggestion is that this bipartite distinction has neglected an important third category: enabling harm. The authors present evidence that judgments about harm can reliably be classified as doing, allowing, or enabling harm, and that this classification has consequences for people’s judgments of moral responsibility. Differentiating “enabled harm” from other cases provides a more fine-grained taxonomy of harm that can inform both normative and empirical work in moral psychology.

Moral Psychology: New Methods Tobias Gerstenberg and David Lagnado consider how moral responsibility is allocated for outcomes that result from the actions of multiple agents. According to most accounts of responsibility attribution, agents should be held responsible to the extent that their actions made a difference to the outcome. Yet in some cases we hold agents responsible for outcomes even when those outcomes do not depend (counterfactually) on their individual actions. For example, we might hold a shooter responsible for the death of a victim even if a second shooter fired simultaneously, and that shooter’s actions were sufficient to bring about the death. Gerstenberg and Lagnado suggest that we can make sense of such cases by considering not only whether an agent’s actions made a difference in the actual situation being considered, but also whether they would have made a difference in other possible situations, such as those in which a second shooter was not present. The authors develop this idea in terms of counterfactuals defined over structured causal models and discuss both published and new results that support their formulation. Judgments about whether an agent intentionally produced an outcome are affected by whether the outcome is morally bad or morally good. For instance, if a person performs an action that he knows will harm the environment, then even if he did not intend to harm the environment people tend to say that he intentionally harmed the environment. Elise Cardinale and colleagues investigate whether this phenomenon occurs in adolescents with psychopathic traits. This population has known deficits in moral cognition, and it has been quite unclear whether these deficits in



T . LOMBROZO ET AL .

moral cognition would lead to atypical performance concerning intentionality. Cardinale and colleagues found no difference between the adolescents with and without psychopathic traits—an interesting result with implications for current accounts of moral evaluation and intentionality attribution. Edouard Machery and Tiziana Zalla explore a related question about the understanding of intentional action among participants with high-functioning autism. In earlier work, normally developing adults were presented with a person deciding to pay extra to get a benefit (a smoothie). As expected, the participants said that the person intentionally paid extra. By contrast, when presented with a case in which a free benefit—a cup—was given to the person, participants tended to deny that the person intentionally got the cup. Machery and Zalla presented this task to participants with high-functioning autism and found that these participants did not distinguish the cases. Machery and Zalla appeal to deficits in social cognition to explain this result. In particular, they suggest that people with high-functioning autism have difficulty recognizing that a person might have an instrumental desire for something that is intrinsically disvalued.

New Topics in Experimental Philosophy A prominent theme in contemporary cognitive science is that people are dualists about mind and brain. At least in many contexts, people regard mind and brain as fundamentally independent kinds of things. Anthony Jack adduces evidence for this view and introduces a novel explanation for folk dualism. He argues that the cognitive tension between reasoning about the mind (“mentalizing”) and reasoning about physical mechanisms reflects a tension at the level of opposing neural systems. In particular, when one of these systems is activated, it actually suppresses activation of the opposing system. Turning to an issue with more immediate practical implications, Sarah-Jane Leslie discusses a series of experimental studies on the use of generic language. Recent work in philosophy, psychology, and linguistics has explored in depth the semantics of generic sentences (e.g. “Tigers are striped.”). Leslie now asks what impact those sorts of sentences have when they are applied to social categories (“Women are . . . ,” “African-American people are . . . ”). She shows that these sentences

INTRODUCTION



communicate a particular conception of social categories involving what psychologists call essentialism. She then argues that this essentialist conception is tied to a detrimental understanding of human abilities. Deborah Tollefsen and colleagues use experimental methods to explore people’s intuitions about joint action. Philosophers distinguish things that human beings do together (having a conversation, singing a duet) from those that people do individually (reading a book, singing a solo). Tollefsen and colleagues use multidimensional scaling to examine the factors that make people regard various actions as similar or different. They obtain a two-dimensional solution, with one dimension corresponding surprisingly neatly to this philosophical notion of “jointness.” In other words, even when people are not being asked questions that explicitly involve the concept of jointness, their spontaneous judgments of similarity show a clear influence of that concept. Jonathan Phillips and colleagues present a series of studies investigating people’s ordinary concept of happiness. The proposal that they develop and test combines elements from two traditions of research on happiness. Like philosophers endorsing descriptive accounts of happiness, as well as many psychologists engaged in the empirical study of happiness, they suggest that attributions of happiness are largely based on the psychological states of the person being considered, not on the normative status of the life that person actually led. Yet more in line with philosophers offering evaluative views of happiness, they also find that such judgments depend on whether people believe that the agent in question ought to have experienced a positive psychological state in a given case. The result is a novel hybrid of descriptive and normative theories of happiness attributions, with interesting implications for both philosophical and psychological work.

Critical and Historical Reflections on Experimental Philosophy James Woodward offers an analysis and review of research on causal judgment, covering both philosophical work from a broadly interventionist perspective and empirical work from a variety of psychological disciplines including cognitive and developmental psychology. Woodward considers how normative accounts of causation have informed and



T . LOMBROZO ET AL .

can inform empirical investigation; he also provides guidelines for the type of research most likely to have interesting philosophical implications. In particular, Woodward discusses research from both surveybased experimental philosophy and measures of causal learning and reasoning from psychology, offering both critical and constructive suggestions about their relative merits. Taylor Shaw Murphy engages in a more historical investigation, exploring an important precursor of contemporary experimental philosophy in a movement that flourished from roughly 1935 to 1965. The experimentalists of this earlier period conducted empirical tests of some of the key theses coming out of ordinary language philosophy, and Murphy describes some of their most striking experimental results. In addition, he explores the metaphilosophical debate that emerged at that time, examining the many ways in which it prefigured the contemporary debate over experimental philosophy. In sum, this volume highlights both the progress and the promise of experimental philosophy, bringing together contributions on established topics alongside interdisciplinary work that combines empirical methods with philosophical issues in new and exciting ways. Perhaps most notable of all, however, is that the integration of philosophy and psychology is so seamless, suggesting that experimental philosophy is truly coming of age.

1 Huck vs. JoJo Moral Ignorance and the (A)symmetry of Praise and Blame David Faraci, University of North Carolina David Shoemaker, Tulane University

The more I studied about this the more my conscience went to grinding me, and the more wicked and low-down and ornery I got to feeling. And at last, when it hit me all of a sudden that here was the plain hand of Providence slapping me in the face and letting me know my wickedness was being watched all the time from up there in heaven, whilst I was stealing a poor old woman’s nigger that hadn’t ever done me no harm, and now was showing me there’s One that’s always on the lookout, and ain’t a-going to allow no such miserable doings to go only just so fur and no further, I most dropped in my tracks I was so scared. Well, I tried the best I could to kinder soften it up somehow for myself by saying I was brung up wicked, and so I warn’t so much to blame; but something inside of me kept saying, “There was the Sunday-school, you could a gone to it; and if you’d a done it they’d a learnt you there that people that acts as I’d been acting about that nigger goes to everlasting fire.” (31.19) It made me shiver. And I about made up my mind to pray, and see if I couldn’t try to quit being the kind of a boy I was and be better. So I kneeled down. But the words wouldn’t come. Why wouldn’t they? It warn’t no use to try and hide it from Him. Nor from ME, neither. I knowed very well why they wouldn’t come. It was because my heart warn’t right; it was because I warn’t square; it was because I was playing double. I was letting ON to give up sin, but away inside of me I was holding on to the biggest one of all. I was trying to make my mouth SAY I would do the right thing and the clean thing, and go and write to that nigger’s owner and tell



D . FARACI , D . SHOEMAKER

where he was; but deep down in me I knowed it was a lie, and He knowed it. You can’t pray a lie—I found that out. (31.20) [After writing the note,] I felt good and all washed clean of sin for the first time I had ever felt so in my life, and I knowed I could pray now. But I didn’t do it straight off, but laid the paper down and set there thinking—thinking how good it was all this happened so, and how near I come to being lost and going to hell. And went on thinking. And got to thinking over our trip down the river; and I see Jim before me all the time: in the day and in the night-time, sometimes moonlight, sometimes storms, and we a-floating along, talking and singing and laughing. But somehow I couldn’t seem to strike no places to harden me against him, but only the other kind. I’d see him standing my watch on top of his’n, ’stead of calling me, so I could go on sleeping; and see him how glad he was when I come back out of the fog; and when I come to him again in the swamp, up there where the feud was; and such-like times; and would always call me honey, and pet me and do everything he could think of for me, and how good he always was; and at last I struck the time I saved him by telling the men we had small-pox aboard, and he was so grateful, and said I was the best friend old Jim ever had in the world, and the ONLY one he’s got now; and then I happened to look around and see that paper. (31.23) It was a close place. I took it up, and held it in my hand. I was a-trembling, because I’d got to decide, forever, betwixt two things, and I knowed it. I studied a minute, sort of holding my breath, and then says to myself: “All right, then, I’ll GO to hell”—and tore it up. (31.24, 31.25)

In these famous scenes from The Adventures of Huckleberry Finn, Huck is trying desperately to decide what to do with Jim, the slave he has been traveling with down the river, the man who is, to Huck’s mind, someone else’s property. When given the opportunity to return Jim to his “rightful owner,” Huck ultimately decides to go against his upbringing, his conscience, and societal norms in not turning Jim in. Consequently, by his own lights he is going to hell for doing what he sincerely believes is the wrong thing. By our lights, though, he is clearly doing the right thing, despite his morally deprived upbringing. Is he thus praiseworthy? Compare the following case: JoJo is the favorite son of Jo the First, an evil and sadistic dictator of a small, undeveloped country. Because of his father’s special feelings for the boy, JoJo is given a special education and is allowed to accompany his father and observe his daily routine. In light of this treatment, it is not surprising that little JoJo takes his father as a role model and develops values very much like Dad’s. As an adult, he

MORAL IGNORANCE , PRAISE AND BLAME



does many of the same sorts of things his father did, including sending people to prison or to death or to torture chambers on the basis of whim. He is not coerced to do these things, he acts according to his own desires. Moreover, these are desires he wholly wants to have. When he steps back and asks, “Do I really want to be this sort of person?” his answer is resoundingly, “Yes,” for this way of life expresses a crazy sort of power that forms part of his deepest ideal.1

In torturing a peasant on a whim, JoJo goes along with his upbringing and conscience, doing what he sincerely believes is the right thing but what we know is the wrong thing. Is he thus blameworthy? Both Huck and JoJo were raised in morally blinkered environments, and they have both consequently come to accept deeply mistaken moral views: Huck thinks he ought to return Jim to his “owner”; and JoJo thinks it is morally permissible (or perhaps obligatory) to beat peasants when he feels like it. There have been (at least) two assumptions that theorists have thought it natural to make in cases such as these. First, it seems plausible to some that moral deprivation of this sort is a responsibility-undermining factor, that those who were raised in such morally blinkered environments are off the hook when it comes to assignments of moral responsibility. So JoJo, we might think, is excused for beating the peasants given the extent of his childhood moral deprivation. Call this the Moral Deprivation– Excuse Thesis (the MDE Thesis).2 Second, it is widely believed that praiseworthiness is the positive analogue of blameworthiness, and thus that factors affecting judgments of moral responsibility should have symmetrical effects on judgments of praise- and blameworthiness. Call this the Symmetry Thesis.3 On a natural interpretation of this thesis, if poor moral upbringing mitigates assignments of JoJo’s blameworthiness, it ought likewise to mitigate assignments of praiseworthiness in cases of equivalent moral deprivation. In this essay, we examine how certain sorts of moral knowledge deprivations in an agent’s upbringing bear on people’s actual assessments of that Susan Wolf, “Sanity and the Metaphysics of Responsibility,” in Gary Watson, ed., Free Will, 2nd edn (Oxford: Oxford University Press, 2003), p. 379. 2 This is the view taken by Wolf, as we will detail below. 3 The Symmetry Thesis is widespread. It is entailed by what Doris and Knobe call the more general assumption of invariance in moral responsibility assessments. For discussion and citation, see John Doris and Joshua Knobe, “Strawsonian Variations: Folk Morality and the Search for a Unified Explanation,” in John Doris, ed., The Moral Psychology Handbook (Oxford: Oxford University Press, 2010). 1



D . FARACI , D . SHOEMAKER

agent’s responsibility. At first blush, the data we have collected appears to cast doubt on both the MDE and Symmetry theses. First, our experimental results suggest that people do not, in fact, view deprivation as wholly morally excusing (though they do not view it as irrelevant, either). Second, our data appear to suggest that people have commitments to distinctly different conditions for blame- and praiseworthy agency.4 This seems to cast doubt on the Symmetry Thesis. In both cases, however, we argue that the correct response to our data is not simply to reject the theses in question. First, our results suggest that the relationship between deprivation and moral assessment is more nuanced than straightforward acceptance or denial of the MDE Thesis can account for. In earlier work, we discovered that while people do not judge JoJo to be wholly blameworthy, they do not judge him to be blameless, either. And the data we present here seems to confirm this for other cases of moral deprivation. This suggests that while the MDE Thesis cannot be accepted as it stands, it has captured an important aspect of blameworthiness judgments. As for the Symmetry Thesis, we argue that it can be interpreted in a way that renders it consistent with our data. In both cases—especially if we are to vindicate our interpretation of the Symmetry Thesis—it is an important question why people judge as they do. In our earlier work, we hypothesized that people judge JoJo as they do because of the difficulty JoJo would have in overcoming his upbringing and doing the right thing. After considering some alternatives, we argue that the very same hypothesis can help vindicate our interpretation of the Symmetry Thesis. This, we take it, both buttresses the plausibility of our claims about difficulty and represents an independent advance in our understanding of the roots of praise and blame.

Blinkered Badness In the previous work mentioned, we explicitly explored folk intuitions on the JoJo case.5 In her original presentation and discussion of the case, 4 Knobe and Doris include this asymmetry in their list, based on early experimental results generated by one of the researchers, but those experiments remained unpublished pending further refinement. The present essay (and the experiment done in its service) is an attempt at said refinement. 5 David Faraci and David Shoemaker, “Insanity, Deep Selves, and Moral Responsibility: The Case of JoJo,” Review of Philosophy and Psychology 1 (2010): 319–32.

MORAL IGNORANCE , PRAISE AND BLAME



Susan Wolf offers a strong version of what we have labeled the MDE Thesis, claiming that people’s pretheoretic intuitions are that JoJo, because of his terribly deprived upbringing, is not morally responsible at all for his actions, despite their flowing from his deep self, i.e. his “true” or “real” evaluating or authenticating character.6 Her diagnosis for this alleged reaction is that JoJo’s deep self is normatively insane, i.e. he lacks the ability to recognize goodness, badness, and the difference between them.7 She then introduces a sanity—normative competence—condition into her theory of moral responsibility to account for such cases. Insofar as Wolf is claiming a result on behalf of our pretheoretic intuitions, we decided to test her claim. Our findings were that, while people did judge JoJo to be less blameworthy than a control (JoJo’s presumably sane but nasty father), they still found him to be seriously blameworthy, assigning him an average blameworthiness score of 5 (where 7 was completely blameworthy and 1 was not at all blameworthy). Insofar as blameworthiness rides piggyback on responsibility (at least prima facie), JoJo is definitely viewed (pretheoretically) as morally responsible to some degree, contrary to Wolf ’s strong version of the MDE Thesis. We took this at the very least to indicate that her adoption of a sanity condition was unmotivated. But—and this is the interesting bit—JoJo was viewed as less blameworthy than he would have been without such a deprived background, so deprivation of this sort does seem to excuse to some extent. But what precisely was the relation of this deprivation to reduced blameworthiness? And could it be overcome? Our initial diagnosis of JoJo’s reduced blameworthiness derived from his moral ignorance, not his normative insanity. It was not that he lacked the capacity to recognize right and wrong; it was, rather, that he had as a child been deprived of exposure to relevant moral alternatives. We attempted to explore these issues via the presentation of a third scenario in which JoJo was eventually exposed to moral alternatives but rejected them in favor of continuing to adhere to Daddy’s value system. Here,

On pp. 379–80, Wolf says “In light of JoJo’s heritage and upbringing . . . it is dubious at best that he should be regarded as responsible for what he does.” Later on p. 380, she says, “Our judgment that JoJo is not a responsible agent is one that we can make only from the inside” (emphasis ours). 7 Wolf, pp. 379–85. 6



D . FARACI , D . SHOEMAKER

surprisingly, there was no statistically significant difference between people’s reactions to the first and second JoJos. This might suggest, then, that it was their unfortunate formative circumstances that (slightly) mitigated their blameworthiness, and not their mere moral ignorance. But we remained unconvinced, arguing instead that the JoJos displayed a more fundamental and insidious type of moral ignorance than is usually discussed—an ignorance that expressions of ill will are wrong, not an ignorance of what specific act-tokens count as expressions of ill will— and that this sort of ignorance may not be displaced by mere exposure to moral alternatives (which would do nothing to counteract the thought that it is generally morally permissible to express ill will).8 What reduces the degree of their blameworthiness, we suggested, was not any sort of incapacity, though; rather, it was the difficulty of overcoming their childhood moral deprivations as adults, a difficulty that nevertheless assumes the possibility of success and so grants a basic normative capacity to them (a necessary condition to their being judged to be seriously blameworthy in the first place). In our current round of experiments, we aimed to do two things. First, we wanted to see if we could duplicate the JoJo-type results in cases in which normative capacities were not at issue, and neither were any thoroughgoing childhood deprivations regarding the general wrongness of expressions of ill will. This would enable us to focus solely on the work being done by ignorance with respect to a specific sort of moral value. Second, we wanted to explore whether structurally identical positive cases would yield analogous results. We will explain this latter point in the next section of the essay, focusing here solely on the former. New subjects were presented, at random, with one of the following two scenarios:9 A. Tom is a white male who was raised in New Orleans. Growing up, he was taught to respect all people equally. Nevertheless, as an adult, he We put this in terms of “ill will” in the original paper, but we recognize this to be a quite ambiguous notion. (For discussion, see David Shoemaker’s “Qualities of Will,” Social Philosophy and Policy 30 (2013): 95–120) We could have put this in less ambiguous terms as follows without losing the point. In standard cases of moral ignorance, the agent knows what properties make an act right or wrong but does not know whether some particular acttype instantiates those properties; JoJo’s ignorance, on the other hand, is about what properties make actions right or wrong in the first place, and so is much more profound. 9 n(A) = 84; n(B) = 84. 8

MORAL IGNORANCE , PRAISE AND BLAME



decided to become a proud racist, someone who believes that all nonwhite people are inferior and that he has a moral obligation to humiliate them when he gets a chance. At the age of 25, Tom moves to another town. Walking outside his home, he sees a black man who has tripped and fallen. In keeping with his moral beliefs, Tom spits on the man as he passes by. B. Tom is a white male who was raised on an isolated island in the bayous of Louisiana. Growing up, he was taught to believe that all nonwhite people are inferior and that he has a moral obligation to humiliate them when he gets a chance. As an adult, he fully embraced what he’d been taught, becoming a proud racist. At the age of 25, Tom moves to another town. Walking outside his home, he sees a black man who has tripped and fallen. In keeping with his moral beliefs, Tom spits on the man as he passes by. For each scenario, subjects were asked to rate Tom’s blameworthiness for spitting on the black man on a scale from 1 (“not all blameworthy”) to 7 (“completely blameworthy”). Our prediction, in keeping with our results from the earlier paper, was that TomB’s upbringing and resultant ignorance with respect to the morality of racism would mitigate attributions of moral blameworthiness. The mean response to TomA was 6.68, a very robust blameworthiness score. The mean response to TomB was significantly lower at 5.4.10 We conclude from this, again in keeping with earlier results, that certain sorts of moral blinders—i.e. childhood deprivations of exposure to moral truth—reduce blameworthiness assessments somewhat, but nowhere near completely.11 And, once again, the fact that subjects viewed TomB as seriously blameworthy reveals that they likely believed him not to lack the relevant normative capacities, i.e., sanity was not an issue. Finally, our 10 Results were subjected to an independent samples t-test: t(166) = 6.54, p < 0.001 (twotailed), SD (ignorant) 1.53, SD (non-ignorant) 0.92, Cohen’s d = 1.01. 11 It might be thought that the vignettes do not do enough to establish sufficiently robust moral ignorance in TomB’s case, as anyone growing up in the modern era will surely have been exposed to alternatives through TV, radio, the Internet, or travel. Given that this might be an assumption of the subjects who read the vignettes, then it might have grounded their being pretty punitive. (Thanks to an anonymous reviewer for pressing this point.) This is a fair point, and it is worth controlling for. Nevertheless, we doubt it is doing much work for subjects, given that we stressed JoJo’s isolation and lack of exposure to alternatives in the original study, and we got nearly identical response levels to our JoJo doppelganger in the latest one.



D . FARACI , D . SHOEMAKER

speculative explanation remains quite plausible. We propose that the reason people judge TomB to be seriously blameworthy, despite the blameworthiness-reducing fact of his deprived childhood, is that, while it is more difficult for him to identify and do the right thing because of that childhood, it is nevertheless not overly demanding to expect him to do so. Some new features we included buttressed this hypothesis by diminishing the plausibility of alternative explanations. First, we had Tom in both scenarios move to another town at the age of 25 in order to avoid the kind of ongoing isolated “preciousness” of the original JoJo case, and also to establish that he is an adult (with a fully-grown brain) who is making genuine moral decisions. His moving to another town would presumably also have given him opportunities to be exposed to moral alternatives. The question, then, was how that exposure would interact with what he had been taught as a child in people’s assessments of him. As it turns out, if he was taught the moral truth and rejected that as an adult, he was viewed as particularly blameworthy for spitting on the black man; he was viewed as less so if he had been taught the racist moral lie but simply did not reject it. Second, we deliberately left open why it is that Tom either rejects or embraces what he has been taught. The reason was to allow for the possibility of any of the wide variety of sources of evaluations that occur in everyday life, and so not to privilege certain sources over others. Sometimes people evaluate after having pored over the reasons on both sides and determining which ones weigh more. Other times people evaluate after having seen a movie on the subject and without further deliberation. Other times people simply follow their intuitive hunches. Presumably all of these are methods that will preserve responsibility in assessors’ eyes. Of course, the extent to which our proposal is plausible depends not only on the paucity of plausible alternatives, but on the inherent plausibility of the proposal itself. As has become clear to us since first introducing the proposal, more details about the nature of the difficulty in question are required before plausibility can be adequately assessed.12 However, we set this matter to the side momentarily in order to introduce the remainder of our data, which are relevant to our explication of the nature of the difficulty in question. 12

Thanks to two anonymous referees and to Shaun Nichols for pressing us to say more about the idea of difficulty in this context.

MORAL IGNORANCE , PRAISE AND BLAME



Thus far, we have been considering the effects of moral deprivation on assignments of blameworthiness. Especially given our interest in the Symmetry Thesis, the obvious question now is whether similar results obtain in cases of praiseworthy action. It is to that question that we now turn.

Blinkered Goodness The Symmetry Thesis avers that blameworthiness and praiseworthiness are structurally analogous. If so, then it seems we should expect that childhood moral deprivations will reduce praiseworthiness in people’s eyes just as they do blameworthiness, at least for actions within the zone germane to the relevant deprivations. The basic thought here is rather compelling. It would seem that degrees of blame- or praiseworthiness both ought to track degrees of childhood-based moral ignorance in the same way: the less you know, the less you are “on the hook” in either case. If the Symmetry Thesis were true, childhood moral deprivations presumably ought to reduce moral responsibility all the way round. The data, however, appear to undermine the Symmetry Thesis, at least when understood in this way. Further subjects were randomly assigned to one of the following two cases:13 C. Tom is a white male who was raised in New Orleans. Growing up, he was taught to respect all people equally. Nevertheless, as an adult, he decided to become a proud racist, someone who believes that all nonwhite people are inferior and that he has a moral obligation to humiliate them when he gets a chance. At the age of 25, Tom moves to another town. Walking outside his home, he sees a black man trip and fall. Usually, Tom would spit on the man. But this time, Tom goes against his current moral beliefs, and helps the man up instead. D. Tom is a white male who was raised on an isolated island in the bayous of Louisiana. Growing up, he was taught to believe that all nonwhite people are inferior and that he has a moral obligation to humiliate them when he gets a chance. As an adult, he decided to become a proud racist, embracing what he was taught. At the age of 25, Tom moves to another town. Walking outside his home, he sees a black man trip and

13

n(C) = 85; n(D) = 83.



D . FARACI , D . SHOEMAKER

fall. Usually, Tom would spit on the man. But this time, Tom goes against his current moral beliefs, and helps the man up instead. This time, subjects were asked to rate Tom’s level of praiseworthiness on a scale from 1 (“not at all praiseworthy”) to 7 (“completely praiseworthy”). The mean response to TomC was 4.28. The mean response to TomD was 5.40.14 There are several surprising conclusions one might draw here. First, TomC, who was raised with moral awareness but adopted racism as an adult, is only viewed as “somewhat praiseworthy” for going against those moral beliefs in doing the right thing. TomA—TomC’s twin who adhered to his adult-formed moral beliefs in spitting on the black man—was viewed as nearly completely blameworthy for doing so. Perhaps, then, the resistance to seeing TomC as more praiseworthy is due to his still being a racist. This cannot explain all the data, however, because TomD is also a racist, and yet he is viewed as significantly more praiseworthy than TomC in going against his moral beliefs. It seems the only thing that could ground the difference in people’s assessments here is the difference in upbringings. But then here is the second surprising conclusion. As discussed above, it seems plausible that TomD’s moral ignorance would be generally mitigating: that if he did not know that his racism was wrong, then he could not know that his going against it was right. But apparently knowledge of the rightness of one’s actions is not viewed as necessary for praiseworthiness; indeed (and this is the truly surprising point) moral ignorance seems to be viewed as a virtue. Not only is the MDE Thesis being denied here, it is being turned on its head. And thus we come to the third surprising feature. While moral ignorance reduces blameworthiness, it seemingly increases praiseworthiness. Importantly, this way of putting it suggests that the Symmetry Thesis as understood thus far is false, or at least highly problematic.

Resilient Symmetry Nevertheless, we would like to explore the possibility of a different interpretation of the Symmetry Thesis that could allow for it to be maintained 14

As before, results were subjected to an independent samples t-test: t(166) = 4.18, p < 0.001 (two-tailed), SD (ignorant) 1.69, SD (non-ignorant) 1.76, Cohen’s d = 0.65.

MORAL IGNORANCE , PRAISE AND BLAME



in light of our results. To this point, we have understood the thesis to imply that factors reducing blameworthiness also reduce praiseworthiness, that negative and positive cases are symmetrical with respect to whether the relevant praise- or blameworthiness is reduced or increased (in comparison to some paradigm control case). It looks as if the Tom cases undermine this symmetry: the ignorance reducing TomB’s blameworthiness score actually increases TomD’s praiseworthiness score. Understood in this way, the relevant comparison point is the baseline at which Tom is not at all praise- or blameworthy. The right question to ask is: Did the feature in question move the degree of _______-worthiness away from or towards that baseline? (See Figure 1.1.) Of course, one could simply accept the asymmetry. One might, for instance, attempt to explain it by drawing an analogy to the famous “Knobe effect.” This effect (named after Joshua Knobe) occurs when subjects are asked whether the side-effects of some agent’s actions were intentional.15 Subjects tend to say “yes” when the side-effects include something harmful, but tend to say “no” when the side-effects include something helpful. This appears to reveal an asymmetry in factual judgments of intentional action. Knobe’s own interpretation is that subjects’ judgments depend on their antecedent assessments of the normative status of the action’s side-effects, so that “bad” effects render the action intentional in subjects’ eyes, whereas “good” effects do not. Applied to our Toms, then, it may appear that a kind of reverse Knobe effect is revealed: moral ignorance via childhood deprivation seems to reduce attributions of responsibility when the actions are bad, whereas it increases such attributions when the actions are good. This could mean that factual judgments of responsible action also are dependent on antecedent assessments of the normative status of the action. Given wide acceptance of the Symmetry Thesis, however, it seems that such a move would be overly hasty if it is possible instead to interpret the thesis in a way that is consistent with our data.16 Indeed, such an 15 See Joshua Knobe, “Intentional Action and Side Effects in Ordinary Language,” Analysis 63 (2003): 190–3; and Joshua Knobe, “The Concept of Intentional Action: A Case Study in the Uses of Folk Psychology,” Philosophical Studies 130 (2006): 203–31. 16 While widely presupposed, the Symmetry Thesis is not accepted universally. For leading examples of asymmetricians, see Susan Wolf, Freedom Within Reason (Oxford: Oxford University Press, 1990); and Dana Nelkin, Making Sense of Freedom and Responsibility (Oxford: Oxford University Press, 2011). We have also already mentioned Doris and



D . FARACI , D . SHOEMAKER

TomA to TomB

1

2

3

Not at all blameworthy

4

5

6

Somewhat blameworthy

1

2

3

4 Somewhat praiseworthy

7 Completely blameworthy

TomC to TomD

Not at all praiseworthy

Moral ignorance moved the degree of blameworthiness from 6.68 to 5.4.

5

Moral ignorance moved the degree of praiseworthiness from 4.28 to 5.4.

6

7 Completely praiseworthy

Figure 1.1 Moral ignorance, praise, and blame.

interpretation exists: we propose, in contrast to the view represented by Figure 1.1, that negative and positive cases are symmetrically structured with respect to the direction in which various mitigating features shift the _______-worthiness judgments in relation to their controls. Understood in this way, the relevant comparison point would be the two endpoints of a continuous scale (see Figure 1.2). So, with respect to the two endpoints (“completely blameworthy” and “completely praiseworthy”), the direction of movement from TomA to TomB, and from TomC to TomD, is symmetrical. What their sort of moral ignorance does, on this understanding of the relation, is move one more in the overall direction of complete praiseworthiness and away from complete blameworthiness. If we understand the Symmetry Thesis in this way, then it might be preserved in light of our results. To do so, though, we must understand Knobe’s arguments against a version of the Symmetry Thesis. Importantly, though, all of these theorists take their burden of proof seriously, and so admit the need to provide arguments against symmetry in blame and praise.

MORAL IGNORANCE , PRAISE AND BLAME

TomA to TomB

7

6

Completely blameworthy

5

4



TomC to TomD

3

Somewhat blameworthy

2

1

2

Neither blameworthy nor praiseworthy

3

4

5

Somewhat praiseworthy

6

7

Completely praiseworthy

Figure 1.2 Symmetrical structure of negative and positive cases.

what the symmetry could consist in, i.e. what the explanation for the symmetrical movement would be. One possibility (continuing on the analogy with intentional action) stems from the work of Chandra Sripada, who has argued in favor of what he calls the “Deep Self Model” (DSM) of moral responsibility to provide a better explanation of the Knobe effect for action intentionality.17 What Sripada argues is that intentionality attaches to agents’ actions to the extent that their sideeffects concord with the values and stable fundamental attitudes—the deep selves—that subjects attribute to them. If there is no concordance, goes the theory, subjects will be less likely to attribute intentionality to the agent’s actions. On this interpretation, then, subjects view the harmful side-effect in the Knobe studies as in concordance with the agent’s deep self and attribute intentionality to it thereby, but they view the helpful side-effect as in conflict with the agent’s deep self and so tend not to attribute intentionality to it thereby. This is how the DSM maintains a unified symmetrical explanation of the Knobe effect. Normative judgments of goodness and badness do not do the relevant work here; the different responses instead flow from perceptions of different structural underpinnings. We might, then, appeal to the DSM to explain our results regarding responsibility. In particular, it will be useful for our purposes to explore whether the DSM alone can explain the results while also preserving the Symmetry Thesis, just as it purports to do with respect to the Knobe effect. One immediate problem with this is in figuring out just how the 17 See Chandra Sekhar Sripada, “The Deep Self Model and Asymmetries in Folk Judgments About Intentional Action,” Philosophical Studies 151 (2010): 159–76. doi: 10.1007/ s11098-009-9423-5.



D . FARACI , D . SHOEMAKER

concordance condition should be specified with respect to responsibility. Here is one plausible possibility: Agents will be viewed as more or less responsible for some action or attitude A solely to the extent that A is viewed as concording more or less with their deep selves.18 Of course, our scales were not put in terms of “responsibility”; rather, they were put in terms of praise- and blameworthiness. But perhaps we might articulate the DSM in terms of those labels as follows: agents will be viewed as more or less _______-worthy for A solely to the extent that A is viewed as concording more or less with their deep selves. Call this the DSMBP. On this application of the DSM, one will be viewed as more blameworthy in correspondence with the perception that A more closely concords with one’s deep self (and A is bad), whereas one will be viewed as more praiseworthy in correspondence with the perception that A more closely concords with one’s deep self (and A is good). Can this explain the difference between our various Toms? At issue is why moral ignorance via childhood deprivation moves one in a positive direction a full step up from control cases in the eyes of subjects. The DSMBP would have us believe that this is because recognition of that sort of moral ignorance affects subjects’ attribution of the relevant actions to the agents’ deep selves. But this does not seem to fit the results of our study. According to the DSMBP, given that TomB is less blameworthy than TomA, it must be that TomB’s action concords less with his deep self—similarly for TomD vs. TomC. But neither view seems motivated, precisely because in each pairing the Toms appear to have identical deep selves. For example, both TomA and TomB fully embrace their racism as adults, and both act on it upon seeing the fallen black man. The only difference is in their upbringings. So why should we (or any subject) therefore think TomB’s action is less reflective of his deep self than TomA’s? One response here would be to suggest that perhaps TomB lacks, or at least is deficient in having, a deep self, given the limited range of moral alternatives to which he was exposed in his upbringing. This might then explain why subjects judge him to be less blameworthy: his action is less concordant with his deep self than TomA’s action insofar as TomB does 18 Indeed, this would be the view Susan Wolf labels the “Deep Self View” of responsibility that she draws from Harry Frankfurt, Charles Taylor, and Gary Watson. See Wolf, pp. 373–9.

MORAL IGNORANCE , PRAISE AND BLAME



not have as robust a deep self as TomA. Unfortunately, making this move undermines the DSMBP explanation of the positive cases, for if a deprived moral upbringing renders one’s deep self less robust or more deficient, and so mitigates attributions of responsibility, then it ought to render TomD less praiseworthy than TomC in the eyes of subjects. But this is the opposite of our results. Perhaps we can get a more plausible version of the DSMBP if we start on the praiseworthy side of the map. How might we explain the differing assessments of TomC and TomD on the DSMBP, given that both Toms were stipulated as identically embracing their racism and so would seem to have identical deep selves? Perhaps the thought is this: TomD’s upbringing somehow made him less committed to the moral beliefs with which he was raised, such that when he goes against them he is more likely to be expressing his actual deep self than is TomC. In other words, given their differences in upbringing, TomC’s action is viewed as more “out of character,” more anomalous, than TomD’s. This thought could then be extended to the negative cases as follows: TomB is less blameworthy than TomA because his upbringing somehow renders his racist action more anomalous than TomA’s. TomB is not “really” committed to his racism, subjects might think, at least in the way that TomA is, so TomB is less blameworthy thereby. This interpretation of the DSMBP preserves the Symmetry Thesis, but how plausible is it? It is unclear to us why subjects would consider TomD’s commitment to his moral beliefs to be any less serious than TomC’s (or that TomB’s is less serious than TomA’s). Indeed, might not their kind of restricted ideological upbringing make them more committed to the belief system into which they had been indoctrinated? Might not this indoctrination create, if anything, a deep self more in line with its exclusively taught, unquestioned principles than an upbringing without it? To the extent that these core “moral” principles were instilled in a way that bypassed the ignorant Toms’ rational, evaluative stance, they are likely to be resistant to such evaluations, much in the way religious belief with its source in childhood indoctrination is often difficult to expunge. To be clear, we do not reject the possibility of reading a deep self view into the results here. It could well be that subjects really are viewing TomD’s commitments as less attributable to his deep self than TomC’s. Nevertheless, given the vignettes as stated, we have no evidence for this approach. Certainly, we welcome any future attempts to see whether



D . FARACI , D . SHOEMAKER

differential deep self commitments are doing any work here. Until then, though, we believe we are licensed in assuming either that the DSMBP supports an asymmetrical approach to blame- and praiseworthiness or that its symmetrical approach is unmotivated and (at least initially) implausible. Focusing on attributions to the deep self alone does not look as if it will help us. In our discussion of the MDE Thesis, we proposed that people’s reactions to JoJo and TomB might stem from an appreciation of the difficulty those agents would have in doing the right thing. We believe that this idea can helpfully be extended to likewise explain our results regarding the Symmetry Thesis: Difficulty Hypothesis: Moral ignorance resulting from childhood deprivation functions symmetrically in both negative and positive cases (moving assessments up the single scale of blameworthiness to praiseworthiness in relation to the control) in virtue of the difficulty agents are viewed as having in overcoming their morally deprived upbringing to grasp the relevant moral reasons.

On this hypothesis, as before, TomB is viewed as less blameworthy than TomA in light of how difficult it would be for him to go for a moral alternative not included in his morally blinkered upbringing. Our further suggestion is that TomD is viewed as more praiseworthy than TomC in light of how difficult it in fact was for him to go for a moral alternative not included in his morally blinkered upbringing. This way of viewing the matter easily explains the results while avoiding the dual implausibility of thinking that subjects think (a) there is a difference in the deep selves for either of the pairings, and (b) childhood indoctrination actually renders one’s deep self more open or oriented toward moral alternatives than nonindoctrinated childhoods. If this is right, then the Symmetry Thesis may not after all be undermined by cases of moral ignorance in upbringing. As noted earlier, however, we need to be clear about just what the Difficulty Hypothesis amounts to. In particular, we need to say something more about the nature of the difficulty in question. For one thing, one might worry that talk of difficulty just collapses into talk of capacity; perhaps, after all, Wolf was right to appeal to capacities. For another, it might be thought that we are suggesting that doing something inherently difficult (like, say, lifting something heavy) is, in itself, sufficient for praise. We agree that this would be implausible. First, even if we grant that talk of difficulty is another way of talking about capacity, at the very least our results suggest that capacity talk must

MORAL IGNORANCE , PRAISE AND BLAME



be scalar. After all, both TomB and TomD are still viewed as _______worthy for what they have done to a significant extent. This means that even though such _______-worthiness is thought to be affected by moral deprivation and ignorance, it is so only to some degree, so that if the Toms are viewed as being incapacitated in some respect it would only be an incapacity by degree. This does not sound, however, like traditional talk of capacity, which is typically taken as either obtaining or not; and it is not clear what the details of this take on capacity would consist in. At the very least, to reintroduce talk of capacity would require some difficult explicatory work about its re-envisioned nature that we are unable to assess sight unseen. Regardless of whether our claims can somehow be adapted to a kind of capacity-talk, our interest is not in capacities themselves, but in the difficulty of exercising certain capacities, where we measure this against a baseline of what comparable agents might be expected to do. Though there is some historical precedence for including difficulty in exercising a capacity in the criteria for moral responsibility, the focus has typically been on volitional capacities.19 It is difficult for the unwilling addict, for example, to resist taking the drug because his counter-desire, his craving for the drug, is so strong. To the extent we cut him some slack, then, we may do so because we think it was just too difficult for him to overcome that volitional obstacle, where the same would be expected of other similarly situated agents. (Of course, this way of putting it might raise capacity talk once more, for perhaps the unwilling addict really is incapacitated with respect to volitional obstacles of that strength, or perhaps he partially lacks a “metacapacity” to exercise his volitional capacities.20) The difficulty we appeal to is analogous to that just mentioned, but differs in that it concerns perceptual, rather than volitional capacities.21 Indeed, there is no reason to think that our Toms are struggling against their own desires and inclinations otherwise to do what is right. Rather, it is just harder 19 See, e.g., a recent thread kicked off by Dana Nelkin on the agency and responsibility blog “Flickers of Freedom”: . 20 One might read Harry Frankfurt in the former way. See his “Freedom of the Will and the Concept of a Person,” in Harry Frankfurt, The Importance of What We Care About (Cambridge: Cambridge University Press, 1988), pp. 11–25. 21 By “perceptual” capacity, we mean whatever capacity allows apprehension of moral qualities in specific cases. We are not assuming that moral knowledge is fundamentally perceptual or empirical.



D . FARACI , D . SHOEMAKER

for our Toms to “see” what the right thing to do is, given their morally deprived upbringings. When they do, it is surprising, for we tend to think that most people from their background would not have seen the light. As an analogy, suppose that I have been shown the famous image of the “duckrabbit” repeatedly since childhood (see Figure 1.3), and I have been taught over and over that what I am looking at is a duck, and only a duck. When, as an adult, I meet you, and you insist that the image can also be seen as a rabbit, it will be no surprise if, given my upbringing, I have a very hard time coming to see it as a rabbit. Certain features of the picture have been drilled into me as exclusively salient (e.g., the bill), so it is quite difficult for me to come to see other features (e.g., the little “rabbit mouth” indentation on the back of the duck’s head) as salient in my assessment of the image. Suppose, however, that someone else present, who was raised just as I was to see the duck, is able to see the rabbit quite quickly once told about it. Especially if we have taken my inability to see the rabbit as representative of the baseline, we are likely to be impressed with this person. We will probably applaud his ability to see past what he is used to. Of course, again, one might suggest that this is best understood in terms of capacities—perhaps he has a stronger “meta-capacity” to exercise his perceptual capacities than I do—but this is not necessary. Regardless of where we come down on capacity-talk, the point remains that this person’s ability to see the rabbit is impressive, and speaks highly of him as a perceiver. Our suggestion is that, on analogy, we are more likely to praise TomD because we are impressed with his ability to “see” past what he is used to, morally speaking. This way of thinking about difficulty also responds to the worry that praiseworthiness might attach to any old difficulty. We are merely advancing the Difficulty Hypothesis to explain cases of moral ignorance given childhood deprivation, and

Figure 1.3 Duckrabbit.

MORAL IGNORANCE , PRAISE AND BLAME



what it appeals to is specifically perceptual difficulty in grasping moral reasons. We take no stand on whether it applies to volitional difficulty, nor other sorts of difficulty for that matter.

Revisiting Huck Our primary aim in this essay has been to bring some empirical results to bear on the MDE and Symmetry theses. While the MDE Thesis looks false in light of our results, there is a weaker, scalar, version of it that may be defensible. And while it may seem as if our results cause real worries for the Symmetry Thesis, it too may be defended. What inclines us toward the latter stance in both cases is the unified defense that may be provided them by the Difficulty Hypothesis. We conclude where we began, with Huckleberry Finn. Huck has been raised to believe that stealing someone’s property is immoral, and that slaves are people’s property. As he grows he continues to mouth and even embrace the racist judgments instilled in him by his family and community since childhood. But when he is finally presented with an opportunity to turn Jim in, he balks, going against his moral beliefs and thus opting for hell instead. He is represented by our TomD, and people view him not only as praiseworthy, but as more praiseworthy than someone just like him but without the childhood moral deprivation. There has been a lot of philosophical ink spilled (or at least a lot of words processed) on the Huck Finn case in recent years.22 Our aim has not been to resolve the many issues raised in these discussions; rather, it has been the far more modest one of providing some much-needed empirical background to them. Far too often, discussion of the case proceeds by stating what our pretheoretic intuitions about Huck are. Ours is the only study we know of that attempts to determine just what those intuitions consist in. People do view him (or someone very much like him) as quite praiseworthy, although not completely so.

22 For a tiny sampling, see Jonathan Bennett, “The Conscience of Huckleberry Finn,” Philosophy 49 (1974): 123–34; Nomy Arpaly, Unprincipled Virtue: An Inquiry Into Moral Agency (Oxford: Oxford University Press, 2003); Nomy Arpaly and Timothy Schroeder, “Praise, Blame, and the Whole Self,” Philosophical Studies 93 (1999): 161–88; Joel J. Kupperman, Character (Oxford: Oxford University Press, 1991); and Craig Taylor, “Moral Incapacity and Huckleberry Finn,” Ratio 14 (2001): 56–67.



D . FARACI , D . SHOEMAKER

Further, we have attempted to show just how much of a role the moral deprivations of someone-like-Huck’s upbringing contribute to people’s assessments of his praiseworthiness. It indeed matters that he has been raised the way he has, moving him up the scale of praiseworthiness a full step over someone who does what he does without such a background. But one surprising feature of the study is that his increased praiseworthiness seems disanalogous to how people view our JoJo lookalike, who is less blameworthy—less responsible?—for his bad deeds than a morally undeprived doppelganger. Indeed, here is the precise rub for determining whether the results of our study negatively impact or ultimately reinforce the Symmetry Thesis: childhood moral deprivations are often thought to mitigate responsibility itself. Being more or less blameworthy may simply be viewed as translating to being more or less responsible. If responsibility is indeed scalar in this fashion, then our results suggest that there is in fact an asymmetry between negative and positive cases: morally deprived upbringings are viewed as decreasing one’s responsibility in the negative realm and increasing one’s responsibility in the positive realm. This would truly be a surprising result. Less surprising, but perhaps no less interesting, would be a different interpretation. Responsibility may not be scalar in the way just suggested; instead, perhaps it is either not scalar at all (one either is responsible for something or one is not), or it is scalar, just not in line with the scalar nature of blame- and praiseworthiness (perhaps it is scalar in line with various capacities one may have or exercise by degrees, where this does not affect degrees of blame- or praiseworthiness). If this is the case, then the Symmetry Thesis might be reinforced. If we think of blame- and praiseworthiness on a single scale, then (where full-blown responsibility would attach to one’s actions at any point on the scale, say) the moral deprivations of childhood could be viewed symmetrically in negative and positive cases as moving one away (by roughly the same amount) from the completely blameworthy endpoint. While we have not taken a definitive stand either way here, our previous discussion on the JoJo case does provide some explanatory ammunition if one adopts the second approach. What could well be moving the assessments of _______-worthiness up the scale away from complete blameworthiness is that people view childhood moral deprivations as making moral perception more difficult for their agents, so that

MORAL IGNORANCE , PRAISE AND BLAME



adhering to, or overcoming, the judgments ingrained by those deprivations will predictably yield assessments at a different degree than their non-deprived counterparts. Just as we cut JoJo some slack for doing what it would have been difficult for him not to see as wrong, we also admire Huck for having done what it was difficult for him to see as right. Or at least we admire Huck in a different way, or to a different degree, than we would those who more easily saw what he did not because of their more privileged upbringing. As his internal monologue suggests, it is a serious and genuine struggle for him to reject “morality” in the way he does. That he did so and got it right (despite what he still thought) is worthy of real praise, apparently. It is a surprise. There is, as always, much more work to be done on these issues. In particular, it would be valuable to know more about the reasons why our positive Toms went against their moral beliefs. Again, we deliberately left this part open so as not to beg the question against any particular theory of relevant moral reasons, but in doing so we also left it open that the Toms might have changed their minds by accident, or under the influence of various non-rational forces. (We doubt this is how the cases were interpreted, but it is certainly possible.) It would also be interesting to know whether it is the mere isolation of their upbringing or the specific moral facts from which they have been deprived that is doing the work on people’s different intuitions. To test this, one might have TomC and TomD raised in different degrees of isolation but now indoctrinated with the same set of moral facts, that people are to be treated equally, etc.; and then have them both act in accordance with their moral beliefs in helping the fallen man up. Whether or not there is a difference in people’s moral intuitions, we would learn something interesting. In addition, the degree of praiseworthiness people would assign may be compared to the degrees assigned in our present cases, so that we may know more about how people assess right-doing in line with moral upbringing. Indeed, finding out how people assess these cases will likely tell us even more about how we ought to view the Symmetry Thesis, and so could ultimately generate real philosophical payoff.23

23 We are very grateful to the folks at Yale’s Experiment Month for carrying out our proposed study on these issues. We are also grateful to Don Callen for serving as advisor to the study, and to Tamler Sommers and Michael McKenna for discussion of some of the ideas herein.

2 The Cognitive Mechanisms of Intolerance Do our Meta-Ethical Commitments Matter? Jennifer Cole Wright, College of Charleston Cullen B. McWhite, College of Charleston Piper T. Grandjean, University of Wyoming

Now more than ever, peaceful coexistence requires open, compassionate dialogue between cultures that have very different sets of beliefs, values, and practices. Unfortunately, such dialogue can encounter intolerance that threatens not only to shut it down but also to generate strong negative reactions—even violence—towards divergent cultures (Haidt et al., 2003; Skitka et al., 2005; Skitka and Mullen, 2002; Wright et al., 2008). Recent scholarship has identified potential sources for this intolerance. Particularly highlighted is the role that people’s moral beliefs and values, especially those strongly held, play in generating negative interpersonal reactions. People report being significantly less supportive of moral diversity than other forms of diversity (Haidt et al., 2003; Wright et al., 2008), especially when encountered in contexts where intimate and/or frequent interaction is likely (e.g., dates, room-mates, workplace). And

Correspondence concerning this chapter should be sent to: Jennifer Cole Wright, Department of Psychology, College of Charleston, 57 Coming Street, Charleston, SC 29424, email: [email protected].

META - ETHICAL COMMITMENTS



even though people are more tolerant of moral diversity in cultures other than their own, this is true only when they believe it to result from different factual—not moral—beliefs (Wainryb, 1993; Wainryb et al., 1998; Wainryb et al., 2004). Believing a disagreement to involve moral beliefs and values is a powerful predictor of intolerance in people of all ages, resulting in less tolerant attitudes towards divergent beliefs and values and less willingness to interact with, help, share resources with, or sit close to those who have them (Wright, 2012; Wright et al., 2008). Similarly, people’s moral conviction predicts their intolerance for divergent beliefs and values, their unwillingness to seek resolution, and their suspicion for legal/ political processes perceived as supporting those divergent beliefs and values (Mullen and Skitka, 2006a, 2006b; Skitka et al., 2005; Skitka and Mullen, 2002).

Meta-Ethical Commitments What is it about people’s moral beliefs and values that drive this lack of tolerance? One plausible explanation comes to us from meta-ethics. Some philosophers hold that people are moral objectivists (Blackburn, 1984; Brink, 1989; Mackie, 1977; Smith, 1994), taking the position 1) that moral judgments are truth-apt, and 2) that their truth is mindindependent (and thus not determined by an individual’s or group’s beliefs, values, customs, etc.). Such a position creates the space for moral beliefs and values to be either (objectively) right or wrong. Given that morality is a normative domain, providing guidance on how things ought to be, this problematizes tolerance: to the extent that people are inclined to think the “right” moral beliefs and values are their own, they will view any that differ as wrong, misguided, even dangerous and immoral (e.g., Rachels and Rachels, 2009; Wong, 1984, 2006; cf. Snare, 1992). Thus, if people are (as many assume) moral objectivists, they are likely to display intolerance of divergent moral beliefs and values. But are people moral objectivists? Though people’s meta-ethical commitments have been a topic of much debate in philosophical circles (Blackburn, 1984; Brink, 1989; Harman, 1975; Mackie, 1977; Rachels and Rachels, 2009; Shafer-Landau, 2003; Smith, 1994; Wong, 1984, 2006), it has received less empirical attention. Indeed, it was not until recently that researchers have begun to systematically explore people’s



J . C . WRIGHT ET AL .

meta-ethical commitments (Goodwin and Darley, 2008, 2010, 2012; Sarkissian et al., 2011; Krettenauer, 2004; Wright et al., 2012; Wright and Sarkissian, 2013; though see also Fishkin, 1984). Of course, given the research highlighting the role morality plays in generating intolerance, it seems reasonable to assume that people are objectivists. And there is plenty of (indirect) evidence for this. People of all ages distinguish between moral and non-moral issues, viewing moral issues as more universally relevant, transcending cultural particulars, and applying more globally to human values/behavior (Nucci and Turiel, 2000; Killen and Nucci, 1995; Nucci, 1981; Smetana, 1981, 1983; Wainryb et al., 1998, 2004; Turiel, 1983, 1998). Moreover, moral considerations are treated as more important than other considerations (Kohlberg, 1969, 1986; Piaget, 1932; Rest, 1979). Moral transgressions are viewed as wrong even in the absence of rules or the presence of social sanction (Smetana, 1981, 1983; Stoddart and Turiel, 1985; Turiel, 1983) and are seen as more serious, less responsedependent, and more severely punishable than social transgressions (Turiel, 1983, 1998; also Davidson et al., 1983; Goodwin and Darley, 2008; Nucci, 1981; Smetana, 1981, 1983; Smetana and Braeges, 1990)— findings that hold cross-culturally (Nucci et al., 2000; Nucci et al., 1983; Song et al., 1987; Turiel et al., 1988). In summary, there is a body of findings that support the hypothesis that people are moral objectivists, viewing divergent moral beliefs and values as violations of objective truths. Yet there is other research that brings this into question. Nichols (2004) found that while most of the people questioned about moral transgressions gave objectivist responses, some did not. Even more interestingly, Goodwin and Darley (2008, 2010, 2012) found that while people gave more objectivist responses for moral issues than other issues, they were nonetheless internally pluralistic in their meta-ethical stance, giving objectivist responses for some moral issues but not others. One potential explanation for this pluralism is that people gave nonobjective groundings for some “moral” issues (identified as such by the researchers) because they did not view them as moral issues. Nonetheless, Wright et al. (2012) found a similar pattern of pluralism when they allowed participants to self-identify moral issues. They also found strong evidence that this pluralism was genuine—i.e., not merely the result of conceptual confusion or different views about morality.

META - ETHICAL COMMITMENTS



Given this surprising meta-ethical variability, what conclusions should we draw about the relationship between people’s meta-ethical commitments and intolerance? Since previous research has failed to adequately disentangle domain classification (moral/non-moral) and grounding (objective/non-objective), we can only hypothesize that people’s metaethical commitments will be strongly related to their tolerance for divergent beliefs and values, along with their willingness to interact with or help the people who endorse them; that, while people will express greater intolerance for divergent moral than non-moral beliefs and values generally, this intolerance will be greatest for objectivelygrounded moral issues. Support for this hypothesis was found by Goodwin and Darley (2012)—asking participants to rate their discomfort for a room-mate with divergent moral beliefs and values revealed a significant correlation (r = .90) between grounding and discomfort. But this leaves an important question yet unanswered: why would people display meta-ethical pluralism (given that it seems an incoherent position)? Given the proposed relationship between grounding and tolerance, our hypothesis is that meta-ethical pluralism serves an important psycho-social function: namely, modulating the level of permissible choice and dialogue about moral issues, both within and between sociocultural groups. Viewing a moral issue as objectively grounded removes it from the realm of legitimate personal/social negotiation (i.e., individual and/or social attempts to condone it will be deemed unacceptable, and censorship/prohibition will be supported). Viewing a moral issue as non-objectively grounded, on the other hand, allows people to acknowledge its moral significance (i.e., that it is not simply a personal matter), while at the same time maintaining room for choice, dialogue, and debate—thus, social censorship/prohibition will be viewed less favorably. The studies reported below are the first that we know of to test this hypothesis.

Study 1 PARTICIPANTS

Seventy-two undergraduate students who were enrolled in Introduction to Psychological Science courses participated in this study for research credit. The data from nine participants were eliminated due to incomplete surveys. Of the remaining sixty-three participants, 78% reported being



J . C . WRIGHT ET AL .

female; 86% Caucasian, 2% African-American, 3% Asian-American, 6% Hispanic. MATERIALS

Participants were given a survey with forty-five issue-statements (e.g. “People should not cheat on their spouse,” “People should wash their bodies regularly,” see Table 2.1), seventeen of which came from Goodwin and Darley (2008) and Wright et al. (2012) and the remainder from Wright et al. (2008). Statement presentation was randomized to control for order effects. Domain classification. For each issue-statement, participants were asked to categorize the issue, choosing the best fit1 from the following: “personal choice/preference,” “social convention/norm,” “moral,” or “scientific fact.” Participants were not given further instructions about these categories; not only does previous research (Wright, 2012; Wright et al., 2012) support the contention that people have a competent conceptual grasp of these categories, but it would have been difficult to provide definitions that would not have influenced participants’ classification. Self-classification was chosen over researcher classification (the method typically employed) because research on the relationship between moral beliefs and values and tolerance (e.g., Wright, 2010; Wright et al., 2008, 2012) has uncovered significant disagreement, between and within age-groups, about what issues belong in the moral domain. This does not appear to be due to conceptual confusion or different views about morality—qualitative analysis of people’s explanations for their classifications suggests they share largely the same ideas about what morality is and what characteristic features it has (at least at its “core”; cf. Haidt, 2012). Rather, disagreement typically involved whether, for a given issue, moral features were present and/or should be prioritized over other features. Therefore, when assessing people’s meta-ethical commitments, it seemed better to not assume the moral status of any issue a priori.

1 Some issues might fit into multiple categories. Therefore, we asked participants to choose whichever category was the best fit, capturing the features that were most salient and/or took priority.

Table 2.1. Domain classification and objectivity grounding percentages, Study 1 Issues Men forcing women to have sex Cheating on lifeguard exam Robbing bank to pay for holiday Incest False testimony for a friend Cheating on spouse Discrimination Eating pets Assisted suicide Death penalty Medical research with animals Protecting the environment Honesty First trimester abortion Gay marriage Not going to war Being born out of wedlock Kids going to school Driving on wrong side of road Driving through red light Keeping bodies clean Shooting gun into crowd Burping/passing gas

Personal

Social

Moral

Scientific fact

Relative

Mixed

Objective

8% 11% 7% 2% 7% 10% 16% 27% 35% 21% 29% 16% 30% 49% 40% 39% 43% 21% 18% 3% 32% 8% 48%

11% 14% 27% 18% 10% 10% 16% 29% 3% 8% 8% 27% 12% 2% 14% 19% 19% 66% 60% 60% 46% 42% 49%

78% 65% 63% 57% 81% 78% 66% 44% 62% 70% 59% 40% 59% 49% 43% 42% 38% 7% 2% 8% 2% 31% 0%

3% 10% 3% 24% 3% 3% 2% 0% 0% 2% 5% 16% 0% 0% 3% 0% 0% 7% 21% 28% 21% 19% 3%

3% 8% 5% 8% 25% 19% 19% 26% 68% 63% 61% 43% 33% 59% 49% 66% 70% 22% 32% 8% 29% 6% 60%

22% 19% 23% 29% 18% 29% 34% 44% 26% 27% 24% 23% 35% 33% 38% 27% 18% 21% 15% 7% 17% 16% 34%

75% 73% 73% 63% 57% 52% 47% 30% 6% 10% 15% 34% 32% 8% 13% 6% 11% 57% 53% 85% 54% 77% 6% (continued)

Table 2.1. Continued Issues Wearing pajamas to seminar Tattoos/piercings Vegetarianism Masturbation Donating money to the poor Believing in God Using recreational drugs Playing violent video games Owning guns Watching pornography Sexual promiscuity Under-age drinking Talking during lecture Parents punishing their children Clouds made of water vapor Jupiter largest planet Earth as center of universe Location of Boston Primary colors red, blue, and yellow 44th President inaugurated Getting aerobic exercise Humans having evolved

Personal

Social

Moral

Scientific fact

Relative

Mixed

Objective

51% 94% 89% 84% 82% 79% 71% 71% 71% 71% 66% 60% 53% 46% 0% 0% 3% 5% 2% 3% 8% 10%

46% 5% 2% 6% 3% 2% 10% 5% 16% 5% 8% 22% 40% 14% 0% 0% 0% 2% 10% 10% 6% 5%

2% 0% 3% 6% 13% 19% 11% 21% 13% 25% 24% 18% 7% 37% 0% 2% 0% 0% 0% 0% 0% 10%

2% 2% 6% 3% 2% 0% 8% 3% 0% 0% 2% 0% 0% 3% 100% 98% 97% 93% 89% 87% 86% 76%

55% 73% 67% 68% 83% 71% 58% 71% 69% 73% 64% 67% 51% 62% 0% 0% 5% 2% 2% 0% 5% 13%

26% 19% 19% 29% 14% 19% 29% 19% 24% 22% 25% 17% 32% 29% 5% 3% 6% 5% 5% 7% 10% 25%

19% 8% 14% 3% 3% 10% 13% 10% 6% 5% 11% 16% 17% 10% 95% 97% 89% 93% 93% 93% 86% 62%

META - ETHICAL COMMITMENTS



Meta-ethical grounding. In line with previous research (Goodwin and Darley, 2008, 2010, 2012; Wright et al., 2012), objectivism vs. nonobjectivism2 was assessed in two related ways. The first was to ask participants whether the issue statement was “true,” “false,” or “just an opinion or attitude”—i.e. a question about the “truth-aptness” of the judgment expressed by the issue statement. This assessed whether participants favored a cognitivist position (i.e., moral judgments can be true or false, like beliefs) or a non-cognitivist position (i.e., moral judgments can be neither true nor false, like attitudes). The second type of assessment was to ask participants to consider someone who had a different judgment than them about the issue statement and then have them select the best characterization of that disagreement, from “the other person would be mistaken,” “it is possible that I would be mistaken, and the other person correct,” or “it is possible that neither I nor the other person would be mistaken—we could both be correct.” This assessed whether participants favored a relativist position (i.e., the truth of the moral judgment is relative to the individual making it and/or the culture to which the individual belongs) or an objectivist position (i.e., the truth is not so relativized). Participants’ overall grounding was calculated by adding these responses together. For the first question, true/false was scored as 1, opinion/attitude as 0; for the second question, either person being mistaken was scored as 1, neither mistaken (both correct) as 0. Together, they gave a possible range of 0–2 (0 = non-objective, 1 = mixed, 2 = objective). Following previous research (Skitka et al., 2005; Wright et al., 2008), participants’ tolerance for divergent beliefs and values was assessed by asking them to consider someone who believed differently than themselves about each issue. They were then asked how willing they would be (0 = not at all; 7 = very willing) to interact with (i.e., “date,” “work with,” and “live in the same town as”) this person, and how willing they would be to help this person, if he/she were to approach them on campus, by “giving him/her change for a campus parking meter” and “dropping something off across campus for him/her.” We labeled the opposite of objectivism “non-objectivism” (as opposed to “relativism,” which we have used elsewhere—Wright et al., 2012) to acknowledge, and remain neutral between, the many different non-objectivist positions (e.g., subjectivism, relativism, noncognitivism, etc.). 2



J . C . WRIGHT ET AL .

PROCEDURE

Participants signed up for the study through an online research system utilized by the psychology department. Upon arrival, informed consents were obtained and participants were given the survey to complete, which took about 40 minutes. Upon completion, they received a debriefing form. RESULTS

Since participants classified the issues for themselves (and thus potentially differently from one another), we calculated the means of objectivity and tolerance for each domain separately for each participant (a technique employed in Wright, 2012; Wright et al., 2008, 2012). This gave every participant a mean for each variable within each of the four domains. Analyses were conducted on these means. Of the forty-five issues, only one was unanimously classified—as factual. Of the remaining forty-four issues, thirteen were dominantly (at least 35% of participants)3 classified as moral, five were dominantly classified as social/conventional norms, thirteen were dominantly classified as personal choice/preference, and seven were dominantly classified as factual. The rest were split between multiple domains (see Table 2.1). Such diversity in domain classification demonstrates the potential methodological advantage of using individual differences (as opposed to assuming domain classification). None of the issues were unanimously given a particular grounding. However, all those classified as social or factual were dominantly (44%+ participants)3 objectively grounded, while those classified as personal issues were dominantly given non-objective groundings. The only domain with significant grounding variation was the moral domain: of the thirteen issues classified as moral, seven were (dominantly) given objective groundings, one was mixed, four were non-objective, and one was split between all three (Table 2.1). Importantly, every participant gave variable groundings to their moral issues. Thus, as in previous studies 3 We calculated the percentage of classification (four response options making 25% the baseline) required to be significantly above chance. With 63 participants, this percentage was 35%. For objectivity groundings, there were three response options (objective/mixed/ relative), so the baseline was 33.3%. The same procedure was used for all classification and grounding comparisons.

META - ETHICAL COMMITMENTS



(Goodwin and Darley, 2008, 2010; Wright et al., 2012), they displayed clear meta-ethical pluralism. But did this meta-ethical pluralism function as we hypothesized? To answer this, we examined participants’ tolerance for divergent beliefs and values as a function of domain classification (personal/social/moral/ factual) and grounding (non-objective/objective). First, we examined the function of people’s meta-ethical commitments specifically: were people more tolerant of disagreement when it involved non-objectively-grounded moral issues than when it involved objectively-grounded issues? A within-participant ANOVA with grounding (non-objective/objective) and interaction type (date/work/live/change/ delivery) revealed main effects on tolerance for both grounding, F(1,52)4 = 228.2, p < 0.001, η2 = 0.81, and interaction type, F(4,208) = 22.9, p < 0.001, η2 = 0.31, as well as a two-way interaction between them, F(4,208) = 8.7, p < 0.001, η2 = 0.14. Participants were significantly less tolerant—that is, less willing to interact with/help someone who disagreed with them about objectively-grounded moral issues than nonobjectively-grounded ones—and this difference was most extreme in those situations (i.e., dating/working with) that required intimate and/ or frequent contact (Figure 2.1). Next, we examined the effect of grounding across the other domains. For instance, while participants clearly viewed non-objectively-grounded moral disagreement as more tolerable than objectively-grounded moral disagreement, how would this tolerance compare to their tolerance for non-objectively-grounded disagreement in other domains? A withinparticipant ANOVA with domain (personal/social/moral)5 and interaction type (date/work/live/change/delivery) revealed main effects on tolerance for both domain, F(2,90) = 4.4, p = 0.014, η2 = 0.1, and interaction type, F(4,180) = 55.7, p < 0.001, η2 = 0.34, but no interaction. Participants were more willing to date, work with, and live in the same town as someone who disagreed with them about (non-objectively-grounded)

4 Degrees of freedom are reduced because there were participants that were dropped from the analysis because they failed to have all the points of comparison (e.g., they had given non-objective but not objective groundings for a particular domain or vice versa). 5 There were not enough issues classified as “factual” that had been given a non-objective grounding to be included in the first analysis. There were also not enough classified as “personal” and given an objective grounding to be included in the second analyses.



J . C . WRIGHT ET AL .

7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 Date

Work with

Live in same town

Non-objective

Give change

Deliver across campus

Objective

Figure 2.1 Difference in tolerance for divergent beliefs and values between nonobjectively- and objectively-grounded moral issues, Study 1.

personal issues than either social or moral issues—though, when it came to helping them, domain did not appear to matter (Figure 2.2). What about those issues that participants had grounded objectively— were they less tolerant of objectively-grounded moral disagreement than other (e.g. factual) forms of objectively-grounded disagreement? A withinparticipant ANOVA with domain (social/moral/factual)5 and interaction type (date/work/live/change/delivery) revealed main effects on tolerance for both domain, F(2,98) = 69.5, p < 0.001, η2 = 0.59, and interaction type, F(4,196) = 23.9, p < 0.001, η2 = 0.33, as well as an interaction between the two, F(8,392) = 2.0, p = 0.041, η2 = 0.04. For objectively-grounded moral issues, people’s tolerance for disagreement plummeted far below their tolerance for other divergent beliefs and values, whether they involved objectively-grounded factual or social issues (Figure 2.3). Together these findings suggest that meta-ethical pluralism does, in fact, perform the psycho-social function hypothesized: while recognizing that the moral status of an issue is by itself related to less tolerance for disagreement, grounding it non-objectively leaves an opening for some diversity. Once a moral issue has been objectively grounded, that opening

META - ETHICAL COMMITMENTS



6.5 6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 Date

Work with

Live in same town

Personal

Give change

Social

Deliver across campus

Moral

Figure 2.2 Difference in tolerance for divergent beliefs and values for nonobjectively-grounded personal, social, and moral issues, Study 1. 6.0

5.0

4.0

3.0

2.0

1.0

0.0 Date

Work with

Live in same town Factual

Social

Give change

Deliver across campus

Moral

Figure 2.3 Difference in tolerance for divergent beliefs and values for objectively-grounded factual, social, and moral issues, Study 1.



J . C . WRIGHT ET AL .

disappears, leaving people not only less willing to interact with anyone having divergent beliefs and values, but less willing to help them as well. A priori coding. Even though we have argued for the methodological value of allowing participants to self-classify issues, we nonetheless felt that it would be worth also examining the data with a pre-assigned domain classification (using, where possible, the categories assigned by Goodwin and Darley, 2008). Specifically, we wanted to verify 1) that participants’ meta-ethical pluralism was still present and 2) that it had the same relationship to participants’ tolerance. We chose eighteen moral and eight social issues (see Table 2.2) and calculated the means for each type of interaction (as previously described). Table 2.2. Grounding for a priori domain classification, Study 1 Domain Issue Moral

Social

Cheat on exam Shooting in crowd Robbing bank Rape Incest False testimony Cheating on spouse Discrimination Protect environment Honesty Eating pets Animal research Abortion Gay marriage Death penalty Assisted suicide Not going to war Donating money Driving through red light Driving on wrong side Going to school Wearing pajamas to seminar Talking during lecture Underage drinking Born out of wedlock Owning guns

Non-objective

Mixed

Objective

8% 7% 6% 4% 8% 24% 20% 18% 41% 34% 27% 61% 55% 48% 59% 65% 66% 82% 7% 31% 24% 54% 54% 63% 66% 69%

17% 18% 20% 23% 25% 20% 28% 35% 25% 34% 45% 25% 31% 38% 30% 25% 27% 15% 7% 13% 21% 28% 30% 21% 21% 23%

75% 75% 75% 73% 66% 56% 52% 46% 34% 32% 28% 14% 14% 14% 11% 10% 7% 3% 86% 56% 55% 18% 17% 15% 13% 8%

META - ETHICAL COMMITMENTS



We conducted a within-participant ANOVA with domain (social/moral), grounding (non-objective/objective), and interaction type (date/work/live/ change/delivery), which revealed main effects on tolerance for all three: domain, F(1,64) = 52.6, p < 0.001, η2 = 0.45, grounding, F(1,64) = 277.3, p < 0.001, η2 = 0.81, interaction type, F(4,256) = 21.1, p < 0.001, η2 = 0.25, along with a significant three-way interaction, F(4,256) = 2.4, p = 0.05, η2 = 0.04. The results were very similar to those found above (Figure 2.4). DISCUSSION

These findings suggest that both the kind of issue people take an issue to be (domain classification) and the type of grounding they give it (objective/non-objective) are important predictors of intolerance for divergent beliefs and values. What is more, it suggests that these are distinct mechanisms that interact with one another. This is a point worthy of reflection. Though many researchers and theorists (including ourselves) have considered “social convention” to 7.0

6.0

5.0

4.0

3.0

2.0

1.0

0.0 Date

Work

Live

Change

Non-objective social

Objective social

Non-objective moral

Objective moral

Delivery

Figure 2.4 Difference in tolerance, a priori categorization, Study 1.



J . C . WRIGHT ET AL .

be synonymous with “non-objective” and “moral” with “objective,” it is clear that participants themselves treat these as orthogonal: there were issues classified as social and yet grounded objectively and, most importantly, issues classified as moral and yet grounded non-objectively. In short, all participants displayed clear pluralism. And this pluralism was, as hypothesized, related to their willingness to interact with or help divergent others. Participants’ non-objectively-grounded moral issues were treated like non-objectively-grounded social issues—they were somewhat less acceptable than matters of personal choice/preference, but nonetheless more acceptable than objectively-grounded moral issues, where disagreement was met with strong intolerance (stronger, even, than for objectively-grounded social issues). This means that, consistent with what advocates for moral relativism have long argued, there is something about an objectively-grounded moral issue (rather than an objectively-grounded social issue or a nonobjectively-grounded moral issue) that brings out the most elevated levels of intolerance towards divergent beliefs and values.

Study 2 While Study 1 examined the effects of domain classification and grounding on participants’ willingness to interact with/help those with divergent beliefs across a variety of contexts, the data was limited in two ways: 1) it did not measure attitudinal tolerance, which previous research (Wright, 2010) suggests can differ from the more behaviorally-oriented measures of tolerance (at least in some age groups), and 2) the helping measures involved actions that were short-lived and minor, requiring little actual contact and lacking “ecological validity” with respect to the collegestudent experience. So, we introduced both a measure of attitudinal tolerance (comfort level with another’s divergent belief) and two new helping measures (including someone in a study group for an exam; letting someone crash on your couch for a week). Additionally, the dominantly non-objectively-grounded moral issues in Study 1 (e.g. assisted suicide, death penalty, gay marriage, etc.) are all issues whose moral status has been a topic of hot debate in the US social and political arena. On the other hand, the dominantly objectivelygrounded issues (e.g., rape, infidelity, discrimination, etc.) are viewed as canonical moral issues—issues for which there is much less disagreement. This raises the possibility that participants mistook consensus for

META - ETHICAL COMMITMENTS



objectivity, or used it as a sort of “proxy.” Of course, the opposite could also be true—participants could have used non-objectivity as a way to acknowledge the existence of diverging opinions, made obvious through the highly publicized social/political debate of the issues. Either way, this would suggest a conflation between grounding and consensus. Consistent with this hypothesis, Goodwin and Darley (2010, 2012) have found a strong correlation (r = 0.84–0.85) between perceived consensus and people’s objectivity ratings across two studies—and a manipulation of perceived consensus influenced people’s objectivity ratings (2012). To further investigate this, we introduced a measure of perceived social consensus. PARTICIPANTS

Participants were 210 students enrolled in the Introduction to Psychological Science course (182 females; 88% Caucasian, 7% African-American, 3% Asian-American, 2% Hispanic) who participated for research credit. MATERIALS AND PROCEDURE

Materials. Participants were given a survey that contained thirty of the forty-five issues used in Study 1 (see Table 2.3), the presentation of which was randomized. Immediately following each issue-statement, participants were asked the same questions as in Study 1. In addition, they were asked a consensus question: “How many people (in general) do you think share your view on this issue?”; an attitudinal tolerance question: “If a person disagreed with you on this issue, how comfortable would you be with that?”; and two new behavioral tolerance (helping) questions: “If a person approached you for help on campus, how willing would you be to help him/her out by: Including him/her in your study group for an exam? Letting him/her crash on your couch for a week?” Procedure. Same as for Study 1. RESULTS

Of the thirty issues, none were unanimously classified into one domain. There were ten issues dominantly (30%+ participants) classified as moral, two dominantly classified as social/conventional norms, eight dominantly classified as personal choice/preference, and four dominantly classified as factual. The rest were split between multiple domains (Table 2.3).

Table 2.3. Domain classification and objectivity grounding percentages, Study 2 Issue Men forcing women to have sex Cheating on a lifeguard exam Incest Cheating on spouse False testimony for a friend Robbing bank to pay for holiday Discrimination Honesty Death penalty Assisted suicide First trimester abortion Gay marriage Shooting gun into crowd Burping/passing gas Driving on wrong side of road Driving through red light Wearing pajamas to seminar Keeping bodies clean Tattoos/piercings Vegetarianism Masturbation

Personal

Conventional

Moral

Scientific fact

Relative

Mixed

%Objective

5% 13% 11% 12% 11% 15% 16% 24% 15% 29% 46% 43% 9% 43% 22% 23% 49% 34% 94% 88% 78%

7% 12% 23% 7% 18% 17% 17% 7% 11% 3% 0% 23% 32% 53% 34% 46% 45% 32% 3% 2% 5%

84% 64% 50% 80% 55% 66% 66% 68% 73% 67% 53% 33% 45% 4% 25% 6% 0% 1% 2% 6% 12%

4% 12% 16% 0% 16% 1% 0% 0% 0% 0% 0% 0% 14% 2% 18% 24% 7% 35% 0% 4% 5%

6% 11% 17% 20% 23% 14% 20% 40% 68% 67% 71% 62% 13% 68% 33% 27% 55% 22% 85% 72% 73%

13% 19% 18% 21% 19% 30% 27% 24% 21% 23% 19% 25% 21% 23% 20% 17% 25% 21% 13% 21% 22%

81% 70% 66% 59% 58% 55% 54% 36% 10% 10% 10% 13% 66% 9% 47% 56% 20% 57% 2% 7% 5%

Watching pornography Donating money to the poor Believing in God Using recreational drugs Playing violent video games Protecting the environment Getting aerobic exercise Humans having evolved 44th President inaugurated

74% 72% 71% 66% 59% 23% 17% 17% 11%

5% 5% 6% 13% 20% 28% 5% 1% 21%

20% 23% 23% 17% 18% 19% 0% 13% 3%

1% 0% 0% 4% 4% 31% 77% 69% 66%

75% 78% 62% 60% 69% 32% 9% 28% 11%

17% 19% 25% 28% 23% 32% 38% 24% 14%

8% 3% 13% 13% 8% 36% 53% 48% 75%



J . C . WRIGHT ET AL .

None of the issues were unanimously given a particular grounding. Of the ten moral issues, seven were dominantly (39% + participants) objectively grounded and three were non-objectively grounded. The two social and factual issues were dominantly objectively grounded; the eight personal issues were dominantly non-objectively grounded. Four of the six split issues were non-objectively grounded and the other two objectively grounded. The same analyses as used in Study 1 were conducted, revealing virtually identical relationships between domain classification, grounding, and behaviorally-oriented tolerance (i.e., interacting/helping divergent others)—therefore, we focused here only on the new measures introduced. We conducted the same analyses as before. First, we examined the function of participants’ meta-ethical commitments by conducting a within-participant ANOVA with grounding (non-objective/objective) and interaction type (tolerance/study-group/couch), which revealed main effects on tolerance for both grounding, F(1,178) = 357.8, p < 0.001, η2 = 0.67, and interaction type, F(2,356) = 208.2, p < 0.001, η2 = 0.54, as well as a two-way interaction between them, F(2,356) = 32.9, p < 0.001, η2 = 0.16. Participants were significantly less tolerant, both in attitudinal tolerance and willingness to help, of disagreement involving objectivelygrounded moral issues than of disagreement involving non-objectively grounded ones—and this difference was most extreme in both their attitudinal tolerance and, once again, in the situation (i.e., sleeping on couch) that required frequent contact in an intimate setting (Figure 2.5). Next, we examined the effect of grounding across the other domains: first for non-objectively-grounded issues and second for objectivelygrounded issues. The first within-participant ANOVA with domain (personal/social/moral) and interaction type (tolerance/study-group/ couch) revealed main effects on tolerance for domain, F(2,266) = 6.4, p = 0.002, η2 = 0.05, and interaction type, F(2,266) = 145.2, p < 0.001, η2 = 0.52, but no interaction. Though there was less distinction made between domains than in Study 1, participants were still slightly more tolerant of disagreement—as well as more willing to allow the disagreeing other into a study group and onto their couch—when that disagreement involved non-objectively-grounded issues that were personal. Paired-sample t-tests further revealed a difference between non-objectively-grounded social and moral issues, but only for attitudinal tolerance, t(147) = 2.6,

META - ETHICAL COMMITMENTS



7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 Tolerance

Study group Non-objective

Sleep on couch Objective

Figure 2.5 Difference in tolerance for divergent beliefs and values between nonobjectively- and objectively-grounded moral issues, Study 2.

p = 0.011—for the helping measures, there was no difference, ts(147) = 1.2–1.5, ns. This suggests that classifying an issue as moral involves increased attitudinal intolerance, even before the increased behavioral intolerance that comes with an objective grounding. The second within-participant ANOVA with domain (social/moral/ factual) and interaction type (tolerance/study group/couch) revealed main effects on tolerance for domain, F(2,262) = 33.4, p < 0.001, η2 = 0.2, and interaction type, F(2,262) = 157.2, p < 0.001, η2 = 0.55, but no interaction. Once again, for objectively-grounded moral issues, participants’ tolerance for disagreement was far less than their tolerance for other (non-moral) objectively-grounded issues (Figure 2.6). Perceived consensus? Did participants’ reports of consensus differ between domains? A within-participant ANOVA with domain (personal/social/moral/factual) revealed a main effect on consensus for domain, F(3,612) = 235.5, p < 0.001, η2 = 0.54. Participants reported the highest level of consensus for factual issues (M = 6.2, SE = 0.04) and the lowest for personal issues (M = 4.7, SE = 0.04). Consensus for social



J . C . WRIGHT ET AL .

6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 Tolerance

Study group

Sleep on couch

Non-objective personal

Non-objective moral

Objective social

Non-objective social

Objective factual

Objective moral

Figure 2.6 Difference in tolerance for divergent beliefs and values for nonobjectively-grounded personal, social, and moral issues and objectively-grounded factual, social, and moral issues, Study 2.

norms/conventions (M = 5.6, SE = 0.06) and moral issues (M = 5.4, SE = 0.04) fell between—and were significantly different from each other, t(204) = 2.3, p = 0.024. Consensus was also significantly correlated with grounding in all four domains, rs(210) = 0.18–0.36, ps < 0.007–0.001. Does this mean that participants used consensus as a proxy for objectivity? To investigate this, we utilized a standard mediation model (Baron and Kenny, 1986) for both willingness to interact with and willingness to help those who disagree. First, we separately regressed willingness to interact and willingness to help onto grounding, revealing that it was predictive of both (interact: B = –0.51, p = 0.001; help: B = –0.39, p = 0.025). Next, grounding was found to be predictive of perceived consensus (B = 0.5, p < 0.001). And finally, when entering grounding and consensus

META - ETHICAL COMMITMENTS



together into each equation (one for willingness to interact and one for willingness to help), grounding ceased to be predictive (interact: B = –0.2, ns; help: B = –0.18, ns), while consensus remained so (interact: B = –0.45, p < 0.001; help: B = –0.44, p = 0.001). Sobel tests reveal that the mediational (indirect) effect of perceived consensus on the relationship between grounding and people’s willingness to interact with/help those who disagree to be significant in both cases, Zs = –3.3 and –2.9, ps < 0.001. This supports the view that perceived consensus was a complete mediator of the relationship between grounding and tolerance—and not the other way around. In other words, rather than participants using perceived consensus as a proxy for objectivity, it is more likely that objectivity served as an indicator of how much consensus they could reasonably expect. DISCUSSION

Study 2 confirmed and expanded upon the findings from Study 1, demonstrating a similar, if attenuated, role for meta-ethical pluralism in people’s attitudinal tolerance compared to that found for more behaviorally-oriented measures of tolerance—a role that was present even when those measures were expanded to include more ecologically valid measures. With respect to the relationship between objectivity and perceived consensus, we found that once we controlled for consensus, grounding no longer directly predicted participants’ attitudinal/behavioral intolerance. Indeed, our mediational analyses suggest that it is more likely that objectivity serves as an indicator of how much consensus can be reasonably expected, which is in turn connected to how acceptable expressions of disagreement would be—and, correspondingly, how much intolerance towards that disagreement it would be acceptable to express.

Study 3 Studies 1 and 2 demonstrated not only that people are meta-ethical pluralists, but also that this pluralism predicts participants’ attitudinal/ behavioral tolerance for divergent beliefs and values. But our hypothesis about meta-ethical pluralism also involved people’s attitudes about the social condoning vs. prohibiting of behaviors, which we had not yet tested directly. Study 3 was designed to do this. We hypothesized that individual and/or social attempts to engage in and/or condone behaviors



J . C . WRIGHT ET AL .

involving objectively-grounded moral issues would be rejected, while censorship/prohibition would be supported. Behaviors involving nonobjectively-grounded moral issues, on the other hand, would be treated as still open for debate and discussion, making censorship/prohibition less acceptable. We also took this opportunity to include participants from Amazon’s Mechanical Turk (M-Turk), which is more representative of the general US population. PARTICIPANTS

We had 311 participants, of which 148 were students enrolled in Introduction to Psychological Science (66% females; 94% Caucasian, 2% African-American) participating for research credit, and 163 were participants from M-Turk (46% females; 80% Caucasian, 11% AfricanAmerican, 4% Asian-American, 2% Hispanic). The college students were dominantly freshmen (18–21 years of age), while 16% of the M-Turk participants were aged 15–24, 50% aged 25–34, 12% aged 35–44, 11% aged 45–54, and the remainder +55. Data from 49 participants (all M-Turk) were eliminated due to incomplete surveys, leaving 262 participants. MATERIALS AND PROCEDURE

Materials. Participants completed a survey containing twenty issue-statements, including new issues (see Table 2.4). Domain classification and grounding were measured using the same questions as before (exception: “scientific fact” was removed as a classification option). We also introduced new tolerance questions: specifically, we asked “how acceptable would it be for someone to engage in [the action mentioned]?” and “how comfortable would you feel if that person was a family member/friend” or “ . . . a member of [your/a different] society?” We also asked questions about societal reactions more generally—e.g., how acceptable would it be “for people to socially shun someone who does this?”, “for [our/a different] society to pass a law prohibiting it?”, “for [our/a different] society to condone/support it?” Finally, there were questions about perceived consensus, this time encompassing both the participant’s own and a different community.

META - ETHICAL COMMITMENTS



Procedure. For the college students, this was the same. The other participants signed up for an HIT online through M-Turk, getting paid $0.25 for the completion of the survey. RESULTS

Of the twenty issues, none were unanimously classified into one domain. There were seven issues dominantly (39% + participants) classified as moral: two as social/conventional norms, and six as personal choice/ preference. The rest were split between multiple domains (Table 2.4). None of the issues were unanimously given a particular grounding. Of the seven moral issues, two were dominantly (56% + participants) objectively grounded; four were non-objectively grounded. The two social issues were dominantly non-objectively grounded; five of the six personal issues were dominantly non-objectively grounded, while one issue (choosing a major) was dominantly objectively grounded. Four of the six split issues were non-objectively grounded; the other two were objectively grounded. For each domain, we collapsed the tolerance variables into two summary variables, the first of which included the questions of attitudinal tolerance for, and social condoning of, divergent belief/values (labeled “support”; Chronbach’s alpha Æ: personal = 0.92, social = 0.96, moral = 0.94) and the second of which included the questions of social shunning/prohibition (labeled “prohibit”; Chronbach’s alpha Æ: personal = 0.88, social = 0.89, moral = 0.79). We then examined the function of participants’ meta-ethical commitments by conducting two paired-sample t-tests (one for each summary variable), revealing that participants were both more supportive of and less willing to prohibit divergent beliefs and values for those moral issues they had grounded non-objectively than for those they had grounded objectively: support, t(135) = 10.4, p < 0.001; prohibit, t(136) = –11.8, p < 0.001 (Figure 2.7). Next, we examined the effect of non-objective grounding across the domains by conducting two within-participant ANOVAs with domain (personal/social/moral)—again, one for each summary variable. This revealed main effects for domain on both types of tolerance: for support, F(2,202) = 61.7, p < 0.001, η2 = .38; for prohibit, F(2,198) = 61.8, p < 0.001, η2 = 0.38. For those issues participants had grounded non-objectively, they were the most willing to support, and the least willing to prohibit, divergent beliefs and values about personal issues—while being the least willing to

Table 2.4. Domain classification and objectivity grounding percentages, Study 3 Total Selling children on Internet Owning slaves Buying forced-labor products Destroying religious symbol Overcharging customers Cheating on spouses Abortions Mother smoking while pregnant False reporting on taxes Incest Cloning body parts Burning flags Engaging in vigilante justice Eating human flesh Piercings and tattoos Choosing a major Driving fuel-inefficient cars Eating factory-farmed animals Failing to recycle Same-sex marriage

Personal

Social

Moral

Relative

Objective

2% 4% 3% 13% 17% 25% 42% 42% 12% 11% 42% 23% 26% 29% 86% 81% 70% 68% 54% 46%

15% 16% 25% 20% 20% 15% 3% 4% 38% 43% 13% 48% 41% 38% 10% 14% 19% 11% 29% 25%

84% 81% 72% 67% 63% 60% 54% 54% 50% 45% 45% 30% 33% 33% 4% 4% 10% 20% 16% 30%

38% 40% 63% 66% 61% 71% 59% 81% 44% 55% 87% 71% 81% 56% 66% 42% 76% 79% 83% 65%

62% 60% 37% 34% 39% 29% 41% 19% 56% 45% 13% 29% 19% 44% 34% 58% 24% 21% 17% 35%

META - ETHICAL COMMITMENTS



7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 Support

Prohibit Non-objective

Objective

Figure 2.7 Difference in tolerance for divergent beliefs and values between nonobjectively- and objectively-grounded moral issues, Study 3.

support, and the most willing to prohibit, divergent beliefs and values about moral issues. And, finally, we examined the effect of objective grounding across domains by conducting two paired-sample t-tests—one for each summary variable—that compared the social and moral domains (there was no factual domain). This revealed once again a significant difference between the social and moral domains—people were more willing to support, and less to prohibit, divergent beliefs and values about objectively-grounded social issues than they were objectively-grounded moral issues (Figure 2.8). Same vs. different cultures. Previous research found people’s judgments about moral transgressions to vary as a function of whether they involved members of the same or different cultures (Sarkissian et al., 2011; Wainryb et al., 1998, 2004). And paired-sample t-tests revealed that our participants were also somewhat more accepting of divergent moral beliefs and values in a different society than in their own, ts (257) = 1.9–2.5, ps = 0.05–0.013. A closer look revealed that this was



J . C . WRIGHT ET AL .

7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0

Support

Prohibit

Non-objective personal

Non-objective moral

Non-objective social

Objective social

Objective moral

Figure 2.8 Difference in tolerance for divergent beliefs and values for nonobjectively-grounded personal, social, and moral issues and objectively-grounded social and moral issues, Study 3.

only true for the non-objectively-grounded moral issues, ts(204) = 2.8–2.9, ps = 0.002–0.004; for objectively-grounded moral issues, there was no difference in participants’ attitudes about the condoning or prohibition of divergent beliefs and values, regardless of whether it involved their own or a different society, ts(181) = 0.05–0.8, ns. Perceived consensus? We tested the mediational relationship discovered in Study 2 using our summary variables and both perceived consensus variables. Starting with consensus within one’s own community, we separately regressed willingness to support and willingness to prohibit divergent moral beliefs and values onto grounding, revealing it to be predictive of both (support: B = –0.49, p = 0.008; prohibit: B = 0.74, p < 0.001). Next, we found grounding to be predictive of consensus (B = 0.34, p = 0.004). And finally, when entering grounding and consensus together into each equation (one for willingness to support and one for

META - ETHICAL COMMITMENTS



willingness to prohibit), grounding ceased to be predictive of willingness to support (B = –0.32, ns) but not of willingness to prohibit (B = 0.59, p < 0.001), while consensus remained so for both (support: B = –0.5, p < 0.001; prohibit: B = 0.44, p < 0.001). Sobel tests reveal the mediational (indirect) effect of perceived consensus (within one’s own community) on the relationship between grounding and people’s willingness to support and their willingness to prohibit divergent moral beliefs to be significant, Zs = –2.5 and 2.5, ps < 0.001. The same pattern was discovered using consensus within a different community: as above, grounding was predictive of willingness to support and to prohibit (support: B = –0.49, p = 0.008; prohibit: B = 0.74, p < 0.001) and grounding was also predictive of consensus (B = 0.38, p < 0.001). And when grounding and consensus were entered together into each equation, grounding ceased to be predictive of willingness to support (B = –0.28, ns), but not willingness to prohibit (B = 0.54, p < 0.001), while consensus remained so for both (support: B = –0.5, p < 0.001; prohibit: B = 0.44, p < 0.001). Sobel tests reveal the mediational (indirect) effect of perceived consensus (within a different community) on the relationship between grounding and people’s willingness to support and their willingness to prohibit divergent moral beliefs to be significant, Zs = –2.9 and 3.0, ps < 0.001. These results suggest that both measures of perceived consensus functioned as complete mediators of the relationship between objectivity and willingness to support divergent moral beliefs and values and as partial mediators of the relationship between objectivity and the willingness to prohibit divergent moral beliefs and values. So, once again, this suggests that one plausible explanation for the relationship between people’s meta-ethical commitments and their willingness to accept/condone divergent beliefs and values is that taking an objectivist stance involves—and possibly even generates—the expectation of stronger and more widespread social consensus, which makes divergent beliefs less acceptable (and makes those who hold them stand out as outsiders). Interestingly, perceived consensus only partially mediated the relationship between grounding and willingness to prohibit. Grounding remained independently predictive, though its effect, after controlling for consensus, was somewhat reduced. This suggests that, when it comes to advocating for prohibition, not only is an objectivist meta-ethical stance related to the expectation for stronger social consensus (making



J . C . WRIGHT ET AL .

disagreement less acceptable), but also it independently supports the acceptability of social prohibition: if something (such as selling children on the Internet) is viewed as objectively wrong to do, then outlawing it seems appropriate, and even desirable, no matter how many people agree.

General discussion Taken together, the studies reported here provide support for the developing view that people are meta-ethical pluralists, treating morality as a complex, heterogeneous set of issues—some grounded by the beliefs, values, and practices of individuals/cultures, and some by mind-independent features of the world (such as unwarranted harm) that run deeper. These studies also provided support for the hypothesis that metaethical pluralism performs a particular psycho-social function. While simply viewing an issue to be moral is enough to generate discomfort and avoidance of divergent others—relative to people who viewed the same issue as one of personal choice/preference—nonetheless grounding it objectively appears to heighten that intolerance, being related to a dramatic drop in private and public support for the divergence and an increased willingness to introduce social prohibition against it. Why this is the case is not fully understood at this point, though perceived consensus appears to play a role. If our mediation model is correct, then one important difference between participants’ non-objective and objectively grounded beliefs is that the latter include a sense of agreement, of having a shared belief system. This suggest that objectivelygrounded moral issues involve (and perhaps even generate) an expectation of consensus—not just that everyone else should agree with me, but that everyone does agree with me. Such an expectation makes anyone who disagrees an outsider: an out-group member worthy of rebuke. Of course, this does not explain why we also found greater intolerance for objectively-grounded moral issues than objectively-grounded social issues. Future research is clearly needed to delve more deeply into the cognitive/affective differences between the two domains. But, meanwhile, it seems safe to say that when it comes to promoting the acceptance of divergent beliefs and values, yes—people’s meta-ethical commitments do matter. Of course, this means they matter just as much if our goal is the opposite (i.e. wanting to lessen people’s tolerance for particular beliefs

META - ETHICAL COMMITMENTS



and values—e.g. gender inequality, domestic violence). It is with this in mind that we would like to end with a story. Every year our university selects a “college reads” book for entering freshman year. This year it was Eating Animals, by Jonathan Safran Foer. This turned out to be fortunate, because it serves as an interesting concrete example of the psycho-social function of meta-ethical pluralism. In the book, Foer does two things: 1) he clearly establishes that our mass-production of animals for consumption is a deeply problematic moral crisis, and 2) he simultaneously maintains that people must be allowed to choose how to respond to this crisis for themselves. While for many card-carrying vegetarians/vegans this latter step feels like an unnecessary (to some, even unacceptable) “pulling of the punches,” our research suggests that Foer’s approach is likely to achieve something that other approaches—i.e. those that unflinchingly hammer on the hard (“objective”) moral facts of the matter—may not, which is appealing to (and helping to shift the thinking of) those who are not already part of the “choir.” Consider: over a thousand copies of the book were purchased for dissemination and discussion throughout the year, by both students and faculty—and thus it was read by many people (younger and older) who do not agree with, but might be influenced by, his view. On the other hand, several more “hard-line” books on the subject (i.e. books that, like Foer’s, established the first point but that then rejected the second point in favor of mandatory social intervention/ legal prohibition) were rejected as too extreme and therefore disrespectful of the wide and varied opinions of the targeted audience. So while the message of these books may be just as important, it is not being heard. If our thesis about meta-ethical pluralism is correct, then perhaps people must first be willing/able to recognize the moral significance of an issue before they can be convinced that it mandates changes in their behavior—and the shifting of an issue into the moral domain takes time; it takes personal/social reflection, discussion, and respect for different perspectives. It is also likely that only after an issue has found solid footing in people’s moral conscience can they begin to collectively recognize, and act in accordance with, its full moral weight.



J . C . WRIGHT ET AL .

References Baron, R.M., and Kenny, D.A. (1986). The Moderator-Mediator Variable Distinction in Social Psychological Research: Conceptual, Strategic, and Statistical Considerations. Journal of Personality and Social Psychology 51: 1173–82. Blackburn, S. (1984). Spreading the Word: Groundings in the Philosophy of Language. New York: Oxford University Press. Brink, D.O. (1989). Moral Realism and the Foundations of Ethics. New York: Cambridge University Press. Davidson, P., Turiel, E., and Black, A. (1983). The Effect of Stimulus Familiarity on the Use of Criteria and Justifications in Children’s Social Reasoning. British Journal of Developmental Psychology 1(1): 49–65. Fishkin, J.S. (1984). Beyond Subjective Morality. New Haven, CT: Yale University Press. Goodwin, G., and Darley, J. (2008). The Psychology of Meta-Ethics: Exploring Objectivism. Cognition 106: 1339–66. Goodwin, G., and Darley, J. (2010). The Perceived Objectivity of Ethical Beliefs: Psychological Findings and Implications for Public Policy. Review of Philosophy and Psychology 1: 1–28. Goodwin, G., and Darley, J. (2012). Why are Some Moral Beliefs Perceived to be More Objective than Others? Journal of Experimental Social Psychology 48: 250–6. Haidt, J. (2012). The Righteous Mind. London: Allen Lane. Haidt, J., Rosenberg, E., and Hom, H. (2003). Differentiating Diversities: Moral Diversity is Not Like Other Kinds. Journal of Applied Social Psychology 33: 1–36. Harman, G. (1975). Moral Relativism Defended. The Philosophical Review 84: 3–22. Killen, M., and Nucci, L. (1995). Morality, Autonomy, and Social Conflict. In M. Killen and D. Hart (eds.), Morality in Everyday Life: Developmental Perspectives. New York: Cambridge University Press, pp. 52–86. Kohlberg, L. (1969). Stage and Sequence: The Cognitive-Developmental Approach to Socialization. In D. Goslin (ed.), Handbook of Socialization Theory and Research. Chicago: Rand McNally, pp. 347–480. Kohlberg, L. (1986). A Current Statement on some Theoretical Issues. In S. Modgil and C. Modgil (eds.), Lawrence Kohlberg: Consensus and Controversy. New York: Taylor and Francis Books, Ltd., vol. 1, pp. 485–546. Krettenauer, T. (2004). Metaethical Cognition and Epistemic Reasoning Development in Adolescence. International Journal of Behavioral Development 28(5): 461–70.

META - ETHICAL COMMITMENTS



Krosnick, J.A., Boninger, D.S., Chaung, Y.C., Berent, M.K., and Carnot, C.G. (1993). Attitude Strength: One Construct or Many Related Constructs? Journal of Personality and Social Psychology 65: 1132–51. Kuhn, D., Cheney, R., and Weinstock, M. (2000). The Development of Epistemological Understanding. Cognitive Development 15: 309–28. Mackie, J.L. (1977). Ethics: Inventing Right and Wrong. New York: Penguin. Mullen, E., and Skitka, L.J. (2006a). When Outcomes Prompt Criticism of Procedures: An Analysis of the Rodney King Case. Analyses of Social Issues and Public Policy 6(1): 1–14. Mullen, E., and Skitka, L.J. (2006b). Exploring the Psychological Underpinnings of the Moral Mandate Effect: Motivated Reasoning, Group Differentiation, or Anger? Journal of Personality and Social Psychology 90: 629–43. Nichols, S. (2004). After Objectivity: An Empirical Study of Moral Judgment. Philosophical Psychology 17: 3–26. Nucci, L. (1981). Conceptions of Personal Issues: A Domain Distinct from Moral or Societal Concepts. Child Development 52: 114–21. Nucci, L., and Turiel, E. (2000). The Moral and the Personal: Sources of Social Conflict. In L. Nucci, G.B. Saxe, and E. Turiel (eds.), Culture, Thought, and Development. Mahwah, NJ: Lawrence Erlbaum Associates, pp. 115–37. Nucci, L., Turiel, E., and Encarnacion-Gawrych, G. (1983). Children’s Social Interactions and Social Concepts: Analyses of Morality and Convention in the Virgin Islands. Journal of Cross-Cultural Psychology 14(4): 469–87. Nucci, L., Saxe, G., and Turiel, E. (2000). Culture, Thought, and Development. Mahwah, NJ: Lawrence Erlbaum Associates. Piaget, J. (1932). The Moral Judgment of the Child. London: Routledge and Kegan Paul. Rachels, J., and Rachels, S. (2009). The Elements of Moral Philosophy. New York: McGraw Hill. Rest, J.R. (1979). Development in Judging Moral Issues. Minneapolis: University of Minnesota Press. Sarkissian, H., Park, J., Tien, D., Wright, J.C., and Knobe, J. (2011). Folk Moral Relativism. Mind and Language 26(4): 482–505. Shafer-Landau, R. (2003). Moral Realism: A Defence. New York: Oxford University Press. Skitka, L.J., Bauman, C.W., and Sargis, E.G. (2005). Moral Conviction: Another Contributor to Attitude Strength or Something More? Journal of Personality and Social Psychology 88: 895–917. Skitka, L.J., and Mullen, E. (2002). The Dark Side of Moral Conviction. Analyses of Social Issues and Public Policy 7: 35–41.



J . C . WRIGHT ET AL .

Smetana, J.G. (1981). Preschool Children’s Conceptions of Moral and Social Rules. Child Development 52: 1333–6. Smetana, J.G. (1983). Social-Cognitive Development: Domain Distinctions and Coordinations. Developmental Review 3: 131–47. Smetana, J.G., and Braeges, J. (1990). The Development of Toddlers’ Moral and Conventional Judgments. Merrill-Palmer Quarterly 36(3): 329–46. Smith, M. (1994). The Moral Problem. Oxford: Blackwell. Snare, F. (1992). The Nature of Moral Thinking. London: Routledge. Song, M., Smetana, J., and Kim, S.Y. (1987). Korean Children’s Conceptions of Moral and Conventional Transgressions. Developmental Psychology 23(4): 577–82. Stoddart, T., and Turiel, E. (1985). Children’s Concepts of Cross-Gender Activities. Child Development 56(5): 1241–53. Turiel, E. (1983). The Development of Social Knowledge: Morality and Conventions. New York: Cambridge University Press. Turiel, E. (1998). The Development of Morality. In N. Eisenberg (ed.), Handbook of Child Psychology, vol. 3: Social, Emotional, and Personality Development (5th edn). New York: Wiley, pp. 701–78. Turiel, E., Nucci, L., and Smetana, J.G. (1988). A Cross-Cultural Comparison about What? A Critique of Nisan’s (1987) Study of Morality and Convention. Developmental Psychology 24(1): 140–3. Wainryb, C. (1993). The Application of Moral Judgments to Other Cultures: Relativism and Universality. Child Development 64: 924–33. Wainryb, C., Shaw, L., and Maianu, C. (1998). Tolerance and Intolerance: Children’s and Adolescents’ Judgments of Dissenting Beliefs, Speech, Persons, and Conduct. Child Development 69(6): 1541–55. Wainryb, C., Shaw, L., Langley, M., Cottam, K., and Lewis, R. (2004). Children’s Thinking about Diversity of Belief in the Early School Years: Judgments of Relativism, Tolerance, and Disagreeing Persons. Child Development 75(3): 687–703. Wong, D.B. (1984). Moral Relativity. Berkeley: University of California Press. Wong, D.B. (2006). Natural Moralities: A Defence of Pluralistic Relativism. New York: Oxford University Press. Wright, J.C. (2012). Children’s and Adolescents’ Tolerance for Divergent Beliefs: Exploring the Cognitive and Affective Dimensions of Moral Conviction in our Youth. British Journal of Developmental Psychology. Wright, J.C. and Sarkissian, H. (2013). Folk Meta-Ethical Commitments. In F. Allhoff, R. Mallon, and S. Nichols (eds.), Philosophy: Traditional and Experimental Readings. Oxford: Oxford University Press.

META - ETHICAL COMMITMENTS



Wright, J.C., Cullum, J., and Schwab, N. (2008). The Cognitive and Affective Dimensions of Moral Conviction: Implications for Tolerance and Interpersonal Behaviors. Personality and Social Psychology Bulletin 34(11): 1461–76. Wright, J.C., Grandjean, P., and McWhite, C. (2012). The Meta-Ethical Grounding of our Moral Beliefs: Evidence for Meta-Ethical Pluralism. Philosophical Psychology 26: 336–61, doi:10.1080/09515089.2011.633751.

3 Doing, Allowing, and Enabling Harm An Empirical Investigation Christian Barry, Australian National University Matthew Lindauer, Yale University Gerhard Øverland, University of Oslo

Introduction Moral philosophers examining the nature and significance of human conduct involving harm have tended to employ a bipartite distinction between doing harm and allowing harm. This distinction has generally been treated as exhaustive—all cases of human conduct involving harm have typically been thought to be descriptively and normatively subsumable under one or the other of these two categories. Enabling harm, in particular, has generally not been treated as an independent category of human conduct. By contrast, cognitive scientists and psychologists have We have benefited from discussions of earlier versions of this chapter with members of the Experimental Philosophy Lab at Yale, the Yale Global Justice Program, and participants in the Workshop on Enabling Harm at the University of Oslo in June 2012. The authors are grateful for comments on this chapter that we received from Steve Guglielmo, Mark Sheskin, Brent Strickland, and three anonymous reviewers, and especially to Joshua Knobe and Tania Lombrozo for their detailed comments on successive written drafts. We would also like to thank Kevin Callender and Elizabeth Roberto for their assistance with the statistical analyses used in the chapter. Barry, Lindauer, and verland are full joint authors of the chapter. The theoretical aspects of the doing–allowing–enabling harm distinction are based on Barry and verland’s previous and forthcoming work. Lindauer had the primary role in conducting the experiments and positioning the chapter in relation to cognitive scientific and psychological research.

DOING , ALLOWING , AND ENABLING HARM



studied the concept of “enabling” as distinct from “cause” and other causal terms. Yet “enabling” and “enabling harm” have not yet been the focus of empirical research on judgments about cases of human conduct involving harm. Empirical studies pertaining to these judgments have generally employed the traditional bipartite doing–allowing harm distinction. In this chapter we marshal new empirical evidence from two studies to argue that well-known cases in the moral philosophical literature are cases of enabling harm, rather than doing or allowing harm. This suggests that enabling harm is judged to be a distinct category of human conduct, not descriptively subsumable under the categories of doing or allowing harm. A third study provides evidence that it is judged to be normatively distinct too. The results of this third study suggest that in assigning compensatory responsibility, enabling harm is regarded as more like doing harm than allowing harm when an agent is culpable, but more like allowing harm than doing harm when an agent is innocent. Given this evidence, we argue that moral philosophers and moral psychologists interested in human conduct involving harm should employ a tripartite doing–allowing–enabling harm distinction, rather than the traditional bipartite doing–allowing harm distinction.

The moral philosophical literature A vast literature in moral philosophy is devoted to examining the meaning and moral significance of the distinction between doing harm and allowing harm. While theorists have employed different terms in characterizing these notions (e.g. Bennett 1995; Donagan 1977; Foot 1994; Quinn 1994), the contributors to this literature have generally taken themselves to be giving competing accounts of the same underlying conceptual distinction. In some instances, agents actively contribute to, do, or bring about harm, whereas in other instances they merely allow or fail to prevent the occurrence of harm. A great many philosophers have held that all instances of human conduct involving harm are accurately described either as instances of doing harm or as instances of allowing harm. The notion of enabling harm has been given very little attention, and where it has been discussed it has generally been thought to be descriptively subsumable under one or the other of these categories.1

1

See, for instance, Foot (1994).



C . BARRY ET AL .

Whether a given case of enabling harm is viewed as a case of doing harm or allowing harm has generally been related to specifics of the philosopher’s account of the doing–allowing distinction. The interest in understanding this distinction between doing harm and allowing harm is due to the fact that it is generally thought to be normatively significant.2 All other things being equal, reasons against doing harm are commonly thought to be more stringent than reasons against allowing harm. They are more stringent in the sense that they constrain agents: prospective doers of harm cannot so easily justify their conduct by appealing to the costs to themselves of refraining from doing harm, or by appealing to the overall good that their conduct will bring about. For example, we generally do not think that we can break someone’s arm to prevent one of our own arms from being broken, or to prevent the breaking of two other persons’ arms, all else being equal. But it seems more acceptable, at the very least, to allow someone’s arm to be broken if it will prevent the same harm from coming to ourselves or to two other people. Reasons against doing harm are also thought to be more stringent than reasons against allowing harm in the sense that they demand more of agents after the harm has eventuated. If, for instance, one has broken another person’s arm, we typically think that one must take on significant cost to compensate this person, whereas if one merely allowed the person’s arm to break, one is far less or perhaps not at all responsible for compensating them. In other words, one has more demanding compensatory responsibilities if one has done harm than if one has allowed the same harm. The main topic of this chapter and the first two experiments that we will discuss is the descriptive classification of well-known cases of doing harm, allowing harm, and (what we will argue are cases of) enabling harm in the moral philosophical literature. While our aim here is not primarily to document or elaborate upon purported normative 2 In this chapter we do not discuss the related act–omission distinction, which is also generally taken to be normatively significant. We remain neutral on the question of whether these distinctions perfectly map onto one another for the purposes of this chapter. If they indeed do, it would seem that the act–omission distinction is unable to distinguish acts that are instances of doing harm from acts that are instances of enabling harm. Since we do not focus on the differences in moral judgments that are made about cases of doing harm, enabling harm, and allowing harm or acts and omissions, we also do not discuss in this chapter the empirical literature on the “omission bias,” describable as the preference for harms caused by omissions over equal or lesser harms caused by acts.

DOING , ALLOWING , AND ENABLING HARM



differences between doing and allowing harm, the existence of a third category of human conduct could turn out to be important for normative assessment. It would be important, for instance, if enabling harm is normatively distinct from doing and allowing harm. We cannot fully argue in this chapter that enabling harm is indeed normatively distinct from doing harm and allowing harm, but there is reason to believe that it may well be. For example, one normative characteristic that reasons against enabling harm appear to have in common with reasons against allowing harm is that they constrain conduct less with regard to shifting cost from one person or group of people to some other people than reasons against doing harm (Hanser 1999; Rickless 2011). At the same time, however, reasons against enabling harm appear to constrain conduct more with regard to shifting cost from ourselves to others than do reasons against allowing harm. If reasons against enabling harm are indeed normatively distinct from reasons against doing harm and allowing harm, then failing to focus on this category could invite quite serious moral misunderstanding. Barry and verland (2012) have argued, for instance, that by failing to recognize the category of enabling harm, participants in debates on global justice have often tended to talk past each other and to make implausible claims about the nature of obligations to address global poverty. The results of the third experiment discussed in this chapter constitute suggestive evidence that enabling harm is regarded as normatively distinct from doing harm and allowing harm when it comes to compensatory responsibility. We hope in further research to investigate how aspects of the normative status of enabling harm are generally understood.

Cognitive scientific and psychological research Cognitive scientists and psychologists have studied the representation of different kinds of causal relations. Wolff (2007) finds evidence for a “dynamics model,” based on the force–dynamics theory originally proposed by Talmy (1988), according to which causal relations are represented in terms of configurations of forces. In particular, “cause” and other causal concepts are applied to causal situations on the basis of what kinds of interaction are represented as taking place between two main types of entity: affectors and patients that affectors act on. According to Wolff, the particular causal concept that is applied to a given representation of causal interaction between an affector and a patient



C . BARRY ET AL .

depends on three factors: 1) whether the patient is represented as having a force-based tendency towards an endstate, 2) whether the forces represented as exerted by the patient and affector are concordant, and 3) whether the patient is represented as making progress towards the endstate.3 For instance, when a patient does not have a tendency towards an endstate, the affector exerts a force on the patient towards the endstate (the forces of the patient and affector are not concordant), and the patient makes progress towards the endstate, the model predicts that the concept “cause” will be applied to the representation. By contrast, when a patient does have a tendency towards the endstate, an affector exerts a concordant force on the patient towards that endstate, and the patient makes progress towards the endstate, the model predicts that the concept “enable” will be applied. Wolff (2007) attempted to show that the dynamics model has better extensional adequacy than competing models, more accurately picking out causal situations from non-causal situations and capturing distinctions among causal situations that the others cannot. Participants were asked to view computer simulations of interactions between affectors and patients—for instance, a boat and a bank of fans in a shallow pool, where a cone served as an endstate—and classify the scenario using various causal terms. The predictions of the model were borne out with respect to the application of the terms “cause,” “enable,” “prevent,” and “despite” in a number of experiments. An additional experiment was conducted to disambiguate the dynamics model from a kinematics model that would make predictions based solely on the visible properties of the interaction events. A further experiment applied the model to social causation by looking at how participants classified simulations involving causal intent and communication in a street intersection, suggesting that “social forces” are treated like physical forces. Other views of causal representation have also been put forward.4 Among the most prominent alternatives to the dynamics model is the

3

Wolff (2007), pp. 87–9. Our discussion of these views has been aided by Lombrozo (2010). Lombrozo finds evidence for causal-explanatory pluralism, according to which theories of causal representation that emphasize physical connections and those that emphasize counterfactual or logical dependence are both psychologically real, but judgments in line with each approach are promoted more and less by different modes of causal explanation. Lombrozo argues for a more unified account of causation which she calls an “exportable dependence” theory. See Lombrozo (2010). 4

DOING , ALLOWING , AND ENABLING HARM



“causal model theory” developed by Sloman and colleagues. According to this theory, causal terms express beliefs about the structure of causal models (Sloman et al. 2009). To assert “A causes B” expresses the belief, roughly, that there is an asymmetric dependence relationship between A and B such that a change in A will bring about a change in B, but a change in B will not bring about a change in A. Asserting “A enables B” implies that A is necessary for B, and also that B has additional causes.5 A third prominent approach is the “mental model theory,” developed by Goldvarg and Johnson-Laird (2001). The mental model theory holds that causal terms and relations are represented as logical relationships between possibilities or “models.” For instance, “A causes B” corresponds to three possibilities: A and B are both present, A is absent and B is present, or both A and B are absent, with the temporal constraint that B does not precede A. “A enables B” has either a weak or strong sense, according to which either all four possibilities might obtain (both A and B present; A present and B absent; A absent and B present; both A and B absent), or, on the stronger interpretation, A is needed for B (and so the possibility that A is absent and B is present is removed).6 Despite the differences between each of these views, they each recognize a psychologically real distinction between the concepts “cause” and “enable.” Nonetheless, they leave open many questions about “enable” and enabling in general. For instance, proponents of these views have suggested that “enable” has both a “help” sense and an “allow” sense roughly akin to “make possible,” but it is not clear how they distinguish enabling from allowing in the sense of “not preventing.” Further, the concept “enable” has largely figured in studies in which authors, implicitly or explicitly, sought to control for the effects of moral evaluation on causal judgment. The distinction between “enable” and other causal terms has yet to be explored with respect to cases of human conduct involving harm. Moral psychological research on the relationship between normative, including moral, judgments and causal judgments has not singled out the concept of enabling harm for investigation (e.g., Hitchcock and Knobe 2009; Knobe and Fraser 2008; Lagnado and Channon 2008). Cushman and colleagues (2008) demonstrate that moral appraisals can affect doing/allowing judgments—morally bad behavior is 5 6

Sloman et al. (2009), p. 21. Goldvarg and Johnson-Laird (2001) focus predominantly on enabling conditions.



C . BARRY ET AL .

more likely to be construed as doing harm than allowing harm. Work in cognitive science and moral psychology directly investigating judgments about doing harm and allowing harm has thus far only employed the traditional bipartite distinction.

Doing, allowing, and enabling harm In this chapter we argue that new experimental data support the view that some instances of human conduct involving harm are not descriptively subsumable under the traditional doing–allowing harm distinction. They are instances of enabling harm, which should be viewed as a distinct category. The cases that we have investigated are variations on well-known cases in the moral philosophical literature on doing and allowing harm that were first employed by Bennett (1995).7 Throughout the chapter we will refer to Push and Stayback cases, which take the following general form: Push: A cart stands at the top of a hill. John pushes it. The cart rolls down the hill and injures Tom, who is sitting at the bottom of the hill. Stayback: A cart is rolling down a hill. John could put a rock in the way of the cart that would stop it, but he does not. The cart rolls down the hill and injures Tom, who is sitting at the bottom of the hill.

Push and Stayback are straightforward in the sense that their classifications under the traditional doing–allowing distinction have been uncontroversial among philosophers. Philosophers generally agree that Push and Push-like cases are cases of doing harm, and Stayback and Stayback-like cases are cases of allowing harm. By contrast, there are what we will refer to as Interpose and Remove cases. They take the following form: Interpose: A cart accidentally starts rolling down a hill. Tom, who is sitting at the bottom of the hill, will not be injured by the cart if he can get out of the way of its path. John puts a rock on the ground. The rock stops Tom, and he is injured by the cart. Remove: A cart is rolling downhill towards a point where there is a rock that would bring it to a stop. John removes the rock. The cart rolls down the hill and injures Tom, who is sitting there.

Interpose and Remove cases and cases like them—cases in which agents either create an obstacle that blocks a causal process that would prevent a 7 We have made changes to the cases so as to make the kind and amount of harm that results in each of them equal.

DOING , ALLOWING , AND ENABLING HARM



harm or remove an obstacle that would block a causal process leading to a harm—are the subject of considerable descriptive disagreement among philosophers who employ the traditional doing–allowing distinction.8 We hold that theorists on either side of these disagreements are mistakenly employing a bipartite distinction where an additional, independent empirical category should instead be employed to accurately classify these cases. Our argument is thus largely in agreement with the general tendency in cognitive scientific and psychological research on causal representation to treat enabling as an independent category worthy of study. Why do Interpose and Remove cases belong to such a category? It is hard to provide an answer to this without giving a detailed account of the tripartite distinction, which is beyond the scope of this chapter. In prior work, Barry and verland (2012) advance a particular view of what separates enabling harm from doing harm and allowing harm. According to this view, the presence or absence of two factors—relevant action and initiation of a complete causal process—determines the appropriate classification of cases of human conduct involving harm under these categories. Relevant action, the first factor, obtains if the question of how an agent is relevant to some harm refers to some action of theirs.9 In Push, for example, the correct answer to the question of how John is relevant to the harms suffered by Tom will refer to his pushing of the cart. The second factor obtains if there is a complete, intact causal process initiated by the agent’s action that links this action to the harm.10 Barry and verland argue that when an agent does harm, both of these factors obtain—we can explain how some action that the agent has taken is relevant to the harm that occurs and there is a complete, intact causal process beginning with the agent’s relevant action

8 A number of philosophers who emphasize an agent’s movements, such as Jonathan Bennett, or actions, such as Warren Quinn, Kadre Vivhelin, and Terrance Tomkow, would say that John does harm to Tom in Interpose and Remove. See Bennett (1995), Quinn (1989), and Vihvelin and Tomkow (2005). Philippa Foot, by contrast, reduces the doing– allowing distinction to that between initiating or sustaining a harmful causal sequence and allowing a harmful sequence to run its course. Consequently, Foot would claim that John merely allows Tom to be harmed in cases such as Remove. It is perhaps less obvious what her account would say about Interpose, but she could say that by interposing a rock that prevents Tom from escaping the cart’s path, John allows that harmful causal sequence to run its course. See Foot (1994). 9 The basic idea of relevant action is due to Judith Jarvis Thomson (1996). 10 Hall (2002, 2004) refers to this kind of causal connection as exhibiting “locality”.



C . BARRY ET AL .

that links this action to the harm. When an agent enables harm, they argue, the agent is connected by relevant action to the harm that occurs but without its being the case that there is a complete causal process between the agent’s action and the harm. In contrast with cases of doing harm and enabling harm, neither of these factors obtains in cases of allowing harm—it is the agent’s omission rather than an action that is relevant to the harmful outcome in these cases, and there is not a complete causal process between an omission and an outcome.11 In this chapter we will not defend this view or discuss how it might relate to the theories of causal representation discussed in the previous section. We mention it here, instead, to emphasize that we do believe that some account can be given of the principles that underlie the differences between cases of doing, allowing, and enabling harm. Our contribution in this chapter involves the use of experimental methods to argue that the very cases that moral philosophers employing the bipartite distinction have disagreed deeply about are more naturally regarded as cases of enabling harm. In making this argument, we go beyond the existing empirical research on enabling by studying what we take to be cases of enabling harm by human agents, and contrasting judgments about these cases with judgments about more straightforward cases. Of course, we believe that there are many different ways of doing, allowing, and enabling harm that are not found in the cases that we study. The value of focusing on the Interpose and Remove cases lies in the fact that they are structurally similar to what have been regarded by philosophers as clear-cut cases of doing harm and allowing harm: Push and Stayback cases respectively. This allows us to compare experimental participants’ classifying judgments about conduct in these cases and conduct in Push and Stayback cases while holding other aspects of the cases fixed. We will now present three key experiments whose results support the claim that moral philosophers, cognitive scientists, and moral psychologists should be investigating a tripartite distinction between doing, allowing, and enabling harm. The first experiment’s results suggest that the traditional bipartite doing–allowing harm distinction cannot easily capture Interpose and Remove cases. The second experiment’s results suggest that Interpose and Remove cases are more naturally classified as

11

This is not to say that there cannot be causation by omission.

DOING , ALLOWING , AND ENABLING HARM



cases of enabling harm. On the basis of these results, we hold that the tripartite distinction between doing, allowing, and enabling harm better encompasses participants’ intuitive classifications of these cases from the moral philosophical literature than the traditional bipartite distinction. Moral philosophers and researchers studying moral judgment should thus explore the nature and significance of the tripartite distinction rather than focusing only on the traditional bipartite distinction. The third experiment’s results suggest that the tripartite distinction is also judged to be normatively significant.

Experiment 1: Doing–Allowing Harm Judgments Regarding Push, Stayback, Interpose, and Remove Cases On our view, certain cases in the moral philosophical literature—Interpose and Remove cases—cannot easily be captured by the traditional bipartite doing–allowing harm distinction. They are neither clear-cut cases of doing harm nor clear-cut cases of allowing harm. In this experiment, we tried to show that this view is supported by the way participants using the traditional doing–allowing harm distinction classify Interpose and Remove cases in contrast to the way they classify Push and Stayback cases. Participants were asked to read Push, Stayback, Interpose, and Remove cases and classify them as cases of either doing harm or allowing harm. We expected that Push cases would be more likely than the other types of cases to be classified as cases of doing harm, and Stayback cases more likely than the others to be classified as cases of allowing harm. We hypothesized, however, that participants’ judgments about Interpose and Remove cases would be intermediate between their judgments about Push and Stayback cases, in the following sense. Interpose and Remove cases would be more likely to be classified as cases of allowing harm than Push cases, but less likely to be classified as cases of allowing harm than Stayback cases.12 If this hypothesis turns out to be correct, we will have

12

Because the traditional doing–allowing harm distinction is bipartite, the dependent variable in this experiment—classification of the case as one of doing harm or allowing harm—is dichotomous. Our hypothesis can thus also be expressed in terms of the comparative likelihood of judging cases as ones of doing harm. We predicted that participants would be less likely to classify Interpose or Remove cases as cases of doing harm than Push cases, but more likely to classify them as cases of doing harm than Stayback cases.



C . BARRY ET AL .

evidence that Interpose and Remove cases differ from Push and Stayback cases across people in the ease with which they can be subsumed under the traditional doing–allowing harm distinction. This evidence would lend support to the view that Interpose and Remove cases are neither cases of doing harm nor cases of allowing harm. We employed two versions of each case. In one version an agent was culpable with respect to the harm suffered by another agent, while in the other version he was innocent. This manipulation of the first agent’s moral status was introduced for two main reasons. First, we wanted to control for potential differences in the moral status of the first agent between the cases that might be inferred by participants if they were to read versions of the cases in which no information about the agent’s moral status was provided. Second, prior work by Knobe and Fraser (2008), Hitchcock and Knobe (2009), and Cushman and colleagues (2008) has demonstrated that causal judgments and judgments about doing and allowing harm can be affected by moral and other normative judgments. We wanted to follow up on this work, even though our main interest in this chapter is in how people classify Interpose and Remove cases as compared to Push and Stayback cases.13

Methods PARTICIPANTS

The participants were 240 users of Amazon’s Mechanical Turk (MTurk). We limited our participant pool to M-Turk users from the United States with a 98% or better approval rate for their work. MATERIALS

The experiment employed a 4 (Push, Stayback, Interpose, Remove)  2 (culpable, innocent) factorial design. In the culpable version of each case, John knows that Tom will be harmed and it will not be costly for John to refrain from performing the action that will contribute to Tom being harmed (Push, Interpose, Remove) or to perform the action that would 13 Foreseeability of harm and cost to self of refraining from taking a harmful action (Push, Interpose, Remove) or engaging in harm-preventing action (Stayback) are used as proxies for culpability. It is conceivable, however, given prior work by Lagnado and Channon (2008), that the effect of foreseeability on causal judgments is driving the effects that we observe and suggest are attributable to moral judgments. We are grateful to an anonymous reviewer for this point.

DOING , ALLOWING , AND ENABLING HARM



prevent Tom from being harmed (Stayback). In the innocent version of each case, John does not know and could not be expected to know that Tom will be harmed. All cases used in Experiments 1, 2, and 3 are listed in Appendix A. Below are two sample cases: Push, Culpable: A cart stands at the top of a hill. John pushes it. John knows that Tom is sitting at the bottom of the hill, and refraining from pushing the cart would not be costly for John. The cart rolls down the hill and injures Tom. Push, Innocent: A cart stands at the top of a hill. John pushes it. John does not know and could not be expected to know that Tom is sitting at the bottom of the hill. The cart rolls down the hill and injures Tom.

The names given to the cases did not appear in the study. Participants were asked to read the case they were assigned and to respond to the prompt “This is most appropriately described as a case of:” by selecting “doing harm” or “allowing harm.” The order of the choices was counterbalanced. PROCEDURE

The experiment employed a between-participant design. Participants were asked to read and respond to only one of the eight cases, to which they were randomly assigned. Responses to each case were capped at n = 30.

Results Figure 3.1 represents the proportion of participants who judged each case to be one of allowing harm rather than doing harm (see Table 3.5 in Appendix B for the proportions of each response in Experiment 1). We used logistic regressions to measure the influence of the independent variables, case type (Push, Stayback, Interpose, Remove) and moral status (coded as 0 = culpable, 1 = innocent), on the dependent variable, classification of the case as one of doing harm or allowing harm. For the first logistic regression, responses to Push were used as the reference group, and for the second logistic regression, responses to Stayback were used as the reference group. We chose to run two regressions with these reference groups so that we could compare judgments about Interpose and Remove cases to judgments about Push cases and judgments about Stayback cases, which would be sufficient to test our main hypothesis. Tables 3.1 and 3.2 show the results of these regressions.



C . BARRY ET AL .

1.00 0.90

Proportion classifying as “allowing harm” rather than “doing harm”

0.80 0.70 0.60 0.50 0.40 0.30 0.20 0.10 0.00 Push

Stayback

Interpose

Remove

Case type Moral status Culpable

Innocent

Figure 3.1 Proportion of participants who classified each case as one of allowing harm rather than doing harm.

Table 3.1 presents the main effect of case type and interaction model results, and Table 3.2 presents the main effect of moral status. Unsurprisingly, participants responded to Push and Stayback cases as we expected. Controlling for moral status, participants were significantly more likely to classify Push cases than Stayback, Interpose, or Remove cases as cases of doing harm, and significantly more likely to classify Stayback cases than Push, Interpose, or Remove cases as cases of allowing harm. Our main hypothesis was confirmed. Controlling for the effect of moral status, judgments about Interpose and Remove cases were intermediate between judgments about Push and Stayback cases. Participants



DOING , ALLOWING , AND ENABLING HARM

Table 3.1. Main effect of case type and interaction models, Experiment 1 DV: Classified as allowing instead of doing Main effects model Innocent vs. Culpable Stayback vs. Push Interpose vs. Push Innocent vs. Stayback a Remove vs. Push Remove vs. Stayback a Interaction model Innocent  Stayback Innocent  Interpose Innocent  Remove Innocent  Interpose a Innocent  Remove a a

Odds Ratio

p-value

95% CI for Odds Ratio Lower

3.49 32.52 2.53 0.078 3.66 0.11 0.41 0.59 1.81 1.45 4.44

Upper

< 0.001 < 0.001 0.021 < 0.001 0.001 < 0.001

1.89 10.68 1.15 0.03 1.65 0.04

6.44 99.04 5.54 0.23 8.09 0.33

0.427 0.512 0.488 0.734 0.184

0.04 0.12 0.34 0.17 0.49

3.74 2.87 9.62 12.20 40.02

Indicates the use of Stayback as the reference group (rather than Push)

responding to an Interpose case were significantly more likely to classify the case as one of allowing harm than participants who received a Push case (OR = 2.53 [1.15, 5.54], p = 0.02), and significantly less likely to classify an Interpose case as one of allowing harm than participants who received a Stayback case (OR = 0.08 [0.03, 0.23], p < 0.001). Participants responding to a Remove case were significantly more likely to classify the case as one of allowing harm than participants who received a Push case Table 3.2. Main effect of moral status, Experiment 1 DV: Classified as allowing instead of doing

Innocent vs. Culpable in Push condition Innocent vs. Culpable in Stayback condition Innocent vs. Culpable in Interpose condition Innocent vs. Culpable in Remove condition

Odds Ratio

p-value

95% CI for Odds Ratio Lower

Upper

3.82

0.029

1.15

12.71

1.56

0.643

0.24

10.05

2.25

0.124

0.80

6.32

6.91

0.001

2.16

22.10



C . BARRY ET AL .

(OR = 3.66 [1.65, 8.09], p = 0.001), and significantly less likely to classify the case as one of allowing harm than participants responding to a Stayback case (OR = 0.11 [0.04, 0.33], p < 0.001). There was an overall main effect of moral status whereby participants assigned to an innocent version of a case were significantly more likely to classify the case as one of allowing harm than participants assigned to a culpable version of a case (OR = 3.49 [1.89, 6.44], p < 0.001). Participants assigned to the innocent version of Push were significantly more likely to classify the case as one of allowing harm than participants assigned to the culpable version of Push (OR = 3.82 [1.15, 12.71], p = 0.03). Similarly, participants assigned to the innocent version of Remove were significantly more likely to classify the case as one of allowing harm than participants assigned to the culpable version of Remove (OR = 6.91 [2.16, 22.10], p = 0.001). There was no main effect of moral status on responses to the Stayback cases or the Interpose cases. No significant case type  moral status interaction effects were found in the analyses conducted using either the Push or Stayback reference groups.

Discussion The results of Experiment 1 confirmed our main hypothesis about the classification of Interpose and Remove cases as compared with the classification of Push and Stayback cases under the traditional doing–allowing harm distinction. Whereas Push and Stayback cases were more likely than any of the other kinds of cases to be given the classification that philosophers have mostly regarded as uncontroversial for them, Interpose and Remove cases were treated differently. An Interpose or Remove case was significantly more likely to be classified as a case of allowing harm than a Push case, but significantly less likely to be classified as a case of allowing harm than a Stayback case. Judgments about these cases, in other words, varied more than judgments about what have been regarded as clear-cut cases of doing harm, on the one hand, and clear-cut cases of allowing harm, on the other. This evidence suggests that the traditional doing– allowing harm distinction is not able to provide easy classification of Interpose and Remove cases, or at least that individual variation with respect to doing and allowing judgments is more sensitive to features of Interpose and Remove cases than to features of Push and Stayback. It is worth noting that there was a significant main effect of moral status on judgments about the Push cases. While the majority of

DOING , ALLOWING , AND ENABLING HARM



participants in each version of Push classified the case they read as one of doing harm, participants assigned to the innocent version of the case were significantly more likely to classify their version as one of allowing harm than participants assigned to the culpable version. This fits with the effect of moral judgments on doing–allowing judgments observed in Cushman et al. (2008) and in Knobe’s work on the influence of normative judgments on what are thought to be non-normative judgments more generally.14 The analogous main effect of moral status obtained for Remove cases. No similar effect of moral status was found for Stayback or Interpose cases. The results of Experiment 1 suggest that the traditional bipartite doing–allowing harm distinction cannot capture Interpose and Remove cases. In the next experiment we investigated whether the tripartite doing–allowing–enabling harm distinction is better able to do so.

Experiment 2: Doing–Allowing–Enabling Harm Judgments Regarding Push, Stayback, Interpose, and Remove Cases In Experiment 1 we found strong evidence to support the claim that the doing–allowing harm distinction cannot accurately classify Interpose and Remove cases. With Experiment 2 we sought to show that our proposed doing–allowing–enabling harm distinction does a better job of capturing these cases. Experiment 2 differs from Experiment 1 only in virtue of adding the category of enabling harm to the options that participants can choose to classify the cases. We hypothesized that participants would generally classify Interpose and Remove cases as instances of enabling harm and, controlling for moral status, would be significantly more likely to do so than participants classifying Push and Stayback cases. If this hypothesis holds, we will have good evidence to support our view that the tripartite doing–allowing–enabling harm distinction is better able to capture these cases than the traditional bipartite doing–allowing harm distinction.

14 Knobe first reported the effect of people’s moral judgments on their judgments about intentional action in Knobe (2003). For an overview of this research program, see Knobe (2010).



C . BARRY ET AL .

Method PARTICIPANTS

The participants were 240 M-Turk users. We limited our participant pool to M-Turk users from the United States with a 98% or better approval rate for their work. Participants from Experiment 1 were not allowed to participate in Experiment 2. MATERIALS AND PROCEDURE

The experiment employed a 4 (Push, Stayback, Interpose, Remove)  2 (culpable, innocent) between-participant factorial design. Participants read only one of the eight cases, to which they were randomly assigned, and were asked to respond to the prompt “This is most appropriately described as a case of:” by selecting “doing harm,” “allowing harm,” or “enabling harm.” The order of the choices was counterbalanced. Responses to each case were capped at n = 30.

Results Figure 3.2 shows the proportion of enabling harm vs. doing harm or allowing harm responses to the four classes of cases (see Table 3.6 in Appendix B for the proportions of each response in Experiment 2). Because we were especially interested in judgments about enabling harm, we grouped doing harm and allowing harm together in our analyses. We used logistic regressions to measure the influence of the independent variables, case type (Push, Stayback, Interpose, Remove), and moral status (0 = culpable, 1 = innocent), on the dependent variable, classification of the case as one of enabling harm vs. either doing harm or allowing harm. In these analyses we also used Push as the reference group for the first logistic regression and Stayback as the reference group for the second logistic regression. We chose to run two regressions with these reference groups so that we could compare judgments about Interpose and Remove cases to judgments about Push and Stayback cases, which would be sufficient to test our hypothesis. Tables 3.3 and 3.4 show the results of these regressions. Table 3.3 presents the main effect of case type and interaction model results, and Table 3.4 presents the main effect of moral status. Enabling harm was the modal response to the Interpose cases (culpable version: 53.33%, innocent version: 76.67%). This was also true for the

DOING , ALLOWING , AND ENABLING HARM



1.00

Proportion classifying as “enabling harm” rather than “doing harm” or “allowing harm”

0.90 0.80 0.70 0.60 0.50 0.40 0.30 0.20 0.10 0.00 Push

Stayback

Interpose

Remove

Case type Moral status Culpable

Innocent

Figure 3.2 Proportion of participants who classified each case as one of enabling harm rather than doing or allowing harm.

Remove cases, though in the culpable version of this case this was only by a margin of one response over doing harm (culpable version: 36.67%, innocent version: 63.33%). Controlling for the effect of moral status, participants responding to an Interpose case were significantly more likely to classify the case as one of enabling harm than participants responding to a Push case (OR = 4.06 [1.85, 8.89], p < 0.001) or a Stayback case (OR = 9.42 [3.93, 22.55], p < 0.001). Participants responding to a Remove case were also significantly more likely to classify their case as one of enabling harm than participants responding to a Stayback case (OR = 4.86 [2.07, 11.41], p < 0.001). The



C . BARRY ET AL .

Table 3.3. Main effect of case type and interaction model results, Experiment 2 DV: Classified as enabling instead of other options Main effects model Innocent vs. Culpable Stayback vs. Push Interpose vs. Push Innocent vs. Stayback a Remove vs. Push Remove vs. Stayback a Interaction model Innocent  Stayback Innocent  Interpose Innocent  Remove Innocent  Interpose a Innocent  Remove a a

Odds Ratio

p-value

95% CI for Odds Ratio Lower

Upper

2.87 0.43 4.06 9.42 2.10 4.86

< 0.001 0.057 < 0.001 < 0.001 0.057 < 0.001

1.61 0.18 1.85 3.93 0.98 2.07

5.10 1.03 8.89 22.55 4.50 11.41

0.11 0.39 0.40 3.59 3.73

0.017 0.271 0.279 0.145 0.125

0.02 0.07 0.08 0.65 0.70

0.67 2.10 2.09 20.04 20.03

Indicates the use of Stayback as the reference group (rather than Push)

comparison between enabling harm judgments about Remove cases and Push cases was marginally significant in the direction predicted by the hypothesis (OR = 2.1 [0.98, 4.5], p = 0.06). There was an overall main effect of moral status whereby participants assigned to an innocent version of a case were significantly more likely to Table 3.4. Main effect of moral status, Experiment 2 DV: Classified as enabling instead of other options

Innocent vs. Culpable in Push condition Innocent vs. Culpable in Stayback condition Innocent vs. Culpable in Interpose condition Innocent vs. Culpable in Remove condition

Odds Ratio p-value

95% CI for Odds Ratio Lower

Upper

7.43 0.80

0.002 0.739

2.08 0.22

26.55 2.97

2.88

0.062

0.95

8.72

2.98

0.041

1.04

8.53

DOING , ALLOWING , AND ENABLING HARM



classify the case as one of enabling harm than participants assigned to a culpable version of a case (OR = 2.87 [1.61, 5.1], p < 0.001). A significant main effect of moral status on participants’ classifications was observed for the Push cases and Remove cases. Participants assigned to the innocent version of Push were significantly more likely to classify the case as one of enabling harm than participants assigned to the culpable version of Push (OR = 7.43 [2.08, 26.55], p = 0.002). Participants assigned to the innocent version of Remove were also significantly more likely to classify the case as one of enabling harm than participants assigned to the culpable version of Remove (OR = 2.98 [1.04, 8.53], p = 0.04). No similar main effect of moral status was found for responses to Stayback or Interpose cases. The only significant interaction effect between John’s moral status and the case that participants received was observed for the Push/Stayback comparison coding (p = 0.017, shown in Table 3.3 as Innocent  Stayback with Push as the reference group). Follow-up analyses showed that participants were significantly more likely to classify the case they were assigned to as one of enabling harm if they were in the Push condition and assigned to the innocent version rather than the culpable version (OR = 3.45, [1.87, 6.37], p < 0.001).

Discussion The results of Experiment 2 confirmed our hypothesis regarding how participants would classify the Interpose and Remove cases and apply the doing–allowing–enabling harm distinction to these cases as compared to the Push and Stayback cases. More participants regarded Interpose and Remove cases as instances of enabling harm than of doing harm or allowing harm. Statistical analyses showed that participants were significantly more likely to classify Interpose cases as instances of enabling harm than Push and Stayback cases—cases which have been treated by philosophers as uncontroversial cases of doing and allowing harm respectively. Participants were also significantly more likely to classify Remove cases as instances of enabling harm than Stayback cases, and marginally more likely to classify Remove cases as instances of enabling harm than Push cases. These results suggest that participants generally regard Interpose and Remove cases as instances enabling harm, and that they apply the doing–allowing–enabling harm distinction discriminately in the way that we expected.



C . BARRY ET AL .

Interestingly, there was a significant main effect of moral status on judgments about the Push cases. Participants assigned to the innocent version of Push were significantly more likely to regard it as an instance of enabling harm than participants assigned to the culpable version. In fact, enabling harm was the modal response to the innocent version of Push (53.33%). There was also a significant difference in enabling harm judgments between the innocent and culpable versions of Remove. These results suggests that the effect of moral judgments on doing–allowing harm judgments that Cushman, Knobe, and Sinnott-Armstrong have provided evidence for can also be observed with respect to doing– allowing–enabling harm judgments. While our main interest in this chapter is not in the effect of moral judgments on classifying judgments, we suggest that this effect is worth studying in future research on the doing–allowing–enabling harm distinction.

Experiment 3: The Doing–Allowing–Enabling Harm Distinction and Compensatory Responsibility We have been arguing that a tripartite doing–allowing–enabling harm distinction does a better job of classifying judgments about well-known cases found in the moral philosophical literature than the traditional bipartite doing–allowing harm distinction. A further question arises as to whether enabling harm is generally regarded as normatively like doing harm or allowing harm, or as distinct from both of these categories. While the main focus of this chapter is on the descriptive classification of cases of human conduct involving harm, we wanted to explore evidence that bears on this question. If enabling harm is regarded as normatively distinct from both doing harm and allowing harm along the lines of some normative factor, this would lend further support to our view that research in moral philosophy and moral psychology should employ the tripartite distinction. We hypothesized that one factor in terms of which people regard enabling harm as normatively distinct from doing harm and allowing harm is the assignment of compensatory responsibility— responsibility to compensate for harm (Goodin 1989). It seemed to us that, like agents who culpably do harm, agents who culpably enable harm have a more stringent responsibility to compensate the victims of harm

DOING , ALLOWING , AND ENABLING HARM



than agents who culpably allow harm. Yet it also seemed to us that, like agents who innocently allow harm, agents who innocently enable harm have less stringent compensatory responsibility than agents who culpably do harm. Using the eight cases from the previous two studies, our third study explored this possibility.

Methods PARTICIPANTS

The participants were 240 M-Turk users. We limited our participant pool to M-Turk users from the United States with a 98% or better approval rate for their work. Participants from Experiments 1 and 2 were not allowed to participate in Experiment 3. MATERIALS AND PROCEDURE

The experiment employed a 4 (Push, Stayback, Interpose, Remove)  2 (culpable, innocent) between-participant factorial design. Participants were asked to read only one of the eight cases, to which they were randomly assigned, and were asked to rate their level of agreement with the statement “John should compensate Tom for the injury” on a seven-point Likert Scale (1 = not at all, 7 = fully agree). Responses to each case were capped at n = 30.

Results Figure 3.3 represents the mean levels of agreement with the statement regarding John’s compensatory responsibility in each of the eight cases (see Table 3.7 in Appendix B for means with standard deviations in response to each case). We used a two-way ANOVA and post hoc tests to determine whether our hypothesis was borne out by the responses. All analyses for Experiment 3 were conducted using the Bonferonni correction for multiple comparisons. As our main hypothesis pertains to differences among culpable cases, on the one hand, and differences among innocent cases, on the other, we focused on the pairwise comparisons in post hoc tests on the interaction of case type and moral status. The two-way ANOVA revealed that there was a significant main effect of case type on responses to the statement regarding compensatory responsibility, F(3, 232) = 17.9, p < 0.001. Controlling for moral status, mean levels of agreement with the statement were significantly higher in



Mean level of agreement with CR statement

C . BARRY ET AL .

6.00

4.00

2.00

0.00 Push

Stayback

Interpose

Remove

Case type Error bars: 95% CI Moral staus Culpable

Innocent

Figure 3.3 Mean levels of agreement with the statement regarding compensatory responsibility in response to each case.

response to Push (M = 5.52, SD = 1.36) than in response to Stayback (M = 3.46, SD = 1.96), Interpose (M = 4.30, SD = 2.44), or Remove (M = 4.18, SD = 2.22), ps < .001. There was also a significant main effect of moral status, F(1, 232) = 143.45, p < 0.001. Controlling for case type, mean levels of agreement with the statement in response to culpable versions of the cases (M = 5.58, SD = 1.69) were significantly higher than in response to innocent versions of the cases (M = 3.15, SD = 1.88), p < 0.001. These main effects were qualified by a significant interaction effect between case type and moral status, F(3, 232) = 7.58, p < 0.001.

DOING , ALLOWING , AND ENABLING HARM



Our hypothesis was confirmed by pairwise comparisons among the culpable and innocent cases, which also help to explain the interaction effect between case type and moral status. Mean levels of agreement with the statement regarding compensatory responsibility were significantly higher in response to the culpable version of Push (M = 6.17, SD = 0.91), Interpose (M = 6.00, SD = 1.70), or Remove (M = 5.87, SD = 1.43) than in response to the culpable version of Stayback (M = 4.27, SD = 1.87), ps < 0.01. There were no significant differences between the mean levels of agreement with the statement in response to the culpable versions of Push, Interpose, or Remove. Mean levels of agreement with the statement were significantly higher in response to the innocent version of Push (M = 4.87, SD = 1.43) than in response to the innocent versions of Stayback (M = 2.63, SD = 1.71), Interpose (M = 2.60, SD = 1.81), or Remove (M = 2.5, SD = 1.46), ps < 0.001. There were no significant differences between the mean levels of agreement with the statement in response to the innocent versions of Stayback, Interpose, and Remove.

Discussion The results of Experiment 3 confirmed our hypothesis that enabling harm is generally regarded as normatively distinct from both doing harm and allowing harm in terms of assigning compensatory responsibility. Among the culpable cases, on one hand, mean levels of agreement with the statement were significantly higher in response to the Push, Interpose, and Remove cases than in response to the Stayback case and not significantly different from one another. Among the innocent cases, on the other hand, mean levels of agreement with the statement were significantly lower in response to the Interpose, Remove, and Stayback cases than in response to the Push case and not significantly different from one another. Interpose and Remove cases, which we hold are cases of enabling harm, are regarded as normatively distinct from Push and Stayback cases, which have been treated by philosophers as straightforward cases of doing and allowing harm.

General Discussion In this chapter we argued that the results of three experimental studies support the view that moral philosophers and moral psychologists should employ a tripartite distinction between doing, allowing, and



C . BARRY ET AL .

enabling harm. In previous work, theorists and empirical researchers studying moral judgment have typically employed the traditional bipartite doing–allowing harm distinction. We have argued that this traditional distinction cannot capture well-known cases in the moral philosophical literature—Interpose and Remove cases—whereas the tripartite doing– allowing–enabling harm distinction is better able to do so. We also provided evidence suggesting that enabling harm is judged to be normatively distinct from doing harm and allowing harm, at least when it comes to assigning compensatory responsibility. In his influential treatment, Jonathan Bennett described three possible outcomes of an investigation of the doing–allowing harm distinction: (1) The ordinary uses of these concepts reflect an underlying jumble, a big mess that resists analysis of any kind; (2) Ordinary uses express a simple and clear distinction; (3) People are guided in the use of these terms by simple and clear distinctions, but sometimes mischaracterize conduct due to distortions in judgment of various kinds.15

The results of our studies suggest that none of the three possibilities that Bennett envisions obtains, because there is no one binary, exhaustive distinction between doing and allowing harm. There is, instead, a tripartite distinction that participants draw between doing, allowing, and enabling harm. Judgments about doing harm and allowing harm, to be sure, are not simply a mess. Participants are more likely to regard Push cases as cases of doing harm and Stayback cases as cases of allowing harm than the comparison class of cases that we have looked at in this chapter. However, things do indeed get messy when one attempts to classify cases like Interpose and Remove as instances of doing harm or allowing harm. The results of the first two studies presented in this chapter suggest that they belong to the distinct category of enabling harm. The doing–allowing–enabling harm distinction—a distinction that gives this category its proper standing— should replace the doing–allowing harm distinction in future research on the nature and significance of human conduct involving harm. As noted above, the deep interest of philosophers in the doing– allowing harm distinction is due to the fact that it is often taken to be normatively significant. What, then, is the normative status of enabling harm? Does it possess normative characteristics that are distinct from 15

Bennett (1995), pp. 99–100.

DOING , ALLOWING , AND ENABLING HARM



doing harm on the one hand, and allowing harm on the other? In previous work, Barry and verland (2012) argue that it is indeed a normatively distinct category. The results of Experiment 3 suggest that experimental participants, too, seem to regard it as normatively distinct, at least when it comes to judgments about whether an agent bears compensatory responsibility for harm. Our primary focus has been on the descriptive classification of what we hold to be cases of doing, allowing, and enabling harm in the moral philosophical literature. We also showed that people judge the cases that we have argued are cases of enabling harm as normatively distinct from cases of doing harm and allowing harm in terms of this normative factor. It is an open question whether people tend to judge these cases as normatively distinct in other respects. We thus conclude by signaling directions for future research in moral philosophy and moral psychology.

Appendix A ALL CASES USED IN EXPERIMENTS

1–3

Push, Culpable: A cart stands at the top of a hill. John pushes it. John knows that Tom is sitting at the bottom of the hill, and refraining from pushing the cart would not be costly for John. The cart rolls down the hill and injures Tom. Push, Innocent: A cart stands at the top of a hill. John pushes it. John does not know and could not be expected to know that Tom is sitting at the bottom of the hill. The cart rolls down the hill and injures Tom. Stayback, Culpable: A cart is rolling down a hill. John could put a rock in the way of the cart that would stop it, but he does not. John knows that Tom is sitting at the bottom of the hill, and it would not be costly for John to put the rock in the cart’s way. The cart rolls down the hill and injures Tom. Stayback, Innocent: A cart is rolling down a hill. John could put a rock in the way of the cart that would stop it, but he does not. John does not know and could not be expected to know that Tom is sitting at the bottom of the hill. The cart rolls down the hill and injures Tom. Interpose, Culpable: A cart accidentally starts rolling down a hill. Tom, who is sitting at the bottom of the hill, will not be injured by the cart if he



C . BARRY ET AL .

can get out of the way of its path. John knows about the cart. He also knows that if he puts a rock on the ground, it will stop Tom from getting out of the cart’s path and Tom will be injured. Not putting a rock on the ground would not be costly for John. John puts a rock on the ground. The rock stops Tom, and he is injured by the cart. Interpose, Innocent: A cart accidentally starts rolling down a hill. Tom, who is sitting at the bottom of the hill, will not be injured by the cart if he can get out of the way of its path. John does not know about the cart and could not be expected to know that if he puts a rock on the ground it will stop Tom from getting out of the cart’s path and Tom will be injured. John puts a rock on the ground. The rock stops Tom, and he is injured by the cart. Remove, Culpable: A cart is rolling downhill towards a point where there is a rock that would bring it to a stop. John knows about the cart. He also knows that if he removes the rock, Tom, who is sitting at the bottom of the hill, will be injured. Not removing the rock would not be costly for John. John removes the rock. The cart rolls down the hill and injures Tom. Remove, Innocent: A cart is rolling downhill towards a point where there is a rock that would bring it to a stop. John does not know about the cart and could not be expected to know that if he removes the rock, Tom, who is sitting at the bottom of the hill, will be injured. John removes the rock. The cart rolls down the hill and injures Tom.

Appendix B Table 3.5. Proportions of participants in Experiment 1 classifying each case as an instance of “doing harm” and “allowing harm” Case Push, Culpable Push, Innocent Stayback, Culpable Stayback, Innocent Interpose, Culpable Interpose, Innocent Remove, Culpable Remove, Innocent

Doing Harm

Allowing Harm

0.83 0.57 0.10 0.07 0.60 0.40 0.63 0.20

0.17 0.43 0.90 0.93 0.40 0.60 0.37 0.80

DOING , ALLOWING , AND ENABLING HARM



Table 3.6. Proportions of participants in Experiment 2 classifying each case as an instance of “doing harm,” “allowing harm,” and “enabling harm” Case Push, Culpable Push, Innocent Stayback, Culpable Stayback, Innocent Interpose, Culpable Interpose, Innocent Remove, Culpable Remove, Innocent

Doing Harm

Allowing Harm

Enabling Harm

0.80 0.33 0.03 0.03 0.33 0.07 0.33 0.13

0.07 0.13 0.77 0.80 0.13 0.17 0.30 0.23

0.13 0.53 0.20 0.17 0.53 0.77 0.37 0.63

Table 3.7. Mean levels of agreement in response to the statement regarding compensatory responsibility and standard deviations for each case Case Push, Culpable Push, Innocent Stayback, Culpable Stayback, Innocent Interpose, Culpable Interpose, Innocent Remove, Culpable Remove, Innocent

Mean

SD

6.17 4.87 4.27 2.63 6.00 2.60 5.87 2.50

0.91 1.43 1.87 1.71 1.70 1.81 1.43 1.46

References Barry, C., and verland, G. (2012). The Feasible Alternatives Thesis: Kicking Away the Livelihoods of the Global Poor. Politics, Philosophy and Economics 11(1): 97–119. Bennett, J. (1995). The Act Itself. New York: Oxford University Press. Cushman, F.A., Knobe, J., and Sinnott-Armstrong, W. (2008). Moral Appraisals Affect Doing/Allowing Judgments. Cognition 108: 281–9. Donagan, A. (1977). The Theory of Morality. Chicago: Chicago University Press. Foot, P. (1994). Killing and Letting Die. In A. Norcross and B. Steinbock (eds.), Killing and Letting Die. New York: Fordham University Press.



C . BARRY ET AL .

Goldvarg, E., and Johnson-Laird, P.N. (2001). Naive Causality: A Mental Model Theory of Causal Meaning and Reasoning. Cognitive Science 25: 565–610. Goodin, R.E. (1989). Theories of Compensation. Oxford Journal of Legal Studies 9: 56–75. Hall, N. (2002). Non-Locality on the Cheap? A New Problem for Counterfactual Analyses of Causation. Noûs 36(2): 276–94. Hall, N. (2004). Two Concepts of Causation. In J. Collins, N. Hall, and L.A. Paul (eds.), Causation and Counterfactuals. Cambridge, MA: MIT Press, pp. 225–76. Hanser, M. (1999). Killing, Letting Die and Preventing People from Being Saved. Utilitas 11: 277–95. Hitchcock, C., and Knobe, J. (2009). Cause and Norm. Journal of Philosophy 106 (11): 587–612. Knobe, J. (2003). Intentional Action and Side-Effects in Ordinary Language. Analysis 63: 190–3. Knobe, J. (2010). Person as Scientist, Person as Moralist. Behavioral and Brain Sciences 33: 315–29. Knobe, J., and Fraser, B. (2008). Causal Judgment and Moral Judgment: Two Experiments. In W. Sinnott-Armstrong (ed.), Moral Psychology. Cambridge, MA: MIT Press, pp. 441–8. Lagnado, D.A. and Channon, S. (2008). Judgments of Cause and Blame: The Effects of Intentionality and Foreseeability. Cognition 108: 754–70. Lombrozo, T. (2010). Causal-explanatory Pluralism: How Intentions, Functions, and Mechanisms Influence Causal Ascriptions. Cognitive Psychology 61: 303–32. Mackie, J.L. (1980). The Cement of the Universe. Oxford: Oxford University Press. Quinn, W. (1994). Actions, Intentions, and Consequences: The Doctrine of Doing and Allowing. [Originally published (1989) in Philosophical Review 98 (3).] In A. Norcross and B. Steinbock (eds.), Killing and Letting Die. New York: Fordham University Press. Rickless, S.C. (2011). The Moral Status of Enabling Harm. Pacific Philosophical Quarterly 92(4): 66–86. Sloman, S.A., Barbey, A.K., and Hotaling, J. (2009). A Causal Model of the Meaning of “Cause,” “Enable,” and “Prevent.” Cognitive Science 33: 21–50. Talmy, L. (1988). Force Dynamics in Language and Cognition. Cognitive Science 12: 49–100. Thomson, J. (1996). Critical Study of Jonathan Bennett’s The Act Itself. Noûs 30 (4): 545–57. Vihvelin, K., and Tomkow, T. (2005). The Dif. Journal of Philosophy 102(4): 183–205. Wolff, P. (2007). Representing Causation. Journal of Experimental Psychology: General 136: 82–111.

4 Attributing Responsibility Actual and Counterfactual Worlds Tobias Gerstenberg, Massachusetts Institute of Technology David A. Lagnado, University College London

Introduction Chelsea played Bayern Munich in the final of the 2012 Champions League—Europe’s most prestigious football competition. Bayern scored a goal in the 83rd minute, but Chelsea’s star player, Drogba, equalized in the 88th minute to carry the teams to extra time. In extra time Bayern’s main penalty-taker, Robben, missed a potentially game-winning penalty. Thus, the match went to a penalty shootout. The score in the shootout was at 4–4. Robben had refused the responsibility to take a penalty in the shootout, so it was left to his teammate Schweinsteiger. Schweinsteiger’s shot hit the post and he buried his face in his jersey. Now Didier Drogba had the chance to secure Chelsea’s first ever victory in the Champions League. He shot, scored, and raised his arms to the sky. Chelsea had won. While Chelsea’s players celebrated their victory, Schweinsteiger and his teammates were left in tears. How do people attribute responsibility to individuals in situations in which several people have collectively contributed to an outcome? We might ask, for example, how much responsibility Robben or Schweinsteiger carry for Bayern’s loss in the final. Thoughts about what could have happened readily come to mind: if only Robben had scored his penalty in extra time, Bayern surely would have won the game. If only Schweinsteiger’s penalty shot had been an inch to the left he would have



T . GERSTENBERG , D . A . LAGNADO

scored, and Drogba, in turn, might not have been able to cope with the pressure. In this chapter, we outline a theoretical framework that links attributions of responsibility to counterfactuals. Counterfactual thoughts are thoughts about possible worlds in which the course of events would have unfolded differently from how it actually did (cf. Roese, 1997; Kahneman and Miller, 1986; Gerstenberg et al., 2013). Most research in psychology has focused on retrospective responsibility attributions, where one is concerned with how much responsibility an individual deserves for an outcome that has already occurred (see, e.g., Gerstenberg and Lagnado, 2010; Lagnado and Channon, 2008; Robbennolt, 2000; Zultan et al., 2012). We might wonder, for example, how responsible Bastian Schweinsteiger is for his team’s loss. However, there is another notion of responsibility that is prospective and captures the degree to which an individual is responsible for a future outcome. For example, the prospective responsibilities of a defender and a striker are different. It is a striker’s responsibility to score goals and a defender’s responsibility to prevent them (see Hart, 2008, for a discussion of different notions of responsibility). In what follows, we will first review previous research that has looked at the relationship between counterfactuals and attributions of responsibility. We will then show how a general framework that conceptualizes attributions of responsibility in terms of counterfactuals defined over causal models illuminates both retrospective and prospective attributions of responsibility (cf. McCoy et al., 2012; Chockler and Halpern, 2004; Lagnado et al., 2013). We will conclude by highlighting what we see as the key contributions of our framework.

Counterfactuals and Responsibility Counterfactual thoughts are commonplace in our everyday lives (Roese, 1997; Kahneman and Miller, 1986). If I had not exceeded the speed limit on my way to the office then I would not have got a ticket. If they had invested some more time doing a proper literature review then they surely would have written a better paper. In this section, we will see that there is a close relationship between counterfactuals and attributions of responsibility. How responsible for an outcome a person is seen to be depends not only on what the person actually did but also on what would have happened if the person had acted differently.

ATTRIBUTING RESPONSIBILITY



Many counterfactuals can be expressed in terms of counterfactual conditionals with an if-part and a then-part. If Bill had not been driving over the speed limit then he would not have got a ticket. Research has shown that attributions of responsibility are sensitive to both parts of the counterfactual (Petrocelli et al., 2011). Imagine that some nifty neuroscientists found out that people’s actions are exclusively determined by unconscious thought processes that are beyond any explicit control. Presumably, this would have rather substantial consequences for our practices of blaming and praising other people for their deeds (Strawson, 2008; Vohs and Schooler, 2008). The principle of alternative possibilities of action is central to many philosophical theories that discuss the relationship between determinism, freedom of will, and moral responsibility (Frankfurt, 1969). Should we really blame someone if they could not have acted differently (i.e. if there is no if)? While the question of the relationship between freedom of will and moral responsibility is intriguing, we will have to put it aside for the purposes of this essay (for more discussion on the topic see Widerker and McKenna, 2006; Nichols and Knobe, 2007; Nichols, 2011). However, even if we assume that people generally have a choice between several courses of action, there is still the question of whether a different action would have actually resulted in a different outcome (i.e. whether the then-part of the counterfactual would have proven true). If, for example, it turns out that the same outcome would have prevailed no matter what a person would have done in a particular situation, our first intuition is that the person’s responsibility for the outcome is at least diminished if not completely absent. Indeed, “it would not have made a difference anyhow” is a popular excuse when having done something wrong (Kerr and Kaufman-Gilliland, 1997; Kerr, 1996). In his culpable control model of blame, Alicke (2000) highlights the importance of both freedom of choice and the relationship between action and outcome for the blame attribution process. In Alicke’s theory, the degree to which a person is blamed for the outcome depends crucially on the perceived control that the person exhibited over her action as well as the outcome. Alicke analyzes control in terms of three structural linkages between mental, behavioral, and consequence elements. First, the mind-to-behavior link captures the degree to which the actor’s behavior was perceived to have been under volitional control. Volitional behavior control is predicted to increase with the belief that the person



T . GERSTENBERG , D . A . LAGNADO

acted purposefully and knowingly. It is diminished by situational and capacity constraints (we will see an example of each of these constraints below). Second, the behavior-to-consequence link characterizes to what extent the person had causal control over the outcome. Alicke throws in a mix of different criteria that are predicted to influence causal control such as uniqueness (How many alternative causes were present?—cf. Kelley’s discounting principle, 1973), proximity (How close was the cause to the effect in the chain of events?) and effective causal control (What would have happened if the cause had been different?). We will describe different factors that diminish causal control below. Third, the mind-to-consequences link describes whether the person had volitional outcome control. How much volitional outcome control a person has is contingent on whether she foresaw the outcome and desired it. Consider a patient who dies as a result of a wrong treatment by a doctor. In this case, the doctor’s volitional outcome control was diminished assuming that the negative consequences of the treatment were in fact not foreseeable and that the doctor did not want the patient to die. Let us now illustrate in some more detail how causal control (the behavior-to-outcome link) can be weakened. We focus on this notion because it is the most relevant to the general framework of responsibility attribution that we will propose in the next section. Consider a slightly adapted scenario from Wells and Gavanski (1989): Bill goes to dinner with Suzy. It is their first date, and Bill, being an old-fashioned gentleman, orders the meal for Suzy. Because the two find themselves in a fancy French restaurant which only serves two different seasonal meals, Bill’s choice is somewhat constrained. Bill orders set menu A for Suzy. Unfortunately the meal contains wine to which Suzy has an allergic reaction, and Suzy dies. There are two possible versions of the story. In story one, Bill and Suzy dine at Restaurant Contrôle Effective. In this restaurant, Suzy would have been fine if Bill had ordered meal B. In story two, Bill and Suzy dine at Restaurant Pas de Contrôle Effective. In this restaurant, meal B also contains wine and Suzy would have also died from eating meal B. So, what is your verdict? In which story is Bill more to blame for Suzy’s death? Is he more to blame for having ordered meal A in Restaurant Contrôle Effective or in Restaurant Pas de Contrôle Effective? In which

ATTRIBUTING RESPONSIBILITY



restaurant does Bill’s choice play a greater causal role? Remember that Bill does the exact same thing in both situations. If you think that the causal significance of Bill’s choice is greater in Restaurant Contrôle Effective, you are in line with what most participants indicated.1 If you wonder why Suzy did not tell Bill that she is allergic to wine, you are probably in line with what most participants must have found rather strange about the scenario. In any case, this example illustrates how causal control is diminished if a person finds himself in a situation of no effective control. No matter whether Bill had chosen meal A or B in Restaurant Pas de Contrôle Effective, the outcome would have been the same—Suzy would have died. However, in Restaurant Contrôle Effective, Bill had effective causal control (as the name of the restaurant subtly suggests). If only he had chosen meal B, Suzy would have been okay. As a further illustration, consider the two penalty shots depicted in Figure 4.1. Let us assume that Bill is now the goalkeeper and Suzy the striker. We know that Bill never saves penalties that are shot very close to the post. However, he gets the penalties that are shot closer to the center of the goal (if he jumps in the correct direction). In both situation (a) and (b), Bill jumped in the wrong direction. In which situation do you think is he more to blame for the goal? The majority of participants in our experiment (67%), who were presented with both situations simultaneously, indicated that Bill is more to blame for not having saved the penalty in situation (a).2 The rest of the participants indicated that Bill is equally to blame in both situations. Hence, none of the participants thought that Bill was more to blame in situation (b) than in situation (a). In situation (a), Bill could have saved the penalty if only he had jumped in the correct direction. However, the penalty in situation (b) was so well placed that Bill could not have saved it even if he had jumped in the correct direction. Hence, in Alicke’s (2000) terms, Bill had effective causal control in situation (a) but not in situation (b). 1 Wells and Gavanski (1989) did not ask for responsibility judgments in the restaurant scenario but in a conceptual replication in Experiment 2. Causal and responsibility judgments were almost identical. 2 Twenty-one participants were recruited online via Amazon Mechanical Turk (see Mason and Suri, 2012). Fourteen indicated that the goalkeeper in (a) is more to blame than the goalkeeper in (b), seven participants blamed both equally, and none of the participants blamed (b) more than (a), w2 ð2; N ¼ 21Þ ¼ 14; p 0.05 > 0.05 > 0.05

Pediatric psychopathic trait rating scale scores, mean (SD) Antisocial Process Screening Device 5.5 (3.9) 28.8 (4.6) Psychopathy Checklist: Youth Version — 24.3 (3.4)

< 0.001 —

Disorders and Schizophrenia for School-Age Children—Present and Lifetime Version (Kaufman et al., 1997). Exclusionary criteria for both groups included psychosis, pervasive developmental disorders, Tourette’s syndrome, mood or anxiety disorders, neurological disorders, IQ < 80, or medical illness severe enough to require treatment. Adolescents were classified as psychopathic following a score  20 on the Antisocial Process Screening Device (Frick and Hare, 2002) and a score  20 on the Psychopathy Checklist: Youth Version (Forth et al., 2003). Participants in the comparison group were required to score < 20 on the Antisocial Process Screening Device. This study was approved by the institutional review board at the National Institute of Mental Health. The parent or legal guardian of each participant provided written informed consent before the study began; participants provided informed assent.

Clinical measures The Antisocial Process Screening Device (APSD; Frick and Hare, 2001). The APSD detects antisocial behavior in youths, including psychopathic traits and conduct and impulsivity problems. The twenty-item parentcompleted scale has a three-factor structure comprised of the following dimensions: Callous/Unemotional, Narcissism, and Impulsivity. Consistent with prior studies (Finger et al., 2008; Marsh et al., 2008; Marsh et al., 2011b) a score of  20 was chosen as a cutoff score for classification of high psychopathic traits.

MORAL STATUS AND PERCEIVED INTENTION



Psychopathy Checklist: Youth Version (PCL:YV; Forth et al., 2007). The PCL:YV was adapted by Hare and colleagues to measure psychopathic traits in adolescents. Based on a semi-structured interview and collateral information, two trained researchers assessed interpersonal, affective, and behavioral features related to psychopathic traits in youths. The two researchers demonstrated good inter-rater reliability (R = 0.91). The PCL-YV is comprised of a twenty-item rating scale. Consistent with prior studies (Finger et al., 2011; Marsh et al., 2011a; White et al., 2012) a score of  20 was chosen as a cutoff score for classification of psychopathic traits.

Experimental task The experimental task was adapted from tasks previously used to assess the side-effect effect in studies of children (Pellizzoni et al., 2009). Three vignettes were generated, each with three variants that featured, respectively, a positive, negative, or neutrally valenced outcome, for a total of nine scenarios. One sample vignette featured one child showing a frog to another child (with three possible outcomes). The variants of this vignette read as follows: This story is about a boy named Andy. Andy has found a frog in his yard. Now, Andy loves frogs. A girl in Andy’s neighborhood named Nicole [positive: loves frogs too; negative: hates frogs; neutral: has never seen a frog before]. Andy wants to bring the frog over to Nicole’s house. If he brings the frog over, she will [positive: be happy; negative: get upset; neutral: see a frog for the first time]. Andy says, “I don’t care that Nicole will [positive: be happy; negative: get upset; neutral: see a frog for the first time]. I am going to bring the frog over because I want to.” So Andy brings the frog over to Nicole’s house and she [positive: is happy; negative: gets upset; neutral: sees a frog for the first time].

Two additional vignettes were created: one featured a child playing a CD that causes another child to be happy, get upset, or tap her foot; and one featured a child who tells ghost stories that cause another child to be happy, get upset, or fall asleep. Because distinctions in moral reasoning between individuals with and without psychopathic traits are most likely to emerge in the context of victim distress, the consequence of all the negatively valenced scenarios was a child becoming upset. Three versions of the task were created, and each child completed only one version. Each version featured one of each vignette (frog, CD, ghost stories), with the valence of the vignette counterbalanced across versions.



E . M . CARDINALE ET AL .

So, for example, one version of the task featured the neutral frog vignette, the negative CD vignette, and the positive ghost stories vignette; another version featured the positive frog vignette, the neutral CD vignette, and the negative ghost stories vignette, and so on. The side-effect effect task was programmed using Eprime, which allowed the order in which participants read the vignettes to be randomized across participants. After reading each vignette, participants answered three questions: (1) Does the agent care [about the outcome of the side-effect]? (2) Did the agent [cause the outcome of the side-effect] on purpose? (3) Was [the outcome of the side-effect] a good thing? Questions were answered using a five-point scale (1 = definitely no; 5 = definitely yes). Participants completed testing on a PC laptop in a single testing session in a private testing room at the National Institute of Mental Health. While they completed the measure, participants were monitored by a researcher who provided participants with clarification as needed. For most children, the measure was administered as part of a larger battery of behavioral testing that included questionnaires and computer tasks assessing learning, reasoning, and reaction times.

Results For each question, a 2 (Group)  3 (Valence) ANOVA was conducted to assess the interaction of moral valence and intentionality across groups. For the “on purpose” question, a main effect of valence emerged, F (2,70) = 30.518, p < 0.001, such that negatively valenced outcomes were viewed as more intentional (M = 3.92, SD = 1.34) than neutral (M = 2.41, SD = 1.69) or positively valenced outcomes (M = 2.16, SD = 1.48) (Figure 5.1). The results of within-subject contrast tests indicated that the pattern of the effect was primarily linear, F(1,35) = 31.360, p < 0.001, rather than quadratic, F(1,35) = 7.837, p = 0.052. The side-effect effect was consistent across groups. Negatively valenced side-effects were viewed as more intentional than positively valenced side-effects by both healthy, t(22) = –5.016, p < 0.001, and psychopathic, t(13) = –3.122, p = 0.008, adolescents. A virtually identical proportion of healthy (65.2%) and psychopathic (64.3%) adolescents viewed the negatively valenced scenario as more intentional than the positively valenced scenario.

MORAL STATUS AND PERCEIVED INTENTION



5

Intentionality

4 3 2 1 Positive

Neutral Healthy

Negative

Psychopathic traits

Figure 5.1 Moral valence and intentionality.

Consistent with this finding, no significant valence by group interaction was observed, F(2,70) = 0.325, p = 0.724. A marginally significant main effect of group was identified, such that psychopathic adolescents viewed all side-effects as more intentional than did healthy adolescents, F (1,35) = 3.698, p = 0.063. Across valences, adolescents with psychopathic traits viewed the agents’ behavior as more intentional than did controls. For positively valenced outcomes, the group difference was M = 0.54, for neutral outcomes M = 0.96, and for negatively valenced outcomes M = 0.48. None of these differences were statistically significant (all ps > 0.10) (Figure 5.1). No main effects or interactions emerged for judgments of whether the agent cared about the outcome. Very little variance was observed in response to this question, either across groups or across valences, with 75% to 90% of participants supplying the extreme answer (1/the agent definitely did not care about side-effect). For judgments of whether the outcome was a good thing, a main effect of valence, F(2,70) = 159.35, p < 0.001, was observed, with higher ratings for positive scenarios (M = 4.64, SD = 0.63) than neutral (M = 3.22, SD = 1.18) or negative (M = 1.30, SD = 0.52) scenarios. A main effect of group was observed for ratings of goodness, F(1,35) = 4.127, p = 0.05, such that psychopathic adolescents rated all outcomes to be better than did healthy adolescents



E . M . CARDINALE ET AL .

5

Goodness

4

3

2

1 Positive

Neutral Healthy

Negative

Psychopathic traits

Figure 5.2 Moral valence and goodness.

(Figure 5.2). No significant interaction was observed in response to this question. We next calculated correlations among the various rating scales across groups. So, for example, we assessed whether ratings of caring or goodness predicted judgments of intentionality across the sample. The only significant correlation among these variables was between assessments of caring and intentionality for positive statements. Across participants, assessments of how much the protagonists cared about the good outcomes positively predicted how intentional those outcomes was judged to be, r(35) = 0.484, p < 0.05 (Bonferroni corrected for multiple comparisons). This effect was driven largely by the responses of the healthy participants, r(23) = 0.63, p = 0.001, rather than the participants with psychopathic traits, r(12) = 0.29, ns. Judgments of the goodness of the various outcomes were not associated with ratings of caring or intentionality.

Discussion We found that the moral valence of an outcome significantly affects judgments of intentionality in both healthy adolescents and adolescents with psychopathic traits. Although psychopathic adolescents exhibit heightened immoral behavior, and although moral reasoning deficits have previously been identified in psychopathy, adolescents with

MORAL STATUS AND PERCEIVED INTENTION



psychopathic traits show an intact side-effect effect. Both healthy adolescents and adolescents with psychopathic traits showed nearly identical patterns of attributing greater intentionality to agents whose actions resulted in morally negative outcomes than to agents whose actions resulted in morally positive outcomes. In both groups, the proportion of adolescents who attributed more intentionality to agents in the negative than the positive outcome scenarios was comparable to the proportions observed in previous experimental studies (Knobe and Burra, 2006; Knobe, 2003; Leslie et al., 2006; Young et al., 2006). In addition, our task included outcomes that were neutral in valence, and judgments of intentionality in response to these outcomes generally fell between those of positively and negatively valenced outcomes, as indicated by a predominantly linear pattern of means across conditions. How can these findings help to explain the mechanisms underlying the side-effect effect? Prior evidence suggests that psychopathy is more likely to alter moral judgments than judgments related to social convention and norms (Blair, 2007). That the side-effect effect is intact in psychopathy is consistent with suggestions that this effect results from comparisons of the agent with a prototypical social actor—a judgment that requires understanding the moral norms of the social group (Knobe, 2010; Pettit and Knobe, 2009). Knobe has proposed that attributions of intentionality to actors in vignettes like those featured in this study rely on the distance between the agent’s concern about the results of his actions and how concerned one would expect a typical person to be in that situation (Knobe, 2010; Pettit and Knobe, 2009). For example, participants in our study read about a boy named Andy who brought a frog over to a frog-loathing neighbor’s house because he wanted to, and who claimed he did not care that the neighbor would get upset. Adolescents both with and without psychopathic traits perceived that Andy upset his neighbor on purpose, despite their agreement that he “did not care” about the effects of his actions. Because all of our participants recognized that upsetting someone else is objectively bad (no group differences in this judgment were observed), participants may have inferred that Andy’s behavior was so far from the anticipated default that in professing not to care, he was in essence actively in favor of upsetting his neighbor. This kind of judgment relies on some minimal moral understanding, insofar as respondents can report that causing someone else to be upset is a bad outcome. However, any moral



E . M . CARDINALE ET AL .

assessments upon which the side-effect effect rest do not seem comparable to moral acceptability judgments that rely upon empathic responses to a victim’s distress. Some explanations for the side-effect effect do emphasize moral emotions that follow negative outcomes, for example, blame (Alicke and Rose, 2010; Alicke, 2008). Are these accounts consistent with the finding that the side-effect effect is intact in psychopathy? Blame is the response to an agent whose actions cause harmful consequences that are seen as foreseeable and avoidable. Like other moral emotions, blame serves useful social purposes. Holding reckless or negligent actors to account for their misdeeds perhaps reduces the likelihood that they will repeat those misdeeds (Alicke, 2008). Blame is thought to result from negative evaluations of an agent’s behavior, which are thought to be rapid and automatic (Bargh and Chartrand, 1999). Once instantiated, blame is suggested to lead to a “blame validation mode” whereby people search for some party to whom they can assign the blame (Alicke and Rose, 2010). The moral tenor of an action can influence blame attributions, as Alicke and colleagues found when they asked participants to consider the case of a student who was in a car accident while either speeding home to hide his cocaine stash, or speeding home to hide a gift for his parents. The student was judged to be more responsible for the car accident in the cocaine scenario, presumably because there were no mitigating circumstances that reduced respondents’ desire to assign blame (Alicke, 1992). Under the blame validation model, the side-effect effect results when respondents blame the harm-doer for the outcome he has caused and then seek to justify their blame by interpreting the behavior as intentional. This model is somewhat more difficult to reconcile with the present findings, primarily because of its reliance on automatic negative evaluations. Abundant evidence exists to suggest that individuals with psychopathic traits show weaker automatic responses to negatively valenced stimuli and information. Psychopathy is associated with weakened changes in autonomic arousal (for example, skin conductance and potentiated startle) to negative stimuli, whereas psychopathic and non-psychopathic responses to positive stimuli are relatively similar (Herpertz et al., 2001; Levenston et al., 2000; Patrick, 1994; Rothemund et al., 2012; Vaidyanathan et al., 2011). If automatic negative evaluations underlie the side-effect effect, one would expect that individuals with psychopathic traits, whose

MORAL STATUS AND PERCEIVED INTENTION



automatic responses to negative stimuli are not as strong as those of nonpsychopathic individuals, would generate a weaker blame response in response to foreseeable harm and would therefore show smaller shifts in their judgments of intentionality. This was not the case in the present study. Does this mean that the types of scenario used to test the sideeffect effect do not generate moral emotions, or that these emotions are irrelevant to judgments of intentionality? One possibility is that emotions are involved in the side-effect effect, but not the sorts of moral emotions that are impaired in psychopathy. If one considers specific types of negative evaluative responses, it is anger that is most closely linked to blame (Malle and Nelson, 2003; Ortony et al., 1988). Anger is the typical emotional response to a negative outcome that is perceived to be caused by a social actor, and this emotion energizes a person to attempt to reverse the negative outcome (Frijda, 1986; Roseman et al., 1994). Angry responses are not impaired in psychopathy—in fact, one of the criteria for assessing psychopathy in children and adolescents focuses on tantrums and episodes of extreme anger (Blair, 2012; Forth et al., 2003). The same is true for individuals with damage to the ventral prefrontal cortex, who are also prone to bouts of extreme anger (Blair and Cipolotti, 2000). This is a patient population in whom various impairments in moral reasoning have been observed (Koenigs et al., 2007; Young et al., 2010), but in whom the side-effect effect is also intact (Young et al., 2006). This leaves open the possibility that blame validation relies not upon generalized negative evaluative responses, but specifically upon anger responses to the misdeeds of agents who did foresee or could have foreseen the harmful consequences of their actions. This explanation could account for observations of an intact side-effect effect both in psychopathy and in patients with ventromedial prefrontal lesions. This explanation is also consistent with previous findings that dissociable emotion systems are linked to different forms of moral judgment (Rozin et al., 1999) and that affective deficits in psychopathy appear to be largely confined to fear (Marsh and Blair, 2008; Marsh and Cardinale, 2012; Marsh et al., 2011b), with affective responding related to emotions such as disgust and anger largely intact. A weakness of this argument, however, is that it presumes that psychopathic individuals would experience anger in response to another person being harmed. In the general population, anger is a common response to harm that befalls others, as demonstrated by moral domain



E . M . CARDINALE ET AL .

research showing anger during judgments of third-party autonomy violations such as theft, poisoning, and assault (Rozin et al., 1999). Judgments of autonomy violations are thought to entail considerations of “harm, rights, justice, freedom, fairness, individualism, and the importance of individual choice and liberty” (p. 575). That first-person harm-based violations evoke anger in psychopathic individuals is clear, with psychopathy reliably associated with angry responses to provocation and goal frustration (Blackburn and Lee-Evans, 2011). There is, by contrast, no evidence that third-person harm-based moral violations result in anger, and indeed, this possibility directly contradicts the available evidence that psychopathy impairs moral judgments about third-party harm.

Responses to neutral outcomes One feature of the current paradigm that may also be more compatible with competence theories as outlined by Knobe (Knobe, 2010) than with blame-validation theories is the responses participants gave to outcomes of neutral valence. Very few studies have examined neutral outcomes in assessments of the side-effect effect (Knobe and Mendlow, 2004; Phelan and Sarkissian, 2008). Outcomes in the present study that were considered neutral were that the target individual in the story would fall asleep, tap her foot, or see a frog for the first time. That these outcomes were in fact approximately neutral in valence is supported by participants’ average response to the question of how good these outcomes were (M = 3.22) which fell roughly between the endpoints of the scale (1 and 5). That respondents judge the intentionality of actions in tasks such as these by comparing the agent’s behavior to that of a prototypical agent requires that agency be construed as a point on a sliding scale ranging from no intentionality to full intentionality. This allows an agent who “does not care” about an outcome to sometimes be viewed as in favor of the outcome and sometimes not, as a function of where on the scale the prototypical agent would be expected to fall. For a neutral outcome, the prototypical agent would be expected to be truly neither in favor of nor opposed to the outcome—that is, to fall upon the midpoint of the scale. Thus, agents in the present neutral-outcome stories would have views consistent with the prototypical agent, and one would predict that they would be ascribed a level of intentionality somewhere between agents who bring about good outcomes and those

MORAL STATUS AND PERCEIVED INTENTION



who bring about bad outcomes. These were in fact the results we obtained, with a predominantly linear pattern of assigned intentionality observed across negative, neutral, and positive outcomes. It is less clear why these results would be obtained under the blame validation model— here one might anticipate a more strongly quadratic pattern, wherein blame responses would elicit increased attributions of intentionality to negative outcomes relative to positive or neutral outcomes, but that intentionality would not be construed differently across neutral and positive outcomes, as neither would elicit a blame response (although it is possible that respondents might perceive causally impinging on another person’s outcomes in any way—negatively or not—to violate that person’s autonomy to some small degree, and might thus perceive blameworthiness). It should be noted that group means suggest the possibility that psychopathic adolescents’ responses followed a more quadratic pattern than did healthy adolescents’, although the absence of a group  valence interaction indicates that the patterns of effects were not significantly different across groups. At the very least, however, it cannot be conclusively stated that psychopathic adolescents’ responses did not follow a quadratic pattern. An investigation of this issue in a study with a larger sample size might permit a more detailed investigation of potential differences in response patterns across groups. Neutrally valenced outcomes also yielded another interesting pattern, albeit not one that reached statistical significance, which is that the largest differences in the mean judgments between groups was usually obtained in response to these questions. Adolescents with psychopathic traits showed a tendency to rate all outcomes to be better, and this difference was most pronounced in response to neutral outcomes. Similarly, adolescents with psychopathic traits as a rule attributed greater intentionality to the protagonists across the vignettes, but the mean group difference was again largest in response to neutral outcomes. It should be noted that a similar response bias toward assigning greater intentionality to the agents across conditions has been observed in very young children (3 years old) (Leslie et al., 2006). This response bias in the youngest children to attempt this task was attributed to their failure to understand the task. Could the same thing be true of psychopathic individuals to a degree? Many have speculated that psychopathic traits leave those who are affected able to simulate normal human responses to social and emotional stimuli without grasping the basis for those



E . M . CARDINALE ET AL .

responses (Blair et al., 2006; Cleckley, 1988). This leaves them prone to base their moral judgments more on social norms than on their low-level affective or empathic responses to moral dilemmas (Blair et al., 1995; Blair, 1995). In response to questions with obvious socially normative responses (is it good to upset someone or to make them happy?) individuals with psychopathic traits can provide approximately normal responses. But perhaps in response to questions with less obvious socially normative value, the mechanisms that individuals with and without psychopathic traits rely upon to answer the question diverge, which results in a divergence of their responses as well.

Further questions Some limitations of this study must be observed in interpreting our results. First, our sample of participants with psychopathic traits was small relative to prior studies of the side-effect effect. It was, however, comparable to sample sizes employed in a number of previous studies of moral reasoning in psychopathy, suggesting that reliable effects can be obtained using this number of participants (Blair, 1995; Cima et al., 2010; Glenn et al., 2009; Gray et al., 2003). Moreover, the reliability of the effect was extremely similar across both groups we tested, and was comparable to the results from larger previous studies of the side-effect effect, suggesting that a larger sample of adolescents with psychopathic traits would be unlikely to yield different results than we obtained. A question that cannot be answered from our data, however, is whether a group of adolescents with disruptive behavior disorders but low levels of psychopathic traits would show the same results as the adolescents we tested. Increasing evidence supports the divergent etiology of conduct problems in these two subsamples of adolescents, and it is possible that patterns of moral reasoning across these groups would diverge as well. Finally, one might argue that a question that emerges from our findings is why we found no group differences in responses to how good it is when the victim of a negative side-effect becomes upset. Given prior findings that psychopathy impairs responses to victim distress, this finding could be seen as aberrant. One possibility is that judgments about the acceptability of victim distress vary depending on how the question is framed. It has been previously observed that asking whether a moral violation is “OK” versus whether it is “bad” yield different response patterns in psychopathic youths (Blair, 1995), perhaps because some phrasings are more

MORAL STATUS AND PERCEIVED INTENTION



likely to be interpreted as referring to social acceptability versus moral acceptability. It is possible that participants interpreted our question about whether the various outcomes were good as socially normative rather than as a moral judgment.

Conclusions Psychopathy—a disorder closely associated with impaired moral judgments and behavior—was found not to affect responses to vignettes testing the side-effect effect. This suggests that information about moral valence does not influence judgments of intentionality in psychopathy, in contrast to previous findings that information about intentionality does affect judgments of morality in psychopathy (Young et al., 2012). This is consistent with an emerging consensus that moral reasoning is unlikely to be associated with a single set of mechanisms, but that different domains of moral judgments are associated with dissociable cognitive processes and neural mechanisms (Parkinson et al., 2011). It is also consistent with the suggestion that whereas the behavior of individuals with psychopathic traits can make them appear “morally blind,” this blindness to issues of morality is domain-specific, reflecting the fact that psychopathy is associated with specific deficits in the integrity of neural systems that support moral reasoning (Blair, 2007).

References Aharoni, E., Antonenko, O., and Kiehl, K.A. (2011). Disparities in the Moral Intuitions of Criminal Offenders: The Role of Psychopathy. Journal of Research in Personalities 45: 322–7. Alicke, M.D. (1992). Culpable Causation. Journal of Personality and Social Psychology 63: 368–78. Alicke, M.D. (2008). Blaming Badly. Journal of Cognition and Culture 1–2: 179–86. Alicke, M., and Rose, D. (2010). Culpable Control or Moral Concepts? Behavioral and Brain Sciences 5: 403–22. Bargh, J.A., and Chartrand, T.L. (1999). The Unbearable Automaticity of Being. American Psychologist 54: 462. Blackburn, R., and Lee-Evans, J.M. (2011). Reactions of Primary and Secondary Psychopaths to Anger-evoking Situations. British Journal of Clinical Psychology 24: 93–100.



E . M . CARDINALE ET AL .

Blair, K.S., Richell, R.A., Mitchell, D.G., Leonard, A., Morton, J., and Blair, R.J. (2006). They Know the Words, but Not the Music: Affective and Semantic Priming in Individuals with Psychopathy. Biological Psychology 73: 114–23. Blair, R.J. (1995). A Cognitive Developmental Approach to Morality: Investigating the Psychopath. Cognition 57: 1–29. Blair, R.J. (2005). Applying a Cognitive Neuroscience Perspective to the Disorder of Psychopathy. Development and Psychopathology 17: 865–91. Blair, R.J. (2007). The Amygdala and Ventromedial Prefrontal Cortex in Morality and Psychopathy. Trends in Cognitive Sciences 11: 387–92. Blair, R.J.R. (2012). Considering Anger from a Neuroscience Perspective. WIREs Cognitive Sciences 3: 65–74. Blair, R.J., and Cipolotti, L. (2000). Impaired Social Response Reversal: A Case of ‘Acquired Sociopathy’. Brain 123: 1122–41. Blair, R.J.R., Jones, L., Clark, F., and Smith, M. (1995). Is the Psychopath Morally Insane? Personality and Individual Differences 19: 741–52. Cima, M., Tonnaer, F., and Hauser, M.D. (2010). Psychopaths Know Right from Wrong but Don’t Care. Social Cognitive and Affective Neuroscience 5: 59–67. Cleckley, H. (1988). The Mask of Sanity: An Attempt to Clarify some Issues about the So-called Psychopathic Personality (5th edn). Augusta, GA: Emily S. Cleckley. Finger, E.C., Marsh, A.A., Mitchell, D.G., Reid, M.E., Sims, C., Budhani, S., Kosson, D.S., Chen, G., Towbin, K.E., Leibenluft, E., Pine, D.S., and Blair, J.R. (2008). Abnormal Ventromedial Prefrontal Cortex Function in Children with Psychopathic Traits during Reversal Learning. Archives of General Psychiatry 65: 586–94. Finger, E.C., Marsh, A.A., Blair, K.S., Reid, M.E., Sims, C., Ng, P., Pine, D.S., and Blair, R.J. (2011). Disrupted Reinforcement Signaling in the Orbitofrontal Cortex and Caudate in Youths with Conduct Disorder or Oppositional Defiant Disorder and a High Level of Psychopathic Traits. American Journal of Psychiatry 168: 152–62. Fisher, L., and Blair, R.J. (1998). Cognitive Impairment and its Relationship to Psychopathic Tendencies in Children with Emotional and Behavioral Difficulties. Journal of Abnormal Child Psychology 26: 511–19. Forth, A.E., Kosson, D.S., and Hare, R.D. (2003). The Psychopathy Checklist: Youth Version. Toronto: Multi-Health Systems. Frick, P.J., and Hare, R.D. (2002). The Psychopathy Screening Device. Toronto: Multi-Health Systems. Frijda, N.H. (1986). The Emotions. Cambridge and New York: Cambridge University Press.

MORAL STATUS AND PERCEIVED INTENTION



Glenn, A.L., Iyer, R., Graham, J., Koleva, S., and Haidt, J. (2009). Are All Types of Morality Compromised in Psychopathy? Journal of Personality Disorders 23: 384–98. Glenn, A.L., Raine, A., and Schug, R.A. (2009). The Neural Correlates of Moral Decision-making in Psychopathy. Molecular Psychiatry 14: 5–6. Gray, N.S., MacCulloch, M.J., Smith, J., Morris, M., and Snowden, R.J. (2003). Forensic Psychology: Violence Viewed by Psychopathic Murderers. Nature 423: 497–8. Herpertz, S.C., Werth, U., Lukas, G., Qunaibi, M., Schuerkens, A., Kunert, H.J., Freese, R., Flesch, M., Mueller-Isberner, R., Osterheider, M., and Sass, H. (2001). Emotion in Criminal Offenders with Psychopathy and Borderline Personality Disorder. Archives of General Psychiatry 58: 737–45. Kaufman, J., Birmaher, B., Brent, D., Rao, U., Flynn, C., Moreci, P., Williamson, D., and Ryan, N. (1997). Schedule for Affective Disorders and Schizophrenia for School-Age Children—Present and Lifetime Version (K-SADS-PL): Initial Reliability and Validity Data. Journal of the American Academy of Child and Adolescent Psychiatry 36: 980–8. Knobe, J. (2003). Intentional Action and Side Effects in Ordinary Language. Analysis 63: 190–4. Knobe, J. (2010). Person as Scientist, Person as Moralist. Behavioral and Brain Sciences 33: 315–29. Knobe, J., and Burra, A. (2006). The Folk Concepts of Intention and Intentional Action: A Cross-cultural Study. Journal of Cognition and Culture 6(1): 113–32. Knobe, J., and Mendlow, G.S. (2004). The Good, the Bad and the Blameworthy: Understanding the Role of Evaluative Reasoning in Folk Psychology. Journal of Theoretical and Philosophical Psychology 24: 252–8. Koenigs, M., Young, L., Adolphs, R., Tranel, D., Cushman, F., Hauser, M., and Damasio, A. (2007). Damage to the Prefrontal Cortex Increases Utilitarian Moral Judgements. Nature 446: 908–11. Koenigs, M., Kruepke, M., Zeier, J., and Newman, J.P. (2012). Utilitarian Moral Judgment in Psychopathy. Social Cognitive and Affective Neuroscience 7(6): 708–14. Leslie, A.M., Knobe, J., and Cohen, A. (2006). Acting Intentionally and the SideEffect Effect. Psychological Science 17: 421–7. Levenston, G.K., Patrick, C.J., Bradley, M.M., and Lang, P.J. (2000). The Psychopath as Observer: Emotion and Attention in Picture Processing. Journal of Abnormal Psychology 109: 373–85. Malle, B.F., and Knobe, J. (1997). The Folk Concept of Intentionality. Journal of Experimental Social Psychology 33: 101–21.



E . M . CARDINALE ET AL .

Malle, B.F., and Nelson, S.E. (2003). Judging Mens Rea: The Tension between Folk Concepts and Legal Concepts of Intentionality. Behavior Science and Law 21: 563–80. Mangan, J. (1949). An Historical Analysis of the Principle of Double Effect. Theological Studies 10: 41–61. Marsh, A.A., and Blair, R.J. (2008). Deficits in Facial Affect Recognition among Antisocial Populations: A Meta-Analysis. Neuroscience and Biobehavioral Reviews 32: 454–65. Marsh, A.A., and Cardinale, E.M. (2012). Psychopathy and Fear: Specific Impairments in Judging Behaviors that Frighten Others. Emotion 12(5): 892–8. Marsh, A.A., Finger, E., Mitchell, D.G.V., Reid, M.E., Sims, C., Kosson, D.S., Towbin, K.E., Leibenluft, E., Pine, D.S., and Blair, R.J.R. (2008). Reduced Amygdala Response to Fearful Expressions in Children and Adolescents With Callous-Unemotional Traits and Disruptive Behavior Disorders. American Journal of Psychiatry 16(5): 712–20. Marsh, A.A., Finger, E.C., Fowler, K.A., Jurkowitz, I.T., Schechter, J.C., Yu, H.H., Pine, D.S., and Blair, R.J. (2011a). Reduced Amygdala-Orbitofrontal Connectivity during Moral Judgments in Youths with Disruptive Behavior Disorders and Psychopathic Traits. Psychiatry Research 194: 279–86. Marsh, AA., Finger, E. E., Schechter, J.C., Jurkowitz, I.T., Reid, M.E., and Blair, R.J. (2011b). Adolescents with Psychopathic Traits report Reductions in Physiological Responses to Fear. Journal of Child Psychology and Psychiatry 52: 834–41. Ortony, A., Clore, G.L., and Collins, A. (1988). The Cognitive Structure of Emotions. New York: Cambridge University Press. Parkinson, C., Sinnott-Armstrong, W., Koralus, P.E., Mendelovici, A., McGeer, V., and Wheatley, T. (2011). Is Morality Unified? Evidence that Distinct Neural Systems Underlie Moral Judgments of Harm, Dishonesty, and Disgust. Journal of Cognitive Neuroscience 23: 3162–80. Patrick, C.J. (1994). Emotion and Psychopathy: Startling New Insights. Psychophysiology 31: 319–30. Pellizzoni, S., Siegal, M., and Surian, L. (2009). Foreknowledge, Caring, and the Side-Effect Effect in Young Children. Developmental Psychology 45: 289–95. Pettit, D., and Knobe, J. (2009). The Pervasive Impact of Moral Judgment. Mind and Language 24: 586–604. Phelan, M.T., and Sarkissian, H. (2008). The Folk Strike Back; or, Why you Didn’t Do it Intentionally, Though it was Bad and you Knew it. Philosophical Studies 138: 291–8. Roseman, I.J., Wiest, C., and Swartz, T.S. (1994). Phenomenology, Behaviors, and Goals Differentiate Discrete Emotions. Journal of Personality and Social Psychology 67: 206.

MORAL STATUS AND PERCEIVED INTENTION



Rothemund, Y., Ziegler, S., Hermann, C., Gruesser, S.M., Foell, J., Patrick, C.J., and Flor, H. (2012). Fear Conditioning in Psychopaths: Event-Related Potentials and Peripheral Measures. Biological Psychology 90(1): 50–9. Rozin, P., Lowery, L., Imada, S., and Haidt, J. (1999). The CAD Triad Hypothesis: A Mapping between Three Moral Emotions (Contempt, Anger, Disgust) and Three Moral Codes (Community, Autonomy, Divinity). Journal of Personality and Social Psychology 76: 574–86. Turiel, E. (1983). The Development of Social Knowledge: Morality and Convention. Cambridge: Cambridge University Press. Uttich, K., and Lombrozo, T. (2010). Norms Inform Mental State Ascriptions: A Rational Explanation for the Side-Effect effect. Cognition 116: 87–100. Vaidyanathan, U., Hall, J.R., Patrick, C.J., and Bernat, E.M. (2011). Clarifying the Role of Defensive Reactivity Deficits in Psychopathy and Antisocial Personality using Startle Reflex Methodology. Journal of Abnormal Psychology 120: 253–8. White, S.F., Marsh, A.A., Fowler, K.A., Schechter, J.C., Adalio, C., Pope, K., Sinclair, S., Pine, D.S., and Blair, R.J. (2012). Reduced Amygdala Response in Youths with Disruptive Behavior Disorders and Psychopathic Traits: Decreased Emotional Response versus Increased Top-Down Attention to Nonemotional Features. American Journal of Psychiatry 169(7): 750–58. Young, L., Cushman, F., Adolphs, R., Tranel, D., and Hauser, M. (2006). Does Emotion Mediate the Relationship between an Action’s Moral Status and its Intentional Status? Neuropsychological Evidence. Journal of Cognition and Culture 6: 291–304. Young, L., Bechara, A., Tranel, D., Damasio, H., Hauser, M., and Damasio, A. (2010). Damage to Ventromedial Prefrontal Cortex Impairs Judgment of Harmful Intent. Neuron 65: 845–51. Young, L., Koenigs, M., Kruepke, M., and Newman, J.P. (2012). Psychopathy Increases Perceived Moral Permissibility of Accidents. Journal of Abnormal Psychology 121(3): 659–67. Zalla, T., Machery, E., and Leboyer, M. (2010). Intentional Action and Moral Judgment in Asperger Syndrome and High-Functioning Autism. Unpublished manuscript, Institut Jean-Nicod.

6 The Concept of Intentional Action in High-Functioning Autism Edouard Machery, University of Pittsburgh Tiziana Zalla, Institut Jean Nicod

Introduction The concept of intentional action plays a central role in social cognition. We distinguish between intentional and non-intentional actions when we explain people’s behavior: we appeal to causes to explain non-intentional actions, but to reasons to explain intentional ones (Bratman, 1987; Malle and Knobe, 1997b). We also rely on this distinction when we evaluate people’s behavior in light of moral or conventional norms: when intentional actions harm others, these actions are judged to be worse than similar non-intentional actions (Lagnado and Shannon, 2008).

The authors wish to thank Pierre Jacob, Paul Egré, and Liane Young for their insightful comments on previous versions of this chapter. The present findings were presented at the Winter Workshop 2008 on Games, Experiments, and Philosophy at the Max Planck Institute of Economics in January 2008 in Jena, Germany. This work was made possible by the Fondation Orange and the Fondation FondaMental. Addresses for correspondence: Edouard Machery, Department of History and Philosophy of Science, University of Pittsburgh, 1017CL, Pittsburgh, PA, 15260, USA; email: [email protected]. Tiziana Zalla, Institut Jean Nicod-CNRS, Ecole Normale Supérieure, 29 rue d’Ulm 75005 Paris, France; email: [email protected].

THE CONCEPT OF INTENTIONAL ACTION IN HFA



While philosophical and psychological research has shed light on the use of the concept of intentional action in typical populations (Malle and Knobe, 1997a; Knobe, 2003, 2006; Nadelhoffer, 2004, 2006; Mele and Cushman, 2007; Nichols and Ulatowski, 2007; Machery, 2008; Mallon, 2008; Phelan and Sarkissian, 2008, 2009; Wright and Bengson, 2009; Guglielmo et al., 2009; Guglielmo and Malle, 2010; Nanay, 2010; Alfano et al., 2012; Rose et al., 2012; Scaife and Webber, 2013), our understanding of the use of this concept among people with atypical social cognition and mindreading abilities, particularly people with autism spectrum disorders, remains incomplete (Phillips et al., 1998; Russell and Hill, 2001; Zelazo et al., 2002; Leslie et al., 2006; Zalla et al., 2009; Zalla et al., 2011; Zalla and Leboyer, 2011; Buon et al., 2013). Autism spectrum disorders (ASDs) are pervasive developmental disorders characterized by abnormal social interactions, verbal and nonverbal communication problems, and restricted stereotypic interests. Within the domain of ASDs, “high-functioning autism” (HFA) commonly refers to individuals meeting criteria for autism with normal intellectual functioning and a history of speech and language delay. Those at the higher-functioning end of the HFA group, sometimes diagnosed with Asperger syndrome (AS), show no evidence of impaired language function, and their intellectual abilities fall within the normal range. As with other individuals within the ASDs, the clinical features of AS and HFA include trouble forming friendships, difficulties with social cognition, inappropriate social interactions, restricted interests, diminished capacity for empathy, and poor communication (Frith, 1989; Frith, 1996; Bauminger and Kasari, 1999). Importantly, and contrary to the more severe forms of autism, individuals with HFA/AS typically pass tests of Theory of Mind (ToM) and thus have some ability to attribute intentions, beliefs, and desires to oneself and to others. However, they often fail more advanced ToM tasks, based on the detection of sarcasm, irony, or bluff (Happé, 1994) or on the recognition of faux pas (BaronCohen et al., 1999; Zalla et al., 2009). The aim of the present work is to extend the research on the similarities and differences between the judgments about intentional action made by adults with HFA/AS and by typical adults. We focus on a particular kind of action: namely, means that are not intrinsically positively valued (what we will call “purely instrumental actions”). We



E . MACHERY , T . ZALLA

show that people with HFA/AS understand the intentional nature of these actions differently from typical individuals. Here is how we will proceed. We first review the existing literature on the folk concept of intentional action, and on this basis we predict that the judgments of people with HFA/AS about the intentional nature of purely instrumental actions will differ from those of typical individuals. We then describe our participants and present our study; we discuss the significance of our findings; and we attempt to explain them.

Purely Instrumental Desires and ASDs The goal of this section is to describe our hypothesis about the concept of intentional action in people with HFA/AS. We first highlight the role played by the ascription of desires in intentionality judgments: when an outcome is morally neutral, people judge that it has been brought about intentionally only if the agent desired bringing it about. We then focus on purely instrumental actions, and we show that the ascription of the desires leading to these actions involves a sophisticated form of mindreading as well as the ascription of superficially conflicting desires: desires for purely instrumental actions (“purely instrumental desires”) are about outcomes that agents do not intrinsically desire or even that they disvalue. These features of the ascription of purely instrumental desires suggest that people with HFA/AS may conceive of the intentional status of purely instrumental actions differently from typical individuals.

Desires and the concept of intentional action Previous research suggests that for an action to be judged intentional it has to have the following five properties (Malle and Knobe, 1997a). The agent has to desire the outcome of this action; she needs to believe that her action will bring about this outcome; she has to intend to perform the action; she needs to be aware of performing it while performing it; finally, she must have some skill to perform it.1 Figure 6.1 summarizes Malle and Knobe’s model.

It also seems that together these five properties are meant to be sufficient for being intentional. 1

THE CONCEPT OF INTENTIONAL ACTION IN HFA

BELIEF

SKILL



DESIRE

INTENTION

AWARENESS

INTENTIONALITY

Figure 6.1 Malle and Knobe’s (1997a) model.

The crucial point for present purposes is Malle and Knobe’s claim that being desired is necessary for an outcome to be judged to have been intentionally brought about. If people do not think that the agent desired the outcome, then they would not judge that this outcome was brought about intentionally. Recent research has, however, cast some doubts on whether desire is really necessary for an action to be judged intentional (Knobe, 2003, 2006; Mele and Cushman, 2007). Knobe (2003, 2006) designed a pair of cases (the harm and help cases) in which an agent brings about a foreseen side-effect, while saying explicitly that he does not care about this side-effect and thus expressing no desire to bring it about. In the harm case, the foreseen side-effect (harming the environment) is plausibly morally bad, while in the help case it is plausibly morally good. Participants were significantly more likely to judge that the chairman intentionally harmed the environment (82%) than they were to judge that she intentionally helped the environment (23%). This effect is now known as “the side-effect effect.” The side-effect effect suggests that when an action is morally wrong, desire is not necessary for it to be judged intentional. It is currently unclear how to revise our understanding of the folk concept of intentional action in light of these findings (Knobe, 2006; Nadelhoffer, 2006; Mele and Cushman, 2007; Nichols and Ulatowski, 2007; Machery, 2008; Mallon, 2008; Phelan and Sarkissian, 2008, 2009; Guglielmo et al., 2009; Wright and Bengson, 2009; Guglielmo and Malle, 2010; Nanay, 2010). Particularly, it is unclear whether in the harm case the CEO has really no desire whatsoever toward harming the environment. Machery (2008) has proposed that people judge negative sideeffects such as harming the environment to be intentional because they



E . MACHERY , T . ZALLA

conceptualize harming the environment as a cost that the agent is willing to incur in order to get a benefit (“the trade-off hypothesis”). Thus, according to the trade-off hypothesis, the side-effect in the harm case (harming the environment) is viewed as the object of a desire.2 Other models, such as Guglielmo and Malle’s (2010) defense of Malle and Knobe’s original model of intentional action, also propose that the relation of the agent to harming the environment is distinct from her relation to helping the environment: the agent has a pro-attitude toward harming the environment, while having no pro-attitude whatsoever toward helping the environment. Obviously, this is not the appropriate occasion to resolve this controversy. It may be that some actions that are morally wrong are indeed judged intentional even if their outcomes are not the objects of any desire whatsoever. But what matters for our study is the following point: morally neutral outcomes (such as paying an extra euro and getting a free cup, as in the case presented in our study) are judged to have been brought about intentionally only if they are thought to be the objects of some kind of desire. People must then be able to ascribe a desire for an outcome in order to judge that a morally neutral action bringing about this outcome is intentional.

Purely instrumental actions An action ç performed by an agent is instrumental only if on this occasion this agent ç-es because she believes that ç-ing brings about an outcome that she wants to bring about. For example, when someone enters her own apartment, she inserts the key in the keyhole, turns the key, and pushes the door. Inserting the key in the keyhole, turning the key, and pushing the door are all instrumental actions: the agent performs them because she believes that doing so brings about the outcome she wants to bring about (entering her apartment). Instrumental desires are desires for the outcomes of instrumental actions. When someone enters her own apartment, she instrumentally desires that the key be in the keyhole, that the key be turned, and so forth. Outcomes that are the

2 Evidence for the trade-off hypothesis is mixed. Mallon (2008; see also Phelan and Sarkissian, 2009; Wright and Bengson, 2009) found that participants judged that some negative side-effects that cannot be interpreted as costs are intentionally brought about (for further discussion, see Nanay, 2010).

THE CONCEPT OF INTENTIONAL ACTION IN HFA



objects of instrumental desires are desired, but not for their own sake: as objects of instrumental desires, they are desired because bringing them about brings about outcomes that are desired for their own sake or intrinsically. Of course, some outcomes that are desired instrumentally may also be intrinsically desired. For instance, one might work out because one wants to lose weight (working out is thus the object of an instrumental desire) and because one enjoys working out (working out is then desired for itself). Other outcomes are only desired instrumentally, either because the agent is neutral with respect to them (she neither desires the outcome to happen nor desires the outcome not to happen) or because the agent disvalues the outcome (everything else being equal she would prefer the outcome not to happen). For instance, one might instrumentally desire to have blood taken, even though everything else being equal one would rather not have a syringe in one’s arm. As noted earlier, we call “purely instrumental desires” those instrumental desires for outcomes one either is neutral about or one disvalues. Thus, people typically have a purely instrumental desire for having a blood test. When people attribute a purely instrumental desire to an agent, their representation of how the agent values the object of the purely instrumental desire in itself diverges from their representation of what she values or intrinsically desires. The agent either does not value or even disvalues the outcome in itself, which is nonetheless represented as being (instrumentally) desired. For instance, when someone needs to get some blood work done, we are likely to ascribe to her a purely instrumental desire to have some blood taken from her body together with an intrinsic preference for not having blood so taken. What matters for our purpose is the idea that the ascription of purely instrumental desires involves some of the more advanced forms of mindreading and that it requires holding two superficially conflicting representations of an agent’s desires: ascribing such a desire involves ascribing a desire for something the agent does not intrinsically desire or even disvalues.

Our hypothesis Because the ascription of purely instrumental desires involves more advanced forms of mindreading, and because it requires holding two superficially conflicting representations of an agent’s desires, some individuals may not find it meaningful to ascribe a desire for an outcome that



E . MACHERY , T . ZALLA

is not positively valued. Because ascribing a desire is necessary to view a morally neutral outcome as having been intentionally brought about, these individuals will not view means that are not intrinsically valued (purely instrumental actions) as being intentionally brought about. We hypothesize that people with HFA/AS are in this situation. While individuals with HFA/AS and typical individuals represent intentions similarly when this involves ascribing a single propositional attitude, the ascription of intentions by individuals with HFA/AS will differ from that by typical individuals when people have to solve an apparent inconsistency between the attitudes of the ascribee, as happens when an agent instrumentally desires the outcome of an action that is intrinsically negatively valued. Thus, we predicted that people with HFA/AS would be less inclined than typical people to treat means that are not intrinsically valued or that are disvalued as intentional. To test this prediction, we presented participants with two cases taken from the experimental-philosophy literature on folk judgments about intentional action: the free-cup and the extra-euro cases (Machery, 2008). These cases were chosen because one of them asks participants to evaluate whether an action that is desired not in itself, but only because it is a means for a desired outcome, is intentional.

Representing intentional action in people with ASDs: previous research To conclude this first section, we examine how our hypothesis fits with previous research on the intentionality judgments of people with ASDs. Although a large number of studies have highlighted impairments in both action execution and action understanding in ASDs (Smith and Bryson, 1994; Hughes, 1996; Russell and Jarrold, 1998, 1999; Zalla et al., 2006; Cattaneo et al., 2007; Fabbri-Destro et al., 2009; Zalla et al., 2010; Stoit et al., 2011), verbally able people with ASDs often exhibit a preserved ability to infer the overall action goal, likely using object semantic knowledge or contextual cues (Zalla et al., 2006; Cattaneo et al., 2007; Fabbri-Destro et al., 2009; Zalla et al., 2010). This could explain why intention attribution is sometimes similar in people with ASDs and typical individuals (Aldridge et al., 2000; Russell and Hill, 2001). According to Carpenter and colleagues (2001), young children with ASDs and typical individuals understand unfulfilled intentions and the goal state of an intended action similarly, suggesting that their

THE CONCEPT OF INTENTIONAL ACTION IN HFA



performance in this domain is more similar to typical children’s than in the domain of belief understanding and joint attention. However, the authors raised the possibility that this similarity may result from their control group’s relatively poor performance. On the other hand, D’Entremont and Yazbek (2007) found that after observing an actor performing intentional and accidental actions on the same objects, typically developing children imitated more intentional actions, whereas children with autism tended to imitate intentional and accidental actions equally often, and to reproduce the action sequence in the same order as the experimenter, without reference to the agent’s intentions. These findings suggest that although individuals with ASDs are able to reproduce movements faithfully, they may be less likely to imitate when they have to identify the model’s goals or intentions (D’Entremont and Yazbek, 2007; Tomasello et al., 2005). Importantly, Phillips and colleagues (1998) have shown that children with autism do not distinguish the unintended from the intended outcomes of their own actions as compared to a control group, when the unintended actions are positively valued. The experimental task involved firing a ray gun to hit one of several colored targets, and participants either hit the chosen target or accidentally hit a target with a different color. The value (and thus desirability) of the outcomes was manipulated in order to make some outcomes intended but not desirable (the participant hit the chosen target, but its value was low), and others desirable but unintended (the participant did not hit the chosen target, but the value of the target she accidentally hit was high). Participants were asked to judge whether they had “meant to” hit the target by comparing their goal and the outcome of their action. Compared with children with similar intellectual abilities, young children with ASDs were less prone to judge that outcomes that are desirable might nevertheless be unintended. More generally, it might be that people with ASDs are less prone to judge that an outcome can be non-intentional when it is positively valued, or to conceptualize an action as intentional when the agent negatively values it.3

3 A reviewer rightly noted that young children with ASDs seem to follow the rule that, if an outcome is desired, then its occurrence was intended, and that this rule differs from the rule we hypothesize our participants tend to follow—viz. if an outcome is not intrinsically desired, then its occurrence was not intended. While this is correct, we are merely



E . MACHERY , T . ZALLA

A previous study using an advanced ToM task, the faux pas recognition task, revealed that judgments about intentional action in people with HFA/AS subtly differ from those of typical individuals (Zalla et al., 2009). In this study, Zalla and collaborators (2009) have shown that individuals with AS have difficulty conceptualizing a faux pas as non-intentional. The faux pas is a particular case of a non-intentional action. As defined by Baron-Cohen and collaborators (1999), a faux pas occurs when a speaker says something that is unpleasant to the listener, although the speaker never intended to do so. Participants were presented with twenty stories describing interpersonal interactions in everyday life situations in which a faux pas occurred, and twenty control stories containing a minor conflict or accident. Here is a typical faux pas story. Jill just moved into a new apartment, and she has bought new curtains. When Jill had just finished hanging the new curtains, her best friend, Lisa, came over to visit the new apartment. After a tour of the apartment, Lisa said: “These curtains are horrible. I hope you are going to get some new ones!” Obviously, Lisa’s statement reflected her mistaken belief that the curtains had been bought by the previous owner, and she did not have the intention to hurt Jill. Interestingly, although individuals with AS generally detected that Lisa had said something awkward, they were generally unable to explain why what Lisa did was awkward. They also failed to understand that Lisa had a mistaken belief and that this belief caused her faux pas. Instead, they tended to view faux pas as having been done intentionally, and they interpreted Lisa’s statement as caused by an intention to hurt Jill. This study showed that, in contrast to typical individuals, individuals with HFA/AS do not conceptualize faux pas as non-intentional, and that their judgments of intentional action differ subtly from those of typical individuals.

Participants A group of twenty-six adults with a clinical diagnosis of HFA or AS (HFA/AS) according to DSM-IV R (American Psychiatric Association, 1994), ASDI (Asperger Syndrome Diagnostic Interview, Gillberg et al.,

discussing Phillips et al. (1998) to show that people with ASDs may apply the concept of intentional action differently from typical individuals.

THE CONCEPT OF INTENTIONAL ACTION IN HFA



Table 6.1. Demographic and neuropsychological data of both groups (mean and standard deviation)

Age (years/months) Gender (F/M) Education (years) Faux pas test (total score) Full scale IQ (WAIS-III) Verbal IQ Performance IQ

HFA/AS (N = 27)

CP (N = 40)

28.3 (6.5) 4/23 14.7 (3.8) 40.1 (5.7) 99.1 (22.8) 100.9 (24.4) 91.1 (20.8)

28.8 (7.9) 8/32 13.3 (3.0) 56.5 (4.1) 99.5 (13.9) 101.4 (13.9) 96.9 (13.3)

2001), and the ADOS (Lord et al., 2000) were recruited from Albert Chenevier Hospital in Créteil, France (Table 6.1). All diagnoses were made by clinicians experienced in the field of autism and were based on observations of the participants. Interviews with parents using the ADI-R (Autism Diagnostic Interview, Lord et al., 1994) confirmed the diagnoses; elevated scores indicated problematic behavior in the three following areas: reciprocal social interaction, communication, and stereotyped behaviors. The cutoff points for the three classes of behavior were reciprocal social interaction 10, communication 8, and stereotyped behaviors 3. All participants scored above the cutoff points. All patients were medication-free at the time of testing and had no history of psychiatric disorder (other than ASD), neurological disorder, or head injury. All participants were native French speakers, had normal or corrected-to-normal vision, and signed informed consent before volunteering for this study, in accordance with the local ethics committee and the Declaration of Helsinki. Forty comparison participants (CPs) were taken from the general population and were chosen to match the clinical group with respect to the characteristics of age, IQ, gender, and education. Participants with HFA/AS and CPs received Verbal and Performance IQs (WAIS-III, Wechsler, 1999) and the Faux Pas Detection Task (max score = 60) (Baron-Cohen et al., 1999) to assess Theory-of-Mind abilities (Table 6.1). Overall, unpaired t-tests revealed that the two groups did not differ on chronological age (t(64) = –0.29; p = 0.77), on education



E . MACHERY , T . ZALLA

(t(64) = 1.4; p = 0.17), or on IQ level (Full-scale, Verbal, Performance: t-test: t(64) = –0.07, p = 0.94; t(64) = –0.9, p = 0.92; t(64) = –1.2, p = 0.24) from the CP group. They also scored significantly lower than the CPs on the faux pas test (t(40) = –4.272, p < 0.0001).

Study Procedure Participants were presented with the French translations of the extra-euro and free-cup vignettes written on a page that remained placed in front of them throughout the questioning so that they did not have to remember the vignettes (see Appendix for the French versions of these vignettes). The Extra-Euro Case Joe was feeling quite dehydrated, so he stopped by the local smoothie shop to buy the largest sized drink available. Before he ordered, the cashier told him that the Mega-Sized Smoothies were now one euro more than they used to be. Joe replied, “I don’t care if I have to pay one euro more, I just want the biggest smoothie you have.” Sure enough, Joe received the Mega-Sized Smoothie and paid one euro more for it. The Free-Cup Case Joe was feeling quite dehydrated, so he stopped by the local smoothie shop to buy the largest sized drink available. Before he ordered, the cashier told him that if he bought a Mega-Sized Smoothie he would get it in a special commemorative cup. Joe replied, “I don’t care about a commemorative cup, I just want the biggest smoothie you have.” Sure enough, Joe received the Mega-Sized Smoothie in a commemorative cup.

The order of vignette presentation was counterbalanced between participants. After reading the extra-euro vignette, participants were asked to answer the following two questions: 1) Did Joe intentionally pay an extra euro? 2) Was Joe’s paying an extra euro blameworthy, praiseworthy, or neutral? For the free-cup vignette, participants were asked to answer the following two questions: 1) Did Joe intentionally obtain the commemorative cup? 2) Was Joe’s obtaining the commemorative cup blameworthy, praiseworthy, or neutral?

THE CONCEPT OF INTENTIONAL ACTION IN HFA



80 70

70%

60 50 40

33.3%

25.9%

30 20 10%

10 0 The extra-euro case

The free-cup case CP

HFA/AS

Figure 6.2 Intentionality—both cases.

Results: intentionality As expected, when we examined French CPs’ answers, we found the same pattern of responses as that previously reported by Machery (2008). 70% of CPs reported that Joe intentionally paid an extra euro, while only 10% said that Joe intentionally obtained the commemorative cup. A å-squared test yielded a highly significant difference between the two cases for the intentionality question (å2(1) = 72.5, p < 0.0001). By contrast, only 33.3% of participants with HFA/AS said that Joe intentionally paid an extra euro to get a mega-sized smoothie, and 25.9% said that Joe intentionally obtained the commemorative cup. This contrast was not significant (å2(1) = 0.86, p = 0.35). The two groups significantly differed in their judgment of intentionality on the extra-euro case (å2(1) = 7.36, p = 0.0067), whereas group difference was not significant on the free-cup case (å2(1) = 1.94, p = 0.16) (Figure 6.2).

Results: praise and blame When asked to judge the moral valence of Joe’s action in the extra-euro case, most CPs judged Joe’s action to be neutral (90%); only three subjects (7.5%) judged Joe’s action as being praiseworthy and one subject (2.5%) as being blameworthy. Similarly, 81.5% of participants with HFA/

 100

E . MACHERY , T . ZALLA

90%

90 81.5%

80 70 60 50 40 30

14.8%

20 10

7.5%

3.7% 2.5%

0 CP neutral

HFA/AS praiseworthy

blameworthy

Figure 6.3 Praise and blame in the extra-euro case.

AS considered Joe’s action to be morally neutral, four subjects (14.8%) as being praiseworthy, and one subject (3.7%) as being blameworthy in the extra-euro case (Figure 6.3). The group difference was not significant in the extra-euro case (å2(2) = 1.04, p = 0.59). Similarly, in the free-cup case most CPs considered Joe’s action as neutral (92.5%), only one subject (2.5%) as being praiseworthy, and two subjects (5%) as being blameworthy. Moral judgment in participants with HFA/AS slightly differed in that in the free-cup case 77.8% of participants with HFA/AS regarded Joe’s action to be morally neutral, while the remaining 22.2% regarded Joe’s action as being praiseworthy (Figure 6.4). The two groups significantly differed in their moral judgments (å2(1) = 4.97, p = 0.02).4

4

We do not know how to explain why people with HFA/AS turned out to be more likely to judge the action praiseworthy in the free-cup case. This is a puzzling result that calls for further investigation. As a reviewer noted, this may contribute to explaining why people with HFA/AS are also more likely to judge this action to be intentional in the free-cup case. However, because the group difference in intentional judgment did not reach significance, we will refrain from speculating about its possible causes. Further studies are needed to investigate a possible relationship between increased attribution of praise and increased attribution of intent for the free-cup case.

THE CONCEPT OF INTENTIONAL ACTION IN HFA

100



92.5%

90 77.8%

80 70 60 50 40 30

22.2%

20 10

5% 2.5%

0 CP neutral

HFA/AS praiseworthy

blameworthy

Figure 6.4 Praise and blame in the free-cup case.

Discussion The present study aimed to investigate how adults with HFA/AS understand the intentional status of those actions that are not intrinsically valuable (e.g. paying an extra euro) but that are desired for purely instrumental reasons. The results showed that while adults with typical development evaluated differently the intentional status of paying an extra euro and of getting a free cup (see also Machery, 2008), adults with HFA/AS treated them similarly, judging both to be non-intentional. These results support the hypothesis that individuals with HFA/AS are less prone to ascribe purely instrumental desires and that, as a result, they are less prone to conceive means as intentionally brought about when they are not intrinsically valued (i.e. when they are the objects of purely instrumental desires). Consider the extra-euro case. To conceptualize paying an extra euro as an intentional action, one needs to represent this action as the object of a purely instrumental desire. The agent intentionally pays an extra euro because he instrumentally desires to pay an extra euro, and he instrumentally desires to pay an extra euro because doing so is a means for an intrinsically valuable outcome, even though everything else being equal



E . MACHERY , T . ZALLA

the agent would prefer not to pay an extra euro. If participants with HFA/AS are less prone to represent intrinsically neutral or negative actions as being desired for instrumental reasons, they should judge that the agent did not intentionally pay an extra euro to get an extralarge smoothie—exactly what we found. By contrast, to characterize the outcome in the free-cup case as being brought about unintentionally, one need not represent the action of getting a free cup as the object of a desire. Thus the fact that people with HFA/AS are less prone to ascribe purely instrumental desires would not prevent them from judging that the agent did not intentionally get a free cup. Why would individuals with AS be less prone to view means as being desired when the agent disvalues them? We have seen that when an outcome is desired purely instrumentally, people need to distinguish what the agent values intrinsically from what she desires. Thus people need to assign two contradictory attitudes toward the same object (e.g., toward paying an extra euro) and to further explain the inconsistency between one of these attitudes (e.g., disvaluing paying an extra euro) and the behavioral outcome (e.g., paying an extra euro). There are two main competing explanations of why people with HFA/AS would differ from typical individuals when confronted with this situation. First, the ability to reason in terms of purely instrumental desires is likely to rely on more advanced capacities that make up the theory of mind, and it is probably the kind of capacities that are tapped into by more advanced ToM tasks. While people with HFA/AS pass the usual ToM tasks, their performances differ from typical individuals’ on these more advanced tasks, which require the ability to process different types of social, semantic, and contextual information and to integrate them to create a coherent scenario. For instance, as previously shown (Zalla et al., 2009), the judgments made by adults with AS about whether an agent committed a faux pas intentionally or accidentally differ from those made by typical individuals, as do their explanations of why a speaker accidentally says something that could hurt her interlocutor. So the performance of people with HFA/AS in the extra-euro and free-cup cases may result from subtle differences between their theory of mind and that of typical individuals. An alternative hypothesis points to differences in the domain of executive functions between typical individuals and individuals with

THE CONCEPT OF INTENTIONAL ACTION IN HFA



ASDs, crucially affecting high-level reasoning and complex problemsolving (Ozonoff et al., 1991; Hughes et al., 1994). A large body of evidence shows that individuals with HFA/AS differ from typical individuals in tasks tapping into planning, inhibition, and cognitive flexibility (i.e. the ability to engage and disengage actions in the service of overall goals) (Hughes and Russell, 1993; Pennington and Ozonoff, 1996). According to the cognitive complexity and control theory (Zelazo et al., 2002), individuals with ASDs differ from typical individuals on mindreading tasks because of a more general difficulty in combining discordant epistemic states into a single system of inferences.5 In usual ToM tasks (e.g. in the usual false-belief task), participants have to combine their own perspective with the perspective of the agent. Similarly, in the extra-euro case, people have to distinguish the intrinsic preference for the non-occurrence of an outcome and an instrumental desire for its occurrence. So the performance of people with HFA/AS in the extra-euro and free-cup cases may result from the differences between their disposition and typical individuals’ disposition to integrate different epistemic perspectives. Finally, one may wonder how these possible explanations fit with Zalla et al.’s (2009) results about the interpretation of faux pas by people with HFA/AS. Our hypothesis that individuals with HFA/AS are less prone to judge that an action that is not intrinsically desired can be intentionally brought about may appear in contradiction with previous findings revealing that they judge that people commit faux pas intentionally. However, this apparent inconsistency can be resolved if one considers that the faux pas case crucially differs from the vignettes used in the present study. In the faux pas case, the speaker violates a moral norm, whereas the actions described by the vignettes do not. As previously suggested (Zalla et al., 2009; Zalla and Leboyer, 2011), we speculate that when an action violates a norm, as a faux pas does, people with HFA/AS often follow the simple rule that if an action violates a norm it is morally wrong.

5

Executive functions play a role in combining representations.



E . MACHERY , T . ZALLA

Conclusion Theoretically, this study addresses important issues about the concept of intentional action in individuals with HFA/AS. Specifically, the present results suggest that the judgments made by adults with HFA/AS about the intentionality of a particular type of actions—viz., those actions that are not intrinsically valuable but that are desired as means for something else—differ from the judgments made by typical adults. To explain these findings, we have hypothesized that in contrast to typical individuals, people with AS do not judge that means that are negatively valued can be desired and as a consequence intentionally brought about. This might be due to a ToM impairment associated with difficulties to ascribe purely instrumental desires or, alternatively, to a domain-general impairment in the computational mechanisms responsible for integrating inconsistent or conflicting perspectives. Further research is needed to assess the explanatory role of these two hypotheses by using a greater number of scenarios per condition as well as specific tasks tapping into the ability to use the concept of instrumental desire to explain intentional behaviors, combined with validated measures of cognitive flexibility and ToM abilities in individuals with ASDs.

Appendix Le verre décoré (free-cup case) Parce que Stéphane a très soif, il entre dans un café et commande un Perrier. Le serveur lui répond que s’il commande un Perrier, son Perrier lui sera servi dans un verre spécialement décoré. Stéphane réfléchit 5 secondes et répond: “Ce verre spécialement décoré n’a pas d’importance, je veux juste un Perrier.” Naturellement, le Perrier lui est servi dans un verre spécialement décoré.

Un euro de plus (extra-euro case) Parce que Stéphane a très soif, il entre dans un café et commande un Perrier. Le serveur lui répond que le Perrier a augmenté d’un euro. Stéphane répond: “Payer un euro en plus n’a aucune importance, je veux juste un Perrier.” Naturellement, le Perrier lui est servi et Pierre paie un euro de plus.

THE CONCEPT OF INTENTIONAL ACTION IN HFA



References Aldridge, M.A., Stone, K.R., Sweeney, M.H., and Bower, T.G.R. (2000). Preverbal Children with Autism Understand the Intentions of Others. Developmental Science 3: 294–301. American Psychiatric Association (2000). Diagnostic and Statistical Manual of Mental Disorders, 4th Edition DSM-IV-TR (Text Revision). Washington, DC: American Psychiatric Association. Baron-Cohen, S., O’Riordan, M., Stone, V., Jones, R., and Plaisted, K. (1999). Recognition of Faux Pas by Normally Developing Children and Children with Asperger Syndrome or High-functioning Autism. Journal of Autism and Developmental Disorders 29: 407–18. Bauminger, N., and Kasari, C. (1999). Brief Report: Theory of Mind in Highfunctioning Children with Autism. Journal of Autism and Developmental Disorders 29: 81–6. Buon, M., Dupoux, E., Jacob, P., Chaste, P., Leboyer, M., and Zalla, T. (2013). The Role of Causal and Intentional Judgments in Moral Reasoning in Individuals with High Functioning Autism. Journal of Autism and Developmental Disorders 43: 458–70. Carpenter, M., Pennington, B.F., and Rogers, S.J. (2001). Understanding of Others’ Intentions in Children with Autism. Journal of Autism and Developmental Disorders 31: 589–99. Cattaneo, L., Fabbri-Destro, M., Boria, S., Pieraccini, C., Monti, A., Cossu, G., and Rizzolatti, G. (2007). Impairment of Actions Chains in Autism and its Possible Role in Intention Understanding. Proceedings of the National Academy of Sciences 104: 17825–30. D’Entremont, B., and Yazbek, A. (2007). Imitation of Intentional and Accidental Actions by Children with Autism. Journal of Autism Developmental Disorders 37: 1665–78. Fabbri-Destro, M., Cattaneo, L., Boria, S., and Rizzolatti, G. (2009). Planning Actions in autism. Experimental Brain Research 192: 521–5. Frith, C. (1996). Brain Mechanisms for Having a “Theory of Mind”. Journal of Psychopharmacology 10: 9–15. Frith, U. (1989). Autism: Explaining the Enigma. Oxford: Basil Blackwell. Gillberg, C., Gillberg, C., Råstam, M., and Wentz, E. (2001). The Asperger Syndrome (and High-Functioning Autism) Diagnostic Interview (ASDI): A Preliminary Study of a New Structured Clinical Interview. Autism 5: 57–66. Guglielmo, S., and Malle, B.F. (2010). Can Unintended Side Effects be Intentional? Resolving a Controversy over Intentionality and Morality. Personality and Social Psychology Bulletin 36: 1635–47.



E . MACHERY , T . ZALLA

Guglielmo, S., Monroe, A.E., and Malle, B.F. (2009). At the Heart of Morality Lies Folk Psychology. Inquiry 52: 449–66. Happé, F. (1994). An Advanced Test of Theory of Mind: Understanding of Story Characters’ Thoughts and Feelings by Able Autistic, Mentally Handicapped and Normal Children and Adults. Journal of Autism and Developmental Disorders 24: 129–54. Hughes, C. (1996). Brief Report: Planning Problems in Autism at the Level of Motor Control. Journal of Autism and Developmental Disorders 26: 99–107. Hughes, C., and Russell, J. (1993). Autistic Children’s Difficulty with Mental Disengagement from an Object: Its Implications for Theories of Autism. Developmental Psychology 29: 498–510. Hughes, C., Russell, J., and Robbins, T.W. (1994). Evidence for Executive Dysfunction in Autism. Neuropsychologia 32: 477–92. ICD-10 (1992). Classification of Mental and Behavioural Disorders Diagnostic Criteria for Research. World Health Organization. Knobe, J. (2003). Intentional Action and Side Effects in Ordinary Language. Analysis, 63: 190–3. Knobe, J. (2006). The Concept of Intentional Action: A Case Study in the Uses of Folk Psychology. Philosophical Studies 130: 203–31. Lagnado, D.A., and Channon, S. (2008). Judgments of Cause and Blame: The Effects of Intentionality and Foreseeability. Cognition 108: 754–70. Leslie, A., Mallon, R., and DiCorcia, J.A. (2006). Transgressors, Victims, and Cry Babies: Is Basic Moral Judgment Spared in Autism? Social Neuroscience 1: 270–83. Lord, C., Rutter, M., and Couteur, A.L. (1994). Autism Diagnostic InterviewRevised: A Revised Version of the Diagnostic Interview for Caregivers of Individuals with Possible Pervasive Developmental Disorders. Journal of Autism and Developmental Disorders 24: 659–85. Lord, C., Risi, S., Lambrecht, L., Cook, E.H., Jr., Leventhal, B.L., DiLavore, P.C., Pickles, A., and Rutter, M. (2000). The Autism Diagnostic Observation Schedule-generic: A Standard Measure of Social and Communication Deficits Associated with the Spectrum of Autism. Journal of Autism and Developmental Disorders 30: 205–23. Machery, E. (2008). The Folk Concept of Intentional Action: Philosophical and Experimental Issues. Mind and Language 23: 165–89. Malle, B.F., and Knobe, J. (1997a). The Folk Concept of Intentionality. Journal of Experimental Social Psychology 33: 101–21. Malle, B.F., and Knobe, J. (1997b). Which Behaviors do People Explain? A Basic Actor–Observer Asymmetry. Journal of Personality and Social Psychology 72: 288–304.

THE CONCEPT OF INTENTIONAL ACTION IN HFA



Mallon, R. (2008). Knobe vs. Machery: Testing the Trade-Off Hypothesis. Mind and Language 23: 247–55. Mele, A.R., and Cushman, F. (2007). Intentional Action, Folk Judgments, and Stories: Sorting Things Out. Midwest Studies in Philosophy 31: 184–201. Nadelhoffer, T. (2004). Blame, Badness, and Intentional Action: A Reply to Knobe and Mendlow. Journal of Theoretical and Philosophical Psychology 24: 259–69. Nadelhoffer, T. (2006). Bad Acts, Blameworthy Agents, and Intentional Actions: Some Problems for Juror Impartiality. Philosophical Explorations 9: 203–19. Nanay, B. (2010). Morality or Modality: What does the Attribution of Intentionality Depend on? The Canadian Journal of Philosophy 40: 28–40. Nichols, S., and Ulatowski, J. (2007). Intuitions and Individual Differences: The Knobe Effect Revisited. Mind and Language 22: 346–65. Ozonoff, S., Pennington, B.F., and Rogers, S.J. (1991). Executive Function Deficits in High-functioning Autistic Individuals: Relationship to Theory of Mind. Journal of Child Psychology and Psychiatry 32: 1081–105. Pennington, B.F., and Ozonoff, S. (1996). Executive Functions and Developmental Psychopathology. Journal of Child Psychology and Psychiatry 37: 51–87. Phelan, M.T., and Sarkissian, H. (2008). The Folk Strike Back; or, Why you Didn’t Do it Intentionally, though it was Bad and You Knew it. Philosophical Studies 138: 291–8. Phelan, M.T., and Sarkissian, H. (2009). Is the Trade-Off Hypothesis Worth Trading For? Mind and Language 24: 164–80. Phillips, W., Baron-Cohen, S., and Rutter, M. (1998). Understanding Intention in Normal Development and in Autism. British Journal of Developmental Psychology 16: 337–48. Rose, D., Livengood, J., Sytsma, J., and Machery, E. (2012). Deep Troubles for the Deep Self. Philosophical Psychology 25: 629–46. Russell, J., and Hill, E.L. (2001). Action-Monitoring and Intention Reporting in Children with Autism. Journal of Child Psychology and Psychiatry 42: 317–28. Russell, J., and Jarrold, C. (1998). Error-Correction Problems in Autism: Evidence for a Monitoring Impairment? Journal of Autism and Developmental Disorders 28: 45–61. Russell, J., and Jarrold, C. (1999). Memory for Actions in Children with Autism: Self versus Other. Cognitive Neuropsychiatry 4: 303–31. Scaife, R., and Webber, J. (2013). Intentional Side-Effects of Action. Journal of Moral Philosophy 10: 179–203. Smith, I.M., and Bryson, S.E. (1994). Imitation and Action in Autism: A Critical Review. Psychological Bulletin 116: 259–73. Stoit, A.M.B., van Schie, H.T., Ruud, R., Meulenbroek, G.J., Bekkering, H., and Buitelaar, J.K. (2011). Internal Model Deficits Impair Joint Action in Children



E . MACHERY , T . ZALLA

and Adolescents with Autism Spectrum Disorders. Research in Autism Spectrum Disorders 5: 1526–37. Tomasello, M., Carpenter, M., Call, J., Behne, T., and Moll, H. (2005). Understanding and Sharing Intentions: The Origins of Cultural Cognition. Behavioral and Brain Sciences 28: 675–735. Wright, J., and Bengson, J. (2009). Asymmetries in Folk Judgments of Responsibility and Intentional Action. Mind and Language 24: 24–50. Zalla, T., and Leboyer, M. (2011). Judgment of Intentionality and Moral Evaluation in Individuals with High Functioning Autism. Review of Philosophy and Psychology 2: 681–98. Zalla, T., Labruyere N., and Georgieff, N. (2006). Goal-Directed Action Representation in Autism. Journal of Autism and Developmental Disorders 36: 527–40. Zalla, T., Sav, A.-M., Stopin, A., Ahade, S., and Leboyer, M. (2009). Faux Pas Detection and Intentional Action in Asperger Syndrome: A Replication on a French Sample. Journal of Autism and Developmental Disorders 39: 373–82. Zalla, T., Labruyere, N., Clément, A., and Georgieff, N. (2010). Predicting Ensuing Actions in Children and Adolescents with Autism Spectrum Disorders. Experimental Brain Research 201: 809–19. Zalla, T., Barlassina, L., Buon, M., and Leboyer, M. (2011). Moral Judgment in Adults with Autism Spectrum Disorders. Cognition 121: 115–26. Zelazo, P.D., Jacques, S., Burack, J., and Frye, D. (2002). The Relation between Theory of Mind and Rule Use: Evidence from Persons with Autism-Spectrum Disorders. Infant and Child Development 11: 171–95.

7 A Scientific Case for Conceptual Dualism The Problem of Consciousness and the Opposing Domains Hypothesis Anthony I. Jack, Case Western Reserve University

Introduction Like a handful of others (Bloom, 2005; Dawkins, 2006; Dennett, 2006; Harris, 2004), I believe that our intuitive dualism causes a lot of problems . . . [T]he mission of social neuroscience, as the offspring of social psychology and neuroscience, is to understand all of human subjective experience in physical terms. The rise of social neuroscience is the demise of the soul. (Greene, 2011)

Greene (2011) argues that one of the most exciting and socially significant products of neuroscience will be a philosophical conclusion. A scientific reductionist approach will be seen to triumph, banishing the misbegotten concept of the soul. This view is certainly tempting to those of us who are enamored with the power and promise of science. But is it right? I do not think so. Philosophy has long suggested difficulties for this brand of scientific triumphalism (Jack and Robbins, 2004). But This work was supported in part by NSF grant 0841313 and by a grant from the University Hospitals Case Medical Center Spitz Brain Health fund, both awarded to the author. The author would like to thank Jared Friedman for help with analysis, and Philip Robbins, Shaun Nichols, Stuart Youngner and two anonymous reviewers for comments on an earlier draft.



A . I . JACK

the account I offer here does not rest on philosophical argument. I am going to show you that the science itself suggests that reductionism will fail in its attempts to elucidate all aspects of the mental. In forwarding this account, I do not disagree with Greene’s claim that intuitive dualism creates problems. I just do not think these problems can be avoided. I believe intuitive dualism tells us something quite fundamental about the nature of human understanding—a feature of our psychology which has important implications for how we should best approach the science of the mind. One reason for supposing this is the long history of influential thinkers who have arrived at the conclusion that scientific thinking fails, in some way or another, to capture important aspects of the mental. Moreover, we must confess that the perception, and that which depends on it, is inexplicable in terms of mechanical reasons, that is, through shapes and motions. If we imagine that there is a machine whose structure makes it think, sense, and have perceptions, we could conceive it enlarged, keeping the same proportions, so that we could enter into it, as one enters into a mill. Assuming that, when inspecting its interior, we will only find parts that push one another, and we will never find anything to explain a perception. (Leibniz, 1714)

As this quote from Leibniz illustrates, the sense of a disconnect between experiential aspects of the mind and mechanical explanations has been present ever since science, as we currently understand it, took shape in the scientific revolution. In the meantime, scientific explanation has advanced considerably—we now appeal to electrochemical mechanics rather than classical mechanics to explain neural function. Philosophical conceptions of the mind–body problem have also become considerably more nuanced, yielding a variety of modern forms most frequently referred to as the problem of consciousness. The core of this problem is widely acknowledged to be captured by the existence of a (real or apparent) “explanatory gap”: a disconnect between our understanding of neural mechanisms and our understanding of experiential mental states (Levine, 2001). The explanatory gap has been made vivid by a variety of thought experiments (e.g. Nagel, 1974; Jackson, 1986; Chalmers, 1995). However, I will not dwell on the details of these arguments, which are in many cases highly nuanced and technical in nature. I think it is tremendously important that we take the problem of consciousness seriously— that we consider the problem in a way that Greene and many other

A SCIENTIFIC CASE FOR CONCEPTUAL DUALISM



scientists eschew. But the way I want to take it seriously here is not the way that may be most familiar—the philosopher’s approach of carefully unpacking the arguments and their consequences. For many scientists, the idea of focusing on an apparently intractable philosophical problem looks like a waste of time—science is about empirical investigations that deliver the answer to a question. This sense of impatience has encouraged many to recast the philosophical problem of consciousness into a more empirically tractable form. Hence, we may aim to use neuroscience to illuminate the biological basis of consciousness (as suggested by a Science magazine editorial, Miller, 2005), or “understand all of human subjective experience in physical terms” (Greene, 2011). The danger here is that in the rush to make the problem tractable we may lose sight of the real problem altogether. I see the scientific significance of philosophy in a different way. On this view, the most scientifically valuable philosophical problems are exactly the ones which appear most intractable. Such problems likely reflect a fundamental tension between two ways of understanding. Cognitive science studies how we understand the world. So, if we have located a genuinely insoluble philosophical problem, we can expect to see the tension reflected in our cognitive and neural structure. The problem of consciousness may be such a genuinely insoluble problem. So here I will resist trying to make the problem tractable, and instead entertain the hypothesis that it is not. Instead of trying to find the biological basis of consciousness, I aim to identify the biological basis for the problem of consciousness. The view that emerges will, I suggest, transform our understanding of the problem and how to deal with it. It will cast the problem of consciousness in a broader context. According to this view, our neural structure constrains our thinking, giving rise to a fundamental division in human understanding. I will suggest that this divide is relevant to a variety of philosophical issues and also to methodological issues that have long plagued the history of psychology. The brain is, of course, a mere organ, a product of evolution. Your arm can move in many ways that allow it to do many different sorts of things. Evolution has created a remarkably flexible and effective structure. But that structure still has constraints. If you have an itchy back, your arm is poorly designed to help with that. You do much better if you give up on contorting your arm, go make nice with someone, and ask them to scratch your back



A . I . JACK

instead. Similarly, I believe that when it comes to understanding the mind, the reductionist approach needs help. The reductionist approach is tremendously flexible and effective. It has solved many problems and it will solve many more. Yet, according to this view, it just is not suited to scratching the itch of consciousness—mechanistic explanations will never enable us to fully understand human experience.

The Divided Mind Systems for social cognition Work in social cognitive neuroscience suggests three major categories of regions involved in social information processing. First, there are systems which respond to social stimuli, regardless of the task (Wiggett et al., 2009). These lie close to occipital and temporal areas involved in basic visual and auditory processing and form parts of the ventral visual stream (Goodale and Milner, 1992). They include some well-delineated functionally specific areas such as the fusiform face area on the ventral surface (Kanwisher et al., 1997) and the extra-striate body area on the lateral surface (Downing, 2005), as well as a complex of regions centered on the posterior portion of superior temporal sulcus. The latter regions are not, as yet, so tightly defined, but they are reliably associated with gaze processing, action observation, and decoding of emotional expressions (e.g. Pelphrey, 2005). Second, there is a system involved in understanding the transitive actions of others (e.g. manual manipulations of objects) which includes inferior parietal sulcus (including the inferior and anterior portion of intra-parietal sulcus) and frontal premotor cortices (including inferior precentral sulcus). This system is known as the mirror neuron system, since it is presumed to be homologous to regions which contain mirror neurons in the monkey. This system includes parts of the dorsal visual stream (Goodale and Milner, 1992) and areas involved in motor planning and visuo-motor transformation (Corbetta et al., 1998). Resting connectivity is increasingly used to identify broad functional networks in the brain. The activity of regions in such a network is highly correlated when participants are not engaged in any task, suggesting that they form part of a functionally coherent network (e.g. Honey et al., 2009). The

A SCIENTIFIC CASE FOR CONCEPTUAL DUALISM



mirror neuron system substantially overlaps a network defined by resting connectivity known as the dorsal attention network (Fox et al., 2006). Third, there is a network of regions known as the mentalizing system which is clearly distinct from the mirror neuron system (Van Overwalle and Baetens, 2009). The mirror neuron system is most engaged by tasks that involve either executing or watching transitive actions, in many cases in the absence of any larger social context and/or without being able to see anything more than the hand/arm of the individual performing the action. In contrast, mentalizing tasks typically involve a richer social context or narrative and are frequently (but not always) textual in nature. Their defining feature is that they encourage the participant to think about the internal mental states of others (beliefs, desires, emotions). The mentalizing system includes two pronounced midline regions (Figure 7.1), dorsal MPFC and medial parietal/posterior cingulate (MP/PC), as well as a right lateralized region near the junction of the temporal and parietal cortices (rTPJ). The latter region lies adjacent to the complex of regions in the superior temporal cortex described as part of the first system. In contrast to this first system, engagement of the mentalizing system is associated with the attribution of mental states regardless of whether the target/stimulus is animate (e.g. humans, animals, faces) or inanimate (e.g. robots) (Waytz et al., 2010). Hence, the first system described is more bottom-up—its activity is relatively = brain areas more engaged by social than mechanical reasoning

= brain areas more engaged by mechanical than social reasoning = brain areas activated by introspecting one’s own emotional states (Ochsner et al., 2004) = grey matter density in this area predicts individual differences in body and emotional selfawareness (Borsci et al., 2009; Gundel et al., 2004) = brain areas involved in representing value and reversal learning

Figure 7.1 Illustration of the relationship between social cognition and other cognitive processes in the medial prefrontal cortex. The figure depicts the medial surface of the left hemisphere. The shading shows the contrast between social and mechanical reasoning tasks, taken from Jack et al. (2012). Color figure available online: .



A . I . JACK

insensitive to the task but highly sensitive to the type of stimulus— whereas the mentalizing system is more top-down—its activity is influenced heavily by the cognitive context and much less by surface characteristics of the stimuli. The functions of the mentalizing network appear to be quite general across different types of thinking about minds, as co-activation of these regions is commonly observed during a variety of tasks involving social cognition (Amodio and Frith, 2006; Mitchell, 2009; Van Overwalle, 2009). These mentalizing regions occupy the majority of a second network that can be defined by resting connectivity, which is known as the default mode network (DMN) (Schilbach et al., 2008; Mars et al., 2012). Although this network is often engaged as a unit, there is evidence for some functional specialization of its regions. The midline regions (Figure 7.1) are of particular significance for understanding the problem of consciousness, since standard formulations of the problem focus predominantly on consideration of our own internal phenomenally conscious states rather than our understanding of the mental states of others. Many studies and meta-analyses have now solidly established that these midline regions are recruited both when we introspect our own current mental states (e.g. emotions) and when we attribute mental states to others (Ochsner et al., 2004; Amodio and Frith, 2006; Saxe et al., 2006; Denny et al., 2012; Schilbach et al., 2012). In contrast, the evidence does not support the view that either of the other two systems described above are involved in thinking about experience.

Systems for physical reasoning How do these systems involved in social cognition relate to those involved in thinking about inanimate objects and physical properties/ processes? There is good evidence for a broad cognitive division. In some cases, this has been established by experiments that use closely matched stimuli designed to examine exactly this distinction (Mitchell, 2002; Martin and Weisberg, 2003; Blos et al., 2012). In other cases, this has been established by meta-analysis (Van Overwalle, 2010). Just as with social reasoning, care must be taken to distinguish different types of physical reasoning and to consider the specificity/generality of function of the brain areas involved by reference to a larger literature. First, with regard to more sensory areas, it is well established that there is an animate/inanimate distinction on the ventral surface, such that

A SCIENTIFIC CASE FOR CONCEPTUAL DUALISM



lateral areas (adjacent to and including the fusiform face area) are more engaged by animate/biological stimuli (including animals as well as human faces and bodies), whereas medial areas (adjacent to and including the parahippocampal place area) respond preferentially to inanimate objects. Competing theories of the functional specialization of the ventral surface also exist (e.g. Tarr and Gauthier, 2000). These predominantly focus on the role of these regions in specific perceptual processes, and hence undermine the notion of a processing distinction that relates to physical versus social reasoning in these areas. Nonetheless, the animate vs. inanimate view remains a dominant hypothesis (Grill-Spector et al., 2004). In particular, more recent findings suggest the animate/inanimate distinction on the ventral surface is innate and is not due to visual processing demands, because it can be observed using auditory stimuli in individuals born blind (Mahon et al., 2009). Second, there is evidence for a domain-general non-social reasoning system which comprises regions in the lateral parietal and frontal cortex. This was evident even in early meta-analyses (Duncan and Owen, 2000). Later reviews extend the picture, demonstrating considerable overlap between regions which are recruited by a variety of non-social tasks, including visual attention, working memory, language, logical reasoning, mathematical reasoning, general problem solving, and causal/mechanical reasoning tasks (Van Overwalle, 2010). Since this network is recruited by such a broad variety of tasks, and because so many of the tasks historically used by neuroscientists were non-social in nature, it came to be labeled the “task positive network” or TPN (Fox et al., 2005). The TPN overlaps two networks which can be identified by resting connectivity: the dorsal attention system and an adjacent network in lateral parietal and frontal cortices called the fronto-parietal control network (Vincent et al., 2008). The task positive network is clearly distinct from the mentalizing system of the default network both in terms of anatomical location and in terms of the tasks which preferentially recruit them (Van Overwalle, 2010). The task positive network substantially overlaps the mirror neuron system (Van Overwalle and Baetens, 2009).

Philosophical brain mapping If we wish to map the problem of consciousness onto the brain, then it appears that current work in neuroscience presents us with two



A . I . JACK

promising candidates. However, neither mapping is without complication. Let us consider them in turn. First, we might seek to map a perceived incongruity in thinking about minded (animate) versus purely physical (inanimate) objects onto the neural division seen in the occipital/temporal cortex. It has been suggested that this division in processing accounts for the explanatory gap and fuels belief in dualism (Fiala et al., 2012). This account helps to make sense of various low-level processing effects which clearly influence our attributions of mindedness, especially in infancy. For instance, increased attributions of mindedness plausibly explain why infants show a greater interest in objects when they have face-like structures or appear self-animated (Johnson et al., 1998). This account also fits well with the observation that infants appear to have a profoundly dualistic view of the world, such that they are not surprised when humans break some laws of physics, whereas they are when inanimate objects do (Bloom, 2004). On the other hand, it is not clear how well this account fares when it comes to explaining the more nuanced dualistic beliefs that are evidenced by adults. Adult reaction times, when making judgments denying experience, are still influenced by cues related to agency (Arico et al., 2011). However, in the actual choices they make, adults show less sensitivity to these surface features than to cognitive context (e.g. cultural beliefs about whether that kind of object has a mind). Further, philosophical arguments concerning the explanatory gap rely predominantly on considerations of one’s own experience.1 However, as previously noted, the sensory-driven regions for agency detection in the occipital and temporal lobes are not implicated in introspection. Fiala et al.’s account seems to represent part of the story, but falls short of offering a full account of the explanatory gap. This neural divide likely plays a role in the tendency to intuitive dualism present in infancy. It is plausible that it also contributes to the intuitions that reinforce intuitive dualism in adulthood. It looks like we can get further with the second mapping (i.e. with the mentalizing and task positive networks) because in this case the social 1 This has not always been the case. It appears the tension with experience is currently perceived as the most compelling; however, Brentano’s problem concerns the naturalization of the intentional. We touch on this briefly again later in the chapter.

A SCIENTIFIC CASE FOR CONCEPTUAL DUALISM



regions are clearly involved in thinking about experiential mental states, both our own and other people’s. Nonetheless, there are a number of potential concerns with this mapping. Let us begin by considering some which I think we can dispose of with relative ease. First, there may be a concern that the TPN is engaged in a variety of forms of reasoning. Hence, like Van Overwalle (2010), we might label it as a system for general reasoning rather than a system for physical reasoning. Undoubtedly the TPN is involved in forms of reasoning other than physical reasoning; however, this does not present a critical objection to our mapping. The critical point is just that reasoning about physical properties and mechanical processes is one of the functions of this network. At the same time, I believe we can be more specific than “general reasoning” when characterizing this network. Anatomical location, overlap in task-related activation, and resting connectivity all suggest the TPN is built upon and reuses primitive systems for perceptual processing and motor control, i.e. systems which originally evolved to perform spatial visual functions such as moving the eyes, allocating attention, visual guidance of action, and motor planning. The frontal regions of the TPN are also heavily engaged by working memory tasks, and their activity increases with the cognitive load of the task. I believe a good characterization of this network is that it is involved in analytic– empirical–critical thinking. This contrasts with the mentalizing network of the DMN which, while most reliably recruited by social tasks, is also implicated in some more synthetic forms of non-social reasoning such as insight problem solving (Subramaniam et al., 2009) and detecting broader patterns (Kroger et al., 2002). Of course, I am pointing here to very broad characterizations of function that describe wide-scale cortical networks. Future research will guide much finer subdivisions of the networks that can in turn guide a more nuanced picture. My claim is merely that this is some coherence to these broad characterizations. Second, a surprising feature of the neural divide we see between the mentalizing network and the TPN is that at least one type of social reasoning appears to fall within the domain of the TPN. There is evidence that mirror neurons specifically code the intentions of actions (Iacoboni et al., 2005). Hence it appears that certain aspects of reasoning about intentional actions may occur in the TPN rather than the DMN. But if they do, then it appears the “intention” that the mirror neuron represents is, in effect, a description of the mechanical manipulation which is to be



A . I . JACK

performed on an object. This fits well with the characterization of the TPN presented above, and is consistent with a broad division between neural mechanisms for thinking about minds vs. mechanical processes. Third, there may be a concern about where thoughts about the body should be situated relative to this neural divide. On the one hand, Descartes’s characterization of the problem as the mind–body problem would seem to suggest that thoughts about our body should be situated with physical reasoning. On the other hand, a great deal of recent research has focused on the notion of embodiment, and it is highly plausible that our interoceptive understanding of our own bodies plays an important role in social cognition. It is important to remember that Descartes was very interested in biology, a discipline which he argued should view the body first and foremost as a mechanism. Hence Descartes’s formulation of the philosophical problem involved a quite specific conception of the body. This is distinct from our first-person understanding of our own bodies. Anatomical location, overlap in task-related activation, and resting connectivity all suggest the DMN is built upon and reuses primitive systems for visceral/emotional self-awareness and self-regulation. The clearest story can be told about the MPFC (Figure 7.1). Anatomical studies in the monkey show that this region lies adjacent to areas which receive visceral inputs from the body (Ongur and Price, 2000). Correspondingly, the density of grey matter in areas of anterior cingulate cortex immediately adjacent to mentalizing regions predicts bodily and emotional self-awareness, depicted in Figure 7.1 (Gundel, 2004; Borsci et al., 2009).

Preparing for the leap Several researchers have drawn a close analogy between folk psychology and folk physics because they both involve high-level processes such as abstraction, inference, model building, prediction, and the postulation of unobservable processes or states (Lewis, 1972; Gopnik, 1996; Saxe, 2005). These types of high-level reasoning capacities are far more developed in humans than in other animals. Correspondingly, there is good evidence that both the TPN and the mentalizing system are highly evolved (Semendeferi et al., 2001; Schoenemann, 2006; Van Essen and Dierker, 2007; Rakic, 2009). Although folk psychology and folk physics may be analogous in this regard, it is evident that people use very different neural systems to engage in these high-level processes, depending on whether

A SCIENTIFIC CASE FOR CONCEPTUAL DUALISM



they are building a model of a mind or a mechanism. Hence it would appear that the critical question for the problem of consciousness is: how should we understand the relationship between folk and scientific understanding? For mathematics and the hard sciences, the evidence suggests that scientific understanding represents a refinement, or cultural recycling (Dehaene and Cohen, 2007), of our folk capacities. But how does it work for the sciences of the mind? Is scientific psychology a refinement of folk psychology? Some schools of psychology, such as the introspectionists and some therapeutic approaches, certainly appear to have this flavor. Yet this hardly seems a plausible model for neuroscience, which shares with biology and the other hard sciences a paradigmatic focus on understanding mechanism (Craver, 2007). Neuroscience looks more like a refinement of folk physics, not folk psychology. Might this be the source of the philosophical problem? Perhaps emphasizing mechanistic explanation promotes a switch in the faculties we use to guide our understanding of the mind, and so generates the sense that more mechanistic accounts are “leaving out experience”? However, there is a problem with this account. Why, when we switch to thinking about the mechanical properties of a brain, do we not also continue to think of it as an experiencing mind? Why should the use of one faculty preclude the use of the other? Indeed, the very fact that these faculties are supported by different neural networks might suggest they are well suited to operating in concert, without interference. An analogy may help here. We know that color and motion processing are associated with distinct brain areas, yet these representational properties are seamlessly integrated in normal perception. Should not it be the same for the experiential and mechanical properties of minds? A better perceptual analogy for the explanatory gap is the rivalry generated by ambiguous figures such as the duckrabbit illusion. We know both things are there, but focusing on one causes us to lose sight of the other. Hence, to have a satisfactory account of the explanatory gap, we need more than mere division. Perceptual rivalry is the result of a reciprocal inhibitory relationship between the two competing representations. Similarly, to explain the explanatory gap, we want evidence for reciprocal inhibition between experiential and mechanical representations.



A . I . JACK

Opposing brain networks The basic methodology for cognitive brain mapping is straightforward. It relies on a principle called cognitive subtraction. The typical approach is to give participants one cognitively complex task and a second control task. The control task may simply be rest—looking at a fixation point. Alternatively, it might be a task which is nearly as complex as the experimental task. Many experiments seek to isolate a single cognitive component and hence employ a control task which involves many of same processes as the experimental task. Measurements of brain activity associated with the control task are then subtracted away from those associated with the experimental task, and the standard inference is that any brain areas which are significantly activated are engaged in the additional cognitive processes associated with the experimental task. However, some time ago experimenters began to notice that some brain areas were breaking the rules of cognitive subtraction. In particular, the DMN (including the mentalizing system) tends to be deactivated during the same broad range of tasks that activate the TPN—i.e. these regions are more active at rest than in many experimental conditions (Shulman et al., 1997). Later, it was discovered that the TPN and DMN are in tension even when participants are at rest (i.e. when they are not given any task). In other words, there is a “spontaneous” or “natural” tendency for the DMN to be suppressed when the TPN is activated, and vice versa (Fox et al., 2005). The term “anticorrelated networks” was coined to describe this phenomenon. Since then, the DMN and anticorrelations have become a major topic of investigation in systems neuroscience, with hundreds of publications describing various aspects of their anatomy, function, relationship to other networks, and how they are affected by neuropsychiatric conditions (Buckner et al., 2008; Broyd et al., 2009; Andrews-Hanna, 2012). This phenomenon represents what may be the hottest topic, in terms of citations and studies, in the field of brain imaging today. So, what does this work suggest is the cognitive significance of the reciprocal inhibitory relationship between these networks? Two major hypotheses have dominated the literature. The first hypothesis gave the networks their names. The idea was that the TPN (“task positive”) is activated and the DMN (“default mode”) is deactivated whenever the participant is engaged in a goal-directed task. According to this hypothesis, the function of the DMN is spontaneous

A SCIENTIFIC CASE FOR CONCEPTUAL DUALISM



cognition, which is in tension with goal-directed cognition. There is little reason to dwell on this hypothesis, first because this characterization of the function of the DMN arguably never made much sense, but more critically because we know the DMN can be activated above resting levels by goal-directed tasks (Iacoboni, 2004). The second major hypothesis is currently dominant in the literature. It is the idea that the TPN is involved in externally directed cognition, whereas the DMN is involved in internally directed cognition, including introspection, mental time travel (including episodic memory recall), and conceptual association. This hypothesis views the tension between the networks as a type of attentional competition. On the face of it, this seems plausible. However, there is a problem with this account. It does not appear that this processing divide would yield the problem of consciousness as a primary concern. A division between internally and externally directed thinking might fit well with the belief “I am not a physical mechanism.” Yet it is much harder to see how we would ever extrapolate this problem to others. On a simple reading of this view we would never attribute the kinds of internal mental states we are acquainted with through introspection to others, because other people are, obviously, external to us. So the most straightforward reading of this account suggests that the most salient disjoin would not be between minds and machines, but rather between one’s own mind and everything external. This would incline us not to dualism but to a type of solipsism where we are convinced that we are the only conscious being. But this is clearly not the view to which most healthy people are inclined.2 In short, the problem with a straightforward reading of the internal vs. external attention account is that it fails to take account of our prolific tendency to attribute experiential mental states to others—a process that we know recruits the same brain regions as introspection. In conclusion, neither of the two major hypotheses that have been put forward to provide a cognitive characterization of the tension between the TPN and DMN appears wholly satisfactory, either as accounts of the imaging data or accounts of the problem of consciousness. If we buy the idea that the tension between these two opposing brain networks accounts for the explanatory gap, then we need to show that there is a 2

Not that I think the view is impossible. My bet is that a tendency to this view is associated with narcissism.



A . I . JACK

tension between thinking about physical mechanisms and thinking about internal mental states, using goal-directed tasks that do not confound the issue of internal vs. external attention.

The opposing domains hypothesis The opposing domains hypothesis holds that the neural antagonism between the TPN and DMN reflects a fundamental cognitive tension between thinking about internal mental states and thinking about physical mechanisms. We recently published evidence which provides strong support for this hypothesis (Jack et al., 2012). In order to distinguish this hypothesis from competing accounts, we compared tasks matched in terms of sensorimotor processing demands and external focus, but which differed in terms of whether participants are thinking about mental states or mechanical processes. Further, to establish generality, we used two different types of social and mechanical reasoning tasks—one presented as short textual narratives, the other as video with sound. We found that both the social tasks (soap-opera-like stories, and videos of two people conversing and misunderstanding each other) activated DMN regions and deactivated TPN regions. Conversely, both the mechanical reasoning tasks (physics problems taken from puzzle books, and clips from the video encyclopedia of physics) activated TPN regions and deactivated DMN regions. In other words, TPN and DMN regions were pushed up and down like a see-saw, depending on whether the task involved thinking about physical mechanisms or internal mental states. Our study revealed that a very large proportion (54%) of the cortical surface is sensitive to the domain of reasoning (i.e. social vs. mechanical), regardless of the modality of presentation (i.e. texts vs. videos). This supports the view that there is a major division in cortical organization between social and mechanical reasoning. However, only more circumscribed regions were pushed all the way above and below resting levels by the different tasks, providing clear evidence for mutual suppression. Why might this be? All tasks recruit a variety of cognitive processes— for instance, our externally focused tasks all involved visual attention, and they all required an occasional manual response (stimuli lasted 20 seconds, followed by a 7-second response window to a yes/no textual question). The sum of activity which is observed will depend on all the processes engaged. So it is hardly surprising that some DMN and TPN regions never fell below resting levels, and that some never rose above

A SCIENTIFIC CASE FOR CONCEPTUAL DUALISM



resting levels. These regions were likely recruited by processes which were common to all the tasks (e.g. external attention), or processes which were not engaged by any of them (e.g. internal attention, strong emotions). The critical question is: what is the functional role of the regions which were pushed up and down? First, the DMN regions which were pushed up and down matched very closely to the classic mentalizing regions (MPFC, MP/PC, rTPJ) which numerous studies have found to be associated with thinking about the internal mental states of ourselves and others. Second, the TPN regions which were pushed up and down matched regions involved in analytic reasoning and working memory, as well as the mirror neuron system. They did not match so well with areas specifically involved in external attention. A third important question is how these DMN and TPN regions line up with regions which are found to be anticorrelated at rest. This question speaks to a central and theoretically significant aspect of the account I offer, namely, the idea that our neural structure constrains our cognition and gives rise to the perceived problem of consciousness. Suppose participants were given two visual attention tasks in alternation: the first task required them to attend to the right visual field while ignoring stimuli in the left visual field; the second required them to attend to the left while ignoring the right. It is plausible that, given the right conditions, this pair of tasks could generate a pattern of reciprocal inhibition between regions similar to that which we observed. However, in this case the positive and negative constraints on cognition are built into the tasks themselves, and do not correspond to cognitive processes which are fundamentally opposed; e.g. it is also possible to spread attention between both visual fields. While such built-in constraints were not evident in the tasks we used, it could be that our culture and education causes us to engage in social and mechanical reasoning in such a way that we inhibit the other type of reasoning. This would be consistent with the suggestion that the problem of consciousness represents a purely cultural construction (Wilkes, 1988). However, if the activations and deactivations match regions which are fundamentally opposed, i.e. which are anticorrelated during spontaneous cognition, then this would suggest that the tension between social and mechanical reasoning arises as a product of our neurobiology. To help establish the cross-cultural validity of our findings, we tested this hypothesis using a large sample in which half the participants were from the USA, and the



A . I . JACK

other half from China. As predicted, there was a very close match between the brain regions which see-sawed during our tasks, and areas of maximal anticorrelation. These findings strongly support our hypothesis concerning the nature of the cognitive tension that is reflected by the neural antagonism between these networks. It is highly implausible that our findings can be better explained by the internal vs. external attention hypothesis. Our social tasks were no less demanding of external attention than our mechanical reasoning tasks, and all the tasks were much more demanding of external attention than the resting baseline condition to which they were compared. At rest, we may presume that the participant predominantly attends to internal information (e.g. memories, imaginings, thoughts, internal sensations, conceptual associations) because their external environment is barren and unchanging. At least, it is just this assumption which has fueled the internal vs. external hypothesis. Yet if that account were true, our externally oriented social tasks should not have activated maximally anticorrelated DMN regions any more than being at rest, and TPN regions should have been more active than they are at rest. Because our tasks were externally focused, philosophers may be concerned that our findings do not speak directly to standard formulations of the problem of consciousness, where the primary focus is on our own experiential mental states. Fortunately, an earlier study by Goldberg et al. (2006) speaks to this lacuna. They demonstrate deactivation during sensorimotor processing of a DMN region which was activated when participants introspected their perceptual experience while being presented with the same stimuli. The region involved, on the superior frontal gyrus, is anatomically very close to the maximally anticorrelated MPFC region in the DMN. For methodological reasons, this study fails to speak decisively about which broad hypothesis best accounts for the tension between the DMN and TPN. However, it does provide empirical support for the claim that the tension extends to the case where participants are focused on their own perceptual experiences. These findings provide strong positive support for my hypothesis concerning the neural origins of the perceived explanatory gap. They do not provide evidence that the mind is non-physical. Thus, they do not challenge physicalism (Jack and Shallice, 2001). But they do suggest that reductive physicalism, the explanatory strategy, is misconceived: our

A SCIENTIFIC CASE FOR CONCEPTUAL DUALISM



neural structure seems to present a barrier to understanding experience in physical terms. According to this view the explanatory gap is genuine, but it is not a feature of the world; it lies in our heads. However, an important question remains. Why does there seem to be a particularly potent philosophical concern about scientific accounts of experiential mental states as opposed, for example, to accounts of intentional action? The mentalizing system is implicated in a broad array of social cognitive tasks, so should not we see a similar tension for all varieties of mental state attribution? Although Descartes’s formulation of the mind–body problem saw human reason as the central aspect that lay beyond mechanical explanation, the march of progress in both psychology and computer science appears to have softened that perception, yielding modern forms of the problem which follow Nagel’s view that “consciousness is what makes the mind–body problem really intractable” (Nagel, 1974). To properly address this issue, we should take a look at the theory which guided this investigation.

The broader picture The phenomenal stance Robbins and Jack (2006) take the view that research in social cognition has been heavily influenced by a model which views social cognition as, first and foremost, a tool for predicting and manipulating others. In our view, this approach fails to take proper account of some very important functions of social cognition, namely, its role in social bonding/affiliation, moral cognition, and pro-social behaviors such as child-rearing. To remedy this omission, we developed a model which built on Dennett’s work. Using his language and framing, we introduced a new construct: the phenomenal stance (Figure 7.2). The phenomenal stance is the stance we adopt when we reflect upon our own experiences and feelings, as well as the experiences and feelings of others. It is a stance that we tend to avoid if we feel no compassion for the other—in which case the tendency is to deny or dismiss their feelings. Entering into this stance towards another deepens our social connection with them, and our sense of moral commitment to them. When this stance successfully guides another’s interactions with us, it makes us feel understood. This notion of intimate interpersonal understanding is, we contend, quite distinct from other



A . I . JACK

Physical Analytic–empirical–critical reasoning Task positive (TPN)

Phenomenal Locus of experience

Moral patient Default (DMN)

Intentional Instrumental social reasoning e.g. prediction and manipulation Co-activation of TPN and DMN

Figure 7.2 Three cognitive stances, their relationships to each other, and the brain networks involved. Bidirectional arrows indicate mutual compatibility; barbell indicates mutual antagonism. Adopting the intentional stance corresponds to a focus on the behavior of an intentional agent. This stance bridges between the phenomenal and physical stances, and involves co-activation of default and task positive regions. Nonetheless, there remains a fundamental tension between representing experiential mental states and representing mechanical processes. The model depicted represents a synthesis of the phenomenal stance model (Robbins and Jack, 2006; Jack and Robbins, 2012) and the opposing domains hypothesis (Jack et al., 2012).

ways of understanding the mind. It is not the same as “I understand what you think and I can guess what you are going to do” (the intentional stance), nor “I understand the neurobiological processes occurring in your brain” (the physical stance). The following quote gives an intuitive sense of it: [W]hen a person realizes he has been deeply heard, his eyes moisten. I think in some real sense he is weeping for joy. It is as though he were saying, “Thank God, somebody heard me. Someone knows what it’s like to be me.” In such moments I have had the fantasy of a prisoner in a dungeon, tapping out day after day a Morse code message, “Does anybody hear me? Is anybody there?” And finally one day he hears some faint tappings which spell out “Yes.” By that one simple response he is released from his loneliness; he has become a human being again. (Rogers, 1980)

A SCIENTIFIC CASE FOR CONCEPTUAL DUALISM



Robbins and Jack (2006) put forward two key claims. The first was that thinking about experience (the phenomenal stance) is tightly linked to moral sentiments—in particular, feelings of moral concern. We suggest this linkage is not present for types of social cognition better characterized by Dennett’s intentional stance. Second, our central hypothesis was that the problem of consciousness is generated by a tension between the phenomenal stance and the physical stance—a tension which either is not present or is considerably less powerful between the physical and the intentional stances. Linking these claims together we predicted that psychopaths would not be able to perceive the problem of consciousness.

Opposing and blended cognitive modes We know beyond any reasonable doubt that there is a major tension in the brain. Aside from the evidence already presented (reviewed at much greater length in the citations that follow), a large number of studies also show that this tension is strongly associated with healthy mental function: the tendency for these networks to suppress each other is diminished in virtually every major mental disorder, including disorders which impact social function such as autism and schizophrenia (Buckner et al., 2008; Broyd et al., 2009; Andrews-Hanna, 2012). Nonetheless, even in the healthy brain this tension is far from absolute. Parts of these networks do co-activate both during spontaneous cognition and during certain tasks. How should we understand this phenomenon? I believe this occurs when the brain is supporting blended cognitive modes. The notion is that these modes are not identical with either the full-blown phenomenal stance or with full-blown varieties of analytical–empirical–critical reasoning, such as mechanistic reasoning. But these blended cognitive modes do borrow aspects from each of the two pure modes. One such mode is creative thinking, or insight problem solving, which involves a combination of logical thought and a more intuitive mode of thinking. We see co-activation of parts of both networks at the moment of insight (Subramaniam et al., 2009). More creative individuals also demonstrate less tension between the networks (Takeuchi et al., 2011)—the only clearly desirable individual difference which appears to be associated with a lessening of this tension. The problem with thinking creatively is that we sometimes come up with flatly foolish ideas. In other words, our critical faculties are partially



A . I . JACK

suspended during creative thinking. I hypothesize that this is an inevitable trade-off of breaching the neural divide to recruit the DMN alongside the TPN. According to my model, the intentional stance also represents such a blended cognitive mode. Both behavioral and neuroimaging evidence is consistent with this view (I will not re-review the evidence here for reasons of space, but see discussion near the end of Jack et al., 2012). According to this view, the intentional stance involves a (limited) appreciation of the internal mental states of others blended with analytic–empirical–critical thinking. This blend both enables us to make predictions about others’ actions and causes us to adopt a more emotionally detached view of them. This blending of the two cognitive modes is reflected in ordinary language: when we refer to someone as “calculating” or “manipulative,” we do not literally mean that they are doing sums or using their fine motor skills. We are, of course, referring to an emotionally distanced and somewhat antisocial mode of social cognition. We likely use these terms because this mode of social cognition involves brain areas associated with mathematical calculation and transitive action. Hence, for instance, when conditions involving deception are compared with non-deceptive conditions, the most consistent differences in brain activity are seen in the TPN (Christ et al., 2009). Individuals with a tendency to Machiavellian thinking also show greater activation of TPN regions during social cognition (Bagozzi et al., 2013).

Incommensurable understanding My laboratory is engaged in testing and extending this model using both neuroimaging and behavioral measures. A central theme that runs through this work is that our model specifically links the phenomenal stance, and the tension with the physical stance, to moral sentiments.3 Hence, for us: (1) Moral sentiments are specifically tied to the attribution of states and properties which escape mechanical explanation, in particular experiential states; (2) both moral sentiments and the

3

Note that the claim is specific to moral sentiments, such as compassion, altruism, moral approbation, and outrage, and thoughts that directly relate to these sentiments. The TPN also clearly plays a role in moral deliberation. I suspect that the TPN is key to moralizing, hence the dissonance this provokes in those who possess genuine moral sensibility.

A SCIENTIFIC CASE FOR CONCEPTUAL DUALISM



attribution of these seemingly non-physical states and properties are associated with activation of the DMN accompanied by deactivation of the TPN. To behaviorally assess the tendency for individuals to adopt the pure phenomenal stance over the blended intentional stance, we use standard well-validated measures of moral concern, the empathetic concern subscale of the Interpersonal Reactivity Index (IRI-EC, Davis, 1980), and the callous affect subscale of the Self-Report Psychopathy scale (SRP-CA, Paulhus et al., forthcoming). To assess the tendency for individuals to adopt pure analytic–empirical–critical reasoning we use two measures: the cognitive reflection test (CRT, Frederick, 2005) primarily measures critical reasoning, and the Intuitive Physics Test (IPT) (Baron-Cohen et al., 2001) measures basic mechanical reasoning. We have found a small (r ~ –0.2) but highly robust and consistent negative correlation between the IRI-EC and the IPT (Jack and Gabriel, forthcoming)4. In contrast, negative correlations were not observed with measures of the intentional stance (e.g. theory of mind accuracy). We do not know of any other theory which would predict this specific negative relationship.5 There is also clear evidence that supports our view that moral concern is more strongly linked to attributions of experience than to attributions of intentional states (Gray et al., 2007; Knobe and Prinz, 2008; Gray et al., 2011; Jack and Robbins, 2012). Work on the phenomenon of dehumanization, in which individuals are viewed as less than human and antisocial behavior towards them is sanctioned, similarly indicates a link between moral concern and the attribution of experiential states. Dehumanization has been shown to be linked to a belief that target individuals

4 There are also robust gender differences associated with each measure (females higher on IRI-EC., lower on IPT); however, it is notable that the negative correlation between these measures is still present even when gender and other demographic variables including education are partialled out. Consistent negative correlations are also observed between IRI-EC and CRT, but are smaller and not significant in all the individual samples. 5 We focus on psychopathy because it represents a different primary deficit (the phenomenal stance) from autism (the intentional stance). The theories of Simon Baron-Cohen and Francesca Happe, both of which are focused on autism, might be translated as hypothesizing a tension between the intentional and the physical stances. James Blair’s work helped to characterize the differences between psychopathy and autism. All of their work was influenced by the work of their shared mentor/collaborator, Uta Frith. The work of all these individuals was influential in the development of our theory.



A . I . JACK

are less capable of experiencing certain kinds of more sophisticated emotional states (Leyens et al., 2001; Castano and Giner-Sorolla, 2006; Čehajić et al., 2009). A number of accounts also highlight the important role that essentialist thinking plays in dehumanization. A human essence, or soul, tends to be attributed to in-group members, but denied to dehumanized out-groups (Smith, 2011). This work has overlap with the observation that undergraduates presented with authoritative scientific denials of the existence of the human soul and free will demonstrate more antisocial behavior (Vohs and Schooler, 2008, Baumeister et al., 2009). We (Jack et al., 2013) have conducted a neuroimaging investigation which distinguishes two forms of dehumanization: animal and machine (Haslam, 2006). This work supports our view that the tension between the DMN and TPN reflects a distinction between thinking about conscious agents for whom we feel moral concern and inanimate objects for which we do not. We found that machine dehumanizing (also known as “objectifying”) involves a lessening of activity in the DMN (corresponding to social indifference), whereas animal dehumanizing involves co-activation of the DMN and the TPN (the signature of the blended cognitive mode of the intentional stance). However, we did find that one region was consistently associated with seeing human: the MP/PC region, which can be seen to the left of Figure 7.1. The MP/PC region is a central node for the DMN, and demonstrates the strongest anti-correlations with TPN regions. The same region activates when we look at pictures of the faces of people we personally know and are close to (i.e. in-group members), but deactivates when we view pictures of famous or unfamiliar faces. I hypothesize that this region plays a key role in generating the belief that the perceived individual possesses a human essence or “soul.” Our tendency to essentialism raises few philosophical qualms when it is applied to the physical stance. Endorsement of essentialism about physical properties motivates physicalism. However, the philosophical problems multiply when essentialist thinking is applied to other modes, such as the intentional stance (Brentano’s problem) or the phenomenal stance (metaphysical dualism). Philosophers continue to debate whether the explanatory gap poses a real challenge to physicalism. I do not believe that it does, provided we suspend our tendency to essentialism for this

A SCIENTIFIC CASE FOR CONCEPTUAL DUALISM



stance. I endorse this approach to metaphysics6 and certain aspects of scientific understanding (more on this shortly). However, I have serious concerns about applying it as a general prescription. In everyday life, I suspect that suspending essentialist thinking represents an unnatural, ineffective, and undesirable way of thinking. We have developed a measure of belief in metaphysical dualism which comprises five items, including “Humans have a soul,” “The mind can be understood completely by thinking of it as like a very complicated computer,” and “Thoughts and feelings are nothing more than the activity of neurons” (last two reverse coded). In a series of five experiments (Jack, forthcoming), we found a highly replicable and robust negative correlation (r ~ –0.34) between belief in dualism and the primary psychopathic trait of callous affect.7 This negative correlation survived after partialling out demographic variables, cognitive measures (e.g. the IPT and CRT) and measures of religious belief. In contrast, correlations between dualism and measures of physical (e.g. IPT) and intentional (e.g. ToM) reasoning were weak or absent. Clearly these findings fit well with the hypothesis (Robbins and Jack, 2006) that psychopaths cannot see the problem of consciousness.8 Taking these findings together with other work on dehumanization and the antisocial effects of denying the soul and free will, they present a powerful picture. When we see persons—that is, when we see others as fellow humans—then our percept is of something essentially non-physical in nature. This feature of our psychology appears to be relevant to a number of other philosophical issues, including the tension between utilitarian principles and deontological concerns about harming persons (Jack et al., 2014), the question of whether God exists (Jack et al., forthcoming), and 6

I view metaphysics as making claims about the structure of the physical world. Hence, I regard metaphysical dualism as a category error. It is driven by a mode of understanding which is not suited to understanding the physical world. I suggest that we ought to recognize the need for distinct ontologies associated with the different stances, and that only one such ontology should be seen as relevant to metaphysics. Beyond these casual remarks, I leave the details to philosophers. 7 Similar but positive correlations exist between dualism and IRI empathetic concern. 8 The effect is more specific to this aspect of the problem of consciousness than we originally anticipated. Psychopaths do not seem to have any problem perceiving an explanatory, epistemological, or conceivability gap. In one experiment we presented scenarios based on Nagel’s bat (1974), Jackson’s Mary (1986), and Chalmers’s philosophical zombie (1995). Virtually everyone perceived these gaps, and there was no clear correlation with psychopathy.



A . I . JACK

the problem of free will.9 So, should we regard our tendency to intuitive dualism as a problem, as Greene (2011) suggests? I believe it does create a problem, because it requires our moral, scientific, and philosophical thinking to be more nuanced and complex. However, I believe we will face problems larger than a mere increase in intellectual demands if we ignore or dismiss this fundamental feature of our psychology. It is tempting to suppose that our different ways of understanding the world can be rationalized into a single coherent world view; hence it is a reasonable hypothesis that a more considered approach might allow for greater reconciliation. However, this hypothesis is not supported by the data. A striking finding that emerges from our work on dualism is that individuals who score higher on reflective reasoning (CRT) do not escape the effect that empathetic concern has on creating a divided metaphysical world view (Figure 7.3). If anything, the effect of empathy on the tendency to believe in dualism is stronger for individuals high in critical thinking than for those who tend to go with intuition.

Concluding remarks The model I present here is a type of dual-process theory. However, it characterizes a very different cognitive divide from classic dual-process theory, which is best known from the work of the Nobel prize-winning psychologist Daniel Kahneman. According to classic dual-process theory (Kahneman, 2003), numerous decision-making phenomena may be seen as reflecting a fight between evolutionarily primitive, unconscious, and automatic processes on the one hand, and on the other a conscious deliberative general-purpose type of reasoning which is more amenable to education. Unlike classic dual-process theory, my model is informed primarily by what cognitive neuroscience tells us about the types of reasoning supported by the networks we see to be in tension in the brain, as well as by some philosophical considerations. According to my view, we possess two mutually exclusive faculties, both of which are conscious, deliberative, and highly evolved, and each of which may be cultivated through distinct cultural learning traditions (Snow, 1959). Yet each is substantially incomplete: one is incapable of comprehending 9 An ongoing project in collaboration with Joshua Knobe and others suggests that antiphysicalist metaphysical beliefs about free will are associated with having stronger moral sentiments about the need for just punishment.

A SCIENTIFIC CASE FOR CONCEPTUAL DUALISM



10.0000

Mean dualism_mean

8.0000

6.0000

4.0000

2.0000

0.0000 0.0000

0.3333

0.6667

1.0000

Crt_correct Error bars: +/- 1 SE SRP_CA_qrts 1.00

2.00

3.00

4.00

Figure 7.3 Graph depicting tendency to believe in dualism as a function of psychopathic callous affect (SRP-CA, split into quartiles) and raw score on the cognitive reflection test (CRT). Data is pooled from mainly online studies, totaling 874 participants. The CRT measures critical thinking by requiring participants to avoid giving an intuitively appealing but incorrect answer. Note that the tendency for less psychopathic individuals to believe in dualism holds regardless of score on the CRT. Hence, the general tendency to inhibit intuitive but incorrect responses does not prevent empathetic concern from influencing one’s metaphysical world view.

human experience and essential aspects of morality, while the other is incapable of comprehending the mechanical and mathematical structure of the physical world. While we can blend these cognitive modes, our neural structure creates interference between them. As a result, blended



A . I . JACK

cognitive modes fail to capture insights that emerge only when each of the pure opposing cognitive modes operates in isolation. According to this view, there is no faculty which can be properly called “general reasoning” because we lack a single integrated capacity capable of generating the full range of human insight. It follows from this view that progress in psychology will not be best achieved by adopting a blended cognitive mode, i.e. the intentional stance, to the exclusion of other perspectives.10 Instead, it appears that a complete understanding requires something more like juggling: we must fully immerse ourselves in distinct perspectives and only then seek to build bridges between the incommensurable conceptual frameworks that emerge. To briefly flesh this out, my view is that our social cognitive machinery creates a virtual world of experiences and persons which we imbue with meaning (Jack, 2011). From the impersonal point of view of the physical stance, this virtual world only exists as a figment of our imagination, and its coherence between individuals is only partial. To us, on the other hand, this virtual world is often more “real,” in the sense of being genuinely important and meaningful, than anything else. The only way to explore this virtual world is to adopt the phenomenal stance—for example, using the first- and second-person approaches initially developed by the introspectionists (Jack and Roepstorff, 2002). We cannot properly explore this world using objective measures, because when we come to interpret the data using either the physical or the intentional stance the very act of scientific interpretation moves us away from providing a description of experience, and blinds us to its nuances.11 Hence, contra Dennett’s heterophenomenological approach, objective measures cannot directly speak to the world of experience and gainsay our introspective enquiries. Similarly, introspective enquiries fail to provide direct evidence concerning the mechanical structure of the

10 For a directly parallel argument about moral thought, concerning the inadequacy of a utilitarian perspective (corresponding to the intentional stance) by itself, see Jack et al. (2014). 11 To illustrate this, the history of failed equations between objective measures and subjective states is long and has generated much controversy. Examples can be found in my cited papers. I suggest the solution is to concede that the notion of an ‘objective measure of awareness’ is an oxymoron. We can make actual progress by carefully establishing convergence between subjective measures which speak to experience, and objective measures which speak to mechanism (Jack and Shallice, 2001).

A SCIENTIFIC CASE FOR CONCEPTUAL DUALISM



mind.12 Nonetheless, these two incommensurable perspectives can be partially reconciled by adopting a more removed perspective that bridges between the personal world of experience and the impersonal world of mechanism. No account of our neural workings will ever tell us what it is like to be in love or see red. But they can tell us what is happening in someone that makes them have the experience of being in love or seeing red. If we want to offer such enlightening scientific accounts, we must not only develop our accounts of mechanism but also more nuanced accounts of experience. For instance, “being in love” is surely too broad and crude a description of experience to find traction with a sophisticated scientific account of the processes involved in human romantic attachment. There is simply no substitute for introspective methods for improving our accounts of experiential phenomena, yet systematic explorations (e.g. Hurlburt et al., 1994) are extremely rare. Without making use of these methods, our understanding of experience will be too crude to be worthy of our scientific efforts, and the scientific accounts we generate will not be able to indirectly inform the alternative nonphysical reality that we can only directly perceive through the phenomenal stance: a world constituted by irreducible persons, experiences, and moral truths. Some psychologists have vehemently resisted the rise of cognitive neuroscience, arguing that the workings of the brain are irrelevant to understanding the mind. These authors have sought to privilege a conceptual framework which emerges from the intentional stance over one that derives from the physical stance. I have often felt embarrassed on behalf of these authors. They fail to realize that their writings are little more than manifestos which promulgate their prejudiced belief in a highly limited conception of the mind (Jack et al., 2006). In this case, embarrassment seems an appropriate emotion, since this prejudice has been in the minority and continues to erode. However, there is another prejudice which still remains in the majority. In scientific psychology, 12 The error here is analogous to imagining that the structure of the filing system apparent from your computer’s user interface is informative about how your files are physically encoded on the hard drive, and vice versa. Just as with experiential and physical perspectives on the mind, the two can be related, but it takes a lot of work to make the link. Is the filing structure you know from everyday experience “real”? In one sense, no; in another it is much more real and certainly more useful than knowledge of the physical encoding.



A . I . JACK

there remains a strong inclination to dismiss an experiential perspective on the mind acquired through introspective methods. This dismissal of the insights that psychology can glean from the phenomenal stance is a more troubling concern. Like other prejudices whose effect is to demean or ignore the humanity of others, the emotion it triggers in those who recognize it is not embarrassment but outrage (Jack, 2013). I have written about this issue, and how we might resolve it, in the past (Jack and Shallice, 2001; Jack and Roepstorff, 2002, 2003; Jack, 2011). However, I have come to understand that the main barrier to progress has been that most psychologists, and some philosophers, simply have not seen the problem (e.g. Greene, 2011). I hope this essay helps to make it more apparent.13

References Amodio, D.M., and Frith, C.D. (2006). Meeting of Minds: The Medial Frontal Cortex and Social Cognition. National Review of Neuroscience 7: 268–77. Andrews-Hanna, J.R. (2012). The Brain’s Default Network and Its Adaptive Role in Internal Mentation. Neuroscientist 18(3): 251–70. Arico, A., Fiala, B., Goldberg, R.F., and Nichols, S. (2011). The Folk Psychology of Consciousness. Mind and Language 26: 327–52. Bagozzi, R.P., Verbeke, W.J.M.I., Dietvorst, R.C., Belschak, F.D., van den Berg, W.E., and Rietdijk, W.J.R. (2013). Theory of Mind and Empathic Explanations of Machiavellianism: A Neuroscience Perspective. Journal of Management 39(7): 1760–98. Baron-Cohen, S., Wheelwegith, S., Spong, A., Scahill, V., and Lawson, J. (2001). Are Intuitive Physics and Intuitive Psychology Independent? A Test with Children with Asperger Syndrome. Journal of Developmental and Learning Disorders 5: 47–78. Baumeister, R.F., Masicampo, E.J., and DeWall, C.N. (2009). Prosocial Benefits of Feeling Free: Disbelief in Free Will Increases Aggression and Reduces Helpfulness. Personality and Social Psychology Bulletin 35: 260–8.

13 I assume that the conflicts that arise between our incommensurable perspectives on the world will always tend to create confusion. Another way to put my take-home message is that neither physicalism nor metaphysical dualism do anything to resolve this. I endorse physicalism, but note that it ignores the problem. Metaphysical dualism acknowledges the problem but responds by setting our confusion in stone. I suggest we accept that we have these incommensurable perspectives and that they are compatible with physicalism, and adopt a methodological dualism in order to chip away at the links between them.

A SCIENTIFIC CASE FOR CONCEPTUAL DUALISM



Bloom, P. (2004). Descartes’ Baby: How the Science of Child Development Explains what Makes us Human. New York: Basic Books. Blos, J., Chatterjee, A., Kircher, T., and Straube, B. (2012). Neural Correlates of Causality Judgment in Physical and Social Context: The Reversed Effects of Space and Time. Neuroimage 63: 882–93. Borsci, G., Boccardi, M., Rossi, R., Rossi, G., Perez, J., Bonetti, M., and Frisoni, G.B. (2009). Alexithymia in Healthy Women: A Brain Morphology Study. Journal of Affective Disorders 114: 208–15. Broyd, S.J., Demanuele, C., Debener, S., Helps, S.K., James, C.J., and SonugaBarke, E.J. (2009). Default-Mode Brain Dysfunction in Mental Disorders: A Systematic Review. Neuroscience and Biobehavioral Reviews 33: 279–96. Buckner, R.L., Andrews-Hanna, J.R., and Schacter, D.L. (2008). The Brain’s Default Network: Anatomy, Function, and Relevance to Disease. Annals of the New York Academy of Sciences 1124: 1–38. Castano, E., and Giner-Sorolla, R. (2006). Not Quite Human: Infrahumanization in Response to Collective Responsibility for Intergroup Killing. Journal of Personality and Social Psychology 90: 804–18. Čehajić, S., Brown, R., and González, R. (2009). What do I Care? Perceived Ingroup Responsibility and Dehumanization as Predictors of Empathy Felt for the Victim Group. Group Processes and Intergroup Relations 12: 715–29. Chalmers, D.J. (1995). Facing Up to the Problem of Consciousness. Journal of Consciousness Studies 2(3): 200–19. Christ, S.E., Essen, D.C., Watson, J.M., Brubaker, L.E., and McDermott, K.B. (2009). The Contributions of Prefrontal Cortex and Executive Control to Deception: Evidence from Activation Likelihood Estimate Meta-Analysis. Cerebral Cortex 19: 1557–66. Corbetta, M., Akbudak, E., Conturo, T.E., Snyder, A.Z., Ollinger, J.M., Drury, H.A., Linenweber, M.R., Petersen, S.E., Raichle, M.E., Van Essen, D.C., and Shulman, G.L. (1998). A Common Network of Functional Areas for Attention and Eye Movements. Neuron 21: 761–73. Craver, C.F. (2007). Explaining the Brain: Mechanisms and the Mosaic Unity of Neuroscience. Oxford and New York: Oxford University Press and Clarendon Press. Davis, M.H. (1980). A Multidimensional Approach to Individual Differences in Empathy. Psychology 10: 85–114. Dehaene, S., and Cohen, L. (2007). Cultural Recycling of Cortical Maps. Neuron 56: 384–98. Denny, B.T, Kober, H., Wager, T.D., and Ochsner, K.N. (2012). A Meta-Analysis of Functional Neuroimaging Studies of Self- and Other Judgments Reveals a



A . I . JACK

Spatial Gradient for Mentalizing in Medial Prefrontal Cortex. Journal of Cognitive Neuroscience 24: 1742–52. Downing, P.E. (2005). Domain Specificity in Visual Cortex. Cerebral Cortex 16: 1453–61. Duncan, J., and Owen, A.M. (2000). Common Regions of the Human Frontal Lobe Recruited by Diverse Cognitive Demands. Trends in Neuroscience 23: 475–83. Fiala, B., Arico, A., and Nichols, S. (2012). On the Psychological Origins of Dualism: Dual-Process Cognition and the Explanatory Gap. In E. Slingerland and M. Collard (eds.), Creating Consilience: Integrating the Sciences and the Humanities. New York: Oxford University Press. Fox, M.D., Snyder, A.Z., Vincent, J.L., Corbetta, M., Van Essen, D.C., and Raichle, M.E. (2005). The Human Brain is Intrinsically Organized into Dynamic, Anticorrelated Functional Networks. Proceedings of the National Academy of Sciences USA 102: 9673–8. Fox, M.D., Corbetta, M., Snyder, A.Z., Vincent, J.L., and Raichle, M.E. (2006). Spontaneous Neuronal Activity Distinguishes Human Dorsal and Ventral Attention Systems. Proceedings of the National Academy of Sciences USA 103: 10046–51. Frederick, S. (2005). Cognitive Reflection and Decision Making. Journal of Economic Perspectives 19: 25–42. Goldberg, I., Harel, M., and Malach, R. (2006). When the Brain Loses Its Self: Prefrontal Inactivation during Sensorimotor Processing. Neuron 50: 329–39. Goodale, M.A., and Milner, A.D. (1992). Separate Visual Pathways for Perception and Action. Trends in Neuroscience 15: 20–5. Gopnik, A. (1996). The Scientist as Child. Philosophy of Science 63: 485–514. Gray, H.M., Gray, K., and Wegner, D.M. (2007). Dimensions of Mind Perception. Science 315: 619. Gray, K., Jenkins, A.C., Heberlein, A.S., and Wegner, D.M. (2011). Distortions of Mind Perception in Psychopathology. Proceedings of the National Academy of Sciences USA 108: 477–9. Greene, J.D. (2011). Social Neuroscience and the Soul’s Last Stand. In A. Todorov et al. (eds.), Social Neuroscience: Toward Understanding the Underpinnings of the Social Mind. New York: Oxford University Press. Grill-Spector, K., Knouf, N., and Kanwisher, N. (2004). The Fusiform Face Area Subserves Face Perception, Not Generic Within-category Identification. Nature Neuroscience 7: 555–62. Gundel, H. (2004). Alexithymia Correlates With the Size of the Right Anterior Cingulate. Psychosomatic Medicine 66: 132–40. Haslam, N. (2006). Dehumanization: An Integrative Review. Personality and Social Psychology Review 10: 252–64.

A SCIENTIFIC CASE FOR CONCEPTUAL DUALISM



Honey, C.J., Sporns, O., Cammoun, L., Gigandet, X., Thiran, J.P., Meuli, R., and Hagmann, P. (2009). Predicting Human Resting-State Functional Connectivity from Structural Connectivity. Proceedings of the National Academy of Sciences USA 106: 2035–40. Hurlburt, R.T, Happe, F., and Frith, U. (1994). Sampling the Form of Inner Experience in Three Adults with Asperger Syndrome. Psychological Medicine 24: 385–95. Iacoboni, M. (2004). Watching Social Interactions Produces Dorsomedial Prefrontal and Medial Parietal BOLD fMRI Signal Increases Compared to a Resting Baseline. NeuroImage 21: 1167–73. Iacoboni, M., Molnar-Szakacs, I., Gallese, V., Buccino, G., Mazziotta, J.C., and Rizzolatti, G. (2005). Grasping the Intentions of Others with One’s Own Mirror Neuron System. PLoS Biology 3: e79. Jack, A.I. (2011). Describing Inner Experience? Proponent Meets Skeptic. Philosophy of Psychology 24: 283–7. Jack, A.I. (2013). Introspection: The Tipping Point. Consciousness and Cognition 22(2): 670–1. Jack, A.I. (forthcoming). Consciousness and Callousness: Empathetic Concern Distinguishes Folk from Scientific Conceptions of the Mind. Jack, A.I., and Gabriel, J.J. (forthcoming). Individual and Gender Differences Demonstrate a Trade-Off between Empathetic Concern and Physical Reasoning Ability. Jack, A.I., and Robbins, P. (2004). The Illusory Triumph of Machine over Mind: Wegner’s Eliminativism and the Real Promise of Psychology. Behavioral and Brain Sciences 27: 665ff. Jack, A.I., and Robbins, P. (2012). The Phenomenal Stance Revisited. Review of Philosophy and Psychology 3(3): 383–403. Jack, A.I., and Roepstorff, A. (2002). Introspection and Cognitive Brain Mapping: From Stimulus-Response to Script-Report. Trends in Cognitive Science 6: 333–9. Jack, A.I., and Roepstorff, A. (2003). Why Trust the Subject? Journal of Consciousness Studies 10: v–xx. Jack, A.I., and Shallice, T. (2001). Introspective Physicalism as an Approach to the Science of Consciousness. Cognition 79: 161–96. Jack, A.I., Sylvester, C.M., and Corbetta, M. (2006). Losing our Brainless Minds: How Neuroimaging Informs Cognition. Cortex 42: 418–21; discussion 422–7. Jack, A.I., Dawson, A.J., Begany, K.L., Leckie, R.L., Barry, K.P., Ciccia, A.H., and Snyder, A.Z. (2012). FMRI Reveals Reciprocal Inhibition between Social and Physical Cognitive Domains. Neuroimage 66C: 385–401.



A . I . JACK

Jack, A.I., Dawson, A.J., and Norr, M. (2013). Seeing Human: Distinct and Overlapping Neural Signatures associated with Two Forms of Dehumanization. Neuroimage 79: 313–28. Jack, A.I., Robbins, P.A., Friedman, J.P., and Meyers, C.D. (2014). More than a Feeling: Counterintuitive Effects of Compassion on Moral Judgment. In J. Sytsma (ed.), Advances in Experimental Philosophy of Mind. London: Bloomsbury, pp. 125–79. Jack, A.I., Nolan, S.N., Boyatzis, R.E., and Friedman, J.P. (forthcoming). A Novel Dual-process Account of Religious Belief. Jackson, F. (1986). What Mary Didn’t Know + Knowledge Argument against Physicalism. Journal of Philosophy 83: 291–5. Johnson, S., Slaughter, V., and Carey, S. (1998). Whose Gaze will Infants Follow? The Elicitation of Gaze-Following in 12-month-olds. Developmental Science 1: 233–8. Kahneman, D. (2003). A Perspective on Judgment and Choice. American Psychologist 58: 697–720. Kanwisher, N., McDermott, J., and Chun, M.M. (1997). The Fusiform Face Area: A Module in Human Extrastriate Cortex Specialized for Face Perception. Journal of Neuroscience 17: 4302–11. Knobe, J., and Prinz, J.J. (2008). Intuitions About Consciousness: Experimental Studies. Phenomenology and the Cognitive Sciences 7: 67–83. Kroger, J.K., Sabb, F.W., Fales, C.L., Bookheimer, S.Y., Cohen, M.S., and Holyoak, K.J. (2002). Recruitment of Anterior Dorsolateral Prefrontal Cortex in Human Reasoning: a Parametric Study of Relational Complexity. Cerebral Cortex 12: 477–85. Leibniz, G. (1714). The Mondadology. Levine, J. (2001). Purple Haze: The Puzzle of Consciousness. Oxford: Oxford University Press. Lewis, D. (1972). Psychophysical and Theoretical Identifications. Australasian Journal of Philosophy 50: 249–58. Leyens, J.-P., Rodriguez-Perez, A., Rodriguez-Torres, R., Gaunt, R., Paladino, M.-P., Vaes, J., and Demoulin, S. (2001). Psychological Essentialism and the Differential Attribution of Uniquely Human Emotions to Ingroups and Outgroups. European Journal of Social Psychology 31: 395–411. Mahon, B.Z., Anzellotti, S., Schwarzbach, J., Zampini, M., and Caramazza, A. (2009). Category-Specific Organization in the Human Brain Does Not Require Visual Experience. Neuron 63: 397–405. Mars, R.B., Neubert, F.X., Noonan, M.P., Sallet, J., Toni, I., and Rushworth, M.F. (2012). On the Relationship between the “Default Mode Network” and the “Social Brain”. Frontiers in Human Neuroscience 6: 189.

A SCIENTIFIC CASE FOR CONCEPTUAL DUALISM



Martin, A., and Weisberg, J. (2003). Neural Foundations for Understanding Social and Mechanical Concepts. Cognitive Neuropsychology 20: 575–87. Miller, G. (2005). What is the Biological Basis of Consciousness? Science 309: 79. Mitchell, J.P. (2002). Distinct Neural Systems Subserve Person and Object Knowledge. Proceedings of the National Academy of Sciences 99: 15238–43. Mitchell, J.P. (2009). Social Psychology as a Natural Kind. Trends in Cognitive Science 13: 246–51. Nagel, T. (1974). What Is It Like to Be a Bat? Philosophy Review 83: 435–50. Ochsner, K.N., Knierim, K., Ludlow, D.H., Hanelin, J., Ramachandran, T., Glover, G., and Mackey, S.C. (2004). Reflecting upon Feelings: An fMRI Study of Neural Systems Supporting the Attribution of Emotion to Self and Other. Journal of Cognitive Neuroscience 16: 1746–72. Ongur, D., and Price, J.L. (2000). The Organization of Networks within the Orbital and Medial Prefrontal Cortex of Rats, Monkeys and Humans. Cerebral Cortex 10: 206–19. Paulhus, D.L., Neumann, C.S., Hare, R.D. (forthcoming). Manual for the SelfReport Psychopathy Scale. Toronto: Multi-Health Systems. Pelphrey, K.A. (2005). Neural Basis of Eye Gaze Processing Deficits in Autism. Brain 128: 1038–48. Rakic, P. (2009). Evolution of the Neocortex: A Perspective from Developmental Biology. National Review of Neuroscience 10: 724–35. Robbins, P., and Jack, A.I. (2006). The Phenomenal Stance. Philosophical Studies 127: 59–85. Rogers, C.R. (1980). A Way of Being. Boston: Houghton Mifflin. Saxe, R. (2005). Against Simulation: The Argument from Error. Trends in Cognitive Science 9: 174–9. Saxe, R., Moran, J.M., Scholz, J., and Gabrieli, J. (2006). Overlapping and Non-Overlapping Brain Regions for Theory of Mind and Self Reflection in Individual Subjects. Social Cognitive and Affective Neuroscience 1: 229–34. Schilbach, L., Eickhoff, S., Rotarskajagiela, A., Fink, G., and Vogeley, K. (2008). Minds at Rest? Social Cognition as the Default Mode of Cognizing and its Putative Relationship to the “Default System” of the Brain. Consciousness and Cognition 17: 457–67. Schilbach, L., Bzdok, D., Timmermans, B., Fox, P.T, Laird, A.R., Vogeley, K., and Eickhoff, S.B. (2012). Introspective Minds: Using ALE Meta-Analyses to Study Commonalities in the Neural Correlates of Emotional Processing, Social and and Unconstrained Cognition. PLoS One 7: e30920. Schoenemann, P.T. (2006). Evolution of the Size and Functional Areas of the Human Brain. Annual Review of Anthropology 35: 379–406.



A . I . JACK

Semendeferi, K., Armstrong, E., Schleicher, A., Zilles, K., and Van Hoesen, G.W. (2001). Prefrontal Cortex in Humans and Apes: A Comparative Study of Area 10. American Journal of Physical Anthropology 114: 224–41. Shulman, G.L., Fiez, J.A., Corbetta, M., Buckner, R.L., Miezin, F.M., Raichle, M.E., and Petersen, S.E. (1997). Common Blood Flow Changes across Visual Tasks: II. Decreases in Cerebral Cortex. Journal of Cognitive Neuroscience 9: 648–63. Smith, D.L. (2011). Less than Human: Why we Demean, Enslave, and Exterminate Others. New York: St. Martin’s Press. Snow, C.P. (1959). The Two Cultures and the Scientific Revolution. New York: Cambridge University Press. Subramaniam, K., Kounios, J., Parrish, T.B., and Jung-Beeman, M. (2009). A Brain Mechanism for Facilitation of Insight by Positive Affect. Journal of Cognitive Neuroscience 21: 415–32. Takeuchi, H., Taki, Y., Hashizume, H., Sassa, Y., Nagase, T., Nouchi, R., and Kawashima, R. (2011). Failing to Deactivate: The Association between Brain Activity during a Working Memory Task and Creativity. Neuroimage 55: 681–7. Tarr, M.J., and Gauthier, I. (2000). FFA: A Flexible Fusiform Area for Subordinate-level Visual Processing Automatized by Expertise. Nature Neuroscience 3: 764–9. Van Essen, D.C., and Dierker, D.L. (2007). Surface-Based and Probabilistic Atlases of Primate Cerebral Cortex. Neuron 56: 209–25. Van Overwalle, F. (2009). Social Cognition and the Brain: A Meta-Analysis. Human Brain Mapping 30: 829–58. Van Overwalle, F. (2010). A Dissociation between Social Mentalizing and General Reasoning. Neuroimage 54(2): 1589–99. Van Overwalle, F., and Baetens, K. (2009). Understanding Others’ Actions and Goals by Mirror and Mentalizing Systems: A Meta-Analysis. NeuroImage 48: 564–84. Vincent, J.L., Kahn, I., Snyder, A.Z., Raichle, M.E., and Buckner, R.L. (2008). Evidence for a Frontoparietal Control System Revealed by Intrinsic Functional Connectivity. Journal of Neurophysiology 100: 3328–42. Vohs, K.D., and Schooler, J.W. (2008). The Value of Believing in Free Will: Encouraging a Belief in Determinism Increases Cheating. Psychological Science 19: 49–54. Waytz, A., Morewedge, C.K., Epley, N., Monteleone, G., Gao, J.H., and Cacioppo, J.T. (2010). Making Sense by Making Sentient: Effectance Motivation Increases Anthropomorphism. Journal of Personality and Social Psychology 99: 410–35.

A SCIENTIFIC CASE FOR CONCEPTUAL DUALISM



Wiggett, A.J., Pritchard, I.C., and Downing, P.E. (2009). Animate and Inanimate Objects in Human Visual Cortex: Evidence for Task-Independent Category Effects. Neuropsychologia 47: 3111–17. Wilkes, K.V. (1988). Yishi, Duh, Um, and Consciousness. In A.J. Marcel and E. Bisiach (eds.), Consciousness in Contemporary Science. Oxford: Oxford University Press.

8 Carving Up the Social World with Generics Sarah-Jane Leslie, Princeton University

Introduction Generics are sentences such as “tigers are striped,” “lions have manes,” and “mosquitoes carry West Nile virus.” Generics express generalizations, but unlike quantified statements, generics do not carry information about how many members of the kind or category have the property. For example, if asked “how many ravens are black?” one could reply “all [or some, or most, etc] ravens are black,” but one cannot felicitously reply with the generic “ravens are black” (Carlson, 1977). Generics have long been the subject of considerable discussion in linguistics and philosophy, as they, unlike quantified statements, are extremely difficult to analyze truth conditionally (e.g., Carlson, 1977; Carlson and Pelletier, 1995; Cohen, 1996, 2004; Greenberg, 2003; Lawler, 1973; Leslie, 2007, 2008; Pelletier and Asher, 1997). As a simple illustration of part of the difficulty, consider the generics “lions have manes” and “lions are male”—the former would seem true, while the latter is false, even though of course it is only ever male lions that have manes. Similarly, we might note that “mosquitoes carry West Nile virus” is true, while “books are paperbacks” is false, even though over 80% of books are paperbacks, yet only 1% of mosquitoes carry West Nile virus.

Preparation of this chapter was supported by NSF grant BCS-1226942. Thanks to Andrei Cimpian, Mark Johnston, Marjorie Rhodes, and two anonymous reviewers for their invaluable help.

CARVING UP THE SOCIAL WORLD WITH GENERICS



This chapter is not, however, concerned with the truth-conditions of generics, but rather with the effects of generic language on social cognition, with a particular eye to its effects on young children. Generics are very frequent in parent–child conversations, and are an important means by which information is communicated throughout development (Gelman, 2003). Recent findings suggest that generics are also a significant source of social information for children—but this information is all too often not the sort one would wish to communicate to one’s children. In particular, generic language may be implicated in the transmission of beliefs that form the backbone of social prejudice, and may also lead children to adopt detrimental conceptions of abilities, which hamper both their motivation and their interest in a range of activities.

Generics as default generalizations Why are generics an important source of information for children? As noted briefly above, generics are truth-conditionally more complex than quantified statements; one might naturally think that generics would be too difficult for children to understand. Perhaps, like certain other complex and sophisticated constructions, generic acquisition would occur only quite late in development. As it turns out, however, generics are quite easy for young children to acquire and process. They are readily produced and understood by preschool-aged children, and the data collected to date suggest that these young children have a remarkably adult-like understanding of generics. For example, preschoolers who know that only “boy” lions have manes will accept “lions have manes” but reject “lions are boys”—despite implicitly understanding that there are at least as many “boy” lions as there are maned lions (Brandone et al., 2012; for a review, see Leslie, 2012). If anything, generics appear to be easier for young children to acquire and process than quantified statements. Several studies indicate that children have considerable difficulty processing category-wide quantified generalizations such as “all girls have curly hair” or “some flowers are blue.” Intriguingly, when preschoolers are confronted with such quantified statements, they do not simply provide random, incorrect answers; instead, it appears that they treat them as though they were generics. That is, preschool children not only consistently evaluate generics just as adults do, they also evaluate kind-wide quantified statements as generics (Hollander et al., 2002; Leslie and Gelman, 2012; Tardif et al., 2011; for a



S .- J . LESLIE

detailed review, see Leslie, 2012). In addition to English-speaking children, such findings have also been documented among Mandarin Chinese- and Quechua-speaking children (Mannheim et al., 2011; Tardif et al., 2011). Adults also display this tendency under certain circumstances—for example, accepting universally quantified statements despite knowing that there are exceptions (e.g. agreeing to “all ducks lay eggs”; Leslie et al., 2011; Meyer et al., 2011). Further, both preschoolers and adults show a robust tendency to recall quantified statements as generics (Leslie and Gelman, 2012). How might these results be explained? It is not clear on the face of it what one should make of the fact that these truth-conditionally puzzling statements are apparently so easily processed by young children— considerably more easily, it would seem, than their quantified counterparts. In Leslie (2007, 2008) I proposed these results could be explained (or rather at that time, I predicted these results, with the exception of the earlier data in Hollander et al., 2002, which was of course an inspiration for the hypothesis) if we suppose that generic sentences are language’s way of letting us give voice to cognitively fundamental generalizations. We observe that infants in their first year of life are able to form general judgments about categories, that is, to form expectations concerning the properties of as-of-yet unencountered instances of the category (see e.g. Baldwin et al., 1993). These general judgments formed by the preverbal infant—we might naturally term these generalizations cognitively fundamental generalizations—represent our most basic form of generalizing, one which we exploit from our earliest days. When we grow up to learn our native language, it is natural to suppose that this language will provide us with some means of giving voice to these generalizations. The hypothesis of Leslie (2007, 2008) is that generic sentences are precisely this means of giving voice to our cognitively fundamental generalizations. If the cognitive system has a basic, default way of forming general judgments then it may sometimes fall back on this means of generalizing when asked to process a more taxing and sophisticated generalization. On this hypothesis, quantified statements are precisely this, namely, more taxing and sophisticated generalizations. Thus, if generics do articulate cognitively fundamental generalizations, it is easy to understand why both adults and children show a tendency to “default” to the generic when asked to consider quantified statements. Most importantly for the

CARVING UP THE SOCIAL WORLD WITH GENERICS



purposes of this chapter, the hypothesis serves to explain why generics play such an important role in the transmission of beliefs from adults to children, and thus why hearing generic language may potentially have a profound effect on how children understand the social world around them.

The Cultural Transmission of Social Essentialism From the preschool years onwards, we tend to manifest a set of beliefs which psychologists term essentialist beliefs. (This essentialism is not the same as what philosophers tend to mean by “essentialism”; for a detailed comparison, see Leslie, 2013, where the psychological notion is dubbed “quintessentialism.”) That is, we believe, implicitly or explicitly, that each animate individual has an underlying nature or essence—an almost substance-like entity that pervades its insides, and which causally grounds its more stable and enduring properties. Further, we believe that certain kinds “carve essence at its joints”—that is, certain kinds (e.g., animal kinds) pick up on genuine differences and similarities in the essences of individuals. Thus we believe that the kind tiger picks out individuals with highly similar essences, and these shared essences explain why tigers share so many outwardly observable properties, such as having stripes and tails and being ferocious (see Gelman, 2003; Leslie, 2013, for more details). Such beliefs have been most extensively studied in the biological domain, yet they can also be found in the social realm. That is, some social categories are also essentialized, meaning that people believe that members of these categories share a fundamental nature that grounds a range of common properties (e.g., Gelman, 2003; Hirschfeld, 1996; Meyer et al., 2013; Prentice and Miller, 2007; Rhodes and Gelman, 2009; Rothbart and Taylor, 1992). For example, gender categories are essentialized from a young age, with people believing that there are deep, inherent, fundamental differences between men and women. Importantly, essentialized social categories are more likely to be the targets of sustained and virulent prejudiced attitudes (see e.g. Haslam



S .- J . LESLIE

et al., 2000, 2002),1 and Keller (2005) offers experimental evidence that suggests there may indeed be a causal link between essentialist beliefs and prejudice. Relatedly, essentialized groups tend to be more susceptible to stereotyping (e.g., Bastian and Haslam, 2006; Prentice and Miller, 2006, 2007; Williams and Eberhardt, 2008; Yzerbyt et al., 2001). (Interestingly, the very nature of generic generalizations may point to one explanatory factor here. Consider the generalizations mosquitoes carry West Nile virus, sharks attack swimmers, pitbulls maul children, ticks carry Lyme disease, and so on. These generics are accepted even though very few members of the kind have the property in question—because, I would argue, the property is striking, it makes its bearers dangerous, and hence is the sort of thing we would wish to avoid. If we essentialize a kind, and its members that manifest such a property, this may suffice for us to generalize the property to the kind. Consider, then, pernicious social generalizations such as Muslims are terrorists and blacks are rapists—do they not have the same character? For more discussion, see Leslie, forthcoming.)

Generics and social essentialism In light of the foregoing, it would seem important to understand the factors that lead us to essentialize certain social categories.2 We can note that only some social categories are seen through an essentialist lens; while race and gender categories are consistently seen this way, professional kinds and sports teams, for example, are not usually essentialized.

1

An apparent exception to this generalization concerns groups such as homosexuals, and certain categories of mental illness, where believing in a natural/biological basis for group membership tends to correlate negatively with prejudice. That is, people who view homosexuality as a chosen lifestyle tend to exhibit more prejudice than people who understand it to be something biologically based and beyond the individual’s control (Haslam and Levy, 2006). However, such examples are clearly complex, since anti-homosexual prejudice generally involves more than viewing it as simply a choice—it involves viewing it as a morally and/or socially objectionable choice. That is, if one assumes that something is morally and/or socially objectionable, then believing it to be grounded in biology works to absolve the individual of responsibility, thereby, presumably, somewhat reducing one’s prejudice or at least some of its expressions. The existence of these more complex patterns of prejudiced thinking, however, do not undermine the basic point here. 2 A more general question is why we essentialize any categories whatsoever, including animal categories. That is, why do we have a tendency from such a young age to view some categories in this way? For discussion of this question see Gelman (2003) and Cimpian and Salomon (forthcoming).

CARVING UP THE SOCIAL WORLD WITH GENERICS



Furthermore, there is considerable cultural and historical variation in which social kinds are essentialized. For example, the different castes in India have traditionally been seen as highly essentialized, especially by members of the upper castes (Mahalingam, 2007; Mahalingam and Rodriguez, 2006). The Ancient Greeks are often said to have believed that there were two fundamental kinds of human being: Greeks and Barbarians, each endowed with their own distinctive nature. We might suppose that class in English society has, at least until very recently, been essentialized, and certainly medieval European notions of the Great Chain of Being involved highly essentialist ways of thinking about the different strata of society. (Nobles supposedly manifested an inner superiority, the clergy had a distinctive ontological status, monarchs were infused with a divine right to rule, and rebellious peasants were often said to be going against nature itself.) Recent work has also found that children growing up in politically conservative communities in the United States have more essentialist beliefs about race than children growing up in more politically liberal communities (Rhodes and Gelman, 2009). Likewise, within Israel, essentialist beliefs about ethnicity are more common in religious communities than in secular communities (Diesendruck and Haber, 2009). It would thus seem that while animal kinds are consistently and uniformly essentialized (Gelman, 2003), social kinds are not. Instead, at some point in their development, children come to single out certain social groups as objects of essentialist beliefs. We might also note that these groups tend to be precisely the ones that are essentialized by adults in the child’s community, which suggests that there must be some mechanism by which social essentialist beliefs are transmitted across generations—some forms of cultural input that lead young children to view some social groupings as reflecting deep, meaningful, and natural distinctions between people. Marjorie Rhodes and I hypothesized that generic language may be one such form of cultural input; and hence that hearing generic language about a diverse, novel social group would lead children to essentialize that group; and further, that parents would produce more generic language when speaking to their children about a social group that they themselves essentialize. (To be clear, this is not to say that there could not be other means by which essentialist beliefs are handed down, only that generic language is one such means.) In the case of animal kinds, Gelman



S .- J . LESLIE

and colleagues (2010) found that hearing generic language about a novel animal kind led both preschool children and adults to essentialize the animal kind more than they otherwise would. However, this leaves open the question of whether the same effects would be found in the case of social groups. Children quickly form essentialist beliefs about animal kinds anyway, even in the absence of any particular linguistic input (see e.g. Gelman, 2003), and furthermore, animal kinds are not selectively essentialized (e.g., we do not essentialize tigers but not lions). Thus, hearing generic language might serve only to speed up the acquisition of essentialist beliefs that would have formed anyway, rather than working to instill selective essentialist beliefs in the first place. To test whether generic language might have direct selective effects in the case of social groups, Marjorie Rhodes, Christina Tworek, and I created a picture book that depicted a group of imaginary people— the Zarpies—who were diverse for race, ethnicity, gender, and age; the group thus could not be mapped onto any familiar, essentialized social group (Rhodes et al., 2012). Each page of the picture book showed a single individual displaying a notable trait (e.g. striped hair) or performing a novel activity (e.g. eating flowers). The accompanying text involved either bare plural generics (“Zarpies eat flowers”), indefinite singular generics (“A Zarpie eats flowers”), specific language with a group label (“This Zarpie eats flowers”), or specific language with no label (“This one eats flowers”).3 Our participants included both four-year-olds and adults, and after they had read the picture book several times they were given a battery of tests designed to assess the extent to which they essentialized Zarpies. For example, one set of test questions involved “switched-at-birth” tasks, 3

Quantified statements were not included as a point of comparison for several reasons. Firstly, parents rarely spontaneously produce kind-wide quantified statements when talking to young children (Gelman, 2003), so it is unlikely that quantified statements actually serve as a mechanism by which social essentialist beliefs are passed along, even if they potentially could be used to do this. Secondly, any finding that suggests that hearing quantified statements leads children to essentialize would have to be understood in light of the fact that young children (and adults) interpret and recall quantified statements as generics (e.g., Hollander et al., 2002; Leslie and Gelman, 2012; Leslie et al., 2011; Mannheim et al., 2011; Tardif et al., 2011). Thus, it is very likely that hearing quantified statements would have similar effects to hearing generic statements, but this would be because the quantified statement would be laid down in memory as a generic (Leslie and Gelman, 2012), and so would function in the same way. Thus one would expect that if generics induce essentialist beliefs, so will quantified statements, but this in itself may not be a particularly interesting fact.

CARVING UP THE SOCIAL WORLD WITH GENERICS



where participants were asked to judge, e.g., whether a baby born to a Zarpie mother (who eats flowers) but raised from birth by a non-Zarpie mother (who eats crackers) would grow up to eat flowers or crackers. Responding that the baby would grow up to eat flowers despite being raised by non-Zarpies would indicate essentialist beliefs about Zarpies. It is important to note that this question, like the others we used, is a very strong test for essentialism, since there is surely a default assumption that people will generally prefer crackers to flowers as things to eat. Our results indicate that hearing generic language—in either bare plural or indefinite singular form—leads both children and adults to essentialize even a wholly novel and diverse social group. If participants heard only specific language about the group, they showed little or no inclination to essentialize; however, if they heard generic language, they showed a marked increase in their tendency to essentialize. (For our four-year-old participants, this tendency emerged both on the basis of only reading the picture book twice prior to testing (Study 2), and on the basis of memory, having read the picture book four times over the course of a week (Study 1); thus the effects of generics emerge rapidly, but also remain over time.) However, if generic language is to be a means by which social essentialist beliefs are transmitted, one would expect not only the above findings, but also that parents should produce more generic language themselves if they essentialize a social group. To test this, we recruited parent–child duos, and presented half the parents with a paragraph designed to lead them to form essentialist beliefs about Zarpies. This paragraph emphasized that Zarpies were very different from other groups of people, but did not mention anything about Zarpies being similar to each other. The other half of the parents were given a paragraph that explained how Zarpies were very similar to other social groups. (A separate group of adults were tested to confirm that these two paragraphs did indeed lead to different levels of essentialist belief about Zarpies.) The parents were then given a version of the original picture book, with no text in it, and were asked to talk through the book with their children. As we predicted, parents who read the essentialist paragraph were significantly more likely to produce generics about Zarpies than the parents who read the non-essentialist paragraph. (Parents very occasionally produced universally quantified statements;



S .- J . LESLIE

however, the (low) rate at which they did so did not differ depending on whether they themselves essentialized Zarpies or not, which further suggests that quantified statements are not, in fact, a significant factor in the transmission of social essentialist beliefs.)

Generics, essence, and explanation Our findings suggest that parents pass on their essentialist beliefs to children by using generic language (Rhodes et al., 2012). One question is why generic language might serve this purpose—why, for example, does hearing generics lead children to essentialize even very diverse, novel groups? A possible answer is suggested in work conducted by Andrei Cimpian and his colleagues. Cimpian and Markman (2009, 2011) told young children novel facts about familiar, essentialized kinds—both natural and social—or specific members thereof, and then asked children to explain why they thought these facts obtained. The facts were either presented in specific form (e.g. “this butterfly has dust on her wings”) or generic form (e.g. “butterflies have dust on their wings”). When asked to explain generic facts, children spontaneously offered kind-based, inherent explanations (e.g. “they need the dust so they can fly”), but when asked to explain the specific facts, they offered more incidental, extrinsically-based explanations (e.g. “she flew through a dusty room”). This suggests that young children expect that generic facts obtain because of common intrinsic features of the members of the kind (Cimpian and Markman, 2009, 2011; see also Cimpian and Erickson, 2012). Thus, if children hear a series of generics about a kind as in Rhodes et al. (2012), it would seem likely that they would come to assume that the members of the kind share a range of deep, non-obvious, inherent similarities, and thereby come to essentialize the kind.4 It should be noted that the semantic truth-conditional profile of generics does not seem to require that there be an inherent, kind-based explanation for the holding of a generic fact. For example, generics such as “barns are red” and “dogs wear collars” arguably would seem to be 4 As noted above, to the extent that children understand kind-wide quantified statements, those statements could well have the same effect and for the same reason; indeed Cimpian and Erickson (2012) found similar effects when children were presented with facts quantified by “most.” However, as the authors note, this finding should again be interpreted in light of the fact that young children understand and remember kind-wide quantified statements as generics (Hollander et al., 2002; Leslie and Gelman, 2012; Tardif et al., 2011).

CARVING UP THE SOCIAL WORLD WITH GENERICS



true generics, even though these generalizations hold because of extrinsic, circumstantial factors, rather than because of something in the nature of the kinds in question. Thus, the data of Cimpian and his colleagues do not reflect that children are appreciating a semantic entailment of the generic, but rather that children are drawing more complex, defeasible inferences. One hypothesis that fits the available data nicely is that, by default, i.e. absent information to the contrary, we understand generics to express generalizations that hold because of common, inherent features of the members of the kind. Sally Haslanger (2011), as I read her, puts forward precisely this hypothesis. In particular, she discusses claims such as “women are submissive,” and makes the intriguing observation that such a claim could be true but still objectionable. Suppose, for example, that society punishes assertiveness in women to such an extent that they rarely, if ever, are other than perfectly submissive. It would be hard to argue that the generic “women are submissive” is then false, but there nonetheless would seem to be something damaging about asserting it. Haslanger proposes that the damage comes, at least in part, from our tendency to suppose that generic generalizations obtain because of inherent features of the kind in question. That is, even if “women are submissive” is made true by purely external sociological factors, it still may by default communicate that there is something in the nature of women that makes them submissive. That is, it may reinforce essentialist beliefs about women, and further communicate that this shared essence causes women to be submissive. This is not, of course, to say that we invariably must interpret “women are submissive” in this way—it would still be possible to interpret it more along the lines of “barns are red,” where the essentialist interpretation is blocked by worldly knowledge. The hypothesis under consideration is instead that we need to have specific worldly knowledge to block such an interpretation; absent that, we will likely understand a generic generalization to be supported by underlying essentialist facts. It is therefore particularly intriguing to consider Haslanger’s hypothesis in light of developmental data, since adults are more likely than preschoolers to have the requisite worldly knowledge that blocks such interpretations. The data from Rhodes et al. are notable in this respect. We introduced an entirely novel social group to four-year-old children with fairly minimal information; yet hearing this social group described



S .- J . LESLIE

in generic terms led these children to suppose that traits such as liking to eat flowers may be heritable, innate, and non-socially determined among the members of this group. (And perhaps even more dramatically, the adults in our study formed the same beliefs about the group as did the four-year-olds—despite having more worldly knowledge at their disposal, which could have served to block the relevant interpretation.)

The insidious generic It is worth reflecting on the subtle nature of the mechanism by which social essentialist beliefs are transmitted. The parents who participated in our third study—in which parents read a paragraph that led them to hold more or less essentialist beliefs about Zarpies, before talking through a wordless picture book with their children—were surely not consciously aware that their essentialist beliefs were leading them to produce more generic language than they otherwise would. As far as they were concerned, they read a paragraph about some group called “Zarpies,” and then talked through a wordless picture book with their children. Having personally coded the transcripts, I found it clear that the parents were really only concerned with keeping their children entertained and with completing the task. Very few parents took the specifically pedagogical attitude of trying to impart information about Zarpies—and why would they, since our Zarpies were clearly a made-up group? The overall sense one gets from the transcribed conversations is that the parents were trying to complete the task without boring their children, so that they could move on to other things. Yet our simple manipulation of essentialist beliefs led parents to more than double their number of generic utterances. Consider, then, a parent who is actively seeking not to impart social essentialist beliefs to his or her child. The mechanism of generic language is so subtle that this parent might nonetheless utter enough generics to instill such beliefs. A parent who is committed to teaching her child that gender differences are not deep or natural might nonetheless have enough seemingly innocuous conversations about boys and girls that the child may come to rapidly essentialize gender (see Gelman et al., 2004, for similar observations). Very young children are adept at (subpersonally) tracking probabilistic information, and so even fairly small elevations in the frequency with which generics are uttered in relation to a given social group may well suffice to induce essentialist beliefs—even

CARVING UP THE SOCIAL WORLD WITH GENERICS



if these small elevations are too subtle for the parents themselves to be able to consciously monitor or even notice. These findings also bear on just how information is communicated in a conversational setting. Since we are dealing here with four-year-old children, the prospects of a parent succeeding at explicitly communicating essentialist beliefs—i.e., by spelling them out—are fairly low. After all, an explicit formulation of essentialism may not be readily comprehensible to children at this age: “there’s something deep and inherent in Zarpies that makes them all alike.” Minimally, such statements are effectively never found in parent-child conversations, even about animal kinds (Gelman, 2003). Generic language may thus provide a vehicle whereby beliefs are communicated without ever being explicitly formulated—perhaps even at an age when the consciously accessible conceptual resources are not even able to explicitly entertain the communicated belief. However implicit and inchoate the essentialist beliefs may be, the seemingly mundane generic sentence is able to communicate them quite effectively.

Generics and Conceptions of Abilities Thus far we have considered essentialist beliefs about social groups, i.e., that certain groups of people share fundamental, inherent natures which ground a range of common features. Of course, only certain features are candidates to be grounded in a person’s essence; such features must be stable, enduring traits—not mercurial, fleeting, or environmentally determined. Thus, just as we might ask what leads us to conceive of a social group as sharing an essence, we can also ask what leads us to think of a given feature as fixed, inherent, or natural. Beliefs about intellectual abilities are an interesting example in this regard, as these abilities can be understood as malleable and highly responsive to circumstance, or, alternatively, as stable and unchanging, and thus essentializable. Let us begin by considering the question: what does it take to be good at a particular, fairly demanding activity—say an intellectual activity, such as doing math? One answer might be that it takes hard work, patience, and dedication. Even if one is initially stumped by mathematics, with enough energy and commitment, one can come to excel at it. Let us call someone who gives such an answer an incremental theorist, thereby reflecting her belief that the ability in question can be



S .- J . LESLIE

incrementally acquired and improved through focus, dedication, and practice. The contrasting answer is that to be good at a particular demanding activity it takes natural, inherent talent—something that is fixed and unchangeable, perhaps grounded in one’s essence. One either “has it” or one does not—mathematical ability is a gift, and hard work in the absence of this gift will be largely futile. In accord with the psychological literature, let us call someone who believes this an entity theorist, to reflect her conception of ability as a fixed, stable entity—something inherent that is either possessed or lacked. (These two answers form a spectrum, of course, but for simplicity of discussion I will speak as though they form instead a dichotomy.) A considerable body of evidence in psychology, collected by Carol Dweck and her colleagues, suggests that it is quite detrimental to performance to be an entity theorist (see Dweck, 1999, 2006 for reviews). That is, across many diverse experimental and real-world conditions, incremental theorists fare far better than entity theorists. In one concrete illustration, Lisa Blackwell, Kali Trzesniewski, and Carol Dweck (2007) examined children who were undergoing the transition from elementary school to middle school. This is a period of time in which students’ grades drop precipitously, and students across the board struggle to keep up with new academic and social challenges. The researchers found, however, that the statistical drop in academic performance was in fact driven by students who were “entity theorists”—students who held the incremental view did not see their grades plummet. They then went on to randomly divide another set of middle-schoolers into two groups. Both groups were given a supplemental course in which they learned about the brain, and also learned practical study skills. The two groups had exactly the same lessons except for one class. In this one class, the control group was taught about memory, and learned some mnemonic skills, while the other group was effectively taught to hold an incremental theory of ability. That is, they had a lesson on how the brain forms new connections as one learns, and that learning thus makes one smarter; they were told that the brain is like a muscle—responsive to effort and training. The students in this second group, but not the ones in the control group, showed notably more improvement in their grades and their general academic performance. This is but one of many empirical demonstrations that people fare better when they hold an incremental rather than an entity theory of

CARVING UP THE SOCIAL WORLD WITH GENERICS



ability. Similar effects have been documented among younger children and among adults, and also among otherwise higher-achieving and lower-achieving individuals; the effects robustly persist whether or not the relevant beliefs were experimentally induced or were held as preconceived notions. At least some of the explanation of these effects would seem to come from differential reactions to failures. If an incremental theorist fails at a given task, her characteristic response will be that she needs to work harder so as to succeed next time. An entity theorist, however, will characteristically take failure to indicate that she does not “have what it takes,” and so hard work would only be a waste of time! Failure for an entity theorist is thus much more distressing and anxietyproducing—it is not something to be easily overcome, a mere temporary obstacle, but rather the crushing “realization” that one does not have the requisite gift. Correspondingly, entity theorists tend to prefer easier tasks where the chances of failure are low and where they are more likely to receive validation, whereas incremental theorists are more likely to seek out challenges so that they can improve their abilities. In general, holding an entity theory is associated with higher levels of anxiety—since failure would be crushingly diagnostic of one’s inherent shortcomings—and this anxiety itself acts to impair performance (Dweck, 1999, 2006, and references therein). In light of this, it is important to understand what might lead children to initially form an entity theory rather than to adopt an incremental view. At least part of the explanation would seem to be due to the sort of praise to which young children are exposed. Suppose a preschooler paints an excellent picture—one might praise her by saying “that is a great painting,” or by saying “you are a great painter.” The former manner of praise is focused on the result, while the latter is focused on the individual—perhaps subtly communicating that the relevant locus of praise is a stable and inherent trait, of the sort that lends itself to entity theoretic thinking. Indeed, recent evidence suggests that children who are praised in this “person-directed” way manifest the same sort of demotivation, distress, and helplessness in the face of subsequent challenges that is characteristic of the entity theorist (Cimpian et al., 2007; see also Kamins and Dweck, 1999; Mueller and Dweck, 1998). The nature of praise may not be the only factor that encourages entity theoretic thinking, however. As noted above, when children are asked to explain a generic fact, they tend to appeal to inherent traits shared by



S .- J . LESLIE

members of the kind; these findings also extend to generics that pertain to abilities. That is, if children are told, e.g., “boys are really good at a game called ‘gorp’,” and asked to explain why they think this might be so, they tend to offer explanations such as “maybe because they are tougher than girls,” or “because they are smart.” In contrast, when they are asked to explain why a particular boy or girl had the property, they tend to offer more effort-based explanations, such as “because she took ballet class and then she practiced a lot, so then she got really good at it.” The same effects were found if children were asked directly whether they thought the fact (individual or generic) obtained because the relevant individual(s) “have to practice this game,” or were rather “just good at it.” Children who were presented with generic facts tended to downplay the importance of effort and practice, and instead preferred more “gift-based” explanations (Cimpian and Markman, 2011).5 The heart of entity theoretic thinking is explaining abilities and success at a given task by reference to stable and fixed traits, rather than by reference to effort and dedication—but this would seem to be precisely what children do when they hear such matters described with generic language. Even though these preschool-aged children often gave effortbased explanations for a given individual’s successes and abilities, when instead they heard such attributions made in generic terms, they adopted more entity-theoretic, essentialist explanations. This suggests that hearing generic language ascribing achievements and abilities to social groups may lead children to adopt a more entity theoretic perspective on the relevant abilities. An even more dramatic illustration of the power of generic language concerning abilities is found in an ongoing study, again by Cimpian and his colleagues (Cimpian, Bian, and Sutherland, in preparation). In this study, preschoolers were asked a series of questions about familiar abilities—questions that were either in generic or individual form. In particular, they were either asked questions such as “are girls good at 5 Interestingly, these effects were not found when children were asked to consider social groups about which they did not have pre-existing essentialist beliefs (boy/girls at a different school). However, the results of Rhodes et al. (2012) suggest that this could potentially be due to the fact that children only heard a small number of generics concerning this social group—it is possible that, had children heard more extensive generic input about these boys and girls at a different school, they may have again come to think of the ascribed abilities in more entity-theoretic terms.

CARVING UP THE SOCIAL WORLD WITH GENERICS



drawing?” or questions about a particular friend of theirs, e.g. “is Hannah good at drawing?”. Thus, children were never actually told anything about group or individual abilities; they were simply asked to consider these questions. Children in both conditions (i.e. generic vs. individual) were then told that a particular individual is good at a novel activity (e.g. “this girl is good at a game called ‘gorp’ ”). The children were asked questions that examined whether they thought that this individual needed to practice to be good at gorp, or whether they though he/she was “just good” at it—that is, the extent to which they held entity vs. incremental conceptions of what it takes to excel at this unfamiliar game. Strikingly, children who had previously answered questions concerning the abilities of groups, rather than individuals, placed much less emphasis on practice, and tended instead to suppose the individual had an effort-independent gift. Thus, simply considering generic questions about familiar abilities led children to adopt entity-theoretic conceptions of a novel ability.

Stereotype threat Suppose you are a member of a group that is associated through cultural stereotypes with subpar performance at a particular activity—in the way that, say, women are often stereotyped as being less good at math than men. A robust and extensively documented phenomenon, known as stereotype threat, is that where there is common knowledge of the stereotype, one’s performance at that activity will be degraded if one’s membership in this group is made salient, even in a subtle way. For example, if women are asked to identify their gender prior to taking a math exam (as has traditionally been the case with the SATs and GREs, for example), they will perform less well on the exam than they would have had gender membership not been made salient (see e.g. Ambady et al., 2001; Aronson et al., 2002; Good et al., 2003; Steele and Aronson, 1995; Steele, 2010; and many others). Membership in the stereotyped group can be made salient in the most subtle of ways—checking a box on a form, or even just being in an environment with subtle cues that make gender or other group membership salient (e.g., there being a poster in the room that says “Math: Got Women?”). Understanding stereotype threat is important for a range of reasons—for example, in many academic contexts where gender gaps are found, gender membership is made subtly salient (such as being asked to identify as a man or



S .- J . LESLIE

woman in the course of the SATs), raising the question of how much of these putative gender gaps are due simply to stereotype threat. Further, stereotype threat is in many respects a self-reinforcing phenomenon— awareness of the cultural stereotype leads women to perform less well in mathematics, which then seems to vindicate the stereotype. (For discussions of how stereotype threat may impact women’s representation and performance in philosophy, see Haslanger, 2008; Saul, forthcoming.) What if one is a member of a group that is positively stereotyped with respect to the relevant ability? For example, if one is again taking a math test, and one is reminded that one is Asian, how does the stereotype that Asians are good at math affect one’s performance? Here, the results are more subtle, and they suggest that the manner in which one’s group membership is highlighted is important. If one’s Asian heritage is subtly highlighted (e.g. by checking a box to identity one’s race/ethnicity), then one’s performance may well be improved relative to how one would otherwise perform (see e.g. Ambady et al., 2001; Shih et al., 2002). If, however, the stereotype is activated in a blatant and direct manner—e.g., if Asian-Americans are told that the test they are taking is to be used to confirm that Asians really are good at math—then performance may be impaired (Brown and Josephs, 1999; Cheryan and Bodenhausen, 2000; Shih et al., 2002). The precise psychological mechanisms—which are likely myriad—by which stereotype threat operates have not yet been fully identified; however, a range of results suggests that stereotype threat may operate in part (though surely only in part) by invoking the characteristic effects of holding an entity theory of the ability whose manifestation is being evaluated. For example, people who hold, or are taught, strongly incremental views about the relevant abilities are less susceptible to stereotype threat (e.g., Aronson et al., 2002; Dweck, 2006; Good et al., 2003), and conversely, women who are told that gender gaps in math performance are biologically based exhibit impaired mathematical performance (Dar-Nimrod and Heine, 2006). Further, if stereotype threat is in part caused by precipitating the typical effects of holding an entity theory, this would explain why performance can be hampered by being blatantly reminded that one is a member of a positively stereotyped group. Since holding an entity theory leads in general to impaired performance (due, among other things, to anxiety about whether one actually has the entity), members of even positively stereotyped groups

CARVING UP THE SOCIAL WORLD WITH GENERICS



would be expected to show worse performance if entity-theoretic thinking is activated in a given context. Since generics concerning abilities provoke entity-theoretic thinking, the question arises whether hearing generics concerning the abilities of groups can induce stereotype threat—perhaps even concerning a wholly novel ability, about which one has no prior beliefs. Cimpian and his colleagues tested precisely this hypothesis. Suppose children are introduced to a novel game, and are either told that girls (or boys) are good at the game, or told that an individual girl (or boy) is good at the game. How might the difference between generic vs. individual language here affect children’s interest and motivation regarding the game and their ability to successfully play the game? The experimenters found that children who heard generic language showed significantly lower motivation, less interest, and impaired performance. This was especially so after children faced challenges in the course of the game or received any negative feedback on their performance (Cimpian, 2010; Cimpian et al., 2012). Notably, these results held regardless of whether children heard a generic concerning their own gender—that is, children displayed lower motivation and worse performance even when they heard that their own gender was good at the game. This again reflects the insidious nature of generics. Suppose a parent decides actively to combat stereotypes associating girls and women with poor mathematical ability by telling her daughter that “girls are really good at math!” Such a parent might think that she will thereby insulate her daughter against stereotype threat in the mathematical context—that her utterance will motivate her daughter to work hard at math, and will reduce her stereotype-based anxiety about math performance, thereby leading her to do better in mathematics. The results discussed here, however, suggest that the parent’s utterance may instead have the opposite effect and activate entity-theoretic thinking, with all its concomitant difficulties.

Conclusion Perhaps this chapter should have been titled “Generics: Just do not use them!” Certainly, the empirical evidence suggests that, especially in conversation with young children, generics have a range of detrimental effects—effect that range from laying the foundations of prejudice to



S .- J . LESLIE

inducing detrimental entity-theoretic conceptions of ability. Should we not simply try to excise them from our linguistic repertoire? The simple answer, of course, is “yes”—it would seem to be potentially very beneficial to cease to use generic language when discussing social phenomena, particularly in conversation with children. Certainly, one should strive to monitor one’s use of generics in this context. The difficulty of trying to eliminate them from our repertoire altogether should not be underestimated, however. As noted at the beginning of this chapter, it seems that generics are our way of articulating our most basic generalizations, and so in this way they reflect something very fundamental in our outlook. To try to excise them from our conversations would thus be, admittedly, a daunting task. To make matters worse, it is not as though one could simply switch to using quantificational language. Category-wide statements that are quantified by “all” and “most” are frequently remembered and interpreted as generics, even by adults (Leslie and Gelman, 2012; Leslie et al., 2011). In the case of young children, this phenomenon becomes even more dramatic. English-speaking three-year-olds interpret and recall even statements quantified with “some” as generics (Hollander et al., 2002; Leslie and Gelman, 2012; Tardif et al., 2011); and even four-year old children whose native languages are less explicit in certain respects than English interpret “some”-statements as generics (Mannheim et al., 2011; Tardif et al., 2011). Thus, when speaking to very young children, one cannot even rely on the more conservative “some” to avoid communicating something generic. Consider also the ambiguous statement “they are good at math,” said while pointing at a small group of girls—this could be interpreted as a communicating a generic (e.g., “girls are good at math”), but one would naturally think that it is more readily understood as meaning that those particular girls in the demonstrated group are good at math. Since this would seem to be the more natural interpretation, especially when the utterance is accompanied by a pointing gesture, this might seem to be a safe form of utterance—not one that will carry the negative effects of generic language. However, recent data suggests that preschoolers do not agree that this is the most natural interpretation of such an utterance! Such ambiguous statements strike adults, but not preschoolers, as likely being about the specific group—preschoolers instead overwhelmingly seem to interpret such utterances as generics (Meyer and Baldwin,

CARVING UP THE SOCIAL WORLD WITH GENERICS



2013). Young children appear to be quite eager to assign generic interpretations, as it were, and so adults may inadvertently communicate generic information even when they explicitly intend otherwise. None of this is to say that it is not worth trying to limit one’s use of generics when describing social phenomena. However, there are no easy substitutions that will not themselves be assigned generic interpretations. The real issue, I submit, is that adults have certain ways of carving up the social world, and these ways are very salient to us. To the extent that our ways of thinking and speaking are shaped by this, children will find a way to glean the relevant information from our utterances, so as to “sync up” with the adults in their community as quickly as possible. The partial but helpful solutions may then not only lie with working to reduce our use of generic language, but also with striving to step outside of the reflex ways in which we, as adults, carve up the social world.

References Ambady, N., Shih, M., Kim, A., and Pittinsky, T.L. (2001). Stereotype Susceptibility in Children: Effects of Identity Activation on Quantitative Performance. Psychological Science 12(5): 385–90. Aronson, J., Fried, C.B., and Good, C. (2002). Reducing the Effects of Stereotype Threat on African American College Students by Shaping Theories of Intelligence. Journal of Experimental Social Psychology 38: 113–25. Baldwin, D.A., Markman, E., and Melartin, R. (1993). Infants’ Ability to Draw Inferences about Nonobvious Object Properties: Evidence from Exploratory Play. Child Development 64: 711–28. Bastian, B., and Haslam, N. (2006). Psychological Essentialism and Stereotype Endorsement. Journal of Experimental Social Psychology 42: 228–35. Blackwell, L., Trzesniewski, K., and Dweck, C.S. (2007). Implicit Theories of Intelligence Predict Achievement across an Adolescent Transition: A Longitudinal Study and Intervention. Child Development 78: 246–63. Brandone, A., Cimpian, A., Leslie, S.J., and Gelman, S.A. (2012). Do Lions have Manes? For Children, Generics are about Kinds not Quantities. Child Development 83(2): 423–33. Brown, R.P., and Josephs, R.A. (1999). A Burden of Proof: Stereotype Relevance and Gender Differences in Math Performance. Journal of Personality and Social Psychology 76: 246–57. Carlson, G. (1977). Reference to Kinds in English. Ph.D. dissertation, University of Massachusetts, Amherst.



S .- J . LESLIE

Carlson, G.N., and Pelletier, F.J. (1995). The Generic Book. Chicago: Chicago University Press. Cheryan, S., and Bodenhausen, G.V. (2000). When Positive Stereotypes Threaten Intellectual Performance: The Psychological Hazards of “Model Minority” Status. Psychological Science 11(5): 399–402. Cimpian, A. (2010). The Impact of Generic Language about Ability on Children’s Achievement Motivation. Developmental Psychology 46(5): 1333–40. Cimpian, A., and Erickson, L.C. (2012). The Effect of Generic Statements on Children’s Causal Attributions: Questions of Mechanism. Developmental Psychology 48(1): 159–70. Cimpian, A., and Markman, E.M. (2009). Information Learned from Generic Language becomes Central to Children’s Biological Concepts: Evidence from their Open-Ended Explanations. Cognition 113(1): 14–25. Cimpian, A., and Markman, E.M. (2011). The Generic/Nongeneric Distinction Influences how Children Interpret New Information about Social Others. Child Development 82(2): 471–92. Cimpian, A., Arce, H.C., Markman, E.M., and Dweck, C.S. (2007). Subtle Linguistic Cues Affect Children’s Motivation. Psychological Science 18(4): 314–16. Cimpian, A., Mu, Y., and Erickson, L.C. (2012). Who is Good at this Game? Linking an Activity to a Social Category Undermines Children’s Achievement. Psychological Science 23(5): 533–41. Cimpian, A., Bian, L., and Sutherland, S. (in preparation). The Origins Of Children’s Beliefs about Achievement: Thinking about the Abilities of Groups causes 4-year-olds to Devalue Effort. Cohen, A. (1996). Think Generic: The Meaning and Use of Generic Sentences. Ph.D. dissertation, Carnegie Mellon University. Cohen, A. (2004). Generics and Mental Representation. Linguistics and Philosophy 27(5): 529–56. Dar-Nimrod, I., and Heine, S.J. (2006). Exposure to Scientific Theories affects Women’s Math Performance. Science 314: 435. Diesendruck, G., and Haber, L. (2009). God’s Categories: The Effect of Religiosity on Children’s Teleological and Essentialist Beliefs about Categories. Cognition 110: 100–14. Dweck, C.S. (1999). Self-theories: Their Role in Motivation, Personality and Development. Philadelphia: Psychology Press. Dweck, C.S. (2006). Mindset: The New Psychology of Success. New York: Random House. Gelman, S.A. (2003). The Essential Child: Origins of Essentialism in Everyday Thought. New York: Oxford University Press.

CARVING UP THE SOCIAL WORLD WITH GENERICS



Gelman, S.A., Taylor, M.G., and Nguyen, S. (2004). Mother–Child Conversations about Gender: Understanding the Acquisition of Essentialist Beliefs. Monographs of the Society for Research in Child Development, 69(1). Gelman, S.A., Ware, E.A., and Kleinberg F. (2010). Effects of Generic Language on Category Content and Structure. Cognitive Psychology 61(3): 273–301. Good, C., Aronson, J., and Inzlicht, M. (2003). Improving Adolescents’ Standardized Test Performance: An Intervention to Reduce the Effects of Stereotype Threat. Applied Developmental Psychology 24: 645–62. Greenberg, Y., (2003). Manifestations of Genericity. Routledge. New York. Haslam, N., and Levy, S.R. (2006). Essentialist Beliefs about Homosexuality: Structure and Implications for Prejudice. Personality and Social Psychology Bulletin 32: 471–85. Haslam, N., Rothschild, L., and Ernst, D., (2000). Essentialist Beliefs about Social Categories. British Journal of Social Psychology 39: 113–27. Haslam, N., Rothschild, L., and Ernst, D. (2002). Are Essentialist Beliefs Associated with Prejudice? British Journal of Social Psychology 41: 87–100. Haslanger, S. (2008). Changing the Culture and Ideology of Philosophy: Not by Reason (Alone). Hypatia 23(2): 210–23. Haslanger, S. (2011). Ideology, Generics, and Common Ground. In C. Witt (ed.), Feminist Metaphysics: Explorations in the Ontology of Sex, Gender, and the Self. Dordrecht: Springer, pp. 179–209. Hirschfeld, L.A. (1996). Race in the Making: Cognition, Culture, and the Child’s Construction of Human Kinds. Cambridge, MA: MIT Press. Hollander, M.A., Gelman, S.A., and Star, J. (2002). Children’s Interpretation of Generic Noun Phrases. Developmental Psychology 38: 883–94. Kamins, M.L., and Dweck, C.S. (1999). Person versus Process Praise and Criticism: Implications for Contingent Self-Worth and Coping. Developmental Psychology 35: 835–47. Keller, J. (2005). In Genes we Trust: The Biological Component of Psychological Essentialism and its Relationship to Mechanisms of Motivated Social Cognition. Journal of Personality and Social Psychology 88: 686–702. Lawler, J. (1973). Studies in English Generics. University of Michigan Papers in Linguistics 1(1). Leslie, S.J. (2007). Generics and the Structure of the Mind. Philosophical Perspectives 21(1): 375–403. Leslie, S.J. (2008). Generics: Cognition and Acquisition. Philosophical Review 117 (1): 1–47. Leslie, S.J. (2012). Generics Articulate Default Generalizations. Recherches Linguistiques de Vincennes 41: 25–44.



S .- J . LESLIE

Leslie, S.J. (2013). Essence and Natural Kinds: When Science Meets Preschooler Intuition. In Gendler, T., and Hawthorne, J. (eds.), Oxford Studies in Epistemology 4: 108–66. Leslie, S.J. (forthcoming). The Original Sin of Cognition: Fear, Prejudice, and Generalization. The Journal of Philosophy. Leslie, S.J., and Gelman, S.A. (2012). Quantified Statements are Recalled as Generics: Evidence from Preschool Children and Adults. Cognitive Psychology 64: 186–214. Leslie, S.J., Khemlani, S., and Glucksberg, S. (2011). All Ducks Lay Eggs: The Generic Overgeneralization Effect. Journal of Memory and Language 65: 15–31. Mahalingam, R. (2007). Essentialism, Power, and the Representation of Social Categories: A Folk Sociology Perspective. Human Development 50: 300–19. Mahalingam, R., and Rodriguez, J. (2006). Culture, Brain Transplants and Implicit Theories of Identity. Journal of Cognition and Culture 6: 453–62. Mannheim, B., Gelman, S.A., Escalante, C., Huayhua, M., and Puma, R. (2011). A Developmental Analysis of Generic Nouns in Southern Peruvian Quechua. Language Learning and Development 7(1): 1–23. Meyer, M., and Baldwin, D. (2013). Pointing as a Socio-Pragmatic Cue to Particular vs. Generic Reference. Language Learning and Development 9(3): 245–65. Meyer, M., Gelman, S.A., and Stilwell, S.M. (2011). Generics are a Cognitive Default: Evidence from Sentence Processing. In Proceedings of the 33rd Annual Conference of the Cognitive Science Society. Boston, MA: Cognitive Science Society. Meyer, M., Leslie, S.J., Gelman, S.A., and Stilwell, S.M. (2013). Essentialist Beliefs about Organ Transplants in the United States and India. Cognitive Science 37 (4): 668–710. Mueller, C.M., and Dweck, C.S. (1998). Intelligence Praise can Undermine Motivation and Performance. Journal of Personality and Social Psychology 75: 33–52. Pelletier, F. and Asher, N. (1997). Generics and Defaults. In van Benthem, J., and ter Meulen, A. (eds.), Handbook of Logic and Language. Cambridge, MA: MIT Press, pp. 1125–79. Prentice, D.A., and Miller, D.T. (2006). Essentializing Differences between Women and Men. Psychological Science 17: 129–35. Prentice, D.A., and Miller, D.T. (2007). Psychological Essentialism of Human Categories. Current Directions in Psychological Science 16(4): 202–6. Rhodes, M., and Gelman, S.A. (2009). A Developmental Examination of the Conceptual Structure of Animal, Artifact, and Human Social Categories across Two Cultural Contexts. Cognitive Psychology 59: 244–74.

CARVING UP THE SOCIAL WORLD WITH GENERICS



Rhodes, M., Leslie, S.J., and Tworeck, C. (2012). Cultural Transmission of Social Essentialism. Proceedings of the National Academy of Sciences (PNAS) 109: 13526–31. Rothbart, M., and Taylor, M. (1992). Category Labels and Social Reality: Do we View Social Categories as Natural Kinds? In Semin, G., and Fiedler, K. (eds.), Language, Interaction and Social Cognition. London: Sage, pp. 11–36. Saul, J. (forthcoming). Unconscious Influences and Women in Philosophy. In Jenkins, F., and Hutchison, K. (eds.), Women in Philosophy: What Needs to Change? Shih, M., Ambady, N., Richeson, J.A., Fujita, K., and Gray, H.M. (2002). Stereotype Performance Boosts: The Impact of Self-relevance and the Manner of Stereotype Activation. Journal of Personality and Social Psychology 83(3): 638–47. Steele, C.M. (2010). Whistling Vivaldi: And other Clues to how Stereotypes Affect Us. New York: Norton. Steele, C.M. and Aronson, J. (1995). Stereotype Threat and the Intellectual Test Performance of African Americans. Journal of Personality and Social Psychology 69(5): 797–811. Tardif, T., Gelman, S.A., Fu, X., and Zhu, L. (2011). Acquisition of Generic Noun Phrases in Chinese: Learning about Lions without an “-s”. Journal of Child Language 30: 1–32. Williams, M., and Eberhardt, J. (2008). Biological Conceptions of Race and the Motivation to Cross Racial Boundaries. Journal of Personality and Social Psychology 94: 1033–47. Yzerbyt, V., Corneille, O., and Estrada, C. (2001). The Interplay of Subjective Essentialism and Entitativity in the Formation of Stereotypes. Personality and Social Psychology Review 5: 141–55.

9 Flavors of “Togetherness” Experimental Philosophy and Theories of Joint Action Deborah Tollefsen, University of Memphis Roger Kreuz, University of Memphis Rick Dale, University of California, Merced

Philosophical Accounts of Joint Action The philosopher’s interest in joint action stems, in part, from a longstanding interest in understanding the distinction between intentional and non-intentional (non-voluntary) actions, between things that happen to us and things that we do. On a causal theory of action, actions are to be distinguished from mere happenings by virtue of the fact that they are caused by antecedent mental events such as beliefs, desires, and intentions. The fact that our actions are often “joint” (or “collaborative”), and intentionally so, raises the question of what antecedent mental states or events cause joint actions. How are the things we do together intentionally (soccer games) different from things which are the result of individual intentional actions (traffic jams), but for which we would not want to say “We did it together”? There is a growing body of literature in philosophy that suggests that what is unique about doing things together is that it involves a sort of irreducible “jointness” that can be characterized as “we-intention.” Just as individual actions are shaped and informed by individual intentions, joint actions are shaped and informed by joint or shared intentions. Thought experiments are often used to solicit intuitions regarding the

FLAVORS OF

“ TOGETHERNESS ” 

existence of shared intentions. To offer the reader some examples of the use of intuitions in this literature, we extensively quote from the works of two prominent theorists in this area. First, consider the following example used by John Searle to motivate this intuition: Imagine that a group of people are sitting on the grass in various places in a park. Imagine that it suddenly starts to rain and they all get up and run to a common, centrally located shelter. Each person has the intention expressed by the sentence “I am running to the shelter.” But for each person, we may suppose that his or her intention is entirely independent of the intentions and behavior of others. In this case there is no collective behavior; there is just a sequence of individual acts that happen to converge on a common goal. Now imagine a case where a group of people in a park converge on a common point as a piece of collective behavior. Imagine that they are part of an outdoor ballet where the choreography calls for the entire corps de ballet to converge on a common point. What exactly is the difference? . . . Intuitively, in the collective case the individual intentionality, expressed by “I am doing act A,” is derivative from the collective intentionality “We are doing act A.” (Searle, 1990, p. 402)

Likewise, we find Margaret Gilbert motivating the existence of shared intention with the following example: Suppose someone tells me “Ralph and Alice are shopping for clothes.” A variety of situations could be at issue. First, Alice could be in one store buying a dress, while Ralph is in a store five miles distant, buying a suit. Or Alice and Ralph may both be in the same store, each one buying what he or she individually needs. Her shopping is being carried out quite independently of his. In both of these cases, it might be said that Ralph and Alice will not be shopping for clothes together. In these cases it would be natural to say: ‘Ralph and Alice are both shopping for clothes.’ Finally, they may be doing something we really would think of as shopping for clothes together. He may sit near by while she looks at dresses, may make suggestions and comments, and may be consulted carefully before the final purchase is made . . . The main point I want to make in relation to the possible situations covered . . . is that the third case, as described, has a very different flavour from either of the first two cases. (Gilbert, 1989, p. 154)

These examples are meant to solicit the intuition that doing something together often means more than simply each individual performing an action at the same time and/or within the same vicinity. According to Searle, Gilbert, and many others, there is a “jointness,” a “we-ness,” a “togetherness” present that accounts for the different flavor of certain social interactions. How to understand the nature of this jointness or the



D . TOLLEFSEN ET AL .

nature of we-intentions that underlie joint action is a topic of much dispute. There are a variety of theories on offer. Our aim here is not to question any particular theory of joint action but to consider the methodology that is universally adopted in this literature—the reliance on “common intuitions” to support analyses of joint action and shared intention. Although philosophers rely on judgments of particular cases in order to build their theories, they do not actually consider the judgments of ordinary people when building these theories. The question remains then whether such theories get at the heart of our concepts of jointness, or simply at the heart of the particular philosopher’s concept of jointness. Do ordinary folk actually make distinctions between doing things and doing things together? And do they make these distinctions on the basis of information about the action type or on the basis of some other information? We believe that experimental philosophy provides a useful method for gauging ordinary people’s intuitions about the concept of joint action, and that doing so provides some empirical support for philosophical theories that appeal to primitive concepts of togetherness, “we-ness,” or “jointness.”

Experimental Philosophy and Collective Intentionality Philosophers interested in joint action are not the only ones whose methodology relies heavily on intuitions. Indeed, most philosophical theories are motivated by appeal to intuitions described as “common.” In recent years there has been a movement in philosophy that is seen by some to significantly challenge this approach. Experimental philosophy refers to a movement that seeks to contribute to fundamental philosophical questions by running empirical studies of the psychological processes underlying people’s intuitions about central philosophical concepts and issues. The idea is to move beyond a mere appeal to the “common” intuitions or to the intuitions of some academic philosophers. By doing so, theories can actually take into account the ways in which real people think, discriminate between cases, and make judgments concerning the nature of personal identity, moral agency, knowledge, justification, and a host of other notions that are of interest to the philosopher.

FLAVORS OF

“ TOGETHERNESS ” 

One particular strand of experimental philosophy, the strand we are most attracted to, attempts to preserve a role for common intuitions. However, the preservation of common intuitions is in the service of a very different sort of methodology from that used by much of twentiethcentury Anglo-American philosophy. Anglo-American philosophy has, at least since the turn of century, been engaged in conceptual analysis which involves a consideration of a number of cases which are then supposed to allow for further refinement of the meaning of certain concepts and inevitably result in a proposed list of necessary and sufficient conditions (involving yet more concepts) for the concept of interest. This has led to a cottage industry of “But what about this!” articles offering counter-examples which challenge either the sufficiency or necessity of the conditions and send the philosopher back to the drawing board. Conceptual analysis has taken a bit of a beating over the past few decades, and there has been a great number of new methodological turns taken—the method of genealogy, for instance. But despite this, conceptual analysis is still prominent in some domains of philosophy.1 This is particularly so in the field of collective intentionality, where concepts like joint action and shared intention are analyzed in terms of necessary and sufficient conditions that appeal to other concepts thought to be more fundamental. Consider Michael Bratman’s (1999) analysis of shared intention: We intend to wash the dishes if and only if: 1. a. I intend that we wash the dishes. b. You intend that we wash the dishes. 2. I intend that we wash the dishes in accordance with and because of 1a and 1b; you intend likewise. 3. 1 and 2 are common knowledge between us.

1 It might be argued that Bratman, Gilbert, and others are not giving a conceptual analysis but the conditions under which joint action or shared intention exist—an ontic analysis. Gilbert, at least, specifically associates her work with conceptual analysis in her early work, On Social Facts (1989). Regardless of what their aim is, they are attempting to provide necessary and sufficient conditions, and the sufficiency and necessity of these conditions is tested by thought experiments and judged according to the intuitions such thought experiments supposedly solicit.



D . TOLLEFSEN ET AL .

As a first approximation, this complex of intentional attitudes seems a plausible candidate. But consider a case in which we each intend to wash the dishes together and we each do so in part because of the other’s intention. However, I intend to wash the dishes with Palmolive and you intend to wash them with Joy. All of this is common knowledge and we will not compromise. Is there a collective intention present? It seems not. In this case we do not have our subplans coordinated in the appropriate way. Recall that one of the jobs that shared intention has is to coordinate our individual plans and goals. In the example above, our individual subplans are in conflict and this would prevent us from achieving our goal of getting the dishes washed. Bratman avoids this counter-example by adding a clause about participants’ subplans. It is not necessary that our subplans match, but they must mesh. So, if my subplan is to wash the dishes with Palmolive and your subplan is to wash them with hot water, and I have no preference about the water temperature, then our subplans mesh though they do not match exactly. But if we have subplans to wash the dishes with completely different types of dish detergent then our subplans do not mesh. Bratman reformulates the account in the following way: We intend to J if and only if: 1. (a) I intend that we J and (b) you intend that we J. 2. I intend that we J in accordance with and because of 1a and 1b, and meshing subplans of 1a and 1b; you intend the same. 3. 1 and 2 are common knowledge. Various counter-examples have been raised to question the adequacy of this account. Consider, for instance, competitive contexts. Although we play a game of chess together, it is not clear that our subplans must mesh. Indeed, it would seem that in many cases they do not mesh because we share distinct goals vis-à-vis winning (Bratman, 1999). Consider, also, joint actions that occur on the fly, so to speak. If a group of people spontaneously run to rescue a person from a burning car, it seems implausible that it involves the sort of complex of attitudes Bratman suggests. Margaret Gilbert has argued that at the core of our notion of sociality (or doing things together in the robust sense) is the concept of a plural subject of belief or action. A plural subject is an entity, or as Gilbert puts it, “a special kind of thing, a ‘synthesis sui generis’ ” (1996, 268) formed

FLAVORS OF

“ TOGETHERNESS ” 

when individuals bond or unite in a particular way. This “special kind of thing” can be the subject to which intentional action and psychological attributes are attributed. We can formulate the conceptually necessary and sufficient conditions for the existence of plural subjects in the following way: Individuals A1 . . . An . . . form a plural subject of X-ing (for some action X or psychological attribute X) if and only if A1 . . . An form a joint commitment to X-ing as a body.

A joint commitment to act as a body is a commitment made by a collection of individuals to perform some present or future action as would a single individual. Joint commitments are formed when each of a number of people expresses his or her willingness to participate in the relevant joint commitment with the others. Each person understands that only when all of the relevant people have agreed to participate in the joint commitment will the joint commitment be formed. Once everyone has agreed, a pool of wills is formed and individuals are then jointly committed. Once the joint commitment is established, each individual is individually obligated to do his or her part to make it the case that he or she acts as part of a body. Consider a case in which a soccer team agrees to play a scrimmage match against another league team. The members of the team do not each individually agree to play a soccer match. If they did, it would lead to a proliferation of soccer matches. Each member, however, agrees to make it the case that they together play soccer and express their willingness to do so on the condition that every other member do the same. This expression of willingness need not be simultaneous. The members may express their willingness over time. Nor do they need to express their willingness verbally. In many cases, silence is an adequate expression of intention. They must, however, in order for the joint commitment to come into existence, communicate in some way and at some point in time their intention to do their part in playing the scrimmage game as a body (or a team) with others. Gilbert’s account has also been subject to a number of counterexamples. Bratman (1999), for instance, has argued against the sufficiency of such an account. Imagine a mobster hit-man who points a gun at the head of his rival and tells him that they are going to take “a little ride” together. No doubt the rival will express his willingness to go, if only to



D . TOLLEFSEN ET AL .

delay the inevitable. And, of course, the mobster has already expressed his willingness. But surely such an expression of willingness to go for “a little ride” together does not constitute a joint action on the part of the mobster and his rival. We would not ordinarily say that they were traveling together or doing anything together. Would we? Bratman simply assumes the answer to this is no. Our discussion so far reflects, in many ways, the analysis and criticism that takes place in the area of collective intentionality. In many circumstances, the discussion revolves around particular examples that are generated to test analyses—washing the dishes, playing a game, constructing a house, mobster hit-men going for a ride—and these are supposed to reveal “common” intuitions and support or attack the necessity and sufficiency of certain features. In all cases, however, the appeal to common intuitions is done from the philosopher’s armchair. Indeed, though the philosopher will often reference “common” intuitions, there has been no attempt to find out if the postulated intuitions are actually the intuitions of the average person, or simply those of the philosopher him- or herself.2 In addition, there have been no explorations of the underlying psychological mechanisms that would constitute or influence such intuitions. The strand of experimental philosophy that we adhere to takes seriously the idea that common intuitions are relevant to philosophy, though they serve a different purpose than those found in traditional philosophical analyses. Knobe and Nichols (2007) provide the following description of the methodology. Typically, one starts out with a fairly superficial characterization of certain patterns in people’s intuitions [patterns revealed through experiments]. Maybe something like this: People are more inclined to regard an agent as morally responsible when the case is described in vivid and concrete detail than they are when the case is described more abstractly. 2

Although those working in collective intentionality have not appreciated the role of the intuitions of the common folk, there have been a number of recent articles in which experimental philosophers have explored the practice of attributing mental states to groups (Phelan, Arico, and Nichols, forthcoming; Knobe and Prinz, 2008). The focus of these experiments has been on people’s willingness (or lack of willingness in some cases) to attribute mental state attributions to groups. We have drawn inspiration from these articles, but our focus here is not on the attribution of mental states to groups but rather on the way that individuals conceptualize joint actions.

FLAVORS OF

“ TOGETHERNESS ” 

The goal, however, is to provide some deeper explanation of why the intuitions come out this way. For example: People are more inclined to regard an agent as morally responsible when they have strong affective reaction to his or her transgressions. And ultimately, the hope is that one will be able to arrive at a more fundamental understanding of people’s thinking in the relevant domain. Maybe something like this: People’s intuitions about moral responsibility are shaped by the interaction of two different systems—one that employs an abstract theory, another that relies on more immediate affective reactions. (Knobe and Nichols, 2007, p. 5)

Experimental philosophy is often accused of making philosophy into a popularity contest: poll the common people, and whatever they say about the concept of knowledge, belief, moral responsibility, and so on is the correct account of that concept. Such a characterization is important to address, but it does miss an important point about experimental philosophy: it is not pitting common intuitions against the philosopher’s intuitions. Rather, the data reveal something about the source of common intuitions and the context in which they arise. The philosopher’s intuitions and analyses can be richly informed by the common origin and flexibility of relevant concepts, and when integrated they form a new foundation for advancing philosophical theory. Experimental philosophy— at least of the type under consideration here—does not want to replace the philosopher with the “man on the street.” Rather, the experimental philosopher is attempting to give philosophers more data to consider when building their theories. It is not that actual percentages themselves are supposed to directly impact our philosophical inquiries. Rather, the idea is that these experimental results can have a kind of indirect impact. First, we use experimental results to develop a theory about the underlying psychological processes that generate people’s intuitions; then we use our theory about the psychological processes to determine whether or not those intuitions are warranted. (Knobe and Nichols, 2007, p. 6)

In what follows we provide some of our data on people’s judgments regarding examples of joint action (examples found within the literature) and reflect on what it might contribute to philosophical inquiries about the nature of joint action.



D . TOLLEFSEN ET AL .

Study 1: “Togetherness” as a Conceptual Feature As we have already noted, numerous examples of individual or joint tasks have been used in the philosophical debate on collective intentionality. Our goal in this chapter is to use experimental methods to test whether naïve participants perceive the “jointness” of these canonical cases. This problem can be cast as a question of conceptual structure, explored in numerous domains in cognitive psychology (e.g., Rips, Shoben, and Smith, 1973). In this area of research, we measure participants’ intuitions about some subset of canonical tasks to assess conceptual or semantic dimensions that underlie these intuitions. For example, if participants compared a set of tasks such as playing poker, dancing a tango, and reading a book (alone), then participants would be expected, over many trials, to evaluate poker and tango as more similar along a target cognitive dimension, namely, that they are joint actions. In order to explore people’s intuitions about joint actions, we employed a paradigm that has been used by experimental psychologists for several decades. Multidimensional scaling (MDS) provides a way of identifying underlying dimensions that affect how people think about concepts. MDS was developed by Shepard (1962) and Kruskal (1964) as a data analysis method based on a geometric model. Concepts are represented as points in a semantic space, and the distance between points reflects the similarity of concepts that they represent. Specifically, MDS attempts to find a best spatial representation that fits all pairs of concepts simultaneously. Although any number of dimensions can be used, solutions are typically mapped onto a two- or three-dimensional space. As an example, consider thinking about the relative distance between a number of cities. Most people would judge New York and Denver as relatively far apart, and New York and Boston as relatively close together. When research participants are asked to make their judgments on a distance scale, and are asked to rate all pairs of a number of cities, the resulting data can be plotted as a two-dimensional solution which captures the best fit of all the rated pairs. When this is done, a rough map of the United States emerges, and the two underlying dimensions can be identified as “east-westness” and “north-southness.” Psychologists have used this method to explore more complex concepts, such as the concept “animal.” When asked to provide ratings for all pairs of a

FLAVORS OF

“ TOGETHERNESS ” 

number of exemplars, a two-dimensional solution capturing “size” and “ferocity” emerges (Henley, 1969). Although the interpretation of the underlying dimensions can be rather subjective (see, for example, Jaworska and ChupetlovskaAnastasova, 2009), it provides a useful starting point in determining whether people spontaneously think about actions in terms of “jointness.” One of the benefits of this method is that it reveals ways people are thinking that are not easily articulated by the subjects themselves. We find this a particularly interesting method for use in the area of collective intentionality. The terms “joint commitment,” “we-intention,” “shared intention” are technical terms. It is unlikely that the ordinary person would say that these concepts play a role in their discrimination of certain types of action or in their own social interactions. This makes the philosopher’s burden even heavier as they may be accused of creating concepts rather than exploring or explaining them. MDS, however, helps to reveal the underlying dimensions that inform, at an unconscious level, how people think about concepts. As we shall see, the data we collected using MDS suggest that there is a “unique flavor” of togetherness that subjects seem to be picking up on. We asked a group of research participants to provide similarity ratings for ten activities that people engage in together. As a first experiment, we wished the first set of instructions to point directly to our target dimension, expecting that such instructions should readily bring about a multidimensional scaling solution with a dimension of “jointness” as a conceptual feature. We included three pairs of related activities that are typically joint or solo actions (“having a conversation” and “talking to oneself”; “singing a duet” and “singing a solo”; and “playing poker” and “playing solitaire”). We also included two other activities that are typically thought of as joint actions (“dancing the tango” and “having sex”), and two activities typically thought of as solo in nature (“driving a car” and “reading a book”). The participants in the study were recruited via Amazon’s Mechanical Turk , a crowdsourcing service that allows “requesters” to recruit “workers” for a variety of online tasks. Participants are paid for their responses, with Amazon functioning as the intermediary between the requester and the workers. Data collected in this way tend to be as reliable as data collected face-to-face from college undergraduates (Munro et al., 2010).



D . TOLLEFSEN ET AL .

We had twenty participants3 make forty-five similarity judgments between all pairs of ten activities described above. To emphasize the concept of “jointness,” the web page viewed by the participants was entitled “Doing Things Together,” and in the instructions they were asked to “rate the similarity of activities that people engage in together.” Ratings were made on a seven-point scale, with endpoints labeled “not at all similar” and “very similar.” The participants were encouraged to look over the first few items before they began to get a sense of the kinds of activity they would be comparing. They were also told that that there were no “right” or “wrong” answers, and that we were simply interested in their intuitions. The data were analyzed using the ALSCAL algorithm in SPSS, with ordinal measurement and Euclidian distance. With this multidimensional scaling method, a two-dimensional solution of participant judgments explained 85.8% of the variance in their responses. If people spontaneously use “jointness” in making judgments about the similarity of activities, then we would expect to see that the stimuli lie along one of the two MDS dimensions in a corresponding manner. Dimension 1 shows just such a distribution. The five joint activities (having sex, dancing the tango, singing a duet, having a conversation, and playing poker) all fall at or below the midpoint on the left side of the solution. Four of the five solo activities (playing solitaire, reading a book, talking to oneself, singing a solo) fall above the midpoint on the right side of the solution, and the fifth, driving a car, is just below the midpoint. It is important to note that the exact coordinates in this multidimensional solution are not meaningful, but only relative coordinates between our items; researchers often rotate these solutions to obtain coordinates that fit more with intuition. In fact, researchers often, for convenience and visualization, label the axes. We have done this in Figure 9.1, for illustration. Within Dimension 1 there is also evidence that people’s judgments reflect a sensitivity to conditions of physical proximity. Those joint actions that require physical touching (having sex and dancing the tango) rather than mere co-location (singing a duet, having a conversation) 3 The participants consisted of fourteen females and five males (and one of unreported gender) living in the United States, and their average age was 36 years. They required an average of 6.5 minutes to complete the task.

FLAVORS OF

“ TOGETHERNESS ” 

1.5 Playing poker 1.0

Playing solitaire

0.5 Dimension 2

Having sex

Having a conversation

0 Dancing the tango

Singing a duet Reading a book

–0.5

Talking to oneself

–1.0 Singing a solo –1.5 –2.5

Driving a car –2.0

–1.5

–1.0

–0.5

0

0.5

1.0

1.5

2.0

2.5

Dimension 1 (jointness)

Figure 9.1 Multidimensional scaling and jointness: Study 1.

are further from the midpoint. Likewise, on the right-hand side the solo actions can be seen as falling according to those that are more or less physically isolating. Playing solitaire and reading a book are actions that are likely to be more physically isolating, whereas singing a solo is usually done in the company of others (an audience). What about Dimension 2? If our participants were thinking about the content of the activities, then we would expect to see clustering based on the three pairs of activities that involved similar actions (i.e., singing, talking, and playing cards). However, an inspection of Figure 9.1 shows that the two card-playing activities are plotted far apart from each other in the upper right quadrant (although both are high on Dimension 2). Singing a duet and singing a solo are plotted in different quadrants (and also vary considerably on Dimension 2). Finally, having a conversation and talking to oneself are also in different quadrants, and appear toward



D . TOLLEFSEN ET AL .

the middle of Dimension 2. Though there are hints of this sensitivity to content, the multidimensional scaling solution did not reveal it as clearly as one would expect. Our results suggest that experimental philosophers can make use of techniques such as MDS and clustering to explore the dimensions that people spontaneously make use of when thinking about members of a category—in this case, different types of human activity. These results are supportive of the intuition that “jointness” plays a role in the conceptualization of activities. In the experiment described next, we wished to test the robustness of this conceptual dimension. Would participants continue to reveal a conceptual structure that had jointness as a dimension if we changed the instructions to be more indirect about this aspect of our stimuli?

Study 2: “Togetherness” Without Prompting? Once again, we employed crowdsourcing to obtain intuitions about the similarity of various activities. As in Study 1, we recruited participants4 via Amazon’s Mechanical Turk service. The participants provided similarity ratings for the same ten activities used in Study 1, but in this case the web page bore a neutral title, “Evaluating Activities,” and they were asked to “rate the similarity of activities that people engage in.” The rest of the instructions and the procedure were identical to Study 1. As before, the data were analyzed using the ALSCAL algorithm in SPSS, with ordinal measurement and Euclidian distance. As in the case of Study 1, a two-dimensional solution of participant judgments explained 85.9% of the variance in their responses. As can be seen by comparing Figures 9.1 and 9.2, the results of both studies are highly similar. Once again, Dimension 1 can be explained in terms of “jointness.” The five joint activities (having sex, dancing the tango, singing a duet, having a conversation, and paying poker) all fall below the midpoint on the left side of the solution. All five of the solo activities (playing solitaire, reading a book, talking to oneself, singing a

4 The participants consisted of eighteen females and five males (and two of unreported gender) living in the United States, and their average age was 40 years. They required an average of 4 minutes to complete the task.

FLAVORS OF

“ TOGETHERNESS ” 

1.5

1.0

Playing poker Playing solitaire

0.5 Dimension 2

Reading a book

Having a conversation

Having sex

0

–0.5

Talking to oneself

Dancing the tango Singing a duet

Singing a solo

–1.0 Driving a car –1.5 –2.5

–2.0

–1.5

–1.0

–0.5

0

0.5

1.0

1.5

2.0

2.5

Dimension 1 (jointness)

Figure 9.2 Multidimensional scaling and jointness: Study 2.

solo, and driving a car) fall above the midpoint on the right side of the solution. Also as in Study 1, it is clear that participants are not making judgments based on the content of the activities. For all three pairs involving singing, talking, and playing cards, the joint activities (singing a duet, having a conversation, and playing poker) are in different quadrants than the solo activities (singing a solo, talking to oneself, and playing solitaire). This is a rather surprising result. The differences and similarities between types of action would seem to be a salient feature, but our results suggest that whether the action was done by a single individual or multiple individuals is more salient. The agent (who did it or does it) and not the action (what is done) seems to be a more salient feature for categorization (and this is so even when subjects are not prompted to categorize solo vs. joint actions).



D . TOLLEFSEN ET AL .

Do the data from Study 2 provide any further insight into Dimension 2? One reviewer has suggested that subjects may be categorizing the activities in terms of the degree to which a successful performance depends upon an individual’s performance. Having a successful conversation and successful sex (whatever that might mean) seems to depend on others in a significant way, whereas talking to oneself and singing a solo is under an individual’s control and so its success is determined by the individual alone. Unfortunately, this does not seem to fit the general layout of Dimension 2. Successful sex does seem to require joint control/ responsibility, but so too do dancing the tango and singing a duet, and yet they are in the bottom quadrants of Dimension 2. Again, our current data do not reveal a salient interpretation of Dimension 2, but reflection on this dimension has suggested a number of other questions for further research. For instance, actions might be categorized in terms of whether they are a means to an end or an end in themselves. Dancing, for instance, might be an end in itself, whereas driving a car is a means to some further end. Are intuitions sensitive to this distinction? There are some activities that are stereotypically solitary—not just done by an individual but done alone. They indicate a sort of antisocial behavior—solitaire and talking to oneself might be such cases. Are people making distinctions between doing things alone and doing things with other people present but individually? Another way of analyzing these data is to employ hierarchical clustering, and this method suggests that the distinction between being alone and completing an action alone may be playing a role in people’s judgments. In this method, each score begins as its own cluster, and then pairs and larger groupings emerge until all the data points are represented (Johnson, 1967). The clusters can be displayed in a dendrogram but can also be described verbally. For the set of six solo and joint action terms, and using a single linkage method, “talking to oneself” and “playing solitaire” emerge as the first cluster. We might characterize these actions as antisocial or solipsistic. A second cluster consists of these two terms along with “singing a solo.” So the three solo activities can be described as clustering together and categorize actions which are done by an individual but are not necessarily antisocial. For the joint actions, “having a conversation” and “singing a duet” cluster together, so two of these also behave in the expected manner. The final action,

FLAVORS OF

“ TOGETHERNESS ” 

“playing poker,” only joins the other items when a single cluster is created. This is probably because playing poker is somewhat different from the other activities, which are cooperative in nature. Even though cooperation is required to play the game, the activity is ultimately competitive, and this may explain why the clustering algorithm does not place this activity with the others. Varying the examples along cooperation vs. coordination, and cooperative vs. competitive contexts, may also reveal the nature of the second dimension. The results of Study 2 make clear that the concept of “jointness” is one that people spontaneously use in thinking about the similarity of activities. Whereas the results from Study 1 could be explained by the fact that we explicitly invoked the idea of “doing things together,” no such explanation is available for the data from Study 2. In the absence of an explicit prompt, and with very neutral instructions (“evaluating activities”), our participants’ intuitions about these activities look very much like the intuitions of others for whom “jointness” was explicitly mentioned. It remains to be seen, however, whether the concept of jointness that the folk use is anything like that of the philosopher. Although clearly people are tracking whether the action is done by an individual alone or jointly (together with someone else) and not the content of the action, the conception of togetherness and aloneness might vary along a number of lines—including physical proximity, how antisocial the action is, and whether the action could be done by more than two people.

Study 3: “Togetherness” Without Proximity? A reviewer of this chapter astutely noted that the items we used for Studies 1 and 2 involved a confound of physical proximity: joint activities correlated almost perfectly with activities that were done in close physical proximity. It is entirely plausible that this may be a criterion for joint action, at least of the canonical. Nevertheless, joint activities may be carried out without close proximitiy, such as conversation or Internet chatting; and solo activities can be done in close proximity, such as standing in a line or riding on a subway. We modified our items and pursued one final study to test whether jointness emerges out of item



D . TOLLEFSEN ET AL .

2.0 Riding a busy subway car

1.6 Driving with one’s family

1.2

Standing in line

Dimension 2

0.8 Dancing the tango

0.4 0

Daydreaming in class

Talking on the phone

–0.4

Reading a book

–0.8 Chatting on the Internet

–1.2

Playing solitaire

–1.6

Playing poker

–2.0 –2.5

–2.0

–1.5

–1.0

–0.5

0

0.5

1.0

1.5

2.0

2.5

Dimension 1 (jointness)

Figure 9.3 Multidimensional scaling and jointness: Study 3.

comparisons, and ran a similar multidimensional scaling solution.5 Items and results are shown in Figure 9.3. Again, a clear pattern emerges across Dimension 1. Activities on the left-hand side of the graph are those done with others and solo activities fell to the right-hand side. Importantly, this pattern holds even for solo activities that are in physical interpersonal proximity (e.g. riding on a subway car) and joint activities done at a distance (chatting on the Internet or phone). Jointness emerges as the dominant feature by which subjects categorized actions. Subjects seem to be tracking the psychological connectivity of actors rather than simply physical

5 The participants consisted of twelve females and sixteen males living in the United States, and their average age was 37.25 years. They required an average of 4.5 minutes to complete the task.

FLAVORS OF

“ TOGETHERNESS ” 

proximity. This is broadly supportive of philosophical theories of joint action that identify a special “flavor” of togetherness.

Philosophical Applications What implications do these results have for philosophical analyses of joint action? One of the major debates concerning the nature of joint action is whether an analysis of joint action can be given that does not appeal to something irreducibly joint. Consider, for instance, Bratman’s analysis in terms of the intentions of individuals and their mutual interdependence. This analysis attempts to reduce joint intention to individual intentions which then can, it is assumed, be understood apart from anything joint. The benefit of such an account is that it avoids a charge of circularity, as one can analyze the notion of joint action and intention in terms that do not themselves appeal to notions such as “intending together.” But attempts at such reductions are often challenged. According to Searle (1990), for instance, collective intentionality is a primitive concept and cannot be reduced to a set of individual intentions. Individual intentions to perform one’s part in a joint action are formed on the basis of we-intentions and not the other way around. We-intentions are “primitive” and cannot be analyzed in terms of something more basic such as individual intentional states. Reference to a primitive notion of “doing things together” arises in the work of Margaret Gilbert as well. I do, of course, posit a mechanism for the construction of social groups (plural subjects of belief or action). And this mechanism can only work if everyone involved has a grasp of a subtle conceptual scheme, the conceptual scheme of plural subjects. Given that all have this concept, then the basic means for bringing plural subject-hood into being is at their disposal. All that anyone has to do is to openly manifest his willingness to be part of a plural subject of some particular attribute. (1989, p. 416)

The formation of plural subjects requires that the members have a basic or primitive concept of doing things together. The data we introduce here suggest that ordinary folk do indeed have a subtle concept of doing things together and that this concept plays a role in their ability to categorize activities. Far from being a figment of the philosopher’s intuitions, the concept of acting jointly seems to exhibit



D . TOLLEFSEN ET AL .

itself in simple comparison tasks performed by the ordinary person. Though our results are not conclusive, they do suggest that philosophical analyses that appeal to primitive understandings of working together or intending together are not psychologically implausible, and so reference to such primitive concepts in philosophical analyses are actually important in understanding the nature of joint action. It should be noted, however, that our studies fail to identify the nature of the psychological connectivity that subjects seem to track. That is, nothing about our studies reveals whether subjects identify the presence of we-intentions (Searle) or shared intentions (Bratman) or joint commitments (Gilbert) or whether such concepts are being used in categorization. There are various ways in which the paradigm we introduce might be used to explore the boundaries of the concept of “jointness.” For instance, variation in the types of joint activity—those involving competitive contexts as opposed to cooperative contexts—might yield insights into whether and to what extent people see competitive contexts as genuinely joint. Philosophical analyses of competitive contexts seem to call for something less stringent than Bratman’s meshing of subplans, yet any competitive game requires some sort of collaboration. Our results provide hints of this, as the only competitive task (poker) seemed to behave differently in our cluster analysis. We also might develop experiments that tease out the second dimension along which people seem to be making comparisons. The extent to which the action is joint is a clear dimension that appears in our data both when subjects are prompted and when they are not, and in contexts of physical proximity and no physical proximity. But the second dimension remains unclear. If we vary the content of the action, the extent to which the action requires ongoing interaction, and the extent to which it requires an organizational context, we might be able to identify this dimension more clearly. Future investigations could uncover other flavors of joint activity. Our results are inevitably preliminary. This application of MDS, in particular, requires getting participants to rate the similarity of every activity pair. This limits the range of stimuli that could be used in our design. For example, with twenty activities we would require 190 comparisons; thirty activities would require 435. There are ways of getting around these and other issues (e.g., Ramsay, 1982; Ashby et al., 1994), which may allow future investigations to explore a wider range of activities. Our goal here, though, was to use the MDS approach “out of the box” and to

FLAVORS OF

“ TOGETHERNESS ” 

showcase the stability of the concept of joint action in some canonical cases. While these results are a first step in this direction, we hope to have shown that experimental philosophy can contribute in a positive way to philosophical theories of joint action. Not only do our results seem to suggest that naïve subjects are working with a basic notion of joint action, they also suggest that the jointness of an action is particularly salient such that comparisons are made with respect to this feature rather than with respect to the content of the action. Philosophical theories of joint action and intention that make reference to a certain distinct “flavor” of jointness or to a primitive concept of doing things together might find that common folk’s intuitions pick up on this distinct flavor as well. Because an activity’s participants are what sustain it, understanding how these participants conceptualize the activity will shed light on the social stability, or evanescence, of the activity itself.

References Ashby, F.G., Maddox, W.T., and Lee, W.W. (1994). On the Dangers of Averaging Across Subjects when using Multidimensional Scaling or the Similarity-Choice Model. Psychological Science 5(3): 144. Bratman, M. (1999). Faces of Intention. Cambridge, MA: Cambridge University Press. Gilbert, M. (1989). On Social Facts. New York: Routledge. Henley, N.M. (1969). A Psychological Study of the Semantics of Animal Terms. Journal of Verbal Learning and Verbal Behavior 8: 176–84. Jaworska, N., and Chupetlovska-Anastasova, A. (2009). A Review of Multidimensional Scaling (MDS) and its Utility in Various Psychological Domains. Tutorials in Quantitative Methods for Psychology 5: 1–10. Johnson, S.C. (1967). Hierarchical Clustering Schemes. Psychometrika 32: 241–54. Knobe, J. and Nichols, S. (2007). An Experimental Philosophy Manifesto. In Knobe, J., and Nichols, S. (eds.), Experimental Philosophy. Oxford: Oxford University Press. Knobe, J. and Prinz, J. (2008). Intuitions about Consciousness: Experimental Studies. Phenomenology and the Cognitive Sciences 7: 67–85. Kruskal, J.B. (1964). Multidimensional Scaling by Optimizing Goodness of Fit to a Nonmetric Hypothesis. Psychometrika 29: 1–27. Munro, R., Bethard, S., Kuperman, V., Lai, V., Melnick, R., Potts, C., Schnoebelen, T., and Tily, H. (2010). Crowdsourcing and Language Studies: The New



D . TOLLEFSEN ET AL .

Generation of Linguistic Data. Proceedings of the NAACL HLT Workshop on Creating Speech and Language Data with Amazon’s Mechanical Turk, 122–30. Phelan, M., Arico, A., and Nichols, S. (forthcoming). Thinking Things and Feeling Things. Phenomenology and Cognitive Science. Ramsay, J.O. (1982). Some Statistical Approaches to Multidimensional Scaling Data. Journal of the Royal Statistical Society. Series A (General), 145: 285–312. Rips, L.J., Shoben, E.J., and Smith, E.E. (1973). Semantic Distance and the Verification of Semantic Relations. Journal of Verbal Learning and Verbal Behavior 12(1): 1–20. Searle, J. (1990). Collective Intentions and Actions. In P. Cohen, J. Morgan, and M.E. Pollack (eds.), Intentions in Communication. Cambridge, MA: Bradford Books, MIT Press. Shepard, R.N. (1962). The Analysis of Proximities: Multidimensional Scaling with an Unknown Distance Function. Part I. Psychometrika 27: 125–40.

10 The Good in Happiness Jonathan Phillips, Yale University Sven Nyholm, Universität zu Köln Shen-yi Liao, Nanyang Technological University

The study of happiness has recently benefited from a wealth of scientific research. One important development in this research has been an emerging consensus on the definition of happiness employed by scientific researchers across a number of different fields. This scientific definition of happiness concerns three core psychological components: high positive affect, low negative affect, and high life-satisfaction (for a review, see Diener et al. 2003; Lucas et al. 2003). On this definition of happiness, whether or not a person is happy depends solely on that person’s psychological states. In contrast to this definition of happiness, a tradition in philosophy has long held that whether or not a person is happy actually depends in large part on whether the person is living a life that is good. An early form of this view can be found in Aristotle (NE 1097b22–1098a20), and it is supported by several contemporary philosophers as well. What both ancient and contemporary versions of these theories have in common is that they hold that happiness is more than just a matter of which psychological states a person experiences. If a person does not have a life that is good, she will not truly be happy, no matter what she may subjectively experience. Given the stark contrast between these two conceptions of happiness, one question that may naturally arise is how people ordinarily understand happiness when they lack the theoretical commitments that exist on both sides of this debate. Is the ordinary understanding concerned For their incredibly helpful contributions to this chapter, the authors would like to acknowledge Josh Knobe, Stephen Stich, Antti Kauppinen, Allan Gibbard, Jason Konek, Steve Campbell, Warren Herold, Randy Nesse, Valerie Tiberius, and Dan Haybron.



J . PHILLIPS ET AL .

with only the psychological states a person experiences in accordance with the scientific definition, or does it place an emphasis on the value of the life the person is living?1 This is the question we attempt to answer. We begin by briefly reviewing some key ideas representative of the philosophical tradition that understands happiness as essentially normative or evaluative. Then we consider the anti-evaluative counter-reaction to this tradition among both philosophers and psychologists who propose to understand happiness in a purely descriptive, psychological way. Turning from these philosophical and psychological accounts, we present six studies that investigate how people ordinarily make judgments of happiness. Five of these studies are reported for the first time, and we summarize one additional study (Study 3) that has been previously published (Phillips et al. 2011). Together, these studies suggest that the ordinary understanding of happiness actually combines both evaluative and descriptive components while differing in important respects from both the evaluative and descriptive theories of happiness that have been previously proposed. Very roughly put, our analysis is that happiness, as it is ordinarily understood, consists in experiencing positive mental states when one ought to experience them. After proposing a general theoretical account of the ordinary understanding of happiness, we end by situating this research within the contemporary philosophical and psychological literature.

The Philosophy of Happiness Philosophy bears a long history of arguments over whether there is anything more to happiness than enjoyable psychological states. On one side, some philosophers (much like contemporary psychologists) have held that all there is to happiness is the possession of certain 1 By “ordinary” we mean to pick out those uses of the word “happiness” that are not used in a technical way. One could stipulate a technical sense of “happiness” to refer to just the psychological states or just the value of a person’s life (as psychologists and philosophers have done), but those uses are not particularly interesting for our purposes. One could, after all, stipulatively use the term “happiness” to refer to almost anything. What we are interested in is the ordinary, non-stipulative use of the word “happiness.” It is, of course, possible that such concepts admit of variation between individuals, cultures, and languages. For our purposes, though, we will set this variance aside, and consider the English word “happiness” as a case study. We take up the question of whether there may be more than one single understanding of “happiness” in Section V.

THE GOOD IN HAPPINESS



psychological states, e.g. pleasure or satisfaction. This position can be thought of as descriptive; it suggests that questions of whether a person is happy can be settled by simply considering an accurate description of the psychological states a person experiences. If the person has the requisite psychological states, then that person is happy. On the other side of this debate, philosophers have also maintained that being happy involves living a life that is normatively good. This second position can be thought of as evaluative; it suggests that questions about whether a person is happy cannot be settled with just a description of the persons’ experiences or life—the important issue is whether that life is genuinely good. If the person’s life is not good, then that person will not truly be happy, no matter how pleasant her psychological states may be. A number of different versions of both of these families of views have been proposed, and it will be helpful to get a rough grasp on how they relate to and differ from one another. At the outset, theories of happiness should be distinguished from theories of well-being. Theories of well-being are concerned with what is ultimately best for a person in a non-instrumental way. Thus, hedonist theories argue that what is non-instrumentally best for a person is some form of hedonic pleasure; objective list theories argue that there is some objective list of what is non-instrumentally best for a person and argue over what should be included on this list, and so on. As such, theories of well-being need not be directly concerned with what happiness is. Accordingly, we focus the remainder of our discussion specifically on theories of happiness, as these theories do directly concern what happiness is.2

Evaluative happiness An early example of a view on which happiness requires more than the possession of particular psychological states can be found in Plato’s Gorgias where Socrates mocks Callicles for believing that the mere satisfaction of desires is happiness. Offering a pointed counter-example, Socrates asks Callicles if he thinks that one could become happy just by

2 It is worth noting that the one way in which theories of well-being may directly concern happiness is if the theory of well-being argues that what is non-instrumentally best for a person is to be happy. This position is clearly compatible with any theory of happiness.



J . PHILLIPS ET AL .

continually scratching an itch (494c–e). Similarly, Aristotle argues that only the most vulgar people suppose happiness to be pleasure—the type of people who would choose lives better suited for grazing cattle (NE 1095b14–1095b21)—and goes on to argue that happiness consists in living a virtuous life (NE 1097b22–1098a20).3 While Plato and Aristotle mark out positions which emphasize virtue, scholars have disputed whether these ancient philosophers were actually concerned with happiness, suggesting that the term eudaimonia would often be better translated as “flourishing” or “well-being” (Cooper 1987). Yet we do not need to look to the ancient debate to find this position; it can also be found among contemporary philosophers who have been quite clear that they are concerned with happiness. To take one example, Philippa Foot has argued that “when we talk about a happiness that is supposed to be a human good, we cannot intend pleasure or contentment alone.” Instead, she continues, “there is a kind of happiness which only goodness can achieve” (2001). In support of this view, Foot asks somewhat incredulously whether a Nazi war criminal who escaped to Brazil after overseeing a death camp could really be happy (2001, 90–96). Her answer, while nuanced, is that it should be clear that he is not truly happy. In another example meant to illustrate the evaluative nature of happiness, Richard Kraut imagines a man who wants most to be loved and admired by his friends (1979). Upon learning about his wishes, his friends collude to trick him into believing that they love and admire him, even though they do not. The man comes to believe that his greatest wishes have been fulfilled, and experiences great pleasure and satisfaction with his life. Considering whether this man is happy, Kraut’s response is as unambiguous as Foot’s.4 They both argue that happiness must be more than the mere sum of a person’s psychological states.

3

There is a question of whether Aristotle’s continued discussion of happiness in Book X of Nicomachean Ethics suggests that the happy life is not necessarily a morally virtuous life (Cooper 1987), but the discussion of that question is outside the purview of this chapter. For our purposes we will identify a broadly “Aristotelian” conception of happiness with one that involves living a morally virtuous life. 4 One important difference between these two theories is that Kraut emphasizes the importance of the man’s own values, whereas Foot does not. However, both views are clearly evaluative in the relevant sense.

THE GOOD IN HAPPINESS



Regardless of the merits of any one of these positions, what holds this diverse group of theories together is the criteria they use to determine whether or not a person is happy: they require an answer to the question of whether that person’s life is good.

Descriptive happiness In contrast, other philosophers have pursued the study of happiness as a purely descriptive project. In this tradition, normative questions about the value of a person’s life are simply irrelevant to whether or not a person is happy. Among the ancients, we can also find views not very dissimilar to this. Consider the Cyrenaics, who held that happiness among other goods could be found only in pleasure itself, even when it was gained at the expense of others’ welfare. In an amusing anecdote meant to illustrate this position, Aristippus and Plato were said to be at the court of Dionysius, who suggested that they amuse themselves by putting on women’s clothing. In the heedless pursuit of pleasure, Aristippus readily humiliated himself, while Plato was predictably much less fun (Giannantoni frs. 25–43).5 In the more contemporary literature on happiness, there are three main families of descriptive theories: hedonist theories, life-satisfaction theories, and hybrid theories.6 Hedonist views of happiness suggest that one is happy if on balance one experiences more pleasure than pain (e.g. Feldman 2010).7 Life-satisfaction views, in contrast, focus on the attitude one has about one’s life as a whole, and equate happiness with a reflective endorsement of the life that one is leading (e.g. Kekes 1982; Suikkanen

5 Perhaps more cynically, Aristippus was also rumored to have spurned his own child, as he was of no use to him in attaining pleasure (Diogenes II 81). Aristippus was only loosely associated with the Cyrenaics and most likely did not contribute to their later written works, as he predated the school by a number of years. Regardless, the example was taken to be illustrative of the Cyrenaics’ position, even among the ancients (Annas 2011), though the truth of the story has also been disputed (Annas 1993). 6 A fourth family may be emotional-state views (Haybron 2008; 2010), though this family of views is relatively new. Like hedonist, life-satisfaction, and hybrid theories, emotional-state theories take a purely descriptive approach to happiness. 7 Fred Feldman’s “Attitudinal Hedonism about Happiness” is actually somewhat nuanced. On his theory, “to be happy at a time is to have a positive net balance of intrinsic attitudinal pleasure over intrinsic occurrent displeasure at that time,” or more simply, “to be happy at a moment is to be taking more pleasure than displeasure in things at that moment” (2010). Thus, whether one is happy is a descriptive psychological question which depends on whether one takes more pleasure than displeasure in things at a given moment.



J . PHILLIPS ET AL .

2011). And hybrid theories attempt to combine elements from both of these views (e.g. Sumner 1996; 2000). For our purposes, what is most relevant about these theories is not the differences between them but what they all have in common. All of these theories identify happiness with some psychological state or set of psychological states. They are all descriptive theories of happiness in the relevant sense because, according to them, we can determine whether or not a person is happy based only on an accurate description of the agent’s psychological states.

The Psychology of Happiness In contrast to the debate within philosophy, the psychological research on happiness has for the most part solely pursued the descriptive side of the project,8 seeking both to define happiness as a set of psychological states and to create methods for measuring whether or not people possess those psychological states (for a review, see Seligman and Csikszentmihalyi 2000). This understanding of happiness combines aspects of hedonic theories which emphasize pleasure and pain, and life-satisfaction theories which emphasize global attitudes about one’s life (Diener et al. 2003). There is quite a widespread consensus that happiness can be best captured by a purely descriptive theory, which is often made explicit (e.g., Gilbert 2006, 36–7). As Gilbert and others make clear, to know whether or not a person is happy one need do no more than come up with unbiased measures of that person’s psychological states, whether he be a mass murder or Mahatma Ghandi.9 In the next section we consider the ordinary understanding of happiness. Before doing so, though, it may be helpful to briefly step back and consider how these various positions relate to one another. Early on, we distinguished theories of happiness from theories of well-being, which are not concerned with the happiness of the agent but rather with what is non-instrumentally best for the agent. Then, focusing on

8

While a number of psychologists have been interested in well-being as opposed to happiness—e.g., Ryan and Deci’s self-determination theory (2000; Deci and Ryan 2012) and Ryff and Singer’s multidimensional account (1998)—few if any have offered genuinely evaluative theories of happiness. 9 For an exception which proves the rule, see Kashdan, Biswas-Diener and King (2008).

THE GOOD IN HAPPINESS



theories of happiness, we considered both evaluative and descriptive theories. Evaluative theories, on the one hand, hold that whether or not an agent is happy depends on whether or not the agent is living a normatively good life. Descriptive theories, on the other, hold that whether or not an agent is happy depends only on the particulars of the psychological states the agent experiences—though different descriptive theories choose to emphasize different psychological states.10 Importantly, these two views of happiness differ along two distinct dimensions. The first is that they differ over which aspect of an agent’s life they focus on when determining whether or not the agent is happy. Evaluative theories of happiness tend to emphasize the agent’s actual life rather than solely the agent’s psychological states when determining whether the agent is happy. Descriptive theories, in comparison, tend to focus solely on the agent’s psychological states. Secondly, these two views of happiness differ in the role they give to normative judgments. Notice that some form of normative judgment may actually play a part in both evaluative and descriptive theories. Evaluative theories of happiness take the relevant normative judgments to be about how the agent’s life is actually going, independent of anything the agent might think or believe. By contrast, descriptive theories of happiness hold that this sort of normative judgment is completely irrelevant to whether or not the agent is happy, as happiness can be determined by simply considering which psychological states the agent experiences. However, a certain sort of normative judgment may still play a role in descriptive theories in that some of the agent’s mental states may involve normative judgments. For example, some purely descriptive theories of happiness hold that an agent must be satisfied with her life to be happy, and this psychological state of life-satisfaction presumably involves a normative judgment that the agent makes about how her life is going. For our purposes, though, it is only evaluative theories of happiness that provide a genuine role for normative judgment, because on these theories one simply cannot determine how happy a person is without first determining whether or not they actually lead a good life. We return to this distinction between these two roles for normative judgments at the end of the chapter, and 10 Almost all further distinctions made by psychologists, e.g. Kahneman’s distinction between “experiencing self ” and “remembering self ” (Kahneman 2011), fall squarely within this framework as specific ways of working out descriptive theories of happiness.



J . PHILLIPS ET AL .

argue that it is key to fully grasping how the ordinary understanding of happiness fits within the long-standing debate over happiness.

Ordinary Happiness The amount of debate over the true nature of happiness is perhaps only rivaled by the even more striking consensus about what people ordinarily mean by the English term “happiness.” One finds the suggestion that “happiness” refers to a descriptive psychological condition not just among psychologists, but among philosophers as well. Fred Feldman effectively illustrates the widely held position: I think “happy” is used in the evaluative sense only by philosophers and others whose good linguistic instincts have been perverted by them . . . As a result of this, I am inclined to treat such uses of “happy” as semi-stipulative, semitechnical uses. It is not clear to me that ordinary speakers of English, completely innocent of philosophy, ever use “happy” to mean the same things as “having sufficiently high welfare.” (2010, 136)11

Indeed, there is no shortage of researchers who have made similar points (Cooper 1987; Crisp 2008; Feldman 2010; Haybron 2008; 2011; Sidgwick 1907/1966).12 Yet despite the widespread agreement about ordinary usage, there has been surprisingly little empirical research on the matter. As of this writing, we are currently aware of only one informal survey outside of the authors’ own research (Haybron 2008).13 We propose to fill this oft-overlooked lacuna with a series of systematic studies that actually investigate how people ordinarily understand happiness.

11 Feldman continues: “A final comment: the OED reminds us that we sometimes speak of ‘happy coincidences,’ where a happy coincidence is just a lucky, or fortunate coincidence. The OED also says that ‘happy’ can be used to mean ‘drunk.’ I am inclined to think that these claims may be right. Perhaps the word does have these two additional senses when used in ordinary English. But again, I am not aware of a decisive test that would put either of these claims beyond question.” 12 There are also a few rare examples of arguments against these claims: see, e.g., Dybikowski (1981) and Kraut (1979). 13 There are, however, now a number of research programs on the topic that have begun: e.g. Vonasch and Baumeister (2012).

THE GOOD IN HAPPINESS



Moral concepts Outside of philosophical and psychological arguments over happiness, there is some independent reason to think that the ordinary concept of happiness might involve normative evaluations. A growing body of research has demonstrated that moral judgments systematically influence how people ordinarily understand a number of concepts that were initially thought to be purely descriptive. To illustrate, consider how it is that people ordinarily go about determining whether or not someone values something. One view would be that people have a purely descriptive understanding of valuing in which they simply consider a description of the agent’s psychological states and decide whether any of those states count as the requisite pro-attitude (some positively valenced psychological state toward something under a particular description, e.g. thinking it good). An alternative view would be that people have an evaluative way of determining whether someone values something. On this sort of view, judgments of whether or not someone values something depend at least in part on whether the thing itself is genuinely good. Attempting to decide between these two possibilities, Knobe and Roedder (2009) asked participants to read a vignette about a person who held a particular psychological attitude toward the action of refraining from premarital sex and asked them whether the person valued that action. Surprisingly, participants’ answer to this question depended on whether or not they happened to believe that refraining from premarital sex was a good thing.14 Surprising or otherwise, the evidence in favor of evaluatively laden concepts does not stop there. To take just a few other examples, moral judgments seem to play a systematic role in the folk concepts of causation (Cushman et al. 2008; Hitchcock and Knobe 2009; Roxborough 14 In this study, participants were asked to read a vignette about a woman named Susan who did not think there was anything wrong with premarital sex but was still conflicted about whether or not she should engage in it. Eventually she decided not to have premarital sex. The researchers asked two different groups of participants, who had radically different views about whether refraining from premarital sex was a good thing, to read this vignette. On average, the participants they recruited from Washington Square Park in New York City did not value refraining from having premarital sex, but those they recruited from a Mormon bible study group did. When both groups were asked whether Susan valued refraining from having premarital sex, the participants from the Mormon bible study agreed that she did, while participants from Washington Square Park disagreed (Knobe and Roedder 2009).



J . PHILLIPS ET AL .

and Cumby 2009), deciding (Pettit and Knobe 2009), freedom (Phillips and Knobe 2009; Young and Phillips 2011), intentional action (Knobe 2003), knowledge (Beebe and Buckwalter 2010), and weakness of will (May and Holton 2012), and new discoveries are continually being made. Given the long-standing philosophical debate over happiness and the widespread impact of normative considerations on other folk concepts, a sustained empirical investigation into whether the ordinary concept of happiness is evaluative seems doubly promising.

Investigating “Happiness” We present six studies that attempt to gradually shed light on the ordinary understanding of happiness. The approach we take to studying the ordinary understanding of happiness involves presenting participants with cases that describe both a person’s life and his or her psychological states. Then, participants are asked for their intuitive assessments of whether that person is happy. By systematically varying features of the person’s life and psychological states, we investigate what is relevant to ordinary judgments of happiness.15 Using this method, six studies provide evidence that the ordinary understanding of happiness reflects aspects of both evaluative and descriptive views. Similar to evaluative views, normative judgments have a substantive role in the ordinary understanding of happiness. Yet, similar to descriptive views, the ordinary understanding is focused on the person’s psychological states and not the overall life they have actually lived. Combining these two aspects, we offer a theory of ordinary happiness on which happiness depends on the normative status of an agent’s positive psychological states.

15 Importantly, this method does not simply ask participants for their explicit theories of happiness (e.g., “Could you be happy if you were not truly leading a good life?”). The approach we take has multiple benefits over the more explicit method. We take it that what is most relevant to the discussion of happiness is not people’s opinion about theories about happiness, but rather how they ordinarily understand happiness. Moreover, we think there is very little reason to suppose that people have good epistemic access to the processes by which they form judgments of happiness. Thus, the approach we take allows us to derive a sense of what is relevant to the ordinary understanding of happiness even when people themselves may not be aware of it. In fact, we end up proposing a theory of the ordinary understanding of happiness that many people may actually find counterintuitive!

THE GOOD IN HAPPINESS



Study 1: Normative evaluation influences judgments of happiness This research began with a study in which participants were told about an agent named Richard. In one condition, Richard was described as a Red Cross field doctor who had both positive and negative psychological states. Richard is a doctor working in a Red Cross field hospital, overseeing and carrying out medical treatment of victims of an ongoing war. He sometimes gets pleasure from this, but equally often the deaths and human suffering get to him and upset him. However, Richard is convinced that this is an important and crucial thing he has to do. Richard therefore feels a strong sense of satisfaction and fulfillment when he thinks about what he is doing. He thinks that the people who are being killed or wounded in the war do not deserve to die, and that their well-being is of great importance. And so he wants to continue what he is doing even though he sometimes finds it very upsetting.

In the other condition, Richard was described as a Nazi death camp doctor who experienced the exact same type of mental states (differences in bold): Richard is a doctor working in a Nazi death camp, overseeing and carrying out executions and non-consensual, painful medical experiments on human beings. He sometimes gets pleasure from this, but equally often the deaths and human suffering get to him and upset him. However, Richard is convinced that this is an important and crucial thing that he has to do. Richard therefore feels a strong sense of satisfaction and fulfillment when he thinks about what he is doing. He thinks that the people who are being killed or experimented on do not deserve to live, and that their well-being is of no importance. And so he wants to continue what he is doing even though he sometimes finds it very upsetting.

After reading about one of these two agents, all participants were asked whether or not they agreed with the simple statement: • Richard is happy. Answers were recorded on a scale from 1 (“disagree”) to 7 (“agree”). Comparing the two conditions, participants were significantly more likely to agree with the statement that Richard is happy when he was described as a Red Cross doctor (mean agreement: 4.63) than when he was described as a Nazi doctor (mean agreement: 3.54) (Figure 10.1).16 16 Participants were 182 undergraduates at University of North Carolina at Chapel Hill taking courses in philosophy, African-American studies, or sports and exercise science. Participants judged the agent to be happier in the good life (Red Cross) condition (M = 4.63,



J . PHILLIPS ET AL .

7

Happiness Attribution

6 5 4 3 2 1

Nazi

Red Cross

Figure 10.1 Agreement ratings for the statement ‘Richard is happy’ for the bad life condition (right) and good life condition (left). Error bars indicate bootstrapped 95% CI.

Although Richard was described as experiencing the same mental states in both conditions, participants overall tended to agree that the Red Cross doctor was happy, but that the Nazi doctor was not. On the face of it, this result undermines the widespread assumption that the ordinary understanding of happiness is only concerned with the mental states an agent experiences. While we proposed that the difference in judgments of happiness was due to the difference in participants’ moral evaluations, one might object that we have not yet provided evidence that it was specifically the moral evaluation of the agent that changed participants’ judgments of happiness. Rather, the difference may have been due to one of the many descriptive differences in the two agents’ lives. Based on the incomplete description of the doctors’ experiences, participants could have reasonably inferred that the two doctors actually experienced quite different sets of psychological states. If this is correct, then we may not yet have evidence that the ordinary understanding of happiness is genuinely evaluative.

SD = 1.217) than in the bad life (Nazi) condition (M = 3.54, SD = 1.486), t(180) = 5.4, p < 0.001, r = 0.372.

THE GOOD IN HAPPINESS



To demonstrate that such differences in judgments of happiness actually result from the moral value of the agent’s life, one would need to find a case in which some participants judge the life to be normatively bad and others judge it to be good, but the life they are judging is descriptively identical.

Study 2: Variance in normative evaluation predicts variance in attributions of happiness In this study, all participants read a single vignette about an agent who experienced both positive and negative mental states. However, the vignette was constructed such that it was likely to elicit a wide range of moral evaluations of the agent. To achieve this, the vignette described a gay man named Bruce who, though conflicted because of a religious upbringing, began dating a man named Andrew. Bruce was brought up in a religious home and used to go to church every Sunday. However, since coming out as gay, Bruce has stopped going to church. He now meets new people at nightclubs, parties, and bars, instead of in church as he did before. One of Bruce’s strongest desires is to find a man to have a long-term relationship with. He thinks that Andrew, whom he just recently met, might be what he is looking for. One of the reasons Bruce thinks this is that they are very compatible, sexually speaking. Occasionally Bruce feels a little guilty about being in a relationship with a man. It goes against his upbringing; he was raised to think that one of the most important things in life for a man is to find a wife. But, most of the time Bruce does not feel these kinds of guilty feelings. And, when he does, Bruce reminds himself that he thinks he really has no reason to. Bruce then also reminds himself that he really likes Andrew and the people they socialize with. And so he does not want to change anything about his lifestyle.

After reading the vignette, participants responded to three items in counterbalanced order: • Bruce is happy. [Scale from 1 (“Disagree”) to 7 (“Agree”)] • How much distress do you think Bruce feels? [Scale from 1 (“None at all”) to 7 (“A lot”)] • Do you think Bruce’s lifestyle is immoral? [Scale from 1 (“No”) to 7 (“Yes”)] As expected, Bruce’s lifestyle elicited a wide range of responses, with some of the participants judging Bruce’s lifestyle to be immoral while



J . PHILLIPS ET AL .

others judged that it was not. Suggestively, participants’ judgments of immorality were highly negatively correlated with their judgments of happiness (r = –0.417). By contrast, judgments of immorality were not at all correlated with judgments of distress (r = 0.173), suggesting that those who found Bruce’s lifestyle to be immoral did not perceive him as feeling more distress.17 More conclusively, when looking at how judgments of immorality and distress predicted judgments of happiness, we found that judgments of immorality predicted judgments of happiness in a way that was independent of perceived distress.18 This overall pattern can be more simply illustrated by dividing participants into two groups: (1) those who reported that Bruce’s life was immoral and (2) those who reported that his life was not immoral. We can then compare the judgments made by the participants in these two groups. As predicted, participants who believed that Bruce was living an immoral life found him to be significantly less happy (Mean agreement: 3.83) than those who did not find his lifestyle to be immoral (Mean agreement: 5.02).19 Moreover, we also compared how the two groups attributed distress to Bruce, to test if those who found his life to be immoral also attributed more distress. We found no significant difference in the groups’ attributions of distress (Figure 10.2).20

17 Participants were seventy-one undergraduates at the University of Michigan. Judgments of immorality were highly correlated with happiness judgments, r = –0.417, p < 0.001, but not with judgments of distress, r = 0.173, p > 0.05. 18 The data were analyzed using a step-wise linear regression, with judgments of immorality and distress entered as independent regressors and happiness judgments as the dependent variable. Unsurprisingly, judgments of distress by themselves predicted judgments of happiness (â = –0.481, p < 0.001). More importantly, when judgments of immorality were included in the model with judgments of distress, judgments of immorality significantly predicted judgments of happiness (â = –0.344, p = 0.001), but we observed very little reduction in the predictiveness of judgments of distress (â = –0.421, p < 0.001) in this second model. These results strongly suggest that the perception of immorality and distress account for independent variance in participants’ judgments of happiness. 19 The five participants who responded “in between” to the question about whether Bruce’s life was immoral were excluded from this analysis. A significant effect of morality on judgments of happiness was observed such that participants who responded that Bruce’s lifestyle was not immoral judged him to be happier (M = 5.021, SD = 1.246) than those who responded that Bruce’s lifestyle was immoral (M = 3.833, SD = 1.505), t(64) = 3.26, p < 0.002, r = 0.395. We are using a median split only to highlight the patterns observed using correlation and regression analyses. None of our conclusions are drawn solely on the basis of these illustrative analyses. 20 t(64) = –1.40, p = 0.165, r = 0.172.

THE GOOD IN HAPPINESS



7

Agreement Rating

6 5 4 3 2 1

Happiness

Distress Immoral

Moral

Figure 10.2 Attributions of happiness (left) and distress (right) for those participants who thought Bruce’s life was not immoral (dark bars) and those who thought his life was immoral (light bars). Error bars indicate bootstrapped 95% CI.

In short, these results suggest that although both judgments of distress and judgments of immorality influenced attributions of happiness, the impact of normative judgments on judgments of happiness was not due to the difference in perceived distress. Participants’ normative evaluations about Bruce’s lifestyle were independently related to their judgments of happiness. This is particularly noteworthy given that all participants read exactly the same story about a single agent whose mental states were always described in exactly the same way. This result provides even better evidence than the previous study that the ordinary concept of happiness is partly evaluative and cannot be explained by perceived descriptive differences in the life the agent was living. Additionally, this study helps to make clear a distinction between two different ways in which normative evaluations might influence judgments of happiness. On the one hand, some theories of happiness (e.g. life-satisfaction theories) have emphasized the importance of whether an agent’s life lives up to her own evaluative standards. On this sort of view, an agent will not count as happy if her life does not measure up to her own evaluative judgments of what makes a life good. On the other hand, other theories (e.g. Foot 2001) have emphasized the importance of



J . PHILLIPS ET AL .

whether or not the life actually is good. On this sort of view, what an agent herself thinks is good may be totally irrelevant—she may simply be mistaken about what a good life actually is. What is relevant on this kind of view is only whether her life truly is good. While these two evaluations often coincide, they can also easily come apart. In Study 2, all participants read the same vignette in which Bruce was described as living a life that he reflectively endorsed. Accordingly, if only the first kind of evaluative judgment (that which Bruce makes about his own life) were relevant to the ordinary understanding of happiness, then we should expect participants to uniformly judge that Bruce is happy. This is not what we observed. Instead, participants’ judgments of happiness were substantially influenced by whether or not they judged Bruce’s life was actually immoral.21

Study 3: The difference between happiness and unhappiness One might still continue to press the objection that we have not yet provided sufficient evidence that the ordinary understanding of happiness is evaluative. In particular, one could object that we measured the perception of only one other psychological state: distress. Even if it were true that participants who thought Bruce was living an immoral life did not find him to be more distressed, they still could have perceived him to experience some other psychological state which we did not measure that actually explains why they judged he was less happy, e.g. dissatisfaction. If this is right, then there is no reason to assume the ordinary understanding of happiness is truly evaluative; we just have not yet found the right descriptive psychological state to measure. To go down this road would be unbearably tiresome. Suppose we do conduct another study and measure judgments of dissatisfaction only to

21 While it is possible that participants’ moral judgments could have influenced how they perceived the agent’s own values (and whether he was living up to those), we think it unlikely that this could explain the pattern we observed, for two reasons. (1) This explanation would suggest that participants who thought the agent’s life was immoral should also believe that the agent was not satisfied with his life and should reflectively want to change it. However, the wording of the vignette specifically cuts against this interpretation, as the agent is described as not wanting to change anything about his life. (2) Even if participants ignored this portion of the vignette and did perceive the agent as differentially living up to his own values, then it is somewhat surprising that those who found his life to be immoral were not also more inclined to judge him as distressed, given that they should perceive the agent as much less satisfied with his life and as not living up to his own values.

THE GOOD IN HAPPINESS



find that moral evaluations do not influence judgments of dissatisfaction but do influence judgments of happiness. We are no better off than when we started. It is open to some other objector to press onward and suggest that we just have not yet measured the relevant psychological state, which she might go on to suggest is actually pain or attitudinal displeasure or discontentedness or angst or ennui. Rather than entering into this tedious back-and-forth, it would be helpful to find a new way forward. In a study previously reported in Phillips et al. (2011), participants read about an agent who was living a normatively good life or an agent who was living a normatively bad life and were then asked to make either judgments of happiness or judgments of unhappiness.22 More specifically, participants in the morally good condition were told about a woman named Maria who was working hard to raise her family, while participants in the morally bad condition were told about a woman named Maria who was doing anything and everything required to climb the social ladder in Los Angeles. After reading about one of these two agents, participants read a description of her mental states. In the happiness conditions she was described as experiencing positive mental states, e.g. pleasure, satisfaction, and enjoyment, and participants were asked the question: • Do you think Maria is happy? [Scale from “No” to “Yes”] In the unhappiness conditions she was described as experiencing negative mental states, e.g. displeasure, dissatisfaction, and regret, and participants were asked the question: • Do you think Maria is unhappy? [Scale from “No” to “Yes”] Participants were assigned to one of four possible conditions: they could have either read about an agent living a morally good life or a morally bad life, and they could have been asked a question about happiness or a question about unhappiness. 22 Even though this study has been reported previously, we summarize the results and experimental methods in detail for two related reasons. First, this study (Study 3) was originally published as an extension of Studies 1 and 2, and cited them as an unpublished manuscript by the second author. As such, the summary of these results here accurately represents the conceptual development of this research. Second, and relatedly, this study also forms a key piece of our theoretical argument and directly motivates the subsequent studies presented in this chapter.



J . PHILLIPS ET AL .

7

Agreement Rating

6

5

4

3

2

1 Happiness

Unhappiness Bad life

Good life

Figure 10.3 Attributions of happiness (left) and unhappiness (right) for the good life condition (dark bars) and the bad life condition (light bars). Error bars indicate bootstrapped 95% CI.

The results of this study showed that while the normative value of Maria’s life influenced participants’ judgments of happiness, it did not influence their judgments of unhappiness. That is, participants tended to say that they thought Maria was happy when she was leading a morally good life, but not when she was leading a morally bad life. However, they equally thought Maria was unhappy regardless of the type of life she was leading (Figure 10.3). This approach to the study of the ordinary understanding of happiness is advantageous in that it allows us to pre-empt the sort of tedious backand-forth that we sought to avoid. The results of this study are problematic for anyone who might have been tempted to argue that we did not specifically measure the relevant descriptive mental state in the previous studies, e.g. pain, displeasure, and so on. If the difference in happiness judgments in this study were due to a difference in the amount of displeasure—for example, that the agent was perceived to have in the bad life condition versus the good life condition—then we should certainly

THE GOOD IN HAPPINESS



expect that the difference in perceived displeasure would equally affect judgments of unhappiness. The same will be true of any other descriptive mental state that one might put forward as a possibility. In short, given that only judgments of happiness differed between the conditions, it seems unlikely that the difference in judgments of happiness is best explained by a perceived difference in the agent’s descriptive mental states.

Study 4: Happiness, unhappiness, and negative mental states A reasonable concern with the study discussed above (from Phillips et al. 2011) is that the happiness and unhappiness conditions differed both in the description of the mental states the agent experienced and in the question that participants were asked. That is, positive mental states are confounded with asking about happiness (in the happiness condition), and negative mental states are confounded with asking about unhappiness (in the unhappiness condition). Accordingly, it remains possible that the difference in the happiness condition but not in the unhappiness condition resulted not from the difference in the questions asked (happiness vs. unhappiness) but from the difference in the descriptions of the agents’ mental states (positive vs. negative). If this turns out to be the case, this previous research may not reveal much about the ordinary understanding of happiness after all. To address this confound, we decided to describe the agent’s mental states in exactly the same way in both the good life and the bad life conditions. More specifically, we tested whether the normative value of the agent’s life would affect judgments of happiness, but not judgments of unhappiness, even when the agent only experienced negative mental states. Participants in the good life condition read about a man named Mark who was volunteering in clinics in rural Africa, while participants in the bad life condition read about a man named Mark who was hopping from nightclub to nightclub across America. Regardless of which life Mark was living, he was always described as experiencing the same negative mental states: Day to day, Mark never feels satisfied. When he reflects on his life, he often feels distressed by the thought that he could do more, especially given his privileged upbringing. Each night, the images from the day haunt him to sleep. Still, each morning he decides to live as the day before, convinced that this is the life that he wants to live.



J . PHILLIPS ET AL .

After reading about Mark, participants were asked to make simple judgment either about happiness or about unhappiness: • Do you think Mark is happy? • Do you think Mark is unhappy? Participants responded to both questions on a scale from 1 (“Not at all”) to 7 (“Very much so”). For the participants who made judgments of unhappiness, there was no significant difference between those who read about Mark living a normatively good life (Mean judgment: 5.00) and those who read about him living a normatively bad life (Mean judgment: 4.96). This is not surprising, given that Mark was always described as experiencing negative psychological states. However, participants did differ in their judgments of whether Mark was happy. Those who read about Mark living a good life judged him to be significantly more happy (Mean judgment: 2.68) than those who read about Mark living a bad life (Mean judgment: 1.63), even though he was always being described as experiencing the same negative psychological states (Figure 10.4). The predicted interaction was significant.23 These results provide new evidence for a discrepancy between the way in which people ordinarily make judgments of happiness and unhappiness. When participants are trying to determine whether a given agent is happy, they seem to not only consider that agent’s psychological state, but also take account of some normative evaluation of that agent. In contrast, when participants are determining whether a given agent is unhappy, this normative evaluation simply does not seem relevant.24 23 Participants were 115 people (forty-seven females, mean age 31) recruited from Amazon’s Mechanical Turk. The data were analyzed using a 2 (Life Value)  2 (Concept) ANOVA. We observed a main effect of Life Value F(1, 111) = 3.98, p < 0.05, η2 = 0.04, a main effect of Concept, F(1, 111) = 11.66, p = 0.001, η2 = 0.10, and, most importantly, a significant Life Value  Concept interaction effect, F(1, 111) = 4.58, p < 0.05, η2 = 0.04. The interaction was decomposed with planned comparisons, which revealed a significant difference in participants’ responses to the question “Do you think Mark is happy?” between the good life (M = 2.677, SD = 1.536) and bad life (M = 1.63, SD = 1.079) conditions, t(56) = 3.0, p < 0.01, r = 0.372, but no significant difference in participants’ responses to the question “Do you think Mark is unhappy?” in the good life (M = 4.963, SD = 1.224) and bad life (M = 5.00, SD = 1.486) conditions, t < 1, r = –0.014. 24 Notice that the key to our argument is the life value by concept interaction effect. In this study, and in all of the following ones, it happens that there was no significant difference in participants’ judgments of unhappiness when reading about a person living

THE GOOD IN HAPPINESS



7

Agreement Rating

6

5

4

3

2

1 Happiness

Unhappiness Bad life

Good life

Figure 10.4 Participant responses to questions about happiness (left) and unhappiness (right) for the good life condition (dark bars) and the bad life condition (light bars). Error bars indicate bootstrapped 95% CI.

Study 5: “Is not happy” vs. “Is unhappy” During a presentation of the previous studies, Stephen Stich pointed out that our way of accounting for this pattern of data makes a counterintuitive prediction. On our account, it seems that participants’ evaluative judgments should not affect their agreement ratings with statements of the form “S is unhappy,” because this statement concerns “unhappiness.” But at the same time, they should influence their agreement with statements of the form “S is not happy,” because this statement concerns “happiness.” This should hold despite the striking similarity between the a normatively good versus normatively bad life. While interesting (and rhetorically convenient), this lack of a difference is not particularly important for our point. What is crucial for the point we want to make is for normative evaluations to have more of an effect on judgments of happiness than they do on judgments of unhappiness. Thus, even if future research uncovers cases in which participants’ judgments of unhappiness also differ between conditions, our point will still stand as long as judgments of happiness are significantly more influenced.



J . PHILLIPS ET AL .

two statements. Admittedly, we found the prediction counterintuitive, but we decided to take up Stich’s challenge. Participants once again read about an agent named Mark who experienced negative mental states (distress, dissatisfaction, etc.) and he was either described as living a morally good life (volunteering in rural Africa) or described as living a morally bad life (hopping from nightclub to nightclub across America). After reading about the agent’s life, participants were asked whether they agreed or disagreed with one of two statements: • Mark is not happy. • Mark is unhappy. Participants rated their agreement on a scale from 1 (“Completely Disagree”) to 7 (“Completely Agree”). Finally, to continue to control for whatever uninteresting differences in perceived mental states there may be, participants also rated how much distress they believed Mark experienced. Despite the minimal difference between the statements, we once again observed the predicted asymmetry between judgments of happiness and judgments of unhappiness (Figure 10.5). Those who rated their agreement with the statement “Mark is unhappy” were equally likely to agree with the statement, regardless of the type of life that Mark was described as living a good life (Mean agreement: 5.73) or bad life (Mean agreement: 5.80). However, there was a significant difference in participants’ agreement ratings for the statement “Mark is not happy.” As predicted, those who read about Mark living a normatively good life agreed less that he “is not happy” (Mean agreement: 5.25) than those who read about Mark living a normatively bad life (Mean agreement: 6.47). The predicted interaction was significant even though we statistically controlled for the amount of distress participants perceived Mark as experiencing.25 25 Participants were seventy people (forty-seven females, mean age 37) spending time on Amazon’s Mechanical Turk website. Four participants were excluded for failing a simple control question. The data were analyzed using a 2 (Life Value)  2 (Concept) ANCOVA with judgments of distress entered as a covariate. We observed a main effect of Life Value, F (1, 65) = 4.96, p < 0.05, η2 = 0.07, no main effect of Concept, p > 0.05, and no main effect of Perceived Distress, p > 0.05. Most importantly, there was a significant Life Value  Concept interaction effect F(1, 65) = 4.12, p < 0.05, η2 = 0.06. This interaction was decomposed with planned comparisons, revealing that participants agreed more that “Mark is not happy” in the bad life condition (M = 6.41, SD = 0.795) than in the good life condition (M = 5.26,

THE GOOD IN HAPPINESS



7

Agreement Rating

6

5

4

3

2

1 Is not happy

Is unhappy Bad life

Good life

Figure 10.5 Participants’ agreement ratings with the statement about happiness (left) and unhappiness (right) for the good life condition (dark bars) and the bad life condition (light bars). Error bars indicate bootstrapped 95% CI.

The results from the previous five studies have shown a highly consistent pattern: when people ordinarily try to figure out whether or not some agent is happy, their judgments are influenced not only by whether the person experiences positive or negative psychological states but also by their normative evaluations of that person.

Study 6: Toward a normative psychological theory of happiness While ordinary judgments of happiness have been shown to be influenced by normative evaluations, a further question remains about exactly which normative judgments are relevant. One possibility is that the relevant normative judgments, as in evaluative theories of happiness, SD = 1.522), t(34) = 2.79, p < 0.01, r = 0.432. By contrast, there was no similar difference in their judgments of unhappiness between the bad life condition (M = 5.84, SD = 1.015) and the good life condition (M = 5.73, SD = 1.033), t < 1, r = –0.054. Including the four participants who failed the control question in this analysis only increased the strength of the observed effects.



J . PHILLIPS ET AL .

primarily concern how the agent’s life is actually going, in a way that is completely independent of what the person thinks or believes. Another possibility is that the relevant normative judgments are about the life that the person believes or understands herself to be living. The key distinction between these two possibilities can be seen in the example Richard Kraut offered about a man who wants most to be loved, admired, and respected by his friends (1979). His “friends” collude in an elaborate plot to give him every reason to believe that he is loved, admired, and respected by them, though this is not actually the case. Unsurprisingly, the man is taken by the deception. In this case, we can imagine two separate “lives” that we might make normative judgments about. Like Richard Kraut suggests, we may judge that the life that the man is actually living is straightforwardly not going well—it is not a good life. This normative judgment is the one that evaluative theories of happiness have focused on. Yet, independent of how the man’s life is actually going, we may also judge that the life that the man understands himself to be living is going well—that life is a good life. These two normative judgments can clearly vary independently of one another. This same distinction may also be applied to the question of whether one is living a morally good or bad life. It is certainly possible, for example, for one to take oneself to be living a life that is leading others to be much better off, but it turns out to be the case that one is actually living a life that is leading others to be significantly worse off. With this distinction in mind, we can ask whether the relevant normative judgments in the previous studies were evaluations of the life that the agent was actually living, external to the agent’s understanding, or whether the relevant normative judgments were evaluations of the life that the agent understood herself to be living. As the previous studies do not distinguish between these two possibilities, we sought to pull them apart with one final study involving cases in which an agent may be understandably misled about how her life is going. Participants read about a nurse named Sarah who worked at a children’s hospital and whose mental states were always described as highly positive. After going to nursing school for several years, Sarah gets a job at the children’s hospital and sees many different children each day. This is the job she has always wanted.

THE GOOD IN HAPPINESS



Almost every single day Sarah feels good and generally experiences a lot of pleasant emotions. In fact, it is very rare that she would ever feel negative emotions like sadness or loneliness. When Sarah thinks about her life, she always comes to the same conclusion: she feels highly satisfied with the way she lives.

However, we varied both whether Sarah understood herself to be living a life that led others to be better off and whether the life she was living actually led others to be better off. Accordingly, Sarah was described as understanding herself to be living a life that was good for others in half of the cases: The reason Sarah feels this way is that she tries to help the sick children by giving them medicine that tastes like candy.

In the other half of the cases, Sarah was described as understanding herself to be living a life that was leading others to be significantly worse off: The reason Sarah feels this way is that she tries to poison the sick children by giving them pesticides that taste like candy.

Along the other dimension, the life that Sarah was actually living did lead others to be significantly better off in half of the cases: Ten years after Sarah died, a new medical discovery was made which suggested that the pesticides [medicine] had made some of the children less sick.

While in the other half of the cases, the life that Sarah was actually living led others to be significantly worse off: Ten years after Sarah died, a new medical discovery was made which suggested that the medicine [pesticide] had made some of the children more sick.

After reading one of the four possible vignettes, participants were asked whether they agreed or disagreed with a single statement about happiness: • Sarah was happy. Participants responded on a scale from 1 (“Completely Disagree”) to 7 (“Completely Agree”). Subsequently, participants were asked a series of control questions to make sure that they were not incorrectly imagining that Sarah actually experienced negative psychological states. These questions ensured that participants believed Sarah did experience the three components of the



J . PHILLIPS ET AL .

psychological definition of happiness: high levels of positive affect, low levels of negative affect, and high life-satisfaction (Diener et al. 2003). The design of this study allowed us to test whether participants’ judgments of happiness were affected by the value of the life the agent understood herself to be living or by the value of the life the agent was actually living. Participants’ judgments of happiness revealed that they were significantly affected by the value of the life the agent took herself to be living, but not by the value of the life the agent was actually living. Accordingly, participants rated Sarah as happier when she understood herself to be living a life that was improving the children’s well-being (Mean agreement: 6.67) than when she understood herself to be living a life that was causing the children to be harmed (Mean agreement: 5.83). By contrast, there was no significant difference in participants’ happiness ratings based on whether the life Sarah led actually did improve the wellbeing of children (Mean agreement: 5.96) or instead significantly harmed their well-being (Mean agreement: 5.83) (Figure 10.6).26 To forestall any concern that this design did not allow us to adequately vary participant’s normative judgments about both the agent’s actual life and the life the agent understood herself to be living, we asked an additional group of participants to read the same vignettes and rate their agreement with two statements that corresponded to the two relevant normative judgments.27 • Ignoring everything that Sarah thought about the life she was living and considering only what actually happened, Sarah was living a good life. 26 Participants were ninety-one people (fifty-six women, mean age 27.8 years, SD = 9.50) recruited from Amazon’s Mechanical Turk. Ten participants were excluded for failing one of the three control questions; the exclusions ensured that participants fully understood that Sarah experienced high positive affect, low negative affect, and high life-satisfaction. The data were analyzed using a 2 (Life Value: good vs. bad)  2 (Mental State Value: good vs. bad) ANOVA. We observed a main effect of Mental State Value, F(1, 77) = 14.08, p < 0.001, η2 = 0.155, but no effect of life value, F < 1, and no interaction effect, F < 1. Planned comparisons were used to investigate the effect of Mental State Value and revealed that participants judged that Sarah was happier when she had good mental states (M = 6.67, SD = 0.853) than when she had bad mental states (M = 5.38, SD = 2.272), t(79) = –3.67, p = 0.001, r = 0.382. Unsurprisingly, including the participants who incorrectly answered the control questions in the analysis did not change any of the observed effects. The main effect of Mental State Value remained, F(1, 87) = 15.57, p < 0.001, η2 = 0.152, and there was once again no main effect of Life Value F < 1, and no interaction effect, F < 1. 27 We would like to thank Tania Lombrozo for helpfully raising this and related points.

Agreement Rating

7

Happiness Assessment

7

6

6

5

5

4

4

3

3

2

2

1

Bad psychology

Good psychology

1

Value of Perceived Life

Bad psychology Good psychology Bad life

Value of Actual Life

Bad psychology

Good psychology

Good life

Figure 10.6 Participants’ agreement ratings with the statement about happiness, value of perceived life, and value of actual life for an agent who understood herself to be helping children (left) or hurting them (right), and who actually was helping children (dark bars) or hurting them (light bars). Error bars indicate bootstrapped 95% CI.



J . PHILLIPS ET AL .

• Considering only what Sarah thought about the life she was living and ignoring everything that actually happened, Sarah was living a good life. Considering first the normative judgments of the agent’s actual life, participants’ responses indicated that their judgments were affected by whether Sarah understood herself to be helping or harming children, such that they more agreed that she was actually living a good life when she took herself to be helping children (Mean agreement: 5.23) than when she took herself to be harming children (Mean agreement: 1.89). While not specifically intended, this is not particularly surprising. The relevant question, however, is whether participants’ normative judgments about Sarah’s actual life were adequately impacted by whether Sarah was actually helping children. Considering the overall pattern, participants’ normative judgments did, in fact, appear sensitive to whether Sarah was actually harming or helping children. However, this was only the case when Sarah understood herself to be helping the children. More specifically, when Sarah understood herself to be poisoning children, participants did not think that having happened to help them significantly made her life any better.28 By contrast, when Sarah understood herself to be helping children, participants agreed that Sarah’s life was significantly better when she actually was helping children (Mean agreement: 6.00) than when she was actually harming them (Mean agreement: 4.57) (Figure 10.6, right).29 This pattern suggests that, at 28

One likely explanation of this is simply that participants’ ratings were so low on the available agreement scale when Sarah understood herself to be poisoning children that it was simply not possible to detect an effect of Actual Life Value. This is often referred to as a “floor effect.” 29 Participants were 105 people (sixty-one women, mean age 30.42 years, SD = 8.61) recruited from Amazon’s Mechanical Turk. Twenty-one participants were excluded for either missing the previously discussed control questions or taking less than 20 seconds to read the vignette (average completion time was over 60 seconds). The data were again analyzed using a 2 (Understood Life Value: good vs. bad)  2 (Actual Life Value: good vs. bad) ANOVA. We observed a main effect of Understood Life Value, F(1, 80) = 86.67, p < 0.001, η2 = 0.522. In addition, we observed a trend of Actual Life Value, F(1, 80) = 2.68, p = 0.105, η2 = 0.032, and a significant interaction effect F(1, 80) = 4.75, p = 0.032, η2 = 0.056. Planned comparisons investigated the interaction effect. When Sarah understood herself to be helping children, participants judged that her life was significantly better when she actually was helping children (M = 6.00, SD = 0.97) than when she was actually harming children (M = 4.57, SD = 1.66), t(32.935) = –3.33, p = 0.002, r = 0.502. However, when Sarah understood herself

THE GOOD IN HAPPINESS



least in the conditions in which Sarah understood herself to be helping children, we also successfully change participants’ normative judgments of Sarah’s actual life. Importantly, however, this difference in the value of Sarah’s actual life was not mirrored in participants’ judgments of whether Sarah was happy in those same conditions. Focusing next on the question about the normative value of the life that Sarah understood herself to be living, participants’ judgments indicated that they did think that the life Sarah understood herself to be living was significantly better when she took herself to be giving children medicine (Mean agreement: 6.33) than when she took herself to be poisoning them (Mean agreement: 5.07). Moreover, participants’ judgments were not affected by whether or not Sarah actually was helping children or actually was harming them.30 This suggests that we were able to adequately vary the value of the life Sarah understood herself to be living. When considered as a whole, the pattern of participants’ normative judgments about the life the agent understood herself to be living bears a striking resemblance to previous participants’ judgments of happiness (Figure 10.6, middle panel). The key question, which we now have the ability to answer, is which of these two normative judgments are relevant for ordinary happiness. Are ordinary judgments of happiness sensitive to the value of the life the person actually lives, or are they sensitive to the value of the life the person understands herself to be living? The data from this study suggest that it is the value of the life the person understands herself to be living that is central to ordinary happiness.

to be harming children, participants’ evaluations of her life did not significantly change if she was actually helping children (M = 1.82, SD = 1.76) or harming children (M = 1.96, SD = 1.89), p = 0.80, r = 0.038. Including the excluded participants in the analysis did not significantly change these effects. 30 The data were again analyzed using a 2 (Understood Life Value: good vs. bad)  2 (Actual Life Value: good vs. bad) ANOVA. We observed a main effect of Value of Perceived Life, F(1, 80) = 12.63, p < 0.001, η2 = 0.136, but no effect of Actual Life Value, F < 1, and no interaction effect, F < 1. Planned comparisons were used to investigate the effect of Understood Life Value and revealed that participants agreed that the life Sarah understood herself to be living was better when she took herself to be helping children (M = 6.33, SD = 0.81) than when she took herself to be poisoning them (M = 5.07, SD = 2.07), t(59) = –3.78, p < 0.001, r = 0.442. Including the excluded participants in the analysis did not change the observed effects.



J . PHILLIPS ET AL .

General Discussion The six studies presented in this chapter provide evidence that normative value is relevant to ordinary happiness, and, more specifically, that the relevant normative judgments focus on the life the agent understands herself to be living. First, we consider one alternative explanation for the patterns we observed. Then, we take up the task of outlining the sort of theory which could adequately capture the way that people ordinarily make judgments of happiness. Finally, we step back and discuss the philosophical and psychological implications of these findings.

The multiple senses objection One way of objecting to the claim that ordinary judgments of happiness are essentially evaluative would be to suggest that the term “happiness” actually has multiple senses, one of which is descriptive (similar to “pleasure”) and one of which is evaluative (similar to “well-being”). Indeed, there has been a large amount of scholarly speculation that this is the case (Feldman 2010; Foot 2001; Haybron 2000; Kraut 1979). According to this objection, there is no single understanding of happiness which focuses on the agent’s psychological states but is also evaluative. Rather, these two aspects represent features of two independent senses of “happiness.” Accordingly, combining them into a single, unified understanding of happiness would be mistaken. While we do not doubt that people sometimes use the term “happy” in a way that seems synonymous with pleasure (e.g. “I’m very happy to meet you”) or with well-being (e.g. “I wish you a happy life”), we think there is good reason to suspect that this could not adequately account for the full pattern of data. One major obstacle is that this objection cannot explain why we observed a difference in judgments about happiness but not judgments about unhappiness. Thus, even if the multiple senses objection could account for participants’ judgments of happiness, it would also need to explain why these independent senses were not at play in judgments of unhappiness. After all, we use “unhappy” in ways that seem synonymous with displeasure (e.g. “Traffic jams make me unhappy”) and with a lack of well-being (e.g. “Their marriage has always been unhappy”).31 31

This objection also makes an empirical prediction concerning the distribution of participants’ responses in the cases in which an agent was described as having positive

THE GOOD IN HAPPINESS



Even more conclusively, recent research (Knobe et al. forthcoming) has developed tests for concepts that have a “dual-character” such that a single concept can be used in both normative and descriptive senses (e.g., mother, rock music, criminal). Employing these tests, researchers have already found evidence that happiness does not fit in the category of dualcharacter concepts. Happiness no more admitted of a dual character than emotions such as sadness, anger, fear, or disgust (Chituc, 2012). This finding directly contradicts what we would expect if happiness genuinely had distinct normative and descriptive senses. Thus, we find it unlikely that an objection based on multiple senses can explain these results.

The good in happiness Even if one accepts that there is a single evaluative understanding of happiness, one might still be at a loss as to how we should make sense of the ordinary understanding of happiness. While it is highly likely that any specific theory proposed at this still early stage will be overturned upon the discovery of new evidence or further theorizing, we propose one way of accounting for ordinary happiness. Let us begin by considering how the present studies relate to the longstanding debate over happiness. On descriptive theories of happiness, whether or not an agent is happy is a matter of figuring out exactly which psychological states that agent experiences and deciding which ones constitute happiness. In contrast, evaluative theories of happiness hold that if an agent is not living a normatively good life, then she will not truly be happy, no matter which psychological states she experiences. When contrasting the two sides of this debate, it may be tempting to take the results of these studies to be evidence against the descriptive theories and for evaluative theories. However, this would be too quick. These two groups of theories differ along two distinct dimensions. First, they differ in whether or not the normative value of an agent’s life is relevant to whether the agent is happy. Second, they also differ in what they focus on

psychological states but a normatively bad life. Specifically, it suggests that the distributions of participants’ responses should be bimodal, indicating that some participants used the purely descriptive sense (thus agreeing that the agent was happy) while others employed a purely evaluative sense (thus disagreeing that the agent was happy), but that very few participants would mix up these two senses. However, the data simply do not bear out this prediction: the distributions of responses were consistently unimodal in these exact conditions. These data are available, upon request, from the first author.



J . PHILLIPS ET AL .

Normative value

No normative value

Psychological

Overall

states

life

?

theories, e.g.

Evaluative Foot; Annas Descriptive theories, e.g.

X

Feldman; Diener

Figure 10.7 Normative value by focus matrix of possible theories of happiness.

when deciding whether someone is happy. Descriptive theories tend to focus primarily on the psychological states that an agent experienced, while evaluative theories tend to focus primarily on the actual life the agent has led (Figure 10.7). The ordinary understanding of happiness seems to deviate in important ways from both evaluative and descriptive theories of happiness. Unlike descriptive theories, normative information about the agent is central to the ordinary understanding. Accordingly, it is unlikely that ordinary happiness is merely a matter of possessing some particular psychological state or set of psychological states (bottom row). Yet, unlike evaluative theories of happiness, the ordinary understanding of happiness does not focus on the overall life the agent has actually lived. Rather, it focuses on the life the agent understands herself to be living, which is quite clearly a matter that is solely determined by the psychological states the agent experiences. Thus it also seems unlikely that ordinary happiness is simply a matter of having led a good life. Instead, ordinary happiness focuses on the psychological states of agents, but does so in a genuinely normative way. Ordinary happiness seems to be a matter of possessing positive psychological states that it is normatively right to experience, given the life one understands oneself to be living.32 32 Important questions clearly remain about what is required for an agent to understand herself to be living a particular type of life. Such an understanding will presumably require more than merely believing that one is living a morally good life—even when one has sufficient reason to believe that one is actually not living that life. Yet, rather than trying to offer some specific account which is unlikely to be correct, we wish to be intentionally imprecise about this issue, and hope that this question will be taken up in future work on this topic.

THE GOOD IN HAPPINESS



It will be important to be more precise about the way in which a psychological state may be right or wrong to possess. On our proposal, a positive psychological state is wrong to possess if one ought not be in that state, given the life one understands oneself to be living. Clearly, one ought not experience satisfaction with one’s life if one understands that life to be one that directly harms children’s well-being by poisoning them. Positive psychological states that are not wrong to possess count (in the relevant sense) as right to possess. Thus, to say that someone is happy would be to say that she experiences positive states such as pleasure or satisfaction, and that given the life she understands herself to be living, it is not the case that she ought not experience those states.33 It should not be too difficult to see how this theory can account for participants’ judgments. Consider the case of the nurse in Study 6 who derives pleasure and satisfaction from trying to save children’s lives. It seems obvious that there is no reason that the nurse ought not to experience satisfaction given that she understands herself to be living a life in which she makes children’s lives better. In contrast, it is clear that the nurse ought not to experience pleasure or satisfaction when she understands herself to be living a life in which she is harming children by poisoning them. This too remains the case whether or not she is actually successful in causing the children harm. Before continuing, it may be helpful to explicitly specify the sort of normative evaluations that we think are at play in ordinary judgments of happiness. On our view, it is specifically moral normativity that is relevant to happiness and not, by contrast, the normativity of practical rationality or prudence. More precisely, then, to be happy in the ordinary way is to experience positive psychological states when it is not the case that one morally ought not have them, given the life one understands oneself to be living. To see why it is specifically moral normativity that matters for the proposed theory of ordinary happiness, consider again the much-discussed case of the Nazi war criminal who experiences pleasure and contentment while living on an Argentinian beach. Suppose 33 One might be tempted by the simpler formulation according to which happiness is a matter of possessing positive mental states that one ought to possess. However, this simpler formulation is overly strict as it rules out amoral cases in which it is neither the case that one ought to possess the mental states nor the case that one ought not to possess them. The pleasure and satisfaction we have in such amoral cases, e.g. when eating ice cream, presumably should count toward our happiness, so we prefer the less restrictive formulation.



J . PHILLIPS ET AL .

that he has the goal of not becoming depressed and committing suicide, which he fears could easily happen. According to practical rationality then, he ought to experience the pleasure and contentment he does when living on the Argentinian beach. After all, he intends to not commit suicide and believes that feeling good will help him achieve his intended goal. Thus, if we are concerned with the normativity of practical rationality or prudence, the Nazi war criminal should count as happy, as it is clearly not the case that he ought not experience the pleasure and contentment he does. By contrast, one certainly might think that he morally ought not to experience those positive states given his understanding of his life. On our account of ordinary happiness, as long as one holds that he ought not to experience those states, the Nazi would not count as truly happy. While the tentative theory we have offered can account for the ordinary judgments of happiness in the six studies we presented, future studies will certainly be what decides between our proposed theory and other alternatives ways of accounting for the ordinary understanding of happiness. Yet regardless of which account is eventually accepted, we take these studies to provide sufficient evidence that any adequate account of ordinary happiness will have to provide some role for normative judgment. Finally, it will be informative to contrast this proposal with other theories that give normative judgments an importantly different role. Recall that descriptive theories of happiness often involve a particular sort of normative judgment. Similar to some theories of well-being (e.g., Tiberius, 2008; Tiberius and Plakias, 2010), life-satisfaction theories of happiness hold that, to be happy, agents must be reflectively satisfied with their own lives (Kekes, 1982; Suikkanen, 2011). While such judgments of life-satisfaction clearly involve normative evaluations, such theories are still purely descriptive because the relevant normative judgments are simply mental states that the agent experiences. To know whether or not someone is happy, all one would need is an accurate list of all the psychological states the person experiences, even if some of the states on this list involve the agent herself forming a normative judgment. In contrast, the evidence we have provided suggests that ordinary happiness is genuinely evaluative in that the relevant normative judgments are about whether the agent ought to be experiencing the positive mental states she experiences. We simply cannot know whether

THE GOOD IN HAPPINESS



an agent is happy by referencing an accurate list of her mental states—we also need to know whether she ought to be experiencing those states, and this is simply not a descriptive question.

Philosophical and psychological implications Even if one were convinced that we have accurately accounted for the ordinary understanding of happiness, one might still wonder why either philosophers or psychologists should be interested in this ordinary understanding. While we think this discovery is relevant for both philosophical and psychological research, we do not think it is relevant to both research projects in the same way. We take up the relevance for these two disciplines in turn. ORDINARY HAPPINESS AND ETHICS

One response to this line of research would be to be skeptical about its relevance for the philosophical discussion. Philosophers, after all, are presumably not interested in what people think happiness is, but rather in what happiness actually is. In making this point clear, one might point out that we have long known that the ordinary understanding of physics differs dramatically from philosophers’ discussions of it (McCloskey and Kohl, 1983; McCloskey et al. 1983), but that this divergence clearly does not give us reason to revise that philosophical discussion! We simply could not more agree with this conclusion, in the case of physics. What is not clear, however, is whether this is an appropriate analogy. The irrelevance of the ordinary understanding of physics follows straightforwardly from considering the evidential relation between the ordinary understanding and the underlying physics itself, which is what philosophers are concerned with. In that case, there is a relatively weak evidential relation between the ordinary understanding and the object of philosophical discussion, and even if there is some evidential relation, it is certainly not as strong as the evidential relation between rigorous scientific experiments and the object of philosophical interest. Accordingly, it is not surprising that contemporary philosophers simply do not show an interest in the ordinary understanding of physics, since this ordinary understanding is a comparatively terrible way of accessing physics itself. By contrast, philosophers who study happiness do seem to show a great deal of interest in the ordinary understanding of happiness. It is instructive to consider why this might be. Given that the philosophical



J . PHILLIPS ET AL .

discussion of happiness is specifically concerned with happiness itself (just as much as the philosophy of physics is concerned with physics itself), one obvious reason may be that philosophers take the ordinary understanding of happiness to have some non-trivial epistemic relation to happiness. Of course, nothing demands that the ordinary understanding have such a strong evidential relation, and perhaps happiness will turn out to be like physics. However, until we are presented with better evidence that this is the case, we are content to throw our lot in with the many other philosophers who also seem to think that the ordinary understanding is relevant to the philosophical discussion in the case of happiness.34 Accordingly, we maintain that, all else being equal, these studies should inform our theorizing about happiness by providing some evidence about happiness itself.35 Yet, even accepting that the ordinary understanding of happiness may be broadly relevant to the philosophical discourse, one may be left wondering what conclusions should be drawn from the present research. We hope this research will inspire further discussion on this exact question, but we take up two specific implications. The most straightforward implication is that, if our proposed theory of the ordinary understanding of happiness turns out to be right, then it provides evidence that the study of happiness may turn out to partly be a branch of normative ethics. Whether somebody is happy, in other words, depends in part on normative questions about what psychological states ought not be had under what understood conditions.

34 Note that nothing we have said suggests that the ordinary conception is the only thing that is relevant to the philosophical discussion. Other concerns, e.g. consistency, should perhaps be weighted even more heavily than the ordinary understanding in theorizing about happiness. After all, if the ordinary understanding of happiness turns out to be internally inconsistent, one might take this to be good evidence that it does not bear a strong evidential relation to happiness. 35 Another reason happiness researchers and philosophers may be concerned with the ordinary understanding of happiness is that people ordinarily care an awful lot about (what they think of as) happiness. In fact, we imagine that this is one of the very reasons that there has been so much interest in studying happiness. To the extent that happiness researchers or philosophers abandon all interest in happiness-as-it-is-ordinarily-understood in favor of some purely technical notion of happiness, call it happiness*, they run the risk of abandoning the reason that this research was motivated in the first place. At the very least, these researchers would now need to provide some independent motivation for caring about happiness* rather than happiness-as-it-is-ordinarily-understood.

THE GOOD IN HAPPINESS



Relatedly, a second implication about the study of happiness is that thoroughgoing research on happiness cannot remain completely neutral on the difficult questions discussed within the field of meta-ethics. To see this, consider cases in which people’s moral disagreements lead them to disagree about whether somebody is happy (e.g. Bruce in Study 2). Metaethical theories according to which there are objective facts about what moral convictions are true imply that some of the participants in our study were simply wrong. These theories imply that there may be an objective fact about whether Bruce is happy—a fact partly settled by some wholly objective moral fact about whether Bruce’s psychological states are the ones he ought to have. By contrast, meta-ethical theories according to which moral facts are subjective, or speaker- or grouprelative, imply that we cannot sensibly talk about Bruce’s happiness without first relativizing our claims to the norms of particular individuals or groups. Accordingly, participants may not actually be disagreeing with each other when they utter seemingly contradictory statements about Bruce’s happiness. Notice that the way that meta-ethical issues bear on Bruce’s happiness is indirect. It is not the case that Bruce’s happiness depends in any direct way on meta-ethical issues; rather, the point here is that whether it is objectively true that Bruce is happy depends directly on meta-ethical issues. By contrast, questions of whether or not Bruce is happy directly depend on issues in normative ethics.36 ORDINARY HAPPINESS AND AFFECTIVE SCIENCE

The studies we have discussed here reveal that the ordinary understanding of happiness combines both descriptive information about an agent’s psychological states and evaluative judgments about whether the agent ought to possess those states. By contrast, the scientific definition of happiness focuses solely on descriptive information about the agent’s psychological states, i.e. some combination of high positive affect, low negative affect, and high life-satisfaction (Diener et al. 2003; Lyubomirsky et al. 2005). 36 This point can be straightforwardly generalized to the superset of findings that illustrate the role of moral judgment in the ordinary understanding of other concepts, e.g., intentional action, causation, or freedom. However, this generalization is conditional on the acceptance that the ordinary understanding bears some non-trivial evidential relation to the object of philosophical interest, and that the role of moral judgment is a genuine part of the ordinary understanding—not simply the result of some bias.



J . PHILLIPS ET AL .

One important implication of this difference between the scientific definition and the ordinary way of understanding happiness is that it indicates that the methods employed by affective scientists cannot investigate the entirety of happiness, as it is ordinarily understood, because this is in part a normative question. This, of course, does not at all suggest that affective scientists should revise their definition to more closely align with the ordinary understanding of happiness. Rather, it suggests that affective scientists and positive psychologists should continue to pursue their research, but with the assumption that the ordinary understanding of happiness differs in an important respect. On this picture, the present research is best understood as a necessary complement to the research in affective science.

Conclusion The study of happiness, from ancient history to contemporary research, has continually presented two different ways of thinking about the topic. On one side are evaluative views of happiness, which hold that being happy requires one to live a good life. On the other side are descriptive views of happiness, which hold that being happy is simply a matter of experiencing the right sorts of psychological states. We have argued that the way we ordinarily understand happiness deviates in important ways from both of these views. The ordinary understanding of happiness turns out to be a surprisingly sophisticated combination of feeling good and being good. Perhaps one reason that these two opposing views have persisted in philosophical discussion is that the two views capture different components of the way we ordinarily understand happiness.

References Annas, J. (1993). The Morality of Happiness. New York: Oxford University Press. Annas, J. (2011). Intelligent Virtue. New York: Oxford University Press. Beebe, J.R., and Buckwalter, W. (2010). The Epistemic Side-Effect Effect. Mind and Language 25(4): 474–98. Chituc, V. (2012). Emotions as Dual Character Concepts. Unpublished manuscript, Yale University.

THE GOOD IN HAPPINESS



Cooper, J. (1987). Contemplation and Happiness: A Reconsideration. Synthese 72 (2): 187–216. Crisp, R. (2008). Well-Being. In E.N. Zalta (ed.), The Stanford Encyclopedia of Philosophy, . Cushman, F.A., Knobe, J., and Sinnott-Armstrong, W. (2008). Moral Appraisals Impact Doing/Allowing Judgments. Cognition 108(1): 281–9. Deci, E.L., and Ryan, R.M. (2012). Self-Determination Theory. In Van Lange, P. A.M., Kruglanski, A.W., and Higgins, E.T. (eds.), Handbook of Theories of Social Psychology: Vol. 1. Thousand Oaks, CA: Sage. Diener, E., Scollon, C.N., and Lucas, R.E. (2003). The Evolving Concept of Subject Well-Being: The Multifaceted Nature of Happiness. Advances in Cell Aging and Gerontology 15: 187–219. Dybikowski, J.C. (1981). Is Aristotelian Eudaimonia Happiness? Dialogue 20(2): 185–200. Feldman, F. (2010). What Is This Thing Called Happiness? New York: Oxford University Press. Foot, P. (2001). Natural Goodness. Oxford: Oxford University Press. Gilbert, D.T. (2006). Stumbling on Happiness. New York: Alfred A. Knopf. Haybron, D.M. (2000). Two Philosophical Problems in the Study of Happiness. The Journal of Happiness Studies 1(2): 207–25. Haybron, D.M. (2008). The Pursuit of Unhappiness: The Elusive Psychology of Well-Being. New York: Oxford University Press. Haybron, D.M. (2010). Mood Propensity as a Constituent of Happiness: A Rejoinder to Hill. The Journal of Happiness Studies 11: 19–31. Haybron, D. (2011). Happiness. In E.N. Zalta (ed.), The Stanford Encyclopedia of Philosophy, . Hitchcock, C. and Knobe, J. (2009). Cause and Norm. Journal of Philosophy 11: 587–612. Kahneman, D. (2011). Thinking, Fast and Slow. New York: Farrar, Straus and Giroux. Kashdan, T., Biswas-Diener, R., and King, L. (2008). Reconsidering Happiness: The Costs of the Eudemonic-Hedonic Distinction. Journal of Positive Psychology 3: 219–33. Kekes, J. (1982). Happiness. Mind 91: 358–76. Knobe, J. (2003). Intentional Action and Side Effects in Ordinary Language. Analysis 63: 190–3. Knobe, J., and Roedder, E. (2009). The Ordinary Concept of Valuing. Philosophical Issues 19:1, 131–47. Knobe, J., Prasada, S., and Newman, G. (forthcoming). Dual Character Concepts and the Normative Dimension of Conceptual Representation. Cognition.



J . PHILLIPS ET AL .

Kraut, R. (1979). Two Conceptions of Happiness. The Philosophical Review 88(2): 167–97. Lucas, R.E., Diener, E., and Larsen, R.J. (2003). Measuring Positive Emotions. In Lopez, S.J., and Snyder, C.R. (eds.), Positive Psychological Assessment: A Handbook of Models and Measures. Washington, DC: American Psychological Association, pp. 201–18. Lyubomirsky, S., King, LA., and Diener, E. (2005). The Benefits of Frequent Positive Affect: Does Happiness Lead to Success? Psychological Bulletin, 131: 803–55. May, J. and Holton, R. (2012). What in the World is Weakness of Will? Philosophical Studies 157(3): 341–60. McCloskey, M., and Kohl, D. (1983). Naive Physics: The Curvilinear Impetus Principle and its Role in Interactions with Moving Objects. Journal of Experimental Psychology: Learning, Memory, and Cognition 9(1): 146–56. McCloskey, M., Washburn, A., and Felch, L. (1983). Intuitive Physics: The Straight-Down Belief and its Origin. Journal of Experimental Psychology: Learning, Memory, and Cognition 9(4): 636–49. Pettit, D. and Knobe, J. (2009). The Pervasive Impact of Moral Judgment. Mind and Language 24(5): 586–604. Phillips, J. and Knobe, J. (2009). Moral Judgments and Intuitions about Freedom. Psychological Inquiry 20: 30–6. Phillips, J., Misenheimer, L., and Knobe, J. (2011). The Ordinary Concept of Happiness (and Others Like it). Emotion Review 3(3): 1–3. Roxborough, C., and Cumby, J. (2009). Folk Psychological Concepts: Causation. Philosophical Psychology 22(2): 205–13. Ryan, R.M., and Deci, E.L. (2000). Self-Determination Theory and the Facilitation of Intrinsic Motivation, Social Development, and Well-Being. American Psychologist 55: 68–78. Ryff, C.D., and Singer, B. (1998). The Role of Purpose in Life and Personal Growth in Positive Human Health. In Wong, P.T.P., and Fry, P.S. (eds.), The Human Quest for Meaning: A Handbook of Psychological Research and Clinical Applications. Mahwah, NJ: Lawrence Erlbaum Associates, Inc. Seligman, M., and Csikszentmihalyi, M. (2000). Positive Psychology: An Introduction. American Psychologist 55: 5–14. Sidgwick, H. (1907/1966). The Methods of Ethics. New York: Dover Publications. Suikkanen, J. (2011). An Improved Whole Life Satisfaction Theory of Happiness. International Journal of Wellbeing 1(1): 149–66. Sumner, L.W. (1996). Welfare, Happiness, and Ethics. New York: Oxford University Press. Sumner, L.W. (2000). Something In Between. In Crisp, R., and Hooker, B. (eds.), Well-Being and Morality. New York: Oxford University Press, pp. 1–19.

THE GOOD IN HAPPINESS



Tiberius, V. (2008). The Reflective Life. New York: Oxford University Press. Tiberius, V., and Plakias, A. (2010). Well-Being. In J. Doris (ed.) The Moral Psychology Handbook. New York: Oxford University Press. Vonasch, A.J., and Baumeister, R.F. (2012). True Happiness Is More Than Feeling Happy and Satisfied: People Think “Bad” People Aren’t Truly Happy. Presented at the annual conference for the Society of Personality and Social Psychology, San Diego, CA. Young, L. and Phillips, J. (2011). The Paradox of Moral Focus. Cognition 119: 166–78.

11 Causal Reasoning Philosophy and Experiment James Woodward, University of Pittsburgh

Introduction The past few decades have seen much interesting work, both within and outside philosophy, on causation and allied issues. Some of the philosophy is understood by its practitioners as metaphysics: it asks such questions as what causation “is,” whether there can be causation by absences, and so on. Other work reflects the interests of philosophers in the role that causation and causal reasoning play in science. Still other work, more computational in nature, focuses on problems of causal inference and learning, typically with substantial use of such formal apparatuses as Bayes nets (e.g. Pearl, 2000; Spirtes et al., 2000) or hierarchical Bayesian models (e.g. Griffiths and Tenenbaum, 2009). All this work—the philosophical and the computational—might be described as “theoretical,” and it often has a strong “normative” component in the sense that it attempts to describe how we ought to learn, reason about, and judge causal relationships. Alongside this normative/theoretical work, there has also been a great deal of empirical work (hereafter “empirical research on causation” or ERC) which focuses on how various subjects learn causal relationships and how they reason and judge causally. Much of this work has been

Thanks to David Danks, Clark Glymour, and Edoaurd Machery, and to three anonymous reviewers for Oxford University Press, for extremely helpful comments on an earlier version of this essay.

CAUSAL REASONING



carried out by psychologists, but it also includes contributions by primatologists, researchers in animal learning, neurobiologists, and, more rarely, philosophers. In my view, the interactions between these two groups—the normative/theoretical and the empirical—have been extremely fruitful. In what follows, I describe some ideas about causal reasoning that have emerged from this research, focusing on the “interventionist” ideas that have figured in my own work. I also attempt to extract some general morals regarding the kinds of interaction between the empirical and the more traditionally philosophical that in my experience have been most fruitful. Along the way (since this is a volume on experimental philosophy, hereafter X-Phi) I will also compare this research with some research strategies employed in X-Phi. I begin with this last topic, turn then to some general themes about experimental methodology and what can be learned from experiment, and turn finally to more detailed illustrations.

X-Phi, Survey Experiments, and ERC In a broad sense, the phrase “experimental [or empirical] philosophy” might be taken to characterize any attempt to bring empirical results to bear on a philosophical issue. With this understanding, a great deal of work in philosophy of science (e.g. attempts to use special and general relativity to settle “philosophical” questions about the nature of space and time) and many projects in ethics and political philosophy pursued in a naturalistic spirit (e.g. Kitcher, 2011) might qualify as “experimental philosophy.” However, this phrase is commonly used much more narrowly in contemporary discussion to encompass experiments carried out by philosophers, often survey-like in character, with adult humans as subjects, and often organized around concerns about the role of “intuitions” in philosophical argument. Even conceived in this narrower way, X-Phi is (obviously) far from homogenous, but one useful distinction1 1 I take this contrast between the negative and positive programs in X-Phi from Alexander et al. (2010), although, as they note, other writers employ a similar distinction. Examples of the positive program include the use of responses to vignettes to support claims about the folk concept of free will (and whether or not it is compatibilist—e.g. Nichols and Knobe, 2007), and the folk concept of intentional action (e.g. Knobe, 2003). Examples of the negative program include Alexander and Weinberg (2007), challenging claims made by analytic epistemologists about the folk concept of knowledge, and Machery



J . WOODWARD

contrasts (i) the use of experiments in a primarily negative or debunking role—to discredit appeals of more traditional philosophers to “intuition” or to claims about “what most people think”—and (ii) their use in the more positive role of providing evidence supporting claims about human psychology, including claims about the characteristics of various “concepts” people possess, patterns of reasoning and judgment in which they engage, and the psychological processes that generate these.2 As several writers (e.g. Alexander et al., 2010) have noted, X-Phi results might be

et al. (2004), challenging the universality of Kripkean intutions about reference. The negative program also includes papers presenting evidence that the intuitive judgments of ordinary people or philosophers are influenced by such normatively irrelevant factors as order effects (e.g. Swain et al., 2008), context effects, and small variations in wording. 2 I concede that some work in X-Phi does not fall neatly into just one of these categories, but still think that the distinction captures something epistemically important. Contrary to what some have claimed, as I see the negative project, it does not require commitment to the idea that X-Phi results measure or detect anything like what armchair philosophers claim to be detecting through their “intuitions,” or even that such intuitions (conceived, at a minimum, as requiring stable response patterns across and within subjects) exist or are sources of evidence about anything of philosophical interest. In particular, one can think of the negative project as resting on the claim that the intuitions of armchair philosophers are in the same epistemic position as the reports of X-Phi subjects; if this is correct, the debunking project can succeed even if what is measured in X-Phi surveys is, philosophically speaking, “noise” (or at least unstable etc. in a way that makes it unsuitable as evidence for traditional philosophical claims), since what this suggests is that the intuitions of the armchair philosopher are equally problematic. By contrast, the positive project takes on more of the commitments of traditional philosophical methodology, replacing the intuitions of single philosophers with the judgments of a larger number of people, but retaining the idea that these are important sources of evidence—e.g. about psychological phenomena or folk concepts of these. This is a source of concern because the problems with the traditional methodology go well beyond its N = 1 commitments. A closely related point is that while the focus of the negative project, on whether the people judge in connection with hypothetical examples as philosophers claim, makes a great deal of dialectical sense, it is less obvious that the positive project is best pursued by looking at folk responses to such examples. For example, if one merely wants to challenge what philosophers claim about how people would judge in connection with true-temp (a scenario due to Lehrer, 1990, in which a device is implanted in S’s brain in such a way that he reliably has beliefs that track the true temperature, without his knowing that or why he is reliable, the question being whether S “knows” the temperature), a survey may be perfectly sensible, since it may show, contrary to what Lehrer claimed, that there is no consensus among the people that this is not a case of knowledge. However, surveying responses to true-temp may be less well motivated if one’s aim is to probe folk conceptions of knowledge, if, as I suspect is the case, the scenario involves factual assumptions that radically depart from those assumed when the folk concept applies. In general, when one engages in the positive project it is unclear why the questions asked and the sources of evidence considered should be restricted in the way that they often are in X-Phi—that is, to verbal responses to hypothetical scenarios and so on.

CAUSAL REASONING



largely successful in their debunking role even if they have important limitations as sources of evidence for these more positive projects. On the other hand, the positive project is attractive for many reasons, not least because it is constructive and does not just consist in amassing negative results about claims of armchair philosophers. In what follows, I will largely focus on the use of X-Phi in the service of this more positive project and how this relates to ERC. I begin with some comparative remarks. Both X-Phi and ERC involve experiments in the minimal sense that variables of interest are manipulated and the results of such manipulations observed. As in X-Phi, ERC uses a number of subjects, rather than a single one or an unsystematically selected small group (the armchair philosopher and colleagues). There are also, however, differences worth remarking. As observed above, a great deal of X-Phi consists in gathering verbal responses of adult humans (often college students) to hypothetical scenarios or vignettes, also described verbally. For example, subjects may be presented with a verbally described scenario and then asked whether they would judge that C caused E in the scenario. Such scenario-based experiments are common in some areas of psychology as well. Although I agree that such experiments can produce valuable results (and examples will be given), it is also widely recognized (both in the X-Phi literature and by other commentators3) that they have a number of potential limitations when treated as sources of evidence regarding positive theses about human cognition: these include lack of experimental control over the way that subjects interpret the verbal scenarios and the likelihood that different subjects may interpret the scenarios differently, with these differences perhaps being correlated with other differences among subjects, possible sensitivity of subject responses to seemingly small variations in wording4, and order and other contextual or conversational effects (as when 3

See, e.g. Scholl (2008). For example, judgments about causation are known to be sensitive to the particular choice of words used in the verbal query employed by the experimenter—see Collins and Shanks (2006) for examples. A closely related problem is that the same verbal query about causation may be interpreted differently by different subjects or may confound independent features that the researcher is attempting to measure. For example, the common practice of asking subjects to rate how much they agree with various causal claims may confound assessments of causal strength with the subject’s degree of confidence that a causal relation of any sort is present, and perhaps also with judgments about how good or paradigmatic an example of causation the claim in question is, to the extent that this is different from 4



J . WOODWARD

subjects attempt to guess what sort of response the experimenter is looking for and then provide it). In my view, the proper response to these limitations is not to eschew survey-style experiments, but rather to recognize that such experiments are most likely to be valuable when (i) combined with other sorts of evidence, including evidence about nonverbal behavior, as will be discussed, (ii) integrated with background knowledge from other sources, and (iii) addressed to appropriate questions, with due regard for the likely variable and contextual nature of the psychological phenomena one is trying to learn about. Part of the appeal of some of the experimental work described/recommended below is that it scores well along these dimensions. One particular way in which some ERC differs from typical X-Phi survey experiments is that in the former, “scenarios” are, so to speak, realized materially rather than just described verbally. For example, subjects may be presented with various physical objects (toy airplanes or blickets and blicket detectors), rather than with just verbal descriptions of these. In some cases subjects may be asked to make verbal judgments about causal relations involving these objects, but they may also be asked to engage in non-verbal behavior—for example, to “activate” the objects (e.g. Gopnik and Sobel, 2000). Why does this matter? To begin with, there is substantial evidence that even adult humans sometimes respond very differently to stimuli that take the form of verbal descriptions of scenarios than to stimuli that are materially realized. As an illustration, subjects often perform differently and sometimes more optimally on statistical inference problems when presented with ecologically realistic data in the form of materially realized frequencies than when given word problems involving probabilities. Saffran et al. (1996) found that eight-month-old children can learn to segment words from fluent speech just based on statistical relationships between neighboring speech sounds, even though intelligent adults are often remarkably bad at verbally presented reasoning problems involving probabilities. As a causal reasoning illustration, Danks et al. (forthcoming) found that although certain causal judgments about verbally described scenarios are influenced by information about norms, this effect largely

assessments of strength and confidence. If researchers are going to continue to use such verbal probes, they need to pay more attention to what the probes measure.

CAUSAL REASONING



disappears when subjects are asked to base causal judgments on materially realized frequencies. A second point is that materially realized experimental designs often naturally allow for additional behavioral (and physiological/neural) measures besides the verbal responses on which X-Phi often focuses.5 Most obviously, researchers can record whether subjects succeed or not at the experimental task, and this can tell us something about the structure and extent of their causal understanding, beyond what is suggested by their verbal behavior. Other measures of non-verbal behavior and processing can also be employed—e.g. reaction times, looking time measures, neuro-imaging techniques, and so on. This helps to deal with some of the problems with exclusive reliance on survey results such as possible sensitivity to the particular choice of verbal description employed. Another important advantage of such “material” designs is that they allow for the use of subjects who cannot speak, such as preverbal children and non-human animals. Including such subjects enables comparisons of causal cognition across species and also allows one to trace developmental trajectories of aspects of causal cognition among humans, as several of the experiments described below illustrate.6 This greatly broadens the range of issues that can be explored.7 Finally, 5 Just to be clear, I am not claiming that the use of verbally described scenarios precludes the use of such other measures. For example, Joshua Greene’s well-known experiments on moral reasoning measure, in addition to subject’s verbal responses to hypothetical scenarios, both neural hemodynamic response via fMRI and reaction times (e.g. Greene et al., 2001). But (i) most experimental philosophers do not employ such alternative measures, and (ii) when the task is one that calls for non-verbal as well as verbal behavior there are typically more quantities to measure—reaction times associated with the non-verbal behavior, various physiological measures associated with this behavior, and so on. 6 Two additional points must be made. First, I am sensitive to the fact that many experimentally inclined philosophers will lack the training and resources to conduct nonsurvey experiments. The obvious solution is to collaborate with those who do have such training and resources. Second, it is of course true, as Edouard Machery has emphasized to me, that it may be technologically impossible or unethical to materially realize some scenarios—for example, one cannot explore experimentally whether subjects will push people in front of real trolleys—and in such cases the use of hypothetical scenarios may seem unavoidable. Even in such cases, however, I think it is often worthwhile to try to experimentally create morally permissible real scenarios that are analogous to the hypothetical ones. For example, one can explore how subjects choose in “real” scenarios in which money or other material goods such as food (rather than trolleys) are diverted from one set of people to another, as in Hsu et al. (2008). 7 Of course, verbal responses to vignettes also sometimes can be evaluated according to whether they are correct or not according to some normative theory. My point is not that



J . WOODWARD

materially realized experiments also naturally allow for a focus on normatively successful (or unsuccessful) performance, according to such fairly uncontroversial criteria for success as whether the subject succeeds in activating some device. This in turn provides information relevant to learning, which is important for reasons to which I now turn.

Methodological Reflections: The Role of Normative Theory, the Importance of Explaining Success, and the Function of Causal Thinking In my view, much of the most successful empirical, descriptive work on causal cognition has been guided by theories that are normative in the sense that they purport to describe how one ought to learn and reason about causal relationships, given epistemic goals such as truth, predictive accuracy, and so on. By “guided” in this context I mean (among other things) that such theories have played an important role both in suggesting experiments and in the interpretation of experimental results. Examples of such work include Gopnik et al. (2004) and Griffiths and Tenenbaum (2009), as well as other research cited below. By contrast, empirical investigations that are not seriously theory-guided at all or that seem motivated mainly by an interest in “refuting” theory-motivated claims seem to me to have produced less interesting (and often less easily interpretable) results.8 Given the usual philosophical tendency to contrast normative and descriptive considerations and the idea that the aspirations of psychology and experimental philosophy are descriptive, such claims about the role of normative theory may seem surprising. In order to motivate them, I begin with the observation that human causal cognition seems to be fairly effective in enabling us to successfully get around in and cope with the world. To the extent that this is so, there must be some story about how we are able to do this (that is, how we only “material” experiments allow for such evaluation (or that all material experiments automatically do), but rather that such experiments often can be designed in such a way that they provide uncontroversial benchmarks for successful performance. 8 Arguably, much of the psychological literature on causal attribution—e.g. Ahn et al. (1995)—illustrates this point. In my view, the absence of a generally accepted normative account of actual causation has had a negative effect on work in this area.

CAUSAL REASONING



succeed), and this is something that it would seem can only be addressed by a combination of normative theorizing and descriptive results. Here an analogy with “normative” or “ideal observer” theories of visual perception is suggestive.9 The visual system is not just a set of mechanisms for producing “visual judgments”; instead, one of the most striking features of the visual system is that, although sometimes subject to visual illusions, it produces outputs that are largely reliable in the sense of enabling us to get around in the world successfully and providing accurate enough information about our local surroundings. Any adequate descriptive theory of the visual system needs to explain this fact—that is, to explain how, from the very limited information that impinges on the retina, the brain is able to reach conclusions that are veridical enough for many practical purposes about the three-dimensional world of medium-sized objects that lies around us. To explain how this is possible, investigators rely on theories that are normative (and typically computational) in character—“normative” in the sense of specifying computations and algorithms that show how it is possible to, e.g., derive accurate information about object segmentation from information available to the visual system concerning shading, edges, and so on. To the extent that the visual system can be shown empirically to operate in accord with principles that we know are reliable in the sense of often issuing in accurate reconstruction of the visual scene before us, we have the basis for an explanation of why the visual system “succeeds.” Obviously, to carry out this project we need, among other things, to think of the visual system and the principles by which it operates in broadly functional terms—the (or a) function or “goal” of the system’s operation is to produce outputs that contain accurate enough information about aspects of the surrounding environment. Analogous projects have been pursued in connection with causal learning and causal reasoning. For example, normative theories of causal learning are available that purport to describe strategies, computations, algorithms, and background assumptions that enable one to reliably infer causal conclusions from various bodies of evidence, such as statistical information gained from passive observation and/or information about the results of interventions. One may use such normative theories

9

This analogy is due to Alison Gopnik.



J . WOODWARD

to explore the extent to which different subjects succeed or fail at various causal inference tasks, and whether, when they succeed, they do so by making use of the sorts of strategies and patterns of reasoning described by the normative theory. Examples of this sort of work include Gopnik et al. (2001) and Sobel et al. (2004). I contend it is often fruitful to investigate causal judgment in a broadly similar way. In this case the relevant normative framework will include a specification of the functions or goals of various forms of causal thinking and how distinctions that people make in causal judgment make sense in the light of these purposes (see Lombrozo, 2011, for a similar approach advocated for explanation). Empirical evidence about causal cognition is both interpreted in the light of this framework and used (in part) to evaluate different hypotheses about the considerations that guide people in making the causal judgments they do. This leads to a different focus and a different set of research questions from those that animate traditional metaphysical explorations or exercises in conceptual analysis, even for those who hold that empirical evidence is relevant to such matters. For example, rather than asking, as metaphysicians might, whether relations of double prevention (cases in which c prevents d which, had it occurred, would have prevented e) are “really” causal, or asking, as a conceptual analyst might, whether it is part of “the concept” of causation that causation requires a connecting process, and then exploring whether empirical results might somehow bear on these matters, we instead focus on questions such as the following: to the extent that people judge that some relations of double prevention are less than paradigmatically causal (in comparison with, e.g., cases of rocks shattering when hit by bottles), what point or goal might be served by this practice? Why, in terms of epistemic and other goals associated with causal judgment, do people care whether a connecting process is present or not when presented with a relation of counterfactual dependence? What factor or factors might people be tracking when they make distinctions among different relations of non-backtracking counterfactual dependence, labeling some as causal and others as non-causal? Do subjects judge that all relations of double prevention are non-causal or only some, and, if the latter, what factors seem to influence this distinction? In pursuing these sorts of functional inquiry, we need not commit ourselves to the idea that either armchair philosophers or the folk (as revealed in surveys) possess “intuitions” that serve as a special source of information or insight about causation. We need not

CAUSAL REASONING



even suppose that there is a single “concept” of causation that we are investigating, as opposed to a range of patterns of reasoning and judgments whose rationales we are attempting to understand. I will add that interventionists about causation such as me (Woodward, 2003) think that the acquisition of information relevant to manipulation and control is among the goals centrally associated with causal thinking, and we thus look (at least in significant part) to these in order to understand empirical patterns in people’s causal judgments.10 I will try to illustrate this idea.

What Can We Expect to Learn?: The Significance of Cognitive Variability and the Enabling Role of Theory Some philosophers (e.g. Paul, 2010) suggest that empirical results from X-Phi (and presumably ERC as well) can provide information about extra-mental phenomena—about causation or knowledge, “as they are in themselves,” rather than merely how they are conceptualized by us. I will focus on the more widely accepted idea that insofar as there is a positive role for X-Phi results, it will consist in telling us something of philosophical interest about mental or psychological phenomena—how we think or reason about causation, free will, reference, and so on. Assuming that this is the goal, what might we reasonably hope to learn from survey and other sorts of experimental results about causal cognition? My first observation is that what we can expect to learn is constrained by the character of the phenomena we are investigating. Although not everyone agrees, I believe that empirical results from X-Phi itself and many other considerations strongly suggest that there often is a great deal of variation across individuals in so-called higher cognition. Moreover, the performance of single individuals often shows strong contextual effects and intra-person variation in the sense that the same

10 I acknowledge (how could I not?) that this normative, “functional” approach to understanding various forms of cognition can be misused, with the invention of ad hoc just-so stories that attempt to rationalize this or that aspect of human thinking as serving some function where there is no independent evidence for this claim. This is a reason for being critical and tough-minded in evaluating such claims, but not for eschewing them altogether.



J . WOODWARD

individual may make different judgments and use different cognitive strategies even in connection with what may look to researchers like closely related tasks. As illustrations of both points, there is good evidence that some significant proportion of experimental subjects (at least in our culture) reason in connection with some problems as though they are good Bayesians, while other subjects use simpler, non-Bayesian strategies (e.g. reinforcement learning) in updating their beliefs.11 In addition, the same subject may also reason correctly, by Bayesian standards, when presented with evidence in certain formats in certain contexts or in connection with certain problems and yet violate Bayesian rationality in other circumstances. For this reason, in my view it is usually not illuminating to ask such broad-brush questions as whether people are Bayesian or not—the correct view of “how Bayesian we are” will need to be far more nuanced. As another example, illustrated by the experiments of Walsh and Sloman and Lombrozo, we may find that a majority of subjects in one experiment judge that certain examples of double prevention do not amount to causation, while a majority of subjects in a different experiment judge that other examples involving double prevention are causal. This does not, in my opinion, show that ordinary people are confused or inconsistent or that the two groups of subjects operate with different concepts of causation. Rather, it shows that their thinking about double prevention is more complex and subtle than anything captured by simply asking whether “our” conception of causation is such that we regard double prevention relations as causal.12 To the extent that the sort of variability and context-dependence just described is a feature of human cognition, it raises obvious questions about how we should think about the psychological phenomena that researchers hope to detect in X-Phi experiments (and in similar

11 See, e.g., Grether and El-Gamal (1995). The positive case that human subjects behave as Bayesians in causal learning tasks is made by Josh Tenenbaum in many papers—see, e.g., Griffiths and Tenenbaum (2009). 12 In a very interesting paper, Genone and Lombrozo (2012) show that with respect to concept reference, subjects are neither pure causal theorists nor pure descriptive theorists but that instead the same individuals rely on both descriptive and causal information to varying degrees in different contexts. My suggestion is that a similar pattern will be found for many other concepts of philosophical interest—reliance on “mixed” or “hybrid” theories, with considerable intra-individual variation depending on context.

CAUSAL REASONING



experiments in cognitive psychology). These questions are particularly pressing because, as already indicated, both in X-Phi and in related discussion there is a tendency to express results as claims about “concepts” (as in “the folk concept of free will is incompatibilist” or “our concept of causation requires the presence of a connecting process between cause and effect”), where the assumed notion of “concept” is what Wilson (2006) calls the “classical” notion. According to this notion, concepts are relatively stable, context-independent, widely shared informational structures with crisp and well-defined application conditions that are readily accessible to users. The central problem with this picture, as Wilson argues, is that a great deal of human cognition seems far more fluid, flexible, and context-sensitive than what is described by the classical theory; strategies and representational structures may be more or less well adapted to local cognitive niches but may perform less well outside of these contexts, with the result that they are sometimes replaced with new strategies and structures in new contexts. Very often the successful deployment of a reasoning strategy or “concept” rests on a background edifice of empirical facts in a way that is very unobvious to users and that may only become apparent when they attempt to move outside the usual range of application of the strategy or concept. One consequence is that it may be difficult to settle, in a non-arbitrary way, what sorts of assumptions or commitments are truly “constitutive” of a concept or when one is dealing with a single concept that has been stretched or modified in an effort to adapt it to new circumstances and when instead we are dealing with several different concepts. This point has obvious implications for such issues as whether we operate with several different concepts of causation, as Hall (2004) has claimed, or only one.13 Wilson’s claims about concepts and the empirical results about variation and context-dependence in cognition, gestured at above, seem to me to be mutually reinforcing. Taken together, they suggest that it may 13 Another consequence is that verbal probes involving science-fictionish scenarios (e.g. “true-temp” as a possible case of knowledge, alleged causal relationships involving magic) that violate background empirical assumptions for the application of a concept may produce results that tell us little about the structure of that concept. To the extent that people operate with non-classical concepts, requiring for their successful application, empirical assumptions of which users may be unaware, X-Phi results are more likely to tell us something about such concepts when non-outlandish scenarios are employed.



J . WOODWARD

not be fruitful to try to express “positive” results in X-Phi (and related results in psychology) as claims about “our concept of X,” at least if the notion of concept in play is anything like the classical one and the candidate Xs are at the level of generality of, e.g., “causation” or “free will.” I would add that there are many interesting issues and problems connected to causal cognition that are not naturally viewed as having to do with our concept of causation in any sense, even if this is construed in a non-classical way—these include, for example, many issues about causal learning and causal reasoning strategies. They also include the role of various default assumptions. The latter may be very important in guiding causal reasoning in particular contexts, even if the assumptions are not “constitutive” of any concept of cause shared by all or most of us.14 The moral that I draw from these observations is not that we should give up on the empirical psychology of higher cognition or that any attempt to generalize in this area is misguided, but rather that we need to reconceive the sorts of claims we can hope to establish. Applied to the subject of causal cognition, I suggest that we should be skeptical of (and not aim at either trying to establish or refute) sweeping generalities such as “according to the folk concept of causation absences are causes” or “the folk have two distinct concepts of causation: dependence and production.” Instead, I suggest that it is more fruitful to explore empirical claims of the following general sorts: • Claims about what subjects of various kinds “can do” or “do with some frequency,” particularly when it can be shown that other sorts of subjects rarely or never do this. For example, as will be discussed, many adult humans are able to distinguish between intervening and conditioning in normatively appropriate ways, learn causal relationships from their own and others’ interventions, and combine evidence from interventions and other sources in making causal judgments. However, by no means all adults do this, and other subjects (e.g. non-human primates or young children) may not be 14 For example, as Kushnir and Gopnik (2007) show, small children tend to assume as a default that many causal relations require spatial contact between cause and effect (or at least they learn such relations more readily), but they also readily give up this expectation when the evidence indicates, and learn causal relationships in which there is no spatial contact. If we simply ask whether it is part of the children’s “concept” that causes must be in spatial contact with their effects, the answer is apparently “no,” but this question misses the role that the default assumption plays in the children’s reasoning.

CAUSAL REASONING



able do the latter at all. Results of this sort about what subjects can do are related to the ideas already discussed about the role of normative theory in explaining successful performance. Again, we need not express such claims as claims about the concepts possessed by most or all subjects—we can instead just talk in terms of abilities and reasoning patterns that many subjects possess, how these conduce to success or failure given the goals they have, and so on. • Claims of the following sort: although subjects do not always do X (where X may be, e.g., judging or reasoning in a certain way), when they do X, this is how most or a substantial number of them do it. For example, by no means all subjects behave as Bayesians when faced with causal inference problems, but there is evidence that some significant number do, and that when they do they are able to make accurate causal judgments by taking base rate information into account in the way prescribed by Bayes’s theorem—see Sobel et al. (2004). • Claims about factors affecting variation in judgment, reasoning, learning, and so on.15 In other words, rather than trying to establish quasi-universal claims such as “the folk concept of causation has feature F” (and perhaps treating apparent deviations from this concept as noise or error), one should instead focus on variations or differences in judgment etc. across different contexts and kinds of subjects and try to understand the factors affecting this variation.

The enabling role of theory Philosophers of science who study experimentation recognize many possible interactions between “theory” and “experiment.” The use of experiments to “test” theories is just one possibility. I observed above that normative/philosophical theories of causal cognition can suggest features F such that if (as an empirical matter) we find them in people’s reasoning, their presence can help to explain why that reasoning is successful, to the extent that it is. In a number of cases, it probably would not occur to anyone to do an experiment exploring whether these

15

As Machery has pointed out to me, a fair amount of work in X-Phi takes this form, although more universalistic claims are also not uncommon.



J . WOODWARD

features are present in the absence of the normative theory.16 This represents another respect in which philosophical or normative theory can play a positive or constructive role in experimental work: theory can enable or motivate or provide a rationale for doing certain experiments and a basis for interpreting their results. The results themselves may be interesting, surprising, and valuable even for those who are skeptical of the motivating theory understood as a highly general hypothesis—it is in this sense that the role of the experiment is not just one of testing. This sort of construal will seem particularly natural if, as urged above, we think of many experimental results as more like demonstrations of what subjects can do or do with some frequency than as tests of hypotheses regarding what subjects always do. An illustration is provided by the results due to Bonawitz et al. showing that four-year-olds are able to use correlations derived from passive observations to design novel interventions, while two-year-olds are not (or at least find this task far more difficult). Such results should seem interesting even to those who are skeptical of interventionism as a general theory of causation or causal judgment.17

Interventionism With this as background, I turn to a more detailed look at some empirical work on causation, beginning with some investigations that can be thought of as suggested or motivated by a broadly “interventionist” account of causation. According to this account, causal claims describe relationships that are potentially exploitable for purposes of manipulation and control. The version of this idea defended in Woodward (2003) proposes the following connection between causation and intervention:

16 Of course I do not mean that in such cases it is logically impossible for anyone to think of the experiment in the absence of the motivating theory—merely that as a matter of empirical fact this usually does not happen. 17 The converse possibility is also worth mentioning. Suppose we find experimentally that people’s causal cognition, when successful, exhibits feature G, where G is some feature that is not assigned a role in any current normative theory. This might motivate us to construct a new normative theory that explains the role of G in successful causal cognition. For what it is worth, the ideas about “causal specificity” developed in Woodward (2010) had this sort of origin. I first noticed that biologists seemed to care about whether causal relations were specific or not without having any idea about whether there was any normative rationale for their doing so. This in turn suggested a search for such a rationale.

CAUSAL REASONING



(M) Suppose C and E are variables. Then C causes E if and only if there is some intervention that changes the value of C, such that if that intervention were to occur, the value of E or the probability distribution of E would change.

This notion corresponds to what Woodward (2003) calls a “total cause,” which has to do with the total or overall effect of one variable on another.18 An intervention on C with respect to a second variable E is an exogenous change in the value of C that is appropriately unconfounded from the point of view of inferring whether there is a causal connection from C to E. Roughly speaking, an intervention on C with respect to E changes C in such a way that any change in E, should it occur, occurs only through the change in C and not in some other way— any “route” by which E changes goes through C. (For details, see Woodward, 2003, pp. 98ff). Randomized experiments are one kind of intervention. As an empirical matter some human actions qualify as interventions, but from the point of view of normative theory what makes a process count as an intervention is the causal structure of that process, whether or not it involves human action. A crucial feature of this framework is that intervening on a variable is different from conditioning on it. In a common cause structure in which C is a common cause of E1 and E2, these last two variables will be correlated. However, intervening to “set” the value of E2, via an intervention I, breaks the causal connection between C and E2, with the result that E1 and E2 are no longer correlated under this intervention. Within the interventionist framework it is facts about how E2 responds to an intervention on E1 that track whether E1 causes E2. (M) should not be understood as the obviously false claim that the only way to learn about causal relationships is by performing interventions. It is consistent with (M) that there are many different possible sources from which one can learn about causal relationships, including learning from passive observation (not involving interventions). Construed as a normative claim, (M) implies that when one learns about a causal relationship (from whatever source), one should think of the 18 This notion of total cause is meant to capture the idea of one variable having a nonnull aggregate or overall impact on another variable over all the various routes or paths connecting to the two variables. A variable can have a causal influence on another along a particular path (and hence be what Woodward, 2003, calls a “contributing cause”) even if it has no aggregate or total influence, as when influences along two paths cancel. The “only if” part of M holds for total causes but not for contributing causes (Woodward, 2003, p. 59.)



J . WOODWARD

content of what is learned in accord with (M). Moreover, our reasoning about causal relationships (from whatever source we learn about them) ought to be guided by the commitments implicit in (M). My interest in what follows is not in defending (M) as a normative theory (for that see Woodward, 2003), but rather in exploring the extent to which, in addition to its normative credentials, (M) might figure in a descriptive empirical theory of how at least some subjects think and judge. For example, the close connections between causal claims and claims about what happens under interventions embodied in (M) naturally suggest empirical questions such as the following: • Can/do subjects learn about causal relationships in a way that respects the normative connection between causation and intervention embodied in (M)? That is, do subjects draw the causal conclusion suggested by (M) when given evidence that Y changes in value under an intervention on X? • Are subjects sensitive to the differences between causal conclusions that are warranted when given evidence that X changes under an intervention and the conclusions warranted when X changes as a result of some process that is “confounded” and hence not an intervention? • Are subjects sensitive to the normative difference between intervening and conditioning? For example, are they more willing to infer that X causes Y when given evidence that Y changes under interventions on X than when given evidence that X and Y are correlated (with no information about interventions)? • Do subjects treat their own actions as “default” interventions in the sense that when such actions produce a change in X that is associated with a change in Y, they tend to infer that X causes Y, in the absence of evidence that their actions are confounded in some way? Questions of this sort can be explored empirically for many different sorts of subjects, including humans of various ages and non-human animals. There is evidence that the answer to all four questions is “yes” for adult human beings. (See e.g. Steyvers et al., 2003; Lagnado and Sloman, 2004; Sloman and Lagnado, 2005.) Somewhat amazingly, at least to me, it also has been claimed on the basis of recent experiments that the answer to the third question is “yes” when the subjects are rats,

CAUSAL REASONING



although the interpretation of this experiment is controversial and there is no evidence that rats use co-variational information to design novel interventions in the way that humans are able to do (Blaisdell et al., 2006).19 These are some simple illustrations of experiments/empirical investigations “suggested” or motivated by a philosophical, normative theory connecting causation and intervention. In the absence of such a normative theory, it is less likely that it would have occurred to anyone to do these experiments.20 At the same time, when we find, experimentally, that people behave in the ways described, this suggests that at least sometimes they are operating with ways of thinking about causation that are connected to intervention in the way that normative theory suggests. This in turn suggests that the connection between causal judgment and our interest in manipulation and control is not just an

19 Standard “associationist” models of learning, such as the Rescorla—Wagner model and its descendants, are insensitive to the difference between intervening and conditioning in the sense that they treat information from both sources equivalently. Since human beings are sensitive to this distinction (and it is normatively correct for them to be so), this is strong prima facie evidence that the structure of human causal knowledge cannot be understood in purely associationist terms. It is an attractive feature of the interventionist framework that sensitivity to the differences among causal structures that imply the same independence and conditional independence relations but differ in what they predict would happen under different possible interventions can be regarded as an important diagnostic of whether subjects possess distinctively causal representations rather than mere representations of patterns of association. 20 David Danks has objected that while the general idea that there is a connection between causation and manipulation (or action) may well have played a role in motivating these and other experiments, more specific ideas about the connection between causation and intervention in the sense of Woodward (2003) or (presumably) of the sort found in Pearl (2000) or Spirtes et al. (2000) were not required. To this I have several responses. First, the connection that I claim is not that possession of a motivating theory was required as a matter of logic for anyone to do the experiments, but rather that possession of the theory as a matter of causal empirical fact prompted or motivated people to do the experiments. And while it is certainly true that the general idea of a connection between causation and manipulation has been around for a long time, I do not think that the impact of wellworked-out and novel computational ideas of the sort found in Spirtes et al. and in Pearl should be underestimated—these helped to turn the vague (and in philosophical circles much criticized) idea that there was such a causation/manipulation connection into something that looked conceptually and mathematically respectable. In my opinion, this in turn influenced experimental practice. Finally, there certainly are experiments that draw on specific ideas about what an intervention involves—for example, experimental demonstrations that subjects are sensitive to whether their manipulations are confounded in making causal inferences, as in Kushnir and Gopnik (2005).



J . WOODWARD

arbitrary association made by some philosophical theory but connects to features actually present in people’s causal judgments.

Toddlers as Agent Causal Learners A more extended illustration of these themes is provided by some of the results of an experiment in which I participated (Bonawitz et al., 2010), but here a bit more background is required. Woodward (2007) observes that once one begins thinking about causation in interventionist terms, it is natural to distinguish conceptually among the following three possibilities: 1) Egocentric causal learning: a subject might be able to learn about causal relationships from her own interventions but not from other sources. This might happen either because (i) the subject is not able to learn causal relationships at all from observing the results of other agent’s interventions or from observation of contingencies not involving interventions, or because (ii) she may learn such relationships but fail to integrate them into a single unified representation, not recognizing, e.g., that the results of the interventions of others provide evidence for what would happen if she were to perform similar interventions. 2) Agent causal learning: the subject grasps that the same relationship she exploits in intervening also can be present when other agents act. Accordingly, the subject learns causal relationships both from observing the results of her own interventions and from observing the results of the interventions of other agents, recognizing, for example, that the results of the interventions of others provide evidence for what would happen if she were to perform similar interventions. But the agent does not similarly learn causal relationships from “passive” observations of contingencies not involving interventions or fails to integrate these with what is learned from observing interventions. 3) Fully (or more nearly fully) causal learning: the subject grasps that the same relationship she exploits in intervening also can be present both when other agents intervene and in nature even when no other agents are involved, and she integrates information from each of these sources into a common unified representation. One empirical test for the presence of this ability is whether a

CAUSAL REASONING



subject will perform appropriate novel interventions to obtain some goal on the basis of an observed association between events not involving interventions; to use Tomasello and Call’s (1997) example, would an ape, observing the wind shaking fruit loose from a tree, be able to infer that if it were to shake the branch, this would cause fruit to fall to the ground? Adult humans are plainly full causal learners. An interesting empirical question, suggested by the distinctions above, is which of the three categories most closely captures the causal cognition of non-human animals or young human children. Some primatologists, including Tomasello and Call, have in effect claimed that non-human primates are not full causal learners—a claim which, if correct, would mark a fundamental difference between human and non-human causal cognition. But what about very young humans—babies and toddlers? Is there a stage in human development in which young children behave as egocentric or agent but not fully causal learners? Bonawitz et al. (2010) reports experimental results addressing this question. Compressing greatly, we compared children in two different age groups (toddlers vs. preschoolers, mean ages in first experiment 47.2 vs. 24.4 months) who observed a block sliding toward a toy airplane which activated when the block touched its base. In the “ghost condition” the block appeared to move spontaneously, with no agent involved, activating the plane on touching it. In another condition (the agent condition) an experimenter moved the block, demonstrating that it would activate the plane on touching. The children were then asked to make the airplane go by themselves. A very large majority of preschoolers were able to use their previous observations of the association between the movement of the block and the activation of the plane in the ghost condition to intervene on their own to activate the plane. By contrast, none of the toddlers was able to do this, although they were able to successfully activate the plane via their own interventions when they observed another agent do this. We interpreted this and other experimental results as prima facie evidence that there is a stage in the development of human causal cognition in which children (in this case toddlers) are able to learn to design their own interventions from observations of the interventions of other agents, but not from the observation of otherwise similar associations not involving interventions. In other words, at this point the



J . WOODWARD

children appear to be in something like an agent causal stage. This is then followed by a stage in which fully causal learning is achieved, as evidenced by the behavior of the older children. This interpretation is consistent with a great deal of other developmental evidence showing that even very young children seem to be sensitive to the difference between agents and non-agents in their environments and are primed for or adept at learning from the behavior of other agents (particularly the intentional or goal-directed actions of other agents) as opposed to other sources of information. At least one member of our collaboration (Meltzoff, 2007) thinks that as soon as children represent their own actions they also represent the actions of others in a common amodal code, and hence in effect they are never in a purely egocentric stage of causal learning. We may think of this as yet another example of an experiment motivated or suggested by a normative theory: unless one has the “philosophical/computational” idea that there is an important connection between causation and intervention and that some human actions are paradigmatic interventions, one is unlikely to undertake the sort of investigation just described. Given this idea, and the accompanying thought that one sometimes learns about causal relationships from interventions, it is natural to ask how and when in development information from interventions is integrated with other sources into more unified causal representations. This experiment may also be used to illustrate how influence can flow in the opposite direction—how empirical results can have implications for philosophical theorizing. Consider an “agency” theory of causation like that defended by Menzies and Price (1993), according to which our concept of causation is derived from our subjective experience of agency, which is then “projected” onto the world. When this is interpreted as an empirical claim, experimental results of the sort described above suggest there is something broadly right about the idea, in the sense that our capacities to act as agents and to manipulate play an important role in the development of our capacities for causal learning and cognition. On the other hand, these experimental results also suggest the need for a somewhat more complex story than the one told by Menzies and Price. To see this, recall a standard philosophical criticism of agency theories: they face problems explaining how the notion of causation comes to be extended to unmanipulable causes such as earthquakes, where no one

CAUSAL REASONING



has the relevant experience of agency—a criticism which Menzies and Price address. However, the experimental results above suggest that a related problem arises, so to speak, much earlier: there is already a problem about how causal learner X moves from her own subjective experience of agency to a recognition that causal relations are present when other agents act and produce effects, since in such cases X presumably does not have the experience of agency. Moreover, there is a parallel problem about cases in which the subject infers the existence of a causal relationship involving causes that are straightforwardly manipulable, but which are not in fact manipulated by any agent, as when wind shakes fruit from a tree. “Projection” can function as a label for our willingness to extend the notion of causation from our own manipulations to such cases, but it provides no insight into how we develop the capacity to do this or the factors influencing such development. Moreover, if Meltzoff is correct, it may be an empirical mistake to suppose that young humans begin in a purely “egocentric” stage centered around their own experience of agency; instead, we may be agent-causal as soon as we are capable of any kind of causal learning. If so, although agency may be crucial to the development of causal cognition, it may be that what is central is not so much the experience of agency, but rather other features associated with the capacity to act effectively as an agent, such as the ability to represent means/ends structures of actions exhibited both by self and others.

Empirical Results Concerning Double Prevention I turn next to a quite different topic: some recent experimental work on double prevention. Again, these are cases in which if d were to occur it would prevent the occurrence of e (which would otherwise occur in the absence of d), and in which the occurrence of c prevents the occurrence of d, with the upshot that e occurs. In Ned Hall’s well-known example (2004), Suzy’s plane will bomb a target (e) if she is not shot down by an enemy pilot (d). Billy, piloting another plane, shoots down the enemy (c), and Suzy bombs the target. In such cases there is overall counterfactual dependence (of a nonbacktracking sort) of e on c. This is taken to be sufficient for causation by



J . WOODWARD

many counterfactual theories, as well as by (M). Nonetheless, many philosophers (and non-philosophers) find it intuitive that cases of double prevention either do not involve causation at all, or at least lack some feature which is central to some other cases of causation—e.g. the presence of a connecting process or transmission of energy/momentum. (Following Hall, 2004, I will describe these as involving “production.”) By contrast, other philosophers such as Schaffer (2000) emphasize that many biological mechanisms (from muscle contraction to gene regulation) and many manufactured artifacts (e.g. many guns) operate by relations of double prevention and that these relations can appear paradigmatically causal. This raises several questions. First, one might wonder as a matter of descriptive psychology to what extent, if any, people regard relations of double prevention as causal. Is it possible that some cases of double prevention are regarded as more paradigmatically causal than others? Second, to the extent that people distinguish between production and “mere” counterfactual dependence, why do they do so? What point or purpose is served by this practice? Walsh and Sloman (2011) presented subjects with a series of scenarios. In one, a coin stands unstably on edge, about to fall tails. Billy and Suzy roll marbles in such a way that, if nothing interferes, each will strike the coin so that it lands heads. Billy’s marble strikes first. In this scenario, 74% of subjects judged that Billy’s marble caused the coin to land heads. In a second scenario, an unstable coin is again on edge, about to land heads. A third party rolls a marble in such a way that if it strikes the coin, it will land tails. However, a book blocks the path of the marble. Frank removes the book, but if he had not, Jane would have. The marble strikes the coin and it lands tails. In this scenario only a minority of subjects (38%) judge that Frank caused the coin to land tails, and virtually no one judges that Jane caused this effect.21

21

Although less than half, 38% is a non-trivial number. How should we think about such subjects? Are they making a mistake, and misapplying “our concept” of causation, the content of which can be determined by the majority who did not regard this relationship as causal? Are they perhaps operating with a different “concept” of causation than the majority? The unsatisfactory character of either of these options illustrates the appeal to the alternative approach recommended above, in which one replaces these questions with such questions as why subjects sometimes distinguish between double prevention and other sorts of causal relations.

CAUSAL REASONING



Walsh and Sloman’s Billy/Suzy scenario is a case of causal pre-emption in which the effect is not counterfactually dependent in any simple, straightforward way on the cause, but in which transmission or a connecting process is present between Billy’s throw and the coin’s falling heads. In the second scenario, the relation of Frank’s action to the coin’s falling tails is double-prevention-like, with the added complication that because Jane’s action is a pre-empted double preventer, the coin’s falling tails is not counterfactually dependent on Frank’s action. Although Walsh and Sloman do not report any results about experiments with a double prevention structure without a second, pre-empted preventer (i.e., the same scenario as above with Jane absent), it is a reasonable guess that fewer subjects would judge that Frank’s action was a cause in such cases than in a variation on the first scenario in which only Billy is present. These results thus seem to support the conclusion that as far as folk thinking about causation goes, cases in which a “connecting process” is present are more likely to be judged as causal than cases of double prevention. Indeed, this is one of the conclusions that Walsh and Sloman draw. I will return to this claim, but I want first to describe some additional experiments, due to Lombrozo (2010), that complicate matters in an interesting way. Lombrozo’s experiments explore people’s causal judgments about double prevention scenarios in contexts involving intentional action, artifacts with designed functions, and biological adaptations. Although Lombrozo finds, in agreement with Walsh and Sloman, that subjects are more willing to describe cases involving production as causal than (at least some) cases involving double prevention, she also finds that subjects distinguish among cases of double prevention, treating cases involving intentional action, designed function, and biological adaptation as more causal than cases of double prevention not involving these features. For example, she presents subjects with a fictitious example in which the tendency of a species of shrimp to reflect ultraviolet light depends, via a double prevention structure, on their diet. Subjects are more willing to judge that this double prevention structure is causal when they are told that this light-reflecting tendency is a biological adaptation than when it has no adaptive significance, and similarly for parallel examples involving effects that are intended or the result of a designed function. Since none of Lombrozo’s cases of double prevention involve an uninterrupted connecting process, or transfer of energy/momentum from the putative



J . WOODWARD

cause to its effect, it cannot be the presence or absence of these features that accounts for subjects’ willingness to regard some cases as more paradigmatically causal than others. Following a suggestion in Woodward (2006), Lombrozo proposes that relations of double prevention are judged as more paradigmatically causal to the extent that they are more stable. Suppose that E counterfactually depends on C, where the dependence is non-backtracking and intervention-supporting in the manner described in (M). Stability has to do with the extent to which this relation of counterfactual dependence would continue to hold as other factors in the background, in addition to C and E, change.22 Both theoretical considerations and casual observation suggest that, other things being equal, dependence relations that are more stable are, as an empirical matter, judged as more paradigmatically causal than less stable relations. This effect is present both for examples that do not involve double prevention and for examples that do,23 and provides a natural explanation of Lombrozo’s results. In particular, as argued both by Lombrozo and in Woodward (2006), it is plausible that cases of double prevention involving biological adaptations, designed functions, and intentional actions tend to be more stable under relevant changes than cases of double prevention lacking these features. For example, double prevention relations involving biological adaptations tend to be more buffered against environmental disruptions (and more 22 For reasons of space, I refer the reader to Woodward (2006) for a more detailed discussion of stability. However, some brief clarificatory remarks may be helpful. First, stability comes in degrees and is relative to particular sets of background conditions. Stability claims thus have the form: relation of counterfactual dependence C is stable under such and such changes in background conditions B (but perhaps not under other changes in B or under changes in other background conditions B0 ). The idea is thus not to dichotomize claims as stable or unstable in some absolute sense or to pick out some single set of background conditions as uniquely appropriate in assessing stability. The account recognizes that virtually all relations of dependence will hold under some circumstances and not under others. Nonetheless, both as a normative matter and as a description of how people think, some sets of changes are more important than others for purposes of assessing stability; which sets of changes have importance depends on a number of considerations having to do both with subject matter and which changes are, as a statistical matter, common or usual. In the examples involving biological adaptiveness in the discussion that follows, biologically normal changes in the environment or in systems within the organism will be particularly important for judgments of stability. 23 Cf. Lewis (1986), who compares the unstable (or, as he calls it, “sensitive”) causal relation between writing a letter of recommendation and the existence of certain people in the distant future with the much more stable relation between shooting at close range and death.

CAUSAL REASONING



stable for that reason) than double prevention relations that arise, so to speak, fortuitously. Somewhat more ambitiously, one might attempt to tell a similar story, also based on stability, about why relationships of counterfactual dependence involving transmission or connecting processes are commonly judged as being more paradigmatically causal than those lacking this feature. But why should stability matter in the way described to judgments of causal status? We can provide at least a partial answer to this question by making use of the strategy recommended above and asking how a concern with stability fits in with goals or functions of causal thinking. According to interventionist accounts, one reason why people care about the difference between causal relationships and purely correlational relationships is that the former are exploitable (at least in principle and often in fact) for purposes of manipulation/intervention and control in a way that the latter are not. Identifying relationships relevant to intervention is, as Lombrozo says, one “function” or normative aim of causal ascription, even if it may not be the only such function. Once this idea is accepted, it is also natural to recognize that different relationships of counterfactual dependence (of the non-backtracking sort) can serve this function or satisfy this aim to different degrees. In particular, the more stable a relationship of counterfactual dependence is, then (other things being equal) the more useful or suitable it is likely to be for purposes of manipulation and control, the more exportable or generalizable it is likely to be to other contexts, and the more it is likely to continue to hold in the present context in the face of changing contingencies. Thus, at least within an interventionist framework, it “makes sense” and is normatively appropriate for subjects to distinguish in the way that they do among relationships of counterfactual dependence in terms of their degree of stability. An additional attraction of this way of looking at things is that it allows us to move beyond the “dull thud of conflicting intuitions” that characterizes much philosophical discussion of double prevention and to ask some more productive questions. Rather than trying to appeal to “intuition” to settle whether double prevention relations are truly causal, we can investigate empirically the judgments people make in double prevention cases and the factors that influence these judgments, and we can also ask whether these judgments have some recognizable normative rationale. This allows us to understand the reasons for (or to what extent



J . WOODWARD

it might be rational for) people to operate with notions of causation that incorporate these features. In other words, rather than arguing about whether causation (or our concept of causation) requires a connecting process, we ask what goals or purposes might be served by distinguishing among relations of dependence according to whether a connecting process is present, whether there might be some point to distinguishing among relations of dependence that lack such a process, and so on. This seems to me to provide one way in which empirical results can be relevant to understanding causal reasoning that does not require problematic assumptions about the role of intuition or, for that matter, strong assumptions about the character of “our concept” of causation.

Conclusion No deep sociological insight is required to recognize that one factor animating current disputes about X-Phi is anxiety about the status and future of philosophy itself. Put crudely, the worry is that philosophy is going to be replaced by something more like experimental psychology, or at least that this is the future for philosophy advocated by some experimental philosophers. At least in the case of causation, this worry strikes me as misplaced; as I have tried to show, there remains lots (that is positive and constructive) for philosophers to do, even if philosophical work on causation becomes more influenced by empirical considerations. In particular, if the overall argument of this chapter is correct, one reason why purely descriptive reports of what the folk or scientists think about causation are not going to replace normative, philosophical theorizing is that we need the normative theorizing to do the descriptive work properly. More generally, although I have been skeptical of the value of asking broad-brush questions about “our concept of cause” and similar matters, I believe philosophers have accumulated a battery of distinctions and an understanding of connections among different features of causal thinking that can play a very useful role in experimental design and interpretation. However obvious these distinctions and connections may seem to philosophers, they are frequently not well understood by experimental psychologists or other non-philosophers, and this can lead to badly designed experiments and results that are difficult to interpret. As a simple illustration, not discussed in any detail above, consider the

CAUSAL REASONING



contrast most philosophers would draw between, on the one hand, (i) the project of capturing a broad, egalitarian notion of what it is for one factor to be causally relevant to another (where this contrasts with the factor being causally irrelevant, although perhaps correlated) and (ii) projects associated with causal selection, where this has to do with understanding the basis on which we select, from among those factors satisfying the broad notion under (i), one or more as, e.g., “a cause,” “the cause” etc. of some effect. The target of many philosophical theories of causation, as well as most of the normative/computational theories of causal inference outside of philosophy, is (i) rather than (ii), while (ii) seems to be the typical target of investigations of causal attribution by psychologists. Whatever the details of one’s views about (i) and (ii) it seems uncontroversial that these are different projects. If so, we should be wary of using empirical results about the considerations influencing causal selection to reach conclusions about the descriptive (much less normative) adequacy of claims having to do with (i). For example, even if it is true that causal selection is not guided by considerations having to do with counterfactual dependence, as suggested in Mandel (2003), it does not follow that counterfactual theories of causation understood as contributions to project (i), interpreted either descriptively or normatively, are thereby refuted. More generally, the philosophical distinction between (i) and (ii) should allow us to see that it should not simply be assumed that results about causal selection and actual cause judgment (and the learning and processing that underlie such judgments) automatically generalize to other sorts of causal claims. Other, similar illustrations abound.

References Ahn, W., Kalish, C., Medin, D., and Gelman, S. (1995). The Role of Covariation versus Mechanism Information in Causal Attribution. Cognition 54: 299–352. Alexander, J., and Weinberg, J. (2007). Analytic Epistemology and Experimental Philosophy. Philosophy Compass 2, 56–80. Alexander, J., Mallon, R., and Weinberg, J. (2010). Accentuate the Negative. Review of Philosophy and Psychology 1: 297–314. Blaisdell, A.P., Sawa, K., Leising, K.J., and Waldmann, M.R. (2006). Causal Reasoning in Rats. Science 311(5763): 1020–2. Bonawitz, E., Ferranti, D., Saxe, R., Gopnik, A., Meltzoff, A., Woodward, J., and Schulz, L. (2010). Just Do It? Investigating the Gap between Prediction and Action in Toddlers’ Causal Inferences. Cognition 115: 104–17.



J . WOODWARD

Collins, D. and Shanks, D. (2006). Conformity to the Power PC Theory of Causal Induction Depends on the Type of Probe Question. The Quarterly Journal of Experimental Psychology 59: 225–32. Danks, D., Rose, D., and Machery, E. (forthcoming). Demoralizing Causation. Genone, J., and Lombrozo, T. (2012). Concept Possession, Experimental Semantics, and Hybrid Theories of Reference. Philosophical Psychology 25: 717–42. Gopnik, A., and Sobel, D. (2000). Detecting Blickets: How Young Children use Information about Novel Causal Powers in Categorization and Induction. Child Development 71: 1205–22. Gopnik, A., Sobel, D., Schulz, L., and Glymour, C. (2001). Causal Learning Mechanisms in Very Young Children: Two, Three and Four-year-olds Infer Causal Relations from Patterns of Variation and Covariation. Developmental Psychology, 37, 620–6. Gopnik, A., Glymour, C., Sobel, D., Schulz, L., Kushnir, T., and Danks, D. (2004). A Theory of Causal Learning in Children: Causal Maps and Bayes Nets. Psychological Review 111: 3–32. Greene, J.D., Sommerville, R.B., Nystrom, L.E., Darley, J.M., and Cohen, J.D. (2001). An fMRI Investigation of Emotional Engagement in Moral Judgment. Science 293: 2105–8. Grether, D., and El-Gamal, M. (1995). Are People Bayesian? Uncovering Behavioral Strategies. Journal of the American Statistical Association 90: 1127–45. Griffiths, T., and Tenenbaum, J. (2009). Theory-Based Causal Induction. Psychological Review 116: 661–716. Hall, N. (2004). Two Concepts of Causation. In Collins, J., Hall, N., and Paul, L.A. (eds.), Causation and Counterfactuals. Cambridge, MA: MIT Press. Hsu, M., Anen, C., and Quartz, S. (2008). The Right and the Good: Distributive Justice and Neural Encoding of Equity and Efficiency. Science 320: 1092–5. Kitcher, P. (2011). The Ethical Project. Cambridge, MA: Harvard University Press. Knobe, J. (2003). Intentional Action in Folk Psychology: An Experimental Investigation. Philosophical Psychology 16: 309–23. Kushnir, T., and Gopnik, G. (2005). Young Children Infer Causal Strength from Probabilities and Interventions. Psychological Science 16: 678–83. Kushnir, T., and Gopnik, A. (2007). Conditional Probability versus Spatial Contingency in Causal Learning: Preschoolers use New Contingency Evidence to Overcome Prior Spatial Assumptions. Developmental Psychology 43: 186–96. Lagnado, D., and Sloman, S.A. (2004). The Advantage of Timely Intervention. Journal of Experimental Psychology: Learning, Memory and Cognition 30: 856–76. Lehrer, K. (1990). Theory of Knowledge. Boulder, CO: Westview Press.

CAUSAL REASONING



Lewis, D. (1986). Postscripts to ‘Causation’. Philosophical Papers, Vol 2. Oxford: Oxford University Press. Lombrozo, T. (2010). Causal-explanatory Pluralism: How Intentions, Functions, and Mechanisms Influence Causal Ascriptions. Cognitive Psychology 61: 303–32. Lombrozo, T. (2011). The Instrumental Value of Explanations. Philosophy Compass 6: 539–51. Machery, E., Mallon, R., Nichols, S., and Stich, S. (2004). Semantics, CrossCultural style. Cognition 92: B1–12. Mandel, D. (2003). Judgment Dissociation Theory: An Analysis of Differences in Causal, Counterfactual, and Covariational Reasoning. Journal of Experimental Psychology: General 132: 419–34. Meltzoff, A. (2007). Infants’ Causal Learning: Intervention, Observation, Imitation. In Gopnik, A., and Schulz, L. (eds.), Causal Learning: Psychology, Philosophy, and Computation. Oxford: Oxford University Press, pp. 37–47. Menzies, P., and Price, H. (1993). Causation as a Secondary Quality. British Journal for the Philosophy of Science 44: 187–203. Nichols, S., and Knobe, J. (2007) Moral Responsibility and Determinism: The Cognitive Science of Folk Intuitions. Noûs 41: 663–85. Paul, L. (2010). A New Role for Experimental Work in Metaphysics. Review of Philosophy and Psychology 1: 461–76. Pearl, J. (2000). Causality. Cambridge: Cambridge University Press. Saffran, J., Aslan, R., and Newport, E. (1996). Statistical Learning by 8-month-old Infants. Science 274: 1926–8. Schaffer, J. (2000). Causation by Disconnection. Philosophy of Science 67: 285–300. Scholl, B. (2008). Two Kinds of Experimental Philosophy and their Methodological Dangers. (accessed 8 November 2012). Sloman, S., and Lagnado, D. (2005). Do We “Do”? Cognitive Science 29: 5–39. Sobel, D.M., Tenenbaum, J.B., and Gopnik, A. (2004). Children’s Causal Inferences from Indirect Evidence: Backwards Blocking and Bayesian Reasoning in Preschoolers. Cognitive Science 28: 303–33. Spirtes, P., Glymour, C., and Scheines, R. (2000). Causation, Prediction, and Search. Cambridge, MA: MIT Press. Steyvers, M., Tenenbaum, J., Wagenmakers, E., and Blum, B. (2003). Inferring Causal Networks from Observations and Interventions. Cognitive Science 27: 453–89. Swain, S., Alexander, J., and Weinberg, J. (2008). The Instability of Philosophical Intuitions: Running Hot and Cold on Truetemp. Philosophy and Phenomenological Research 56: 138–55.



J . WOODWARD

Tomasello, M., and Call, J. (1997). Primate Cognition. New York: Oxford University Press. Walsh, C., and Sloman, S. (2011). The Meaning of Cause and Prevent: The Role of Causal Mechanism. Mind and Language 26: 21–52. Wilson, M. (2006). Wandering Significance: An Essay on Conceptual Behavior. Oxford: Oxford University Press. Woodward, J. (2003). Making Things Happen: A Theory of Causal Explanation. New York: Oxford University Press. Woodward, J. (2006). Sensitive and Insensitive Causation. Philosophical Review 115: 1–50. Woodward, J. (2007). Interventionist Theories of Causation in Psychological Perspective. In Gopnik, A., and Schulz, L. (eds.) Causal Learning: Psychology, Philosophy and Computation. New York: Oxford University Press, pp. 19–36. Woodward, J. (2010). Causation in Biology: Stability, Specificity, and the Choice of Levels of Explanation. Biology and Philosophy 25: 287–318.

12 Experimental Philosophy 1935–1965 Taylor Shaw Murphy, Washington University, St. Louis

Introduction In the heydays of linguistic philosophy, a group of Norwegian philosophers saw that philosophers often relied on intuitions when speaking about the conceptual commitments and linguistic uses of ordinary people, despite enduring disagreement on the matter. Then there was the critical insight: “When philosophers offer conflicting answers to questions that have empirical components, empirical research is needed” (Naess, 1953a, p. vii). Calling themselves empirical semanticists, this group stood opposed to the “anti-empirically oriented armchair philosopher” and pointed “to the possibility of an ‘experimental philosophy’ ” (Tennessen, 1964, p. 290; Naess, 1938a, p. 161). They argued that it is “hardly sufficient that a single person registers his own reactions to this or that sentence, or makes pronouncements based on intuitions, or undertakes scattered observations of others’ usage” (Gullvåg, 1955, p. 343). And so off they went, testing the claims made by philosophers

Special thanks to Adam Morton and Jeff Pelletier’s summertime seminar on empirical semantics and experimental philosophy at the University of Alberta in 2009. Here I became aware of the manuscripts, seminar notes, and original studies that were left by Herman Tennessen, which served as valuable material for understanding the research at Oslo and at Berkeley. Some of this material has been placed online on my website, and is otherwise available on request. I would also like to thank the anonymous reviewers and editors for their insightful and helpful comments.



T . S . MURPHY

and investigating how concepts are understood, defined, and used by ordinary people. Empirical semantics is deeply similar to today’s experimental philosophy; it was linguistic philosophy’s experimental philosophy. Although the empirical semantics movement is not historically continuous with today’s experimental philosophy movement, empirical semanticists engaged with much the same subject matter, had similar motivations and aims, and encountered characteristic problems and objections at the interface between analytic philosophy and experimental psychology. This chapter spells out these connections and situates empirical semantics within the context of experimental philosophy and analytic philosophy. Arne Naess led the empirical semantics movement. The most renowned Norwegian philosopher today, Naess was honoured with a state funeral and is well remembered both for pioneering the ecological movement and for his activism in the international peace movement (Stadler, 2009). Naess is also credited with bringing social science methods into Norway, and empirical semantics is considered an intellectual forerunner to sociolinguistics and corpus linguistics (Chapman, 2008, 2011; Thue, 2009). Along with Naess, the central empirical semanticists were Herman Tennessen, a psychologist with interests in logical analysis; Ingemund Gullvåg, a logician; and Harald Ofstad, a legal scholar turned moral philosopher (Ofstad, 1951, pp. 42–3; Tennessen, 1962). This chapter is organized into three sections, roughly corresponding to three periods of empirical semantics research. I begin with the original inspiration for empirical semantics (around 1935–38), and survey its reception over the years. It grew from Naess’s own interests and his encounters with those he met in Vienna. Naess had an ongoing interest in behavioural psychology and the concept of truth, and gained further motivation to study the notion of truth experimentally in conversation with Tarski and others at the Vienna Circle. The overarching factor was the apparent reliance on intuition when assessing the way that terms are defined, conceived, and used by ordinary people. Naess’s work over the years was met with resistance, largely due to antipsychologistic (or antinaturalistic) sentiments and arguments that became central to analytic philosophy when it parted ways with experimental psychology around the turn of the century. Empirical semanticists pointed to the

EXPERIMENTAL PHILOSOPHY

– 

broadly empirical traditional conception of philosophy that existed prior to the rise of analytic philosophy. While the distinction between analytic conceptual analysis and synthetic sciences such as psychology was sensible, it was often used to depreciate empirical research on traditional philosophical topics. My account of what transpired in Vienna and Naess’s intellectual development at this stage is drawn largely from his recollections, which were penned after World War II. In the next section of the chapter I survey the work done by empirical semanticists in Oslo (around 1939–56). I sketch some basic ideas and methods and place this work in its social and political context. Despite the emphasis in sociopolitical topics, there was a vibrant research program on topics in analytic philosophy. The experimental studies here focused on understanding how people conceive of truth, democracy, synonymy, consciousness, and “testability” among physicists, to name just a few. I concentrate on one study of typological concepts for illustration, and note some contrasts with experimental philosophy. Perhaps most striking is that there was no attempt to access analyticities or conceptual truths from the experimental data—it was regular science and not, as it were, prior to it. However, it also embodied a unique analytical and philosophical approach of its own. For my sources here, I began with Tennessen’s course textbooks and accounts by the empirical semanticists about their own research, as well as some archival material that includes original studies and data (as cited below). For the illustration on typological concepts, I partially translated some of Tennessen’s work from Norwegian and used Tennessen’s English summary (now available online). Lastly, I turn to the period of work that followed at Berkeley (around 1957–61). Whereas the work in Oslo consisted in efforts to understand how people think about specific topics, at Berkeley the focus turned toward explanation of certain patterns of intuitive judgments about language found in analytic philosophy. As destiny would have it, there was a clash of sorts between empirical semanticists and ordinarylanguage philosophers, though this clash was also quite constructive. Although logical analysis could be distinguished from empirical semantics, ordinary-language philosophy certainly appeared to be about the same subject matter: ordinary language. Some of this work is obscure because it suddenly stopped around 1960, and much of it was done in connection with J.L. Austin (who died in February 1960). This section



T . S . MURPHY

draws heavily on archival material from this period, including seminar notes written at Berkeley, and drafts of work (some of which were not published). Again, much of this work is now online or available on request.

Experimental Philosophy at the Vienna Circle Much happened in this period, so a brief timeline of Naess’s activities is in order (see Stadler, 2009). Naess spent his student years in Paris studying philosophy, psychology, mathematics, and astronomy. In 1933 he completed two Master’s theses, one on truth and one on behavioral psychology. This work included some quantitative analysis of the usage of evidential expressions in science, such as “show,” “prove,” and “demonstrate” (Naess, 1933; cf. Overton, 2013). When Naess arrived in Vienna in 1934, he was invited to participate in Moritz Schlick’s Vienna Circle seminars. Here he continued to develop his interests in close conversation with Rudolf Carnap, Charles Morris, Otto Neurath, and Alfred Tarski, among others. Naess wrote his dissertation on the empirical study of scientific behaviour and presented this to the Vienna Circle in March 1936. Naess was also inspired to conduct experiments on the concept of truth, and presented some results at the Third International Congress for the Unity of Science (July 1937, Paris). In parallel he did research with the Viennese psychologist Egon Brunswick—also a participant at the Vienna Circle. Naess was to continue this research in exile in 1938–39 at Berkeley, where he worked in Edward C. Tolman’s psychology laboratory. According to Naess, one central idea of his was influenced by how the Vienna Circle seminars were structured. Participants worked toward gaining agreement on precise formulations of their philosophical positions, revising their formulations when there were diverging interpretations. As he later recalled it, this process led him to think that “we were not quite clear in our heads—that we in a sense were only vaguely aware of what we might be talking about” (Naess, 1993, p. 263). “Their quest for clarity and cordial cooperation in pursuit of knowledge led me to appreciate that ‘What do I mean?’ is an open question” (Naess, 2005a, p. lxiii f). Naess observed considerable shallowness of semantic intention and that there were often surprisingly diverse interpretations for each other’s formulations. This led him to argue extensively against the

EXPERIMENTAL PHILOSOPHY

– 

assumption that propositions were precisely grasped in actual thinking and understanding in his dissertation, Cognition and Scientific Behaviour (1936). According to Naess, Vienna Circle participants often appealed to the ordinary, common-sense meanings and uses of words. He recalls finding it perplexing how “the logical empiricists [could] boast about a scientific attitude when they relied so much on intuition when speaking about the use of words” (2005b, p. 199). He was confident that empirical methods could be of use here, as well as for studying language more generally: I believed that one could purge logical empiricism of its antiempirical tendencies by a program for purely empirical studies of linguistic usage. Precisely such research, without further intentions, seemed to me necessary (1) to counterbalance a form of “logical analysis” that strictly speaking was not logical, and (2) to create the preconditions for the construction of a system of exact concepts intended to cover all empirical fields of importance in the philosophical discussion. (Naess, 2005b, p. 203)

In the Vienna Circle, Naess and Tarski discussed Tarski’s recent work on the concept of truth in formalized languages. One of Tarski’s desiderata linked his account to the ordinary concept of truth—his “material adequacy” condition. As Linsky put it: The requirement of material adequacy is simply the requirement that the definition, once achieved, shall correspond more or less closely with that concept of truth which all of us have in mind before we ever undertake the task of explication. (Linsky, 1952, p. 1)

Tarski believed that his account (roughly of the form, “p” is true iff p) does indeed “do justice to our intuitions” about truth and conforms with “common-sense usage,” thereby fulfilling his material adequacy criterion (Tarski, 1944). Naess doubted that this was really the common-sense notion of truth, as would be revealed by systematic questioning of nonphilosophers. However, Naess also had a larger concern, for the Vienna Circle participants were not alone in their habit of referencing the views of non-philosophers as decisive in rejecting another’s formulation or position.



T . S . MURPHY

Naess’s extensive investigations were published in his book “Truth” as Conceived by Those who are Not Professional Philosophers and in Theoria (Naess, 1938a, 1938b). Here Naess reports on the extent to which philosophers take “a standpoint to what the non-philosophers mean, stating that the theories of truth adhered to by their opponents contradict the basic structure of truth revealed among the non-philosophers” (Naess, 1938a, p. 165). He concluded that no agreement was coming any time soon from the armchair: Under no conditions can we attribute any value to statements on these matters deduced from general philosophical views or from “intuition.” If one wishes to know something about the matter, the traditional methods of attack must be radically and definitely abandoned. (Naess, 1938a, p. 93)

In his effort to settle things, Naess interrogated and surveyed ordinary people; he asked them to explicitly state what they think truth is, to state what is common to all that is true, to make synonymity judgments, to evaluate others’ definitions; and he tried a variety of other techniques (see Appiah, 2008; Chapman, 2008, 2011; Stadler, 2009). Naess also investigated other factors such as age, gender, suggestibility, and education.1 Naess identified some thirty-seven truth theories, including those centering on what is provable, what is arrived at from one’s senses, what is learned, what serves life, what cannot become otherwise, what agrees with all the evidence, and what is agreed upon by consensus. He took this to show that there was just no thing that deserved to be called the common-sense theory or pre-philosophic conception of truth; so much, then, for the material adequacy condition:2

1 As Ernest Nagel elegantly summarized: “It also contains many delicious morsels: for example, Dr. Naess found that school children at the age of puberty are capable of discussing the problem of truth with as much aplomb, though without the technical language, as philosophers with reputation; that the theory of truth as adaequatio rei et intellectus was propounded to him, except for the jargon, by a school-girl of sixteen; that his women test-persons had a greater tendency than men to believe in “absolute truth“; and that the criticisms by his test-persons of statements by their fellows, when these statements were torn completely out of their context, were not unlike those made by professional philosophers upon the writings of their colleagues.” (E. Nagel, 1939, p. 78) 2 On Tarski, Naess pointed out that the group in this vicinity is unified by its function in conversation as a means of affirming something stated or as otherwise avoiding

EXPERIMENTAL PHILOSOPHY

– 

It is therefore nonsensical to speak of the common sense view of the truth-notion. Equally nonsensical it is to speak of the view of the man in the street, of the uneducated, of the prephilosophic mind etc. No philosopher speaks of the philosophic view of the truth-notion . . . this would not, however, be any more ridiculous than to speak of the common sense view. (Naess, 1938a, p. 85)

Instead of research on the ordinary conception of truth, Naess proposed some alternative projects: the development and acceptance of scientific hypotheses, the function of maxims and statements in ideological currents, formalization of evidential expressions in science, and conceptual clarification when it is useful to do so. As Naess later summarized his conclusions: In “Truth” as Conceived by Those who are Not Professional Philosophers, I tried to show the inadequacy of intuitive methods employed by philosophers for the purpose of determining how “true” and related terms are conceived, defined and used by ordinary people. The exclusive use of intuitive methods for these purposes tends to result in an underestimation of the diverse trends of reflection among those who are not learned . . . Dialogues with those who are philosophically uneducated convinced me that acceptance of intuitions reported by the philosophically sophisticated about the verbal and conceptual habits of others leads to confusion and error. (Naess, 1953a, p. vii)

It is worth highlighting the use of “intuition” by Naess here, as it differs from some current ways of using the term. Now, it was not uncommon to treat conformity with ordinary usage and “our intuitions” as decisive. Naess was certainly not the only one to notice this, nor was he alone in surveying such appeals and proposing empirical investigations. Richard Rudner (an American philosopher at Washington University in St. Louis) did so as well. Rudner surveyed a number of philosophers—Carnap, Goodman, Moore, Hempel—and asked how the intuitions that they appeal to could be justified, weighed against each other, and systematized (Rudner, 1950). There is an important difference here between appealing to intuitions as such and treating them as evidence of the correctness of one’s analysis,

redundancy. Tarski responded that participants likely misunderstood, and proposed another test (Tarski, 1944, p. 360).



T . S . MURPHY

versus a method of reflecting from the armchair on what others would ordinarily say or think. Rudner (and Goodman) may have been concerned with the former, but Naess only had the latter use in mind. Naess did not view common-sense usage or shared intuitions as having any distinctive, epistemically significant role in philosophical analysis; instead he worried that the reliance on intuitions in making these claims was a source of fruitless controversy. As he put it: It is not necessary to depart from philosophical pastures in order to see the need for trying out empirical procedures to discover the linguistic uses and conceptual commitments of the man on the street . . . [For example, see the disagreement in] articles in recent volumes of the periodicals Mind, Analysis and Philosophical Review. I do not contend that these philosophers in all cases should have investigated conventional usage by other means than intuition. I merely suggest that empirical procedures should be applied to empirical questions. When philosophers offer conflicting answers to questions that have empirical components, empirical research is needed . . . If intuitions are used, procedures should be devised by which intuitive results of different, presumably competent people can be compared. If the intuitive results seem to conflict or are difficult to delimit and express, one should look for methods by which to avoid at least some of the intuitive components of the procedure. . . . The kind of activity today referred to by names such as “logical analysis” and “conceptual clarification” is only partly deductive and axiomatical in character. Much of it seems to me to rest on intuitions about one’s own and others’ uses of terms and to contain recommendations or preferences in matters of terminology. The intuitional approach is excellent so long as the agreement in results is of the intersubjective, intercultural kind that characterizes some of the results in the formal or factual sciences. Such agreements, however, have not been obtained. (Naess, 1953a, pp. vii–x)

The role of agreement with ordinary usage and our intuitions does not appear to amount to anything more than Aristotle’s use of endoxa as reasonable starting points in dialectical arguments (Hintikka, 1999). Naess did not regard intuitions and ordinary usage as anything more than conventional points of departure for one’s explications, construction of axiomatical systems, or other theorizing.3

3 For this reason, it is incorrect to count Naess as an early proponent of “experimental philosophy,” characterized as advocating empirical studies of non-philosophers’ “intuitions about philosophical cases,” or as subscribing to the view that philosophers appeal to intuitions as such as evidence (Cappelen, 2012, p. 219).

EXPERIMENTAL PHILOSOPHY

– 

Reception of empirical semantics in analytic philosophy Reviews of Naess’s work were strikingly hostile, particularly at first. Briefly: J. Moore criticized Naess for not having “formulated his conclusions in any systematic fashion,” adding that “there are fewer misprints than are usually found in works of this character” (Moore, 1939). Malisoff began his review by stating that “this may be described as a psychological study” (Malisoff, 1939). Nagel predicted that Naess “will no doubt remain an outcast from the philosophic community and will have to find what solace he can in being a ‘mere’ scientist” (Nagel, 1939). Later reviews of his work barely differ in what they consider important to mention. Strawson also complained about Naess’s writing aptitude and was concerned that his (not inaccurate) summary was “parodying the author” (Strawson, 1954). Hutten wrote that it is a “social and psychological study about how people use words; it hardly touches upon the logical or philosophic issues involved” (Hutten, 1953). Chisholm stressed that although it is important for linguistics it does not have any clear relevance to philosophical questions (Chisholm, 1953). These reviews are short and do not engage much with the work. To put these statements in context, consider the period of 1880–1920 as the time of an academic “power struggle” between philosophers and experimental psychologists (Kusch, 1995, 2011). While the turn of the century is remembered as a point of departure between philosophy and experimental psychology, it was also, as Sober put it, “a time of exile: while the psychologists were leaving, philosophers were slamming the door behind them” (Sober, 1978, p. 165). The rise and expansion of experimental psychology took place in philosophy departments and presented the vision of a new academic role that is part philosopher and part experimental scientist. This motivated many philosophers to argue for a strict separation between pure philosophy and experimental psychology, in a process of “role purification” (Kusch, 1995). Carnap’s vision is emblematic: “Now we shall eliminate the psychological questions also, not from the region of knowledge, but from philosophy. Then, finally, philosophy will be reduced to logic alone (in a wide sense of this word)” (Carnap, 1935, p. 33). Psychologism, Carnap said, consists in the conflation of the task of logical analysis with the empirical questions of psychology.



T . S . MURPHY

These early debates were couched in terms of “psychologism” or “psychologicism,” and climaxed in a controversial petition in 1913— one that was expressly “directed against the filling of chairs of philosophy with representatives of experimental psychology” (quoted in Kusch, 1995, p. 191). The outcome of these debates was a much-emphasized separation of “pure” philosophy from experimental psychology as distinct fields of study, with Frege and Husserl receiving credit for making the critical distinctions (see Kusch, 1995, 2011). Frege argued for a sharp distinction between logic/mathematics and psychology: mathematics and logic are neither parts of psychology nor are their objects defined, illuminated, justified, or proven true through psychology. One must, after all, distinguish between ideas of numbers and the numbers themselves. Frege acknowledged that knowledge of vague psychological processes may be of some interest, but rejected psychological interpretations of the analytic/synthetic and a priori/a posteriori distinctions; the difference is in how they are justified or proven true. In particular, they are justified without reference to matters of fact, psychological or otherwise. And at any rate, psychological laws do not evaluate thinking habits for their truth or falsehood—an independent, prior criterion is needed to evaluate them and to distinguish between being true and merely being taken-as-true. All good points to be sure, but this conception and these distinctions became central to philosophy, properly conceived.4 Naess recalls that his work was met with hostility and that he was often accused of psychologism. At the Third International Congress for the Unity of Science, Carnap even warned Naess not to present on his experimental studies (Naess, 1981, pp. 144–5). The use of questionnaires was scorned by “genuine” philosophers, Naess says, so much so that his department chair at Oslo threatened that he would not vote for Naess’s tenure if he published his study on truth (Naess, 1983, p. 311). Naess published it anyway and his chair did not vote for his tenure. Naess credits Morris’s distinction between “pragmatics” and “semantics” with providing an easy way of dodging the potential relevance of his

4

Empirical semanticists had much to say about Frege and Husserl. Frege’s views about vaguely grasped propositions receive treatment in Gullvåg’s formalization of depth of intention as discussed later in the chapter, using Barwise and Perry’s situation semantics (Gullvåg, 1983). See also Naess on Husserl’s apodictic evidence of ideal laws from the perspective of empirical semantics (Naess, 1954).

EXPERIMENTAL PHILOSOPHY

– 

empirical work (Morris, 1935; Naess, 1993). Morris introduced the trio of empirical, pragmatic, and formal dimensions of meaning, writing that “the meaning of a term is completely specified when it is known what objects the term designates, what expectations it produces in the persons for whom it has meaning, and what its connections are with other terms in the language of which it is part” (Morris, 1935, p. 278). The formal and pragmatic dimensions of meaning divided along the same lines as pure philosophy and experimental psychology; whereas semantics refers to the logical connections among terms in a language and is the domain of analyticity and the a priori, pragmatics is an empirical psychology of language which must not be confused with philosophy. Sellars’s manner of discussing pragmatics and his concomitant concern with what counts as philosophy provides an idea of how many would have approached Naess’s work in what might be called “pragmatic semantics” (Apostel, 1953). It is hardly necessary to point out that the additional tools for which we are looking are not to be found in the development which has come to be known as “pragmatics,” for this is, on the whole, a branch of empirical science, a focusing of psychology and sociology on the phenomena subsumed under the empirical concept of language . . . Classical empiricism . . . confused the grammar of philosophical predicates by attempting to identify them with psychological predicates. In many cases the grammar was so seriously confused that certain of the more consequent empiricists can hardly be called philosophers. (Sellars, 1947, pp. 645–6)

In response to this attitude Naess was quick to point out that “the term semantics is a catchword that does not convey any definite meaning,” and that his work was not to be conflated with what is “legitimately done in pure logical analysis” (Naess, 1953a, p. i). Although Naess’s empirical semantics was generally met with resistance and scepticism about its relevance for analytic philosophy, it was viewed quite amicably among his closer Vienna Circle peers.5 Tarski, for instance, said of material adequacy that it “can be settled scientifically, though of course not by a deductive procedure, but with the help of the statistical questionnaire method. As a matter of fact, such research has

5 The views of many Vienna Circle members such as Carnap, Schlick, Nagel and Neurath (who were in agreement with Naess’s approach) did not gain much credence in German academic philosophy because of this contrasting attitude (Kusch, 1995, pp. 222–6).



T . S . MURPHY

been carried on [by Naess]” (Tarski, 1944, p. 360). Carnap advocated for an empirical approach to the application conditions or intension of natural language terms, and endorsed Naess’s work as exemplary (Carnap, 1955a).6 Carnap thought that a theory of pragmatics was needed not just for psychology and linguistics, but also for analytic philosophy due to the latter’s focus on natural language (Carnap, 1955b; cf. Lutz, 2009). Carnap viewed knowledge about these “pragmatical concepts” as instrumental in inspiring and informing one’s explications. He also saw it as instrumental in furnishing a practical justification for an explication, as one may attend to the function that these concepts are serving. Empirical semanticists agreed with Carnap’s assessment (e.g., Naess, 1953a; Tennessen, 1960a). It is perhaps no surprise that ordinary-language philosophers discovered a way to reformulate psychologism in their own terms. Ryle did this explicitly when he introduced a distinction between use and usage (Ryle, 1953). In the revolt against psychologism, he says, linguistic “vogues” have evolved: first from talk of concepts to talk of meanings, and now to talk of uses. Ryle attributes psychologism to these misleading verbal vogues.7 Yet use has a critical advantage in that it contrasts with misuse and so is clearly “normative” (i.e., evaluative), whereas usage is merely a descriptive type of linguistic anthropology and sociology and is of “no philosophical interest.” If usage is not in accordance with use the folk are mistaken. And analysis of use is not informed by analysis of usage anyway, as descriptions of usages presuppose descriptions of uses. To be sure, Naess and Tennessen did refer to themselves as specifically analyzing “usages” (Naess, 1949; Tennessen, 1949a). Tennessen saw Ryle’s distinction as “fruitful, thought-economical” but did not find a difference in the use/usage of “use” and “usage” in ordinary language (“Ryle’s Dilemma”; Tennessen, 1965). But as Tennessen’s colleague Max Wright quickly pointed out, this interpretation presupposed the answer. Either there is no difference and the results show it, or there is a 6 Naess (1953b) also thanks Carnap for input into experimental hypotheses. Probably this input was given at the 1937 Congress in Paris, where Naess presented some preliminary results. 7 That is to say, when philosophers have remarked such things as “the concept of P” or “the meaning of “P,” they have never been talking about some concept or term that stands in relation to P, but P itself. Ryle’s take on this phenomenon is similar to that of Timothy Williamson and Herman Cappelen.

EXPERIMENTAL PHILOSOPHY

– 

difference and the results show that the folk are mistaken; “the appeal was, in any case, philosophically pointless” (Wright, 1967). Eventually, three reviews appeared that critically engaged with empirical semantics, and these are familiar in their arguments and concern with what counts as philosophy.8 Since these reviews present arguments in some detail, it is useful to highlight how they pose their concerns. Apostel was worried about the ordinary concept of synonymy (Apostel, 1953). There is just no clear way to avoid the apparent circle when moving from descriptive “occurrence synonymy” (or similar types) to the more ambitious “normative synonymy” (synonymy according to a rule)—especially concerning the meaning of “synonymy” itself. Crockett brought up a similar point about the missing bridge to analyticity, which he says Naess “quite naturally wishes” to cross by “counting noses.” The question remains, however, as to whether Naess has made any positive contribution to analytic philosophy . . . What is the philosophical point of these surveys? [It] is not at all clear that the description of a stock use of an expression is assisted by counting the noses of those who employ it in this way. . . . Let us suppose that in a questionnaire, call it QS1A, one hundred per cent of the subjects say that the following sentences, chosen by them from other similar sentences, express the same assertion: [P, Q]. Then we may say that [P] and [Q] are QS1A-synonymous, and this will be a shorthand way of referring to the above results. Naess, quite naturally, wishes to say more than this, and what he wishes to say is that these tests results are relevant confirmatory evidence for the synonymity of these expressions. Here we need a clear-cut hypothesis of the meaning of “synonymity” as it is used in the preceding sentence, and Naess’s failure to provide such a hypothesis in this and other cases makes one wonder what can be the usefulness of his techniques. (Crockett, 1959, pp. 109–110)

In another review, Toulmin complained that Naess’s studies were only able to handle “descriptive” statements and raised the possibility of error, thereby questioning the project’s philosophical relevance. What do the studies prove about the correctness of basic math, or equivalently, the use of language and the nature of our concepts?

8 Apostel, Crockett, and Toulmin all have positive things to say in the rest of their reviews, as does Mates (Mates, 1958a). Of course, Quine influentially noted that it is not at all clear how empirical investigation can solve disputes about meaning, and his influence shows up in some of the reviews (cf. Naess, 1957; Quine, 1951).



T . S . MURPHY

Exactly what Mr. Naess takes to be their relevance, is to one’s sorrow, left unclear—“The question of relevancy is complicated,” he says. What makes it so puzzling and tantalizing is Naess’s vagueness about the point of the investigation for philosophy. . . . One must hope that, before Naess gets too immersed into the practical work of framing and using more and more similar questionnaires, he will sit down and tell us what exactly they are designed to prove. Until that is done, it will be easy for philosophers to ignore his work. “Even if 25.8% of persons are found to give the sum of two and two as five,” they will argue, “that would leave the correctness of the formula ‘2 + 2 = 4’ in formal arithmetic unaffected; surely also, the fact that quite a number of people were prepared to give some sense to the statement ‘Jones knows the time assiduously’ would not destroy the familiar, established use of terms, which rules out the collocation of such a verb and adverb?” And it would be a pity if Naess’s work were to be entirely ignored, for, reading through the paper, one certainly feels that the reactions of his answerers proves something about the nature of our concepts . . . At the moment, all one can do about the larger aspects of his work is to suspend judgment. (Toulmin, 1956, p. 118)

Indeed, history repeats itself for those in analytic philosophy who undertake empirical investigation into concepts, which are presupposed in evaluating or interpreting people’s performance (Alexander et al., 2010; Kauppinen, 2007; Machery, 2008). These concerns pose problems for empirical semantics and experimental philosophy to the extent that the aim is to access some ambitious form of a priori analyticity or conceptual truth by means of experimental psychological investigation. But in the case of empirical semantics, there was just no attempt pursue such a “mentalist” project (Alexander et al., 2010). Naess, after all, rejected that goal in his work on the notion of truth, and his students and collaborators agreed. Empirical semanticists were well aware of the above difficulties and had nothing against the “laudable effort to stamp out every trace of psychologism” (Naess, 1954, p. 55). What empirical semanticists deplored was the persistent concern for what counts as philosophy, and they always insisted that the philosopher who uses the methods of science “need not stop being a philosopher for that reason” (Naess, 1961a, p. 173). In their analysis of some of Hume’s texts, they close with the observation that the isolation of philosophy from psychological research is “one of the paradoxes of contemporary philosophy” (Naess and Naess, 1960, p. 146). Throughout much of the work of empirical

EXPERIMENTAL PHILOSOPHY

– 

semantics they advocated a return to a traditional and broadly empirical conception of philosophy. Naess sums up this view well: There is a tendency to look upon deductive and axiomatical procedures as somehow more philosophical than empirical ones, and this has undermined the position of the broad empirical traditions (Aristotle, Ockham, Locke, Berkeley, Hume, Bentham, John Stuart Mill), which in my view deserve a strong representation in contemporary culture. The charge of psychologism against thinkers of this tradition is well founded, but has been largely misapplied. It has discouraged research into genuinely empirical components of question complexes of a mixed formal and empirical character. . . . Very roughly, one may distinguish a deductive, an intuitional, and an empirical component in the writings of analytical philosophers. Even in those cases in which deductions and intuitions can help us considerably, consistent neglect of the empirical component will bring research toward stagnation. If empirical studies are neglected, we shall see much intelligent debate along intuitionist lines, but less of that process that many of us find so inspiring in the history of philosophy and science: the development of new branches of reliable knowledge as a result of combined philosophical and scientific efforts. . . . Critics who would assume that the methods described in this book aim at solving questions that the intuitively and deductively operating logician has not been able to solve, mistake the intention. (Naess, 1953a, pp. iii–iv)

In The Function of Moral Philosophy: A Plea for Integration of Philosophical Analysis and Empirical Research (1958), Ofstad begins with the classical conception of philosophy: Among the ancient Greeks the philosophy of morals did not exist as a specific discipline. Socrates was not only a philosopher of morals, he was a psychologist too, and also a sociologist and a political scientist . . . The attempt to answer such questions led them into empirical as well as analytical problems. They accepted no definite limits for their speculations in these areas. Why should they? It is we who have tried to distinguish carefully between questions of analysis of language and those of an empirical nature, and split up the study of man into a number of different sciences. [The] view that the philosopher, qua scientist, cannot assert pure norms or value-statements . . . has dominated important parts of AngloAmerican and Scandinavian moral philosophy for nearly fifty years, and so it may be useful to take it up for evaluation . . . It has stimulated contacts with such other branches of philosophy as logic and semantics, but the connections with psychology and the social sciences have been almost broken. The training of moral philosophers might be changed so that their education would qualify them for taking part in team-work with logicians, semanticists, psychologists and social scientists. (Ofstad, 1958, pp. 35–7)



T . S . MURPHY

Ofstad then surveys research programs in various disciplines: metaethics, communication and argumentation, moral deliberation and decision-making processes, beliefs and ethical behaviour, and so on, still insisting on methodological pluralism: Whether research of this kind is called “moral philosophy” or not, seems, however, rather unimportant. The important thing is that there ought to be a close connection between such studies and investigations which are more central to moral philosophy . . . For the philosophical significance of such work, it is important to preserve all the subtleties which are compatible with the exploitation of the research-instruments developed within the social science. (Ofstad, 1958, p. 40)

In today’s experimental philosophy, one also finds a nearly identical dialectic, exemplified by Knobe’s response to Kaupinnen—that is to say, the response that such difficulties, while interesting, are largely a red herring; where this is accompanied by the advocation of a traditional, broadly empirical conception of philosophy (Knobe and Nichols, 2008; Knobe, 2007; Knobe et al., 2012). Again, the issue for empirical semanticists was not that the questions that were raised were not interesting or difficult ones (given relevant aims), but that these objections have been misapplied and so have discouraged otherwise valuable empirical research.

Empirical Semantics Despite some resistance, Naess was appointed as chair of philosophy at the University of Oslo in 1939, and here he continued to work on empirical semantics with his students and collaborators. This work culminated in his monograph Interpretation and Preciseness (Naess, 1953a), and later, Communication and Argument (Naess, 1966). Empirical semantics was highly influential in Norway; for nearly twenty-five years, an introductory version of Interpretation and Preciseness had served as the obligatory text for graduate students who intend to take any other major examination at “any Norwegian university, at most advanced great-schools, some teacher colleges and at all the military staff colleges Norway” (Tennessen, 1962, p. 1). After introducing some relevant terminology and an overview of the methods, I survey the work done in this period.

EXPERIMENTAL PHILOSOPHY

– 

Empirical semantics views communication in terms of a sender, signal, and receiver—particularly the interpretation of the signal by the receiver (or sender), where interpretations are modeled in set theoretic terms. Interpretations can be discovered experimentally with the use of questionnaires, such as through judgments of the form “Q may be synonymous with P” and “when you read P, did you take this to mean Q?” An interpretation Q of P is a precization (is more precise) when the synonymic alternatives to Q are a proper subset of the synonymic alternatives to P. An expression is a plausible interpretation of another roughly when it would be judged as potentially synonymous by many interpreters. Importantly, precizations and interpretations may be depicted in treelike maps, which encode direction and depth of precizations (Figure 12.1). Strictly speaking, interpretations are properties of individuals, plausible interpretations are properties of groups, and the preciseness of expressions are given by social usages (Gullvåg, 1983). Definiteness or depth of intention refers to the precization operative in an interpreter/speaker, as evinced by the point at which the individual The person P decided freely in situation S

Absence of compulsion . . .(15). . .

Rationality

Power

. . .(23). . .

. . .(32). . .

Indeterminacy

Self-expression

. . .(48). . .

. . .(46). . .

(6.14) P, in S, where P decided D1, would very likely have decided D2 instead if P had made an effort to decide D2. . . .(4). . . (6.14.1) P, in S, where P decided D1, would very likely have decided D2 instead if P had thought more about the consequences of D1 and D2, or studied the whole case more thoroughly, or been more aware of this or that, or if P had studied him/herself by the use of such sentences as “It is my duty to decide D2.” . . .(4). . .

Figure 12.1 Tree-like diagram of precizations, modeled from Ofstad’s precizations of the sentence scheme “The person P decided freely in situation S” (Ofstad, 1961). Numbers below branches refer to the count of precizations that have been omitted. Also omitted is “freedom as virtue” in the top branch.



T . S . MURPHY

becomes indecisive among more precise formulations of what they mean or understand. Is noonish 14 or 15 minutes past noon? One may be indecisive about the matter and unable to answer the question, and this indecisiveness is the hallmark of indefiniteness. The concept was influenced by Pierce and his definition of “vagueness” (Gullvåg and Naess, 1996; Peirce, 1902, p. 748). It is more precisely a type of “process vagueness” found in the interpreter’s head (cf. Sorensen, 1990). Depth of intention was later explained in terms of the conceptual framework underlying an individual’s ability to discriminate in perception and thought and to access finer distinctions between types of situations (Gullvåg, 1983; Tennessen and Gullvåg, 1959). Much use was made of the concepts of definiteness of intention and preciseness in diagnosing and explaining fruitless disagreement, and they were central to how empirical semanticists thought about conceptual change and scientific development. It also underpinned their views of the value, role, and limits of philosophical analysis (Gullvåg, 1988; cf. Howe, 2010; Naess, 1936; Tennessen, 1973). The centrality of definiteness of intention and their empirical approach to language encouraged empirical semanticists to interpret statements about conceptual or analytic truths as optative rather than, say, indicative, constative, apodictic, or anamnestic in character, though they were also fervent methodological pluralists. Empirical semantics research would typically begin with a survey of uses alongside definitions and commentaries, dubbed “occurrence analysis” and “metaoccurrence analysis” respectively. The distinction between occurrence and metaoccurrence analysis distinguishes the projects of understanding how a term is used versus understanding how it is defined and conceived of by its users, so as to be able to diagnose discrepancies. This first stage was coupled with a detailed discussion of the interpretation of certain texts of interest (“elementary analysis”). There was a tendency to map out the space of possible interpretations by substituting precizations of component parts of expressions and then narrowing the list down. The goal was often not just to find out what any individual actually thought but to map out the space of what one could possibly think or mean more precisely.9 The whole process can be done from armchair up to this point in the procedure. 9 Indeed, when Naess talked of the “possibility of an ‘experimental philosophy’ ” in his study of truth he referred to a developmental psychology of conceptual systems—one which

EXPERIMENTAL PHILOSOPHY

– 

Once some hypotheses were formed about usages/underlying concepts and any psychological or social factors of interest, these were tested with questionnaires, with interviews, or simply by revisiting the original occurrences. Usually a panel of interpreters coded occurrences under the usage rules or underlying concepts. Quite often a goal was to understand how people interpret and understand each other in fairly general terms (e.g., how does considering someone as out-group or in-group affect one’s direction of interpretation of key words in political argumentation?). The result of the investigation would be a map of precizations and underlying concepts, plus the effects of such factors as personality and philosophical positions. This served as a basis for making further recommendations and evaluations, such as for facilitating good political or scientific discourse through an increase in precision, for estimating the convincingness of arguments and appeal of political slogans (“market analysis”), or for diagnosing unnoticed ambiguities and conflations that arise from indefiniteness of intention. The scope and detail of this work is impressive. Naess’s analysis of Zaslavski’s usage of “democratie” in La démocratie soviétique considered all 192 occurrences (Naess, 1953a, pp. 300–49). Tennessen’s study on “the system of private enterprise” surveyed two years of annual newspapers in Norway for occurrences before constructing “the longest questionnaire ever given in Norway” (Naess, 1964, p. 7). Under the six conceptions of private enterprise identified, they coded 7,667 occurrences and analyzed them with respect to political party and profession (Tennessen and Gullvåg, 1959, p. 23). Siri and Arne Naess classified 661 sentences in Hume’s Treatise of Human Nature into normative, declarative (analytic, synthetic), and other linguistic categories (Naess and Naess, 1960). Tennessen, Ofstad, Gullvåg, and Bay (1950) investigated nationalism and its relationship with psychological, sociological, and economic factors. They tested sixty-three hypotheses in ten surveys, most with about eighty questions and 500 questionnaires per survey. Analyzing the wealth of data at the time proved quite difficult, as one can imagine.

begins with the “embryonic form” of philosophical positions found among non-philosophers (Naess, 1938a, p. 161).



T . S . MURPHY

Empirical semantics research at Oslo Naess’s appointment as chair of philosophy at Oslo was soon followed by a five-year Nazi occupation. This, of course, impacted everyone working with Naess during his wartime seminars. Indeed, some of Naess’s students did not survive the war and their studies were published posthumously.10 It is in this context that a “unique interdisciplinary milieu developed, combining an emphasis on general theory and methodology with a strong concern for social and political problems” (Bay, 1958, p. vii). This attitude is reflected in the subject matter and the goals of many of their studies. For example, at the end of Tennessen’s investigation of the attitudes of lawyers to the trials of Quislings11 in Norway, he concluded: The emotional impetus behind this kind of work is linked to the hope that our efforts will incorporate research with a relatively direct struggle for humanist ideals. Regardless of whether we call our surveys opinion polls or attitude analysis or scientific studies in the areas between sociology, social psychology, semantics, and ethics, the hope is that if these examinations were performed on a larger scale, they would effectively contribute to the eradication of antagonism throughout society. (Tennessen, 1950, my translation)

Naess and his students helped establish the Institute for Social Research in Norway in 1950, and Naess led UNESCO’s Philosophical Analysis of Fundamental Concepts project to study concepts of democracy, nationalism, and liberty. Still, throughout this period there was also a vibrant empirical semantics research program in philosophy, and this can be divided into some rough categories (Table 12.1). In philosophy of language, there was an 10 In a 1945 letter to Otto Neurath, Naess wrote of his wartime experiences: “I am still somewhat groggy and disheartened because of lost friends and collaborateurs, but I hope soon to recover. The very brilliant young philosopher Ludvig Lövestad died this year. He was my close friend in all kinds of work, also the ‘illegal.’ He was tortured to death, remaining silent about my hiding-place. Another close friend and collaborateur in philosophy, Wickström-Nielsen, was killed when jumping from parachute. He came from England and jumped with documents and Russell’s new book on Truth etc. and Lundberg’s new book on the methods of sociology. Also other young people who wished to go on with philosophy and mathematics are missing. This field got an exceptionally hard blow.” (Quoted in Stadler, 2009, p. 20.) 11 Quislings were members of Vidkun Quisling’s collaborationist party during the Axis occupation of Norway in World War II.

EXPERIMENTAL PHILOSOPHY

– 

Table 12.1. Overview of topics and sample of works in empirical semantics. Asterisks signify experiments that use questionnaires and interviews. For some other bibliographies and overviews, see Ofstad (1951), Naess (1953a, pp. viii–ix), and the journals Synthese, Theoria, The Philosophical Review, and Inquiry for the 1940s through the 1960s Sample of empirical semantics research Philosophy of language • Verification of statements on ordinary language (Gullvåg, 1955; Mates, 1958b). • * “True,” “perfectly certain,” and “extremely probable” (Naess, 1953b). • * “Or” (Naess, 1961b). • * Common sense theories and “truth” (Naess, 1938a, 1938b). • * The intuitive concept of synonymity (Naess, 1956, 1957). • * Ordinary-language philosophers’ claims about ordinary language (Austin and Naess, 1964; Tennessen et al., 1964; Tennessen, 1959a, 1965). Ethics, action theory, and freedom of the will • Aesthetics and ethics in Kierkegaard’s Either/Or (Ofstad and Löfgren, 1965). • “What is virtue?” in Plato’s Meno (Grimm, 1962, 1964). • Free will and “The person P decided freely in the situation S” (Ofstad, 1953, 1961, 1967). • Verifiability and objectivity of descriptive and normative claims, relationship to “reality” (Naess, 1959; Ofstad, 1951; Wickström-Nielsen, 1948). Philosophy of science • Evidential expressions in science (Naess, 1933). • Normative, analytic, and synthetic sentences in Hume’s Treatise (Naess and Naess, 1960). • “Consciousness” in the psychology of perception (Fluge, 1945, in Norwegian). • * “Testability” in physics (Lövestad, 1945; see also Naess, 1953a, pp. 374–6). • * Concepts of type/typicality (Tennessen, 1949b, in Norwegian). Social, political, and legal philosophy • Foundations for the science and empirical semantics of law (Aubert, 1943). • Concepts of legal norm (Ofstad, 1949, 1950). • Critical examination of Nietzsche’s “Will to power” philosophy (Haaland, 1947; see also Naess, 1953a, p. 266). • Examination of dialectical materialism, by Arne Torvik (Ofstad, 1951, p. 41). • * Businesspersons’ views of white-collar crime’s status as “crime” (Aubert, 1952). • * “The system of private enterprise” (Tennessen, 1949c, 1959b). • * Lawyers’ attitudes to prosecution of Quislings in Norway (Tennessen, 1950). (continued )



T . S . MURPHY

Table 12.1. Continued Sample of empirical semantics research UNESCO and Norwegian Institute for Social Research • Gandhi’s ethics (Naess, 1958). • Freedom and liberty (Bay, 1958). • * Concepts of democracy (Naess et al., 1956; Naess, 1953a, pp. 300–49; Rokkan and McKeon, 1951). • * Nationalism (Bay et al., 1950). Philosophy and education • Precization in education, and educational development of the concept of truth (Grimm, 1955). • * The effect of philosophy education on students’ philosophical positions (Fain and Kaelin, 1960).

effort to understand statements about ordinary-language rules, de-precization in ethical statements, and studies on particular terms such as “true,” “or,” and “synonymity”. In ethics and action theory there are interests in whether people “produce” decisions, “ought” and “can,” the freedom of the will, and relationships between interpretations of descriptive and normative statements. In philosophy of science, topics included consciousness and perception, the “testability” of physical laws, and the concept of “type” in psychology. They even tested whether scientists end up at observation sentences through repeated questioning (they do not). In social and political philosophy they investigated concepts of crime, dialectical materialism, Nietzsche’s “Wille zur Macht,” and interpretations of “legal norm” in law. The diverse themes in empirical semantics span traditional and contemporary philosophical topics, not unlike experimental philosophy today. Tennessen’s “Concepts of Type” aptly illustrates the process and style of empirical semantics research (Tennessen, 1949b). This study begins by stating that its motive is to take a comparative and evaluative approach to typological methods, hypotheses, and research programs, using tools from experimental psychology. Tennessen surveys occurrences of and commentaries on “type,” “typical,” and closely related terms using sources from philosophy and psychology, encyclopaedias, and newspapers. Then he surveys the range of possible precizations of the type concept using a schematic formula and substituting

EXPERIMENTAL PHILOSOPHY

– 

interpretations of key components. This is coupled with an analysis of specific typologies in psychology to understand how they function in research. There is also a wider interest in non-cognitive aspects of meaning, such as the relationships among binaries such as female/ male, young/old, plant/animal, passive/active, and their relationship to the types of integration disorders distinguished in psychology. After formulating hypotheses about some underlying concepts or usages, Tennessen reports on the 669 questionnaires that were completed by students, farmers, and professionals. Tennessen identifies four main type concepts, which seem to be roughly as follows: (1) a characteristically descriptive individual of the type or exemplar; (2) typicality in the sense of statistical mode or set of most expected features; (3) those distinguishing properties with high sensitivity and specificity; (4) a “class” that is individuated according to some constitutive qualities of that class. Tennessen subsequently evaluates the work of typologists and philosophers. For instance, he criticizes Hempel and Oppenheim for failing to distinguish between (2) and (4) in their logical analysis. He also recommends that typologists in psychology use precization (3), given their explanatory and diagnostic aims, and shows how a number of them unknowingly shift between usages. Although Tennessen closes by saying this methodology is a valuable and much needed contribution to philosophical and logical analysis, he places special emphasis on its limited role. In particular, this strategy makes the process of discovery explicit by including the techniques for surveying the possible directions of precization. It also uses experimental methods to test the descriptive adequacy of one’s delimited usages or underlying concepts. This is meant, among other things, to fix the mysterious absence of a methodology section (the “method of revelation”) in logical analysis publications (Tennessen, 1949a). One program in today’s experimental philosophy focuses on understanding how people think about specific philosophical topics of interest, and this aim is well represented in the research done at Oslo. However, the studies performed in empirical semantics are unlike experimental philosophy in their concentration on language systems, at least when theorizing about their work. But in practice they did not sharply distinguish between meanings and concepts, nor did they distinguish between predication of a term and application of a concept, nor the use of a term and deployment of a concept, etc. Where these studies



T . S . MURPHY

are experimental, they are in the tradition of behaviourist psychology and social science. The measurements are between variables such as personality and philosophical views and their effects on questionnaire results and other behaviour. The theories contain no cognitive architecture, identification of cognitive systems, or information-processing models, and this is typical of the time. This is also true of the work done at Berkeley (discussed below). There is also a complete absence of any attempt to establish analyticities or conceptual truths. Empirical semanticists were happy to keep their results descriptive and understand various parts of the world, and never aimed directly at the more ambitious kind of philosophical upshot that was expected of them. In general, they did not find projects of refining intuitions or increasing definiteness of intention in conceptual analysis worthwhile, except as a means to greater clarity and precision for some specified purpose. This overall attitude is clearly expressed by Naess on the notions of synonymy and analyticity: Benson Mates contends “that one is justified in saying that there are ‘intuitive’ notions of analyticity and synonymy.” This empirical hypothesis about the existence of certain phenomena or kinds of phenomena is tenable, so far as I can see. Mates has an intuitive notion of synonymity; I have had several in my life, and there is reason to believe that all of them have much in common. On the other hand, there is no reason to believe that the various intuited entities are identical or near identical . . . If both of us should assert that our own intuitions are more adequate or more nearly similar to the intuitions of respectable authorities, there would be disagreement between us. But by what kind of discussions or research can a disagreement about the nature of an intuited entity be settled? Fortunately, collective research does not seem to presuppose that all intuitions are shared by all researchers, or that they should even know of the differences, or that the intuited entities should be definite in outline and content. Thus, there may in the years to come be much fruitful research concerning synonymity by researchers with partially different intuitive notions of synonymity . . . It is our contention that sound methodology does not require strict conformity of research terminology to prior intuitions. One may even say that strict conformity is not possible because of the indefiniteness of the intuitions. (Naess, 1957)

Indeed, he argued that intuitions about universal interchangeability salva veritate are not universally held and that there are lax and broad notions of synonymy. Naess hypothesized that the thought that this intuition is universal owes to overestimation of the definiteness of intention

EXPERIMENTAL PHILOSOPHY

– 

in unqualified statements, thereby mistaking unqualifiedness for generality or absoluteness (as with “lying is wrong”). Indefiniteness was also a major point of contention between empirical semanticists and ordinarylanguage philosophers. Although use/usage is a distinction between correct language use and actual performance, when taken as an empirical hypothesis about ordinary language it turned out to be below the definiteness of intention of its users (Tennessen, 1965). Whereas Ryle saw the ordinary concept of voluntary as somehow shaped by its role in blaming (as shown by illustrations of presumably competent usage), empirical semanticists would see this as unduly precise, i.e., as more definite than or even contrary to how ordinary folk define “voluntary” or conceive of voluntariness (Mates, 1958b).

Empirical Semantics at Berkeley The short period from 1957 to 1961 at Berkeley was another highly fruitful episode in empirical semantics. Much work originally done in Norwegian at Oslo was also published in English by Tennessen and Naess (the two main experimentalists).12 Whereas empirical semanticists had previously focused on logical analysis and the social and political applications of their techniques to specific concepts, at Berkeley the focus shifted to ordinary-language philosophy and to explaining why people have the intuitions about linguistic expressions that they do. One goal in this section is to outline this explanatory project. The work done at Berkeley is also a little piece of lost history that is interesting in its own right, and it has some parallels with debates in experimental philosophy. The main cohort of empirical semanticists went to Berkeley in 1957.13 Tennessen took a professorship at Berkeley in the Department of Speech (now named the Department of Rhetoric). Naess also became a part-time visiting professor in philosophy, and Gullvåg a visiting scholar. Here they worked alongside philosophers David Rynin, Isabelle Hungerland, Benson Mates, and John Searle. Everyone here was critical of ordinary-

12 Much of this was facilitated by Naess’s editorship at Synthese, and Naess’s new journal, Inquiry. 13 This is aside from Ofstad, who went to the University of Stockholm in 1955.



T . S . MURPHY

language philosophy.14 John Austin was a visiting professor in 1958 and inspired much of the empirical study that transpired in these years (Tennessen, 1959a). Upon arrival, Tennessen and Gullvåg participated in Mates’s 1957 seminar on Austin’s “A Plea for Excuses” (1956). In this seminar they concentrated on ordinary-language philosophy and their own methodology. Notes from the seminar show that they focused intensively on how Austin was arguing for his positions: Each set of hypotheses, it will be noticed, contains statements about usage and statements about actions: e.g., that it is generally not permissible to use adverbs in descriptions of normal actions, and some (or most) actions are neither voluntary nor involuntary. One occasionally feels that, for Austin, the two kinds of statements, which he simply juxtaposes, have some close logical relation: one kind is evidence for the truth of the other kind, for instance. (Tennessen et al., 1964, pp. 106–7)

They also found it striking that “Austin’s favourite method appears to be telling little stories and asking people what they should say in the described situation; by means of sets of stories, he finds himself able to elicit general agreement as to ‘what we should say when’ ” (Tennessen et al., 1964, p. 103). The main worry in Mates’s seminar was the lack of explicitness about the way in which these vignettes were paired with usage hypotheses. Nearly a decade earlier, Tennessen ran some experiments to investigate problems with assessing usage from the armchair, and found a tendency to conflate evidence with illustration (Tennessen, 1949a).15 As Naess summarized it:

Their colleague Stanley Cavell described this period as one of being engaged in “all but continuous argument, sometimes consisting of friendly exchange sometimes of (temporarily, but you couldn’t be sure) estranging dispute” (Cavell, 1999, p. xxiii). “[I was surprised by the] outrage [OLP] produced in my older colleagues. Outrage is what it was. This was evident in my colleague Benson Mates’s contempt, echoed in his older friend David Rynin’s exasperation” (Cavell, 2010, p. 372). 15 Some of the experiments went as follows: “The word x seems to be used in different ways. Occasionally it is used in the sense of y, as for instance in the sentence: ‘ . . . ’. We inserted a sentence which made it seem preposterous to believe that the word was used as indicated in the text. In spite of this, there was a tendency among the respondents to agree to the subsumability. Some of the questionnaires contained questions of the following kind: ‘Do you think this x is a good or bad example of y being used in the sense of z?’ ” (Tennessen and Gullvåg, 1959, p. 3). 14

EXPERIMENTAL PHILOSOPHY

– 

This mechanism radically destroys the function of definitions. Instead of giving us precise and tenable hypotheses for language usages to be tested by observing usages of the rule in a language, a definitoform sentence is looked upon as a formulation, the meaning of which is to be understood by means of the definiendum within the language. As a result, there is a tendency to accept uncritically and without any qualifications whatever subsumptions are explicitly or implicitly asserted. (Naess, 1953a, pp. 278–9)

They apparently achieved similar results with Austinian vignettes, and so stressed that this distinction needs to at least be made fully explicit in philosophical writing and seminars such as his (Tennessen, 1959c, 1959d). In order to explore Austin’s method more carefully, they surveyed possible interpretations of Austin’s approach. They stressed that what is important is not just what Austin meant but any other interpretations as well, as was standard practice. Some examples: (2) Usually, use of any modifying expression is not permissible. (2a) It would sound odd. (2b) It is not in accordance with communication norms in English. (2c) It would be meaningless. (2d) It would be false . . . (5a) Most people do not ever use “voluntary” and “involuntary” in describing more than a small number of cases. (5b) Can not meaningfully be used outside of such cases. (5c) Most people do not use the terms so that they ever say it was voluntary or involuntary. (Tennessen et al., 1964, pp. 101–2; results in Tennessen, 1965)

Their diagnosis of the bleak prospects of the more ambitious claims of ordinary-language philosophy, as they saw it, turned on an ambiguity about what is “correct.” Of course it is true that, if an action has a certain characteristic, then some statement is “correct”, i.e., true, namely, one which attributes that characteristic to that action. But on any other meaning of “correct,” this does not follow, nor does the converse relation hold: that a locution is “permissible” does not allow us to infer anything about the world . . . Simply put: we feel it necessary to distinguish between what one would say and what one could say, and to insist that knowing the former does not give us complete information about the latter. (Tennessen et al., 1964, pp. 107–8)

While the Mates seminar went on, Stanley Cavell was expressing high praise for ordinary-language philosophy (OLP) just a few doors away. So



T . S . MURPHY

Rynin arranged a debate at the 1957 Pacific APA, and they told Cavell he would have to defend OLP against Mates (Cavell, 1958; Mates, 1958b).16 Mates’s paper draws from the seminar notes, but the target shifts to Ryle and his apparently irritating use/usage distinction (Ryle, 1953). Mates argues that Ryle’s claims about use are descriptive after all, and accuses OLP of conflating semantics and pragmatics (conversational implicatures).17 He then distinguishes between extensional and intensional methods of studying ordinary language, and complains that OLP focuses only on the extensional method. The extensional method is occurrence analysis, and the intensional method is metaoccurrence analysis with a Socratic twist. The intensional method involves asking participants what they mean and presenting cases that might make them revise and precisify their accounts. Ordinary-language philosophy does not account for the conflict between methods, Mates charges. Both methods are incorporated into empirical semantics (Hungerland, 1960; Tennessen, 1959e). The Socratic dialogue was first proposed by Toulmin to transgress merely “descriptive” statements by investigating the semantics and entailment relations endorsed by the folk’s own lights (cf. Naess, 1961b; Toulmin, 1956). The Socratic dialogue model is revived in Kauppinen’s discussion of experimental philosophy for the same purpose (Kauppinen, 2007).

Experiments on language judgments By 1959, at least fifteen projects and 4,500 questionnaires and interviews had been administered at Berkeley on topics including the analytic/synthetic distinction, definiteness of intention, verbal rigidity and argumentation patterns in religious and political discourse, and “P can decide to do x” (see Tennessen, 1959a, p. 287). Tennessen used these to argue “against any tendencies to narrow down the field of permissible communication by employing rigid, a priori norms or rules for ‘what can possibly be said

Although it launched his career, Cavell remembers it thus: “Rynin issued this invitation—summons rather—coming upstairs and down the hall from his to my office in Dwinelle Hall, at the end of a conversation that he began by noting that since I arrived in town I had been saying a lot of extravagant things about this new work on ordinary language . . . The impression of anger in such exchanges never left me.” (Cavell, 2010, pp. 372–3.) 17 This appears to be the first time conversational implicatures, though not named as such, are explicitly used as arguments against ordinary-language philosophy. 16

EXPERIMENTAL PHILOSOPHY

– 

and meant’” because “the whole thing is most often a question of general (including hermeneutical) imagination” (Stern, 1969; Tennessen, 1959a, pp. 276–7, 1961). Tennessen’s target is exemplified by his appraisal of the Russell–Strawson controversy over Russell’s theory of descriptions. It has always been clear that whatever advantages this proposal might have, they have nothing to do with analyses or hypotheses about (common or “ordinary”) language usages. None the less, the following passage is found in Strawson’s “On Referring” (p. 330): Now suppose some one were in fact to say to you with perfectly serious air: “The king of France is wise.” Would you say, “That”s untrue”? I think it’s quite certain that you wouldn’t. Strawson is wrong: Of about 1,500 informants tested in some recent experiments no one seemed to act in accordance with Strawson’s predictions . . . Strawson, one might say, has opened the door a crack to the vast field of empirical investigations of language, taken a peep in, and, after (almost immediately) having shut the door, he reports: “Russell is wrong: The Theory of Descriptions is fatally incorrect because one would not (could not? should not? ought not to?) utter, and/or mean: ‘The present King of France is wise is false’!” However, in his own attempt at a “solution to this puzzle” Strawson seems absolutely uninterested in what “one” would or would not say. (Tennessen, 1960b, pp. 187–8)18

By dividing up the mechanisms underlying intuitions about language and explaining the tendency to reject statements for purely language reasons, Tennessen and collaborators aimed to show that these judgments were quite irrelevant as to a statement’s tenability. After surveying plausible interpretations of “what should we say” and similar locutions, they identified three general lines of defense: ineffability, infrequency, and impermissibility. A sample of each follows. Under the ineffability approach, a tenuous connection was found in the “verbal rigidity” hypotheses of Vygotsky, Piaget, and Frazer, according to which children and “primitive societies” cannot separate word forms from word meanings. Tennessen found that children just have a strong tendency to adhere to the permissibility rather than potentiality direction of interpretation of “can” in such questions as “can you call a cat a dog?” (Tennessen, 1959a, pp. 266–72). The order of questions was sufficient to make the difference, and once clarified there was no

The “1,500” number presumably comes from its regular inclusion in studies at Berkeley. 18



T . S . MURPHY

evidence of verbal rigidity for English speakers. However, there was an analogous form of rigidity on the side of experimenters, as they were apparently unaware of these language ambiguities and did not imagine such alternative interpretations. The infrequency approach draws on what is “never said” to formulate language rules. This is what Austin reportedly suggested at Oslo (October 1959), in a debate with Naess over some of Tennessen’s experiments on “voluntary yawning”: AUSTIN :

The subjects gave wrong answers concerning their own use of expressions, e.g.: when saying they would never use “he yawned voluntarily” as a description of a perfectly ordinary of event of yawning because it is perfectly obvious that such yawnings are voluntary. Actually the subjects would not say it because it cannot be said. NAESS : The subjects interpreted “he yawned voluntarily” as synonymous with “he was not forced to yawn” and thus conceived it as obvious that he was not forced to (and therefore not worth while saying). AUSTIN : But then they do not know well enough the expression “voluntary.” It is too difficult a word, maybe. Better use “clumsy.” NAESS : Suggestion to HT: New experiments . . . AUSTIN : . . . Rules (grammatical or others) do not exist as rules. We say there is a rule against saying x when x is never said. What is against the language system cannot be true or false or obvious. Thus “He yawned voluntarily” cannot be true or false or obvious. NAESS : Tennessen investigates communication rather than language systems. What happens if something actually is said (uttered) which “never is said”? . . . AUSTIN : Suppose that someone yawned in a standard way and there is nothing exceptional (i.e., the case is a standard case). If we ask people to describe completely as possible what happens, they will never add “clumsily” or “voluntarily.” If we say “but did he or did he not yawn clumsily (or voluntarily)” they would find the terms inapplicable, the sentences neither true nor false. (Austin and Naess, 1964)19

19

The full transcript and other materials are available on the author’s website.

EXPERIMENTAL PHILOSOPHY

– 

These potential problems and predictions were the subject of further experiments, of course. In one case, they hosted fake discussion groups on social problems among students (accompanied by a hidden tape recorder), and aimed to have the locution “voluntary yawning” occur as naturally as possible (Tennessen, 1965, pp. 234–6). Contrary to Austin’s predictions, the statements were accepted as meaningful.20 Lastly, consider the impermissibility approach. Tennessen and collaborators predicted a dissociation between two sources underlying the evaluation of a statement: (1) a grammatical and idiomatic/literal direction of interpretation, and (2) a tenability direction of interpretation. So some experiments began with either (1) a “logical-maniacal” lecture on how people often assert nonsense and contradictions, or (2) a “commonsensical” lecture on how the most important thing is to understand what someone means. Participants then classified sentences as tautologies, contradictions, or nonsense, or as conveying factual synthetic statements. Participants were also asked to provide their reasons, and these were coded as language or tenability reasons for rejecting the statement. These experiments showed that participants could adopt and switch between the two interpretive attitudes and exposed tendencies for pseudo-disagreement when these distinct sources or attitudes explain the disagreement (Figure 12.2A; Tennessen, 1959a, pp. 280–4, 1959f ). So Tennessen identified two mechanisms or processes, one fast and intuitive and the other requiring a bit of reflection and imagination, and these underlie the difference between language and tenability directions of interpretation. Tennessen took the upshot of his studies to show that so far there are no empirical grounds for a theory of linguistic necessity or for linguistic restrictivism; they are either empirically unsupported or they split between two sources of judgments about the acceptability of a statement. The reasons and features these sources track do not tell us anything about each other; if a statement is ungrammatical, counterintuitive, or goes against ordinary usage in one way or another, this tells 20 Austin planned to study the interviews apparently supported occurrence analysis in dictionaries (Austin and Naess, 1964). Indeed, Austin saw his approach as one that would be absorbed into a larger scientific enterprise (see Naess, 1961a, p. 197). Unfortunately Austin passed away shortly after these debates. At the time it was an open secret at Oxford that Austin was seriously considering moving to Berkeley, having reportedly remarked that he “could build an empire there” (T. Nagel, 2009), and having expressed concern with a lack of a next generation of like-minded philosophers at Oxford (Chapman, 2009).



T . S . MURPHY

A

Would you say...

Trivial

B

“It is bedtime, I am alone, I yawn: but I do not yawn involuntarily (or voluntarily!) nor yet deliberately. To yawn in any such peculiar way is not to just yawn.” (Austin)

on tati rpre

T0

Inte

Language condition

Language

Yes 5

No 48

Total 53

Tenability Total

14

9

23

18

57

Audacious

D

C

Occurrence frequency

Plausible Interpretations

Trivially true ............................. Audaciously false audacity significance

Tenability Condition

Yes 19

No 5

Total 24

Tenability

4

31

35

Total

23

36

Language

T1 T2 ... Tn-1 Tn

untenabilty inadvisability

tenability advisability

insignificance triviality

Figure 12.2 Example of one of Tennessen’s experiments and his rough explanation. (A) Example of one question evincing a dissociation between language and tenability reasons for evaluating a statement. Rows: Type of reason given by the participants. Columns: Whether the participant accepts or rejects the sentence. (B) Expressions and arguments may be interpreted in trivial or audacious directions, trading on plausibility. (C) Logical oddities as trivialities due to their wide acceptance or as audacities due to their wide non-acceptance, where the range of intuitively plausible interpretations are a function of occurrence frequency. (D) Significance is the maximization of tenability and audacity and provides a norm for interpretation and a gauge of significance when proposing a new theory. (A adapted from Tennessen, 1959a; B–D adapted from 1959f, 1959g.)

us nothing about its tenability, and vice versa. Tennessen insisted on Carnap’s principle of tolerance for matters linguistic and conceptual, and defended the viability of Russell’s approach against Strawson-style ordinary-language refutations (Tennessen, 1960a). More interesting to Tennessen was what the ineffability, infrequency, and impermissibility approaches do get right, and to this end he offered some tentative “explanations of the fact that there are linguistic expressions, locutions, formulations which intuitively or discoursively sound odd or even ‘logically odd’” (Tennessen, 1959f, p. 369). Tennessen viewed this as part of his “attack on the method of revelation” (the use of intuition) in philosophy. He said: A particularly interesting situation arises, when [historically] philosophically interesting problems—linguistic or non-linguistic—have not yet been tackled by the scientists within any ramiculated branch of existing science disciplines.

EXPERIMENTAL PHILOSOPHY

– 

The prim and proper philosopher, then, who insists on an a priori attitude, has to choose between keeping his hands clean, at the cost of ignorance on relevant matters, or to engage in empirical research himself. It seems that confronted with this choice-situation, most analytical philosophers, and in particular the so called “ordinary language” oriented philosophers, choose ignorance as the lesser of the two evils. The present paper is partly meant as an attempt to indicate what may be gained for philosophy by choosing the more earthly, a posteriori, attitude, employing empirical investigations after the pattern of the social (and other “soft”) sciences, and developing the available methods and techniques to fit within a philosophical frame of reference. (Tennessen, 1960a, pp. 496–7)

Tennessen’s tentative proposal was that the quick, unreflective judgments about linguistic or “logical” oddness can be explained by a person’s linguistic and conceptual habits plus the factors governing the remarkability of an expression (Tennessen, 1959f, 1959g, 1963). For example, “P is cultivating weeds” was labeled as “logically odd” by Nowell-Smith, despite this being a perfectly natural way to describe a normal and important activity undertaken by many anti-weed-spray producing companies (Tennessen, 1959f). It is just because NowellSmith is not a worker at such a company and that this is never remarkable that a tension is intuited by him, and this explanation holds quite generally as a first approximation. Here is the gist of Tennessen’s theory. The “audacity” or “triviality” of a statement is a measure of how widely accepted the statement is among one’s audience. In general, interpretations of statements, arguments, and theories tend to leave open a continuum between audaciously false and trivially true directions of interpretation—as he previously noticed in his previous work on nationalism, “the system of private enterprise,” and elsewhere (Figure 12.2B). The range of fast and intuitive interpretations is a function of that expression’s remarkability among situations that a person regularly encounters, so that these intuitive semantic judgments are a reflection of one’s linguistic and conceptual habits. However, the whole range of plausible interpretations is not immediately obvious and may require deliberation and a little imagination (Figure 12.2C). What makes a “hypothesis” or “proposal” significant is that it is sufficiently audacious while still tenable (Figure 12.2D). Tennessen used the assumed rule of significance as a benchmark for successful interpretation; if someone propounds a contradiction or obvious absurdity and significance is assumed, the interpreter is forced to



T . S . MURPHY

engage in a roundabout interpretation by figuring out what the sender is “up to” and intends to convey. He applied this schema to various cases and accounted for witty sayings and double entendres as a mismatch between intuitive and reflective interpretations, to explain the value in audacity for special emphasis, and to give reasons why one would ever remark a truism, obvious falsehood, or something completely irrelevant (Tennessen, 1959a, 1960b, 1965). The difference between mere sensationalism and significance was accounted for in terms of a difference between prima facie and actual tenability (Barnes and Robinson, 1972; Tennessen, 1959g, 1973, 1984). Something as audacious as “photons are both particles and waves” or “neuroscientists discover free will is an illusion” is truly significant only if it is tenable in spite of its audacity; only if it does not receive its audacity by trading on untenable implications or rigging spurious interpretations. Tennessen and Naess were very interested in how this view could apply to theory construction in social science (see Tennessen, 1960a, pp. 496–7). It seems that this was the last major period of development in empirical semantics. The Department of Speech that was home to Rynin, Tennessen, Hungerland, and others underwent restructuring in around 1960 and the Norwegian empirical semanticists all moved to different institutions. Many of the books and collections of experiments that Tennessen repeatedly cites throughout his studies were no longer pursued after 1959–60 and were never completed.21 This is perhaps unfortunate, as manuscripts show that they were aiming to develop a theory of “contextual pragmatic implications,” which they identified as an area for which a scientific theory is lacking (Tennessen, 1959e, 1959h). Hungerland made some key insights into conversational implicatures and the manuscripts show enthusiasm for systematizing these (Hungerland, 1960).22 To sum up, the empirical semanticists at Berkeley viewed ordinarylanguage philosophers as appealing to ordinary usage and tested these claims, often with negative or uncertain results. The constructive project

21

This is aside from Objectivity, which was published by a small San Francisco publisher. 22 Hungerland argued that a satisfactory account of contextual implication depends crucially on what one can infer about the speaker’s beliefs given that the norms of conversation, whatever they are, are still not violated (see also Chapman, 2008, 2009).

EXPERIMENTAL PHILOSOPHY

– 

to explain sources of language intuitions immediately followed. Much earlier was the discovery that the interpretation mechanism tends to treat evidence and illustration of usage indiscriminately. At Berkeley they explored a dissociation between language and tenability-based judgments in evaluating a statement, plus the various other studies on usage frequency and remarkability (see Tennessen, 1959a, 1959f). These were used to provide some tentative explanations for the patterns of counterintuitivity in terms of usage and a distinction between fast intuitive language judgments and reflective interpretations, which was then applied to conversational and scientific contexts. By better understanding the sources of the intuitions in these cases, the hope was to clarify when they have a legitimate role in argumentation and theorizing.

Summary and Conclusion Empirical semantics had an interesting history from its launch in Vienna, its development and applications at Oslo, and through its disagreement with the methods of ordinary-language philosophy at Berkeley. Naess was motivated by philosophers’ appeal to the way terms are conceived, defined, and used by ordinary people, as exemplified by Tarski’s material adequacy condition. For Naess and the other empirical semanticists, intuitions about common sense and the pretheoretic views of ordinary people were not sufficient, especially given enduring disagreement on such matters. The conclusion was underwhelming: there is just as much variation and indefiniteness in the minds of ordinary people. The empirical techniques were valuable for other reasons, and so Naess continued his work with collaborators at Oslo. Empirical semanticists had success in projects with UNESCO on concepts with social and political significance, in the study of science and law, and in their inquiries into questions of traditional philosophical interest. Later work at Berkeley involved developing theories to understand disagreements arising from the claims of ordinary-language philosophers. The research done at Oslo and Berkeley was extensive and broad in its scope. It was very constructive as well; part of this was of necessity due to the lack of pre-existing methods and theories for their purposes, part of this was due to a commitment to pluralism about the methods and subject matter of philosophy, and part of this owed to the view that long-term constructive



T . S . MURPHY

cooperation of many workers is just as important in philosophy as it is in science (Naess, 1953a). I have highlighted the lamentable dialectic that surrounded empirical semantics throughout the period of logical analysis and linguistic philosophy. Empirical semanticists advocated a return to the traditional conception of philosophy that accepted the philosophical legitimacy of both empirical and analytical questions. Their kind of approach was marginalized, often inadvertently to be sure, due to the expectation that a priori analytical results should follow from or be the aim of scientific investigation in matters philosophical—as if, for instance, in developing a scientific theory or explanation of how people define, conceive, and use certain terms and their cognates, it would not be philosophy proper unless it also went toward proving some contentious logical, semantic, or conceptual truth without begging the question. That was the worry with analyticity and synonymy, with deriving use from usage, the correctness of math from math performance, or other forms of psychologism. Although some of the dialectic persists today (Kauppinen, 2007; Knobe, 2007), the philosophical climate and attitude has definitely improved. Compared to experimental philosophy, there is an absence of any “positive” or “negative” mentalist program, where these both involve taking intuitions to have some distinctive evidential role in conceptual analysis or discovering analyticities, and the negative program casting doubt on such mentalist programs (Alexander et al., 2010; Cappelen, 2012; Machery, 2008). Whether or not this is an accurate picture of today’s experimental philosophy movement, it was at any rate not a topic of interest in the eyes of empirical semanticists. And there are many reasons why, including their view of indefiniteness of intention in semantics, their survey of possible universal normative conclusions that might be drawn from descriptions of language, and their broadly Carnapian attitude to theory construction—not to mention their attention to differences in directions of interpretation and preciseness (compare Chalmers, 2011). Today there are two broad explanatory goals found in experimental philosophy that are concerned with understanding how people think about philosophical topics and explaining why they think the way they do about them. Both were well represented in empirical semantics, though of course in the form of a behaviourist psychology that

EXPERIMENTAL PHILOSOPHY

– 

emphasizes language usage. Much of the research at Oslo was conceptdriven research on how closely related terms are defined, conceived, and used by people, and was motivated by an interest in improving political discourse and contributing to science and philosophy. The work done at Berkeley exemplified the interest in explaining why people have the intuitions they do about the acceptability of a statement. Although many experiments directly tested the claims made by philosophers about ordinary language and thought, this tended to function as a rhetorical point of departure for subsequent theorizing and explanation. At the center of empirical semantics and experimental philosophy is the use of the latest tools from psychology and social science, and a return to a traditional conception of philosophy as one that engages with both analytical and empirical questions. Empirical semantics had a small following and faced some difficulties with experiment construction and interpretation of the evidence, and most of the actual experiments in philosophy were done in connection with just a few philosophers, among them Naess and Tennessen. In contrast, today’s experimental philosophy has much wider appeal and has far better tools at its disposal, and the experiments and researchers already outnumber the work in empirical semantics by a wide margin. Despite empirical semantics’ successes, Ernest Nagel was correct when he predicted that Naess would “no doubt remain an outcast from the philosophic community and will have to find what solace he can in being a ‘mere’ scientist” (Nagel, 1939). Needless to say, experimental philosophy has much brighter days to look forward to.

References Alexander, J., Mallon, R., and Weinberg, J. (2010). Accentuate the Negative. Review of Philosophy and Psychology 1(2): 297–314. doi:10.1007/s13164-0090015-2. Apostel, L. (1953). Book Review: Interpretation and Preciseness. Synthese, 9(6): 413–16. Appiah, K.A. (2008). Experimental Philosophy. Proceedings and Addresses of the American Philosophical Association 82: 7–22. Aubert, V. (1943). Rettsvitenskapens logiske grunnlag [Logical Foundation of the Science of Law]. Tidsskrift for Rettsvitenskap 56: 174–200.



T . S . MURPHY

Aubert, V. (1952). White-Collar Crime and Social Structure. American Journal of Sociology 58(3): 263–71. Austin, J.L. (1956). A Plea for Excuses. Proceedings of the Aristotelian Society 57: 1–30. Austin, J.L., and Naess, A. (1964). On Herman Tennessen’s “What Should We Say”. In H. Tennessen (ed.), Eighteen Papers on Language Analysis and Empirical Semantics. Edmonton: University of Alberta Press, pp. 143–50). Barnes, J., and Robinson, R. (1972). Untruisms. Metaphilosophy 3(3): 189–97. doi:10.1111/j.1467-9973.1972.tb00823.x. Bay, C. (1958). The Structure of Freedom. Stanford: Stanford University Press. Bay, C., Gullvåg, I., Ofstad, H., and Tennessen, H. (1950). Nationalism: A Study of Identification with People and Power. Oslo: Institute for Social Research. Cappelen, H. (2012). Philosophy Without Intuitions. Oxford: Oxford University Press. Carnap, R. (1935). Philosophy and Logical Syntax. London: K. Paul, Trench, Trubner and Co., Ltd. Carnap, R. (1955a). Meaning and Synonymy in Natural Languages. Philosophical Studies 6(3): 33–47. doi:10.1007/BF02330951. Carnap, R. (1955b). On some Concepts of Pragmatics. Philosophical Studies: An International Journal for Philosophy in the Analytic Tradition 6(6): 89–91. doi:10.2307/4318235. Cavell, S. (1958). Must we Mean what we Say? 1. Inquiry 1(1–4): 172–212. doi:10.1080/00201745808601279. Cavell, S. (1999). The Claim of Reason: Wittgenstein, Skepticism, Morality, and Tragedy. New York: Oxford University Press. Cavell, S. (2010). Little Did I Know: Excerpts from Memory. Stanford: Stanford University Press. Chalmers, D.J. (2011). Verbal Disputes. Philosophical Review 120(4): 515–66. doi:10.1215/00318108-1334478. Chapman, S. (2008). Language and Empiricism—After the Vienna Circle. Basingstoke: Palgrave Macmillan. Chapman, S. (2009). Paul Grice: Philosopher and Linguist. Basingstoke: Palgrave Macmillan. Chapman, S. (2011). Arne Naess and Empirical Semantics. Inquiry 54, 18–30. doi:10.1080/0020174X.2011.542946. Chisholm, R.M. (1953). Review: Semantics and the Philosophy of Language by Leonard Linsky. Philosophy and Phenomenological Research 14(1): 115–21. doi:10.2307/2104025. Crockett, C. (1959). An Attack upon Revelation in Semantics. The Journal of Philosophy 56(3): 103–11.

EXPERIMENTAL PHILOSOPHY

– 

Fain, H., and Kaelin, E.F. (1960). Student Philosophical Opinions: A Survey. Inquiry 3(1–4): 137–52. doi:10.1080/00201746008601309. Fluge, F. (1945). Medvetsproblemet: et bidrag til spørsmålet om psykologiens teoretiske grunnføresetnader [The Problem of Consciousness: A Contribution to the Question of Psychology’s Foundational Assumptions], vol. 2. Oslo: Oslo University. Grimm, L. (1955). On the Application of the Concept of Precisation. Synthese 9 (1): 104–20. doi:10.1007/BF00567397. Grimm, L. (1962). Definition in Plato’s “Meno”. Oslo: Universitetsforlaget. Grimm, L. (1964). Definition and hypothesis in Plato’s Meno (II). Inquiry 7(1–4): 227–30. doi:10.1080/00201746408601414. Gullvåg, I. (1955). Criteria of Meaning and Analysis of Usage. Synthese 9(1): 341–61. Gullvåg, I. (1983). Depth of Intention. Inquiry 26(1): 31–83. doi:10.1080/ 00201748308601984. Gullvåg, I. (1988). Remarks on Wittgenstein’s “Über Gewissheit” and a Norwegian Discussion. Inquiry 31(3): 371–85. doi:10.1080/00201748808602162. Gullvåg, I., and Naess, A. (1996). Vagueness and Ambiguity. In Dascal, M., Gerhardus, D., Lorenz, K., and Meggle, G. (eds.), Philosophy of Language: An International Handbook of Contemporary Research. Berlin: Walter de Gruyter, vol. 2, pp. 1407–17. Haaland, A. (1947). Exposition and Critical Examination of Nietzsche’s “Will to Power” Philosophy, vol. 5. Oslo: Skrivemaskinslua. Hintikka, J. (1999). The Emperor’s New Intutions. The Journal of Philosophy 96 (3): 127–47. doi:10.2307/2564660. Howe, B. (2010). Was Arne Naess Recognized as the Founder of Deep Ecology Prematurely? Semantics and Environmental Philosophy. Environmental Ethics 32(4): 369–83. Hungerland, I.C. (1960). Contextual Implication. Inquiry 3(1–4): 211–58. doi:10.1080/00201746008601311. Hutten, E.H. (1953). Book Review: Semantics and the Philosophy of Language by Leonard Linsky. The British Journal for the Philosophy of Science 4(15): 229–35. doi:10.2307/685900. Kauppinen, A. (2007). The Rise and Fall of Experimental Philosophy. Philosophical Explorations 10(2): 95–118. doi:10.1080/13869790701305871. Knobe, J. (2007). Experimental Philosophy and Philosophical Significance. Philosophical Explorations 10(2): 119–21. doi:10.1080/13869790701305905. Knobe, J., and Nichols, S. (2008). An Experimental Philosophy Manifesto. In J. Knobe and S. Nichols (eds.), Experimental Philosophy. Oxford: Oxford University Press, pp. 3–14.



T . S . MURPHY

Knobe, J., Buckwalter, W., Nichols, S., Robbins, P., Sarkissian, H., and Sommers, T. (2012). Experimental Philosophy. Annual Review of Psychology 63(1): 81–99. doi:10.1146/annurev-psych-120710-100350. Kusch, M. (1995). Psychologism: A Study in the Sociology of Philosophical Knowledge. London: Routledge. Kusch, M. (2011). Psychologism. In E.N. Zalta (ed.), The Stanford Encyclopedia of Philosophy. . Linsky, L. (1952). Semantics and the Philosophy of Language. Urbana, IL: University of Illinois Press. Lövestad, L. (1945). The Structure of Physical Laws. Theoria 11(1): 40–70. doi:10.1111/j.1755-2567.1945.tb00003.x. Lutz, S. (2009). Ideal Language Philosophy and Experiments on Intuitions. Studia Philosophica Estonica 2, 117–39. Machery, E. (2008). The Folk Concept of Intentional Action: Philosophical and Experimental Issues. Mind and Language 23(2): 165–89. Malisoff, W.M. (1939). Book Review: “Truth” as Conceived by Those who are Not Professional Philosophers. Philosophy of Science 6(3): 379. Mates, B. (1958a). Book Review: Interpretation and Preciseness. The Philosophical Review 67(4): 546–53. doi:10.2307/2182950. Mates, B. (1958b). On the Verification of Statements about Ordinary Language. Inquiry 1(1–4): 161–71. doi:10.1080/00201745808601278. Moore, J.S. (1939). Book Review: “Truth” as Conceived by Those who are Not Professional Philosophers. The American Journal of Psychology 52(3): 489–90. doi:10.2307/1416771. Morris, C.W. (1935). Philosophy of Science and Science of Philosophy. Philosophy of Science 2(3): 271–86. Naess, A. (1933). “You Assert This?”: An Empirical Study of Weight Expressions. In Krabbe, E.C.W., Dalitz, R.J., and Smit, P.A. (eds.), Empirical Logic and Public Debate: Essays in Honour of Else M. Barth. Amsterdam and New York: Rodopi, pp. 121–32. Naess, A. (1936). Erkenntnis und wissenschaftliches Verhalten [Cognition and Scientiic Behavior] (Vol. Skrifter Utgitt av Det Norske Videnskaps-Akademi i Oslo. II. Hist.-Filos. Klasse. 1936). Oslo: Kommisjon Hos Jacob Dybwad. Naess, A. (1938a). “Truth” as Conceived by Those who are Not Professional Philosophers. Oslo And London: Kommisjon Hos Jacob Dybwad. Naess, A. (1938b). Common Sense and Truth. Theoria 4: 39–58. Naess, A. (1949). Toward a Theory of Interpretation and Preciseness. Theoria 15(1–3): 220–41. Naess, A. (1953a). Interpretation and Preciseness: A Contribution to the Theory of Communication, vol. 1. Oslo and London: Kommisjon Hos Jacob Dybwad.

EXPERIMENTAL PHILOSOPHY

– 

Naess, A. (1953b). An Empirical Study of the Expressions “True,” “Perfectly Certain” and “Extremely Probable.” Oslo and London: Kommisjon Hos Jacob Dybwad. Naess, A. (1954). Husserl on the Apodictic Evidence of Ideal Laws. Theoria 20(1–3): 53–63. doi:10.1111/j.1755-2567.1954.tb00387.x. Naess, A. (1956). Logical Equivalence, Intentional Isomorphism and Synonymity as Studied by Questionnaires. Synthese 10(1): 471–9. Naess, A. (1957). Synonymity as Revealed by Intuition. The Philosophical Review 66(1): 87–93. Naess, A. (1958). A Systematization of Gandhian Ethics of Conflict Resolution. The Journal of Conflict Resolution 2(2): 140–55. Naess, A. (1959). Do we Know that Basic Norms Cannot be True or False? Theoria 25(1): 31–53. doi:10.1111/j.1755-2567.1959.tb00290.x. Naess, A. (1961a). The Inquiring Mind. Inquiry 4(1–4): 162–89. Naess, A. (1961b). A Study of “Or”. Synthese 13(1): 49–60. Naess, A. (1964). Foreword to “The System of Private Enterprise”. In H. Tennessen (ed.), Eighteen Papers on Language Analysis and Empirical Semantics. Edmonton: University of Alberta. Naess, A. (1966). Communication and Argument, Elements of Applied Semantics (A. Hannay, trans.). Oslo and London: Universitetsforlaget, Allen and Unwin. Naess, A. (1981). The Empirical Semantics of Key Terms, Phrases, and Sentences: Empirical Semantics Applied to Nonprofessional Language. In A. Drengson (ed.), The Selected Works of Arne Naess. Dordrecht: Springer Netherlands, pp. 1562–82. Naess, A. (1983). How My Philosophy Seemed to Develop. In A. Drengson (ed.), The Selected Works of Arne Naess. Dordrecht: Springer Netherlands, pp. 2189–202. Naess, A. (1993). Logical Empiricism and the Uniqueness of the Schlick Seminar: A Personal Experience with Consequence. In A. Drengson (ed.), The Selected Works of Arne Naess. Dordrecht: Springer Netherlands, pp. 1765–83. Naess, A. (2005a). Author’s Introduction to the Series. In A. Drengson (ed.), The Selected Works of Arne Naess, vol. 1. Dordrecht: Springer Netherlands. Naess, A. (2005b). How Can the Empirical Movement Be Promoted Today? A Discussion of the Empiricism of Otto Neurath and Rudolf Carnap. In A. Drengson (ed.), The Selected Works of Arne Naess. Dordrecht: Springer Netherlands, pp. 1666–719. Naess, A., Christophersen, J.A., and Kval, K. (1956). Democracy, Ideology and Objectivity: Studies in the Semantics and Cognitive Analysis of Ideological Controversy. Oslo: Oslo University Press. Naess, S., and Naess, A. (1960). Psychological Research and Humean Problems. Philosophy of Science 27(2): 134–46.



T . S . MURPHY

Nagel, E. (1939). Book Review: “Truth” as Conceived by Those who are Not Professional Philosophers. The Journal of Philosophy 36(3): 78–80. doi:10.2307/ 2017766. Nagel, T. (2009). Lecture. Jon Searle, 50 Years at Berkeley. . Ofstad, H. (1949). Alf Ross’s begrepsbestemmelse av begrepet “rettsregel”: En elementær analytisk og kritisk undersøkelse [Alf Ross’s concept of the determination of the term “legal rule”: An elementary analytical and critical examination], vol. 13. Oslo: Skrivemaskinslua. Ofstad, H. (1950). The Descriptive Definition of the Concept “Legal Norm” proposed by Hans Kelsen: An Elementary Analytical and Critical Investigation. Theoria 16(3): 211–46. Ofstad, H. (1951). Objectivity of Norms and Value-Judgments According to Recent Scandinavian Philosophy. Philosophy and Phenomenological Research, 12(1): 42–68. doi:10.2307/2103367. Ofstad, H. (1953). Analyses of: “P Could Have Decided Differently in the Situation S,” A field for Interdisciplinarian Research. In Proceedings of the XIth International Congress of Philosophy, vol. 14, Amsterdam, pp. 128–35. Ofstad, H. (1958). The Functions of Moral Philosophy: A Plea for an Integration of Philosophical Analysis and Empirical Research. Inquiry 1(1–4): 35–71. doi:10.1080/00201745808601270. Ofstad, H. (1961). An Inquiry into the Freedom of Decision. Oslo: Norwegian Universities Press. Ofstad, H. (1967). Recent Work on the Free-Will Problem. American Philosophical Quarterly 4(3): 179–207. Ofstad, H., and Löfgren, A.A. (1965). Morality, Choice and Inwardness. Inquiry 8 (1–4): 33–73. Overton, J.A. (2013). “Explain” in Scientific Discourse. Synthese 190(8): 1383–405. doi:10.1007/s11229-012-0109-8. Peirce, C.S. (1902). Vagueness. In Baldwin’s Dictionary of Philosophy and Psychology, vol. 2. London: Macmillan. Quine, W.V. (1951). Main Trends in Recent Philosophy: Two Dogmas of Empiricism. The Philosophical Review 60(1): 20–43. doi:10.2307/2181906. Rokkan, S., and McKeon, R. (1951). Democracy in a World of Tensions. Chicago: University of Chicago Press. Rudner, R. (1950). Counter-Intuitivity and the Method of Analysis. Philosophical Studies: An International Journal for Philosophy in the Analytic Tradition 1(6): 83–9. doi:10.2307/4318095. Ryle, G. (1953). Ordinary Language. The Philosophical Review 62(2): 167–86. Sellars, W. (1947). Epistemology and the New Way of Words. The Journal of Philosophy 44(24): 645–60. doi:10.2307/2019792.

EXPERIMENTAL PHILOSOPHY

– 

Sober, E. (1978). Psychologism. Journal for the Theory of Social Behaviour 8(2): 165–91. doi:10.1111/j.1468-5914.1978.tb00398.x. Sorensen, R. A. (1990). Process Vagueness. Linguistics and Philosophy 13(5): 589–618. doi:10.1007/BF00627293. Stadler, F. (2009). Arne Naess—Dogmas and Problems of Empiricism. In J. Manninen and F. Stadler (eds.), The Vienna Circle in the Nordic Countries. Dordrecht: Springer, pp. 11–32. Stern, K. (1969). Linguistic Restrictionism and the Idea of “Potential Meaning”. The Monist 53(2): 246–61. Strawson, P.F. (1954). Book Review: Semantics and the Philosophy of Language by Leonard Linsky. Philosophy 29(109): 180–1. doi:10.2307/3747774. Tarski, A. (1944). The Semantic Conception of Truth and the Foundations of Semantics. Philosophy and Phenomenological Research 4(3): 341–76. Tennessen, H. (1949a). The Fight Against Revelation in Semantical Studies. Synthese 8(1): 225–34. Tennessen, H. (1949b). Typebegreper [Concepts of Type]. Filosofiske Problemer 12: 231. Tennessen, H. (1949c). “Private Enterprise.” A Semantical Study. In The Psychological and Sociological Implications of Economic Planning in Norway. Oslo: University Press. Tennessen, H. (1950). Holdninger til rettsoppgjøret (1945–1948) belyst ved intervjuer av 150 Oslo-jurister [Attitudes to the Trials (1945–1948) Illustrated by Interviews of 150 Lawyers in Oslo], vol. 15. Oslo: Klubben. Tennessen, H. (1959a). What Should We Say? Inquiry 2(1–4): 265–90. Tennessen, H. (1959b). “The System of Private Enterprise,” An Empirio-Semantical Analysis of a Slogan. Synthese 11(1): 72–83. Tennessen, H. (1959c). Evidence and Illustration. Synthese 11(3): 274–6. Tennessen, H. (1959d). Note on Confusion of Evidence and Illustration in Descriptive Definitions. The Journal of Philosophy 56(18): 733–5. doi:10.2307/2022186. Tennessen, H. (1959e). Controversies: Factual and Fictitious as Revealed by Empirio-Semantical Investigations, sections 28–32 (manuscript). Alberta: University of Alberta. Tennessen, H. (1959f). Logical Oddities and Locutional Scarcities. Synthese 11(4): 369–88. Tennessen, H. (1959g). On Worthwhile Hypotheses. Inquiry 2(1–4): 183–98. Tennessen, H. (1959h). Language and Reality, manuscript to be included in Controversies. Alberta: University of Alberta. Tennessen, H. (1960a). Permissible and Impermissible Locutions: “Principle of Tolerance” and “Ordinary Language Philosophy”. Synthese 12(4): 495–508.



T . S . MURPHY

Tennessen, H. (1960b). Vindication of the Humpty Dumpty Attitude towards Language. Inquiry 3(1–4): 185–98. doi:10.1080/00201746008601317. Tennessen, H. (1961). Whereof one has been Silent, thereof one May Have to Speak. The Journal of Philosophy 58(10): 263–74. Tennessen, H. (1962). Introduction to Empirical Semantics. Alberta: University of Alberta Press. Tennessen, H. (1963). Language and Absurdity. Alberta: University of Alberta. Tennessen, H. (1964). Eighteen Papers on Language Analysis and Empirical Semantics. Alberta: University of Alberta Press. Tennessen, H. (1965). Ordinary Language in Memoriam. Inquiry 8(1–4): 225–48. Tennessen, H. (1973). On the Conceptual Absurdity of “Conceptual Absurdity”. Philosophical Forum 5(74): 584–91. Tennessen, H. (1984). What is Remarkable in Psychology? In J.R. Royce and L.P. Mos (eds.), Annals of Theoretical Psychology. Dordrecht: Springer, pp. 273–8. Tennessen, H., and Gullvåg, I. (1959). Logical Analysis and Definiteness of Intention. Oslo: Universitetsforlaget. Tennessen, H., Ofstad, H., Gullvåg, I., and Bay, C. (1950). Nationalism: A Study of Identifications with People and Power. I. Problems and Theoretical Framework, vol. 1. Oslo: Institute for Social Research. Tennessen, H., Mates, B., and Gullvåg, I. (1964). Some Vague and Preliminary Reflections on Austin’s “A Plea for Excuses.” In H. Tennessen (ed.), Eighteen Papers on Language Analysis and Empirical Semantics. Edmonton: University of Alberta Press, pp. 99–110. Thue, F.W. (2009). Empiricism, Pragmatism, Behaviorism: Arne Naess and the Growth of American-Styled Social Research in Norway after World War II. In J. Manninen and F. Stadler (eds.), The Vienna Circle in the Nordic Countries. Dordrecht: Springer, pp. 219–29. Toulmin, S. (1956). Book Review: An Empirical Study of the Expressions “True,” “Perfectly Certain” and “Extremely Probable” by Arne Naess. The Philosophical Review 65(1): 116–18. doi:10.2307/2182195. Wickström-Nielsen, O. (1948). Forholdet mellem dekriptive og normative utagns gyldighet og om utsagns forhold til “virkeligheten” [The Relation between the Verifiability of Descriptive and Normative Statements and the Relation between Statements and “Reality”], vol. 7. Oslo: Skrivemaskinslua. Wright, M. (1967). Language and the Use of Language. Inquiry 10(1–4): 439–46. doi:10.1080/00201746708601511.

Index ability beliefs 220–1, 223–6 Adalio, C. 151 adolescents 131–47 Adolphs, R. 149, 151 affective science 289–90 Aharoni, E. 133, 147 Aldridge, M.A. 158, 169 Alicke, M. 93–5, 106–7, 125, 134–5, 142, 147 Ambady, N. 223–4 American Psychiatric Association 160, 169 Amodio, D. 178, 200 Andrews-Hanna, J.R. 184, 191, 200–1 anger 133–4, 143–4 Annas, J. 257 n. 5, 284 anticorrelated 184 antisocial 133, 136, 192–3 Antonenko, O. 147 Arico, A. 180, 200, 202 Aristotle 253, 256 Aronson, J. 223–4 Arpaly, N. 25 n. 22 Ashby, F. 250 Asher, N. 208 Asperger syndrome 152–68 attitudinal tolerance 42–3, 46, 49, 51 Au, W. 117 autism spectrum disorder 152–68 Baetens, K. 177, 179, 206 Bagozzi, R.P. 192, 200 Baldwin, D. 210, 226 Bandura, A. 117, 125 Banzhaf, J. 118, 125 Bargh, J. 142, 147 Baron, R. 48 Baron-Cohen, S. 160–1, 169, 171, 193, 200 Barry, C. 65, 69, 87 Bastian, B. 212 Baumeister, R.F. 194, 200, 260 n. 13 Bauminger, N. 153, 169 Bechara, A. 151

Bechlivanidis, C. 126 Beebe, J. 262 Bengson, J. 153, 155–6, 172 Bennett, J. 25 n. 22, 63, 68, 69 n. 8, 86 Bernat, E. 151 Birmaher, B. 149 Blackburn, R. 144, 147 Blackburn, S. 29 Blackwell, L. 220 Blair, K. 148 Blair, R.J. 133, 141, 143, 146–8, 150–1 blame 93–6, 98–9, 103–4, 107, 113–15, 123, 134–5, 142–5 blameworthiness 9–27 Bloom, P. 173, 180, 201 Blos, J. 178, 201 Bodenhausen, G. 224 Borsci, G. 177, 182, 201 Bradley, M. 149 Braeges, J. 30 Brandone, A. 209 Bratman, M. 235–8, 249–50 Brent, D. 149 Brewer, M. 96, 97, 101, 102, 103, 106, 124, 125, 128 Brink, D. 29 Britt, T.W. 129 Brown, R. 224 Broyd, S.J. 184, 191, 201 Bruun, S.E. 117, 127 Bryson, S.E. 158, 171 Buckner, R.L. 184, 191, 201, 206 Buckwalter, W. 262 Budhani, S. 148 Buon, M. 153, 169 Burra, A. 132, 141, 149 callous affect 193, 195, 197 callous/unemotional traits 136 Campbell, J. 104, 125 Cane, P. 116, 125 capacities, perceptual 23–5 capacities, volitional 23, 25 Cardinale, E. 133–4, 143, 150



INDEX

Carlson, G. 208 Carpenter, M. 158, 169, 172 Castano, E. 194, 201 categorization 245–50 Cattaneo, L. 158, 169 causal concepts 65–7, 302–7, 314, 320 causal network 107–8, 110–11, 125 causation 67, 72, 91–130 and agency 314–15 and stability 318–19 interventionist theories 303, 308–9, 311–12, 319 normative theories 295–321 Čehajić, S. 194, 201 Chalmers, D.J. 174, 195, 201 Chan, K.K. 130 Channon, S. 72 n. 13, 92, 99, 107, 127, 170 Chartrand, T. 142 Chen, G. 148 Chen, X. 125 Cheng, P. 97, 126 Cheryan, S. 224 childhood moral deprivations 15, 17, 20–7 Chituc, V. 283 Chockler, H. 92, 109–13, 115, 121, 126 Christ, S.E. 192, 201 Chupetlovska-Anastasova, A. 241 Cima, M. 146, 148 Cimpian, A. 212 n. 2, 216–17, 221–2, 225 Cipolotti, L. 143 Clark, F. 148 Cleckley, H. 146, 148 Clore, G. 150 Cognitive Reflection Test (CRT) 193, 195–7 Cohen, A. 149, 208 Cohen, L. 183, 201 collective intentionality 233–6, 238, 240–1, 249 Collins, A. 150 compensatory responsibility 64–5, 82–9 conceptual analysis 302 concordance condition 19–20 conditional 93, 100 consciousness 174–6, 178–9, 183, 185, 187–9, 191, 195 Cooper, J. 256, 260 Corbetta, M. 176, 201

counterfactual 92–3, 96–7, 100–1, 103–10, 112, 115, 124–5 Craver, C.F. 183, 201 credit 97–8, 103–4 Crisp, R. 280 criticality 116–24 Csikszentmihalyi, M. 258 Cumby, J. 262 Cushman, F. 72, 77, 149, 151, 153, 155, 171, 262 D’Entremont, B. 159, 169 Damasio, A. 149, 151 Damasio, H. 151 Darley, D. 30–2, 35–40, 43 Darley, J. 107, 115, 120, 126 Dar-Nimrod, I. 224 Davidson, P. 30 Davis, M.H. 193, 201 De Cremer, D. 118, 126 Deci, E. 258 n. 8 deep self 11, 19–22 default mode network (DMN) 178, 181–2, 184–8, 190, 192–4 Dehaene, S. 183, 201 dehumanizing 194 Denny, B.T. 178, 201 Diener, E. 253, 258, 278, 284, 289 Dierker, D.L. 182, 206 Diesendruck, G. 213 Difficulty Hypothesis 22–5 disagreement 29, 32, 35, 37–8, 42, 46–7, 49, 56 divergent beliefs 28–31, 35, 37–42, 47–9, 51, 53–6 divergent cultures 28 Doherty, K. 129 doing and allowing harm 62–89 Donagan, A. 63 Doris, J. 9 n. 3, 10 n. 4, 17 n. 16 dorsal attention network 177 double prevention 302, 304, 315–19 Downing, P.E. 176, 202, 207 Drown, D. 113, 129 dual-character concepts 283 dualism 173–4, 180, 185, 194–7, 200 dual-process theory 196 duckrabbit 24 Duncan, J. 179, 202 Dweck, C. 220–1, 224 Dybikowski, J. 260 n. 12

INDEX

Eberhardt, J. 212 efficacy 117 enabling harm 62–89 Erickson, L. 216 executive function 166, 167 experimental philosophy 234–5, 238–40, 244, 251 experimental psychology (contrasted with experimental philosophy) 300, 320 explanatory gap 174, 180, 183, 185, 188, 189, 194 Fabbri-Destro, M. 158, 169 Faraci, D. 10 n. 5 faux pas 153, 160–2, 166–7 Feldman, F. 257, 260, 282, 284 Felsenthal, D. 118, 126 Fiala, B. 180, 200, 202 Fincham, F.D. 97, 101, 106–7, 124, 126 Finger, E. 136–7, 148, 150 Fisher, L. 133, 148 Fishkin, J. 30 Flesch, M. 149 Flor, H. 151 Flynn, C. 149 Foell, J. 151 folk physics 182–3 folk psychology 182–3 Foot, P. 63, 69, 256, 267, 282, 284 Forsyth, D.R. 120–1, 126 Forth, A. 133, 136–7, 143, 148 Fowler, K. 150–1 Fox, M.D. 177, 179, 184, 202, 205 Frankfurt, H. 23 n. 20, 93, 126 Fraser, B. 72 Frederick, S. 193, 202 Freese, R. 149 Frick, P. 146 Frijda, N. 143, 148 Frith, C. 153, 169, 178, 200 Frith, U. 153, 169 fronto-parietal control network 179 Gabriel, J.J. 193, 203, 205 Gauthier, I. 179, 206 Gavanski, I. 94, 95, 129 Gelman, S. 209–14, 216, 218, 219, 226 generics 208–27 Gerstenberg, T. 92, 101, 102, 103, 111, 124, 126, 127, 128, 130



Giammanco, C.A. 126 Gilbert, D. 258 Gilbert, M. 233, 235 n. 1, 236–7, 249–50 Gillberg, C. 161, 169 Giner-Sorolla, R. 194, 201 Glenn, A. 134, 146, 149 Glymour, C.N. 129 Goldberg, I. 188, 200, 202 Goldvarg, E. 67 Good, C. 223–4 Goodale, M.A. 176, 202 Goodin, R. 82 Goodwin, G. 30–2, 35–40, 43 Gopnik, A. 182, 202 Graham, J. 149 Gray, H.M. 193, 202 Gray, K. 193, 202 Gray, N. 146, 149 Greenberg, Y. 208 Greene, J.D. 173–5, 196, 200, 202 Gregory, C. 107, 126 Grill-Spector, K. 179, 202 Gruesser, S. 151 Guglielmo, S. 153, 155–6, 169–70 Gunasegaram, S. 97, 128 Gundel, H. 177, 182, 202 Haber, L. 213 Haidt, J. 28, 32, 149, 151 Hall, J. 151 Hall, N. 69 n. 10 Halpern, J. 92, 105, 109, 110, 111, 112, 113, 115, 121, 126 Hamilton, V.L. 116, 120, 127 Hanser, M. 65 Happé, F. 153, 170 happiness 253–90 Hare, R. 136–7, 148 Hart, H. 92, 105–6, 116, 120, 127 Haslam, N. 194, 202, 211–12 Haslanger, S. 217, 224 Hauser, M.D. 148–9, 151 Haybron, D. 257 n. 6, 260, 282 hedonism 255, 257 Heider, F. 105, 127 Heine, S. 224 Henley, N. 241 Hermann, C. 151 Herpertz, S. 142, 149 heterophenomenology 198 hierarchical clustering 246



INDEX

Hill, E.L. 153, 158–60, 171 Hilton, D.J. 99, 127–8 Hirschfeld, L. 211 Hitchcock, C. 72, 105, 109, 126–7, 262 Hollander, M. 209–10, 214 n. 3, 216 n. 4, 226 Holton, R. 262 Honey, C.J. 176, 203 Honoré, T. 105–6, 127 Huckleberry Finn 8, 25 Hughes, C. 158, 167, 170 Hurlburt, R.T. 199, 203 Iacoboni, M. 181, 185, 203 ICD-10, 170 ideal observer theories 301 ill will 12 Imada, S. 151 indoctrination 21–2 instrumental action 154, 156, 158, 165–8 instrumental desire 154, 156–8, 165–8 intentional stance 180, 190–5, 198–9 intentionality (intentional action) 17, 19, 99, 104, 107, 125, 131–47, 152–68 intolerance 28–57 introspection 180, 183, 185, 198 intuition, role in philosophical theory 296, 302, 319–20 Intuitive Physics Test (IPT) 193, 195 Iyer, R. 149 Jack, A.I. 173–4, 177, 186, 188–95, 198–200, 203, 205 Jackson, F. 174, 195, 204 Jaggi, C. 129 Jarrold, C. 158, 171 Jaspars, J.M. 97, 101, 106–7, 124, 126 Jaworska, N. 241 Jenkins, H.M. 97, 127 Johnson, S. 180, 204, 246 Johnson-Laird, P. 67 joint action 232–51 joint commitment 237, 241, 250 jointness 232–4, 240–5, 247–8, 250–1 JoJo 8–14, 22, 26–7 Jones, L. 148 Josephs, R. 224 Jurkowitz, I. 150

Kahneman, D. 92, 105, 127, 196, 204, 259 n. 10 Kamins, M. 221 Kanwisher, N. 176, 202, 204 Kasari, C. 153, 169 Kaufman, J. 136, 149 Kaufman-Gilliland, C.M. 93, 127 Kekes, J. 257, 286 Keller, J. 212 Kelley, H.H. 94, 97, 127 Kenny, D. 48 Kerr, N.L. 93, 117–19, 127 Kiehl, K. 147 Killen, M. 30 Kincannon, A. 106–7, 129 Knobe, J. 9 n. 3, 10 n. 4, 17–19, 72, 77, 93, 105, 107, 127–8, 131–2, 134–5, 141, 144, 149–50, 152–6, 170–1, 193, 196, 204, 238–9, 261–2, 283 Koenigs, M. 133–4, 143, 149, 151 Kohl, D. 287 Kohlberg, L. 30 Koleva, S. 149 Komorita, S. 125 Koralus, P. 150 Kosson, D. 148, 150 Kraut, R. 256, 260 n. 12, 276, 282 Krettenauer 30 Kroger, J.K. 181, 204 Kruepke, M. 149, 151 Kruskal, J. 240 Kunert, H. 149 Kupferman, O. 126 Kupperman, J. 25 n. 22 Lagnado, D.A. 72 n. 13, 92, 99, 101–2, 107, 111, 113, 116, 119, 121, 126–7, 130, 152, 170, 310, 322–3 Lang, P. 149 Latané, B. 110, 115, 120, 126, 128 Lawler, J. 208 Leboyer, M. 151, 153, 167, 169, 172 Lee-Evans, J. 143 Leibenluft, E. 148, 150 Leibniz, G. 174, 204 Leslie, A. 132, 141, 145, 149, 153, 170 Leslie, S. 208, 209–12, 214, 216 n. 4, 226 Levenston, G. 142, 149 Levine, J. 174, 204 Levy, S. 212 n. 1 Lewis, D. 111, 128, 182, 204

INDEX

Leyens, J.-P. 194, 204 life-satisfaction 253, 255, 257–9, 267–9, 274, 278, 285–6, 289 Lombrozo, T. 66 n. 4, 99, 104, 107, 125, 128, 133, 135, 151 Lord, C. 161, 170 Lowery, L. 151 Lucas, R. 253 Lukas, G. 149 Lyubomirsky, S. 289 McCloskey, M. 287 McClure, J. 99, 127–8 McCoy, J., 92, 128 MacCulloch, M. 149 McGeer, V. 150 Machery, E. 151, 155, 158, 163, 165, 170–1, 294–5, 299, 307, 322–3 Machiavellian 192 Machover, M. 118, 126 McKenna, M. 93, 129 Mackie, J. 29 Mahalingam, R. 213 Mahon, B.Z. 179, 204 Malle, B. 131–2, 134, 143, 149–50, 152–6, 169, 170 Mallon, R. 153, 155–6, 170–1 Mandel, D.R. 98–100, 124, 128 Mangan, J. 131, 150 Mannheim, B. 210, 214 n. 3, 226 Markman, E. 216, 222 Mars, R.B. 178, 204 Marsh, A. 133–4, 136–7, 143, 150–1 Martin, A. 178, 205 Mason, W. 95, 128 May, J. 262 Mele, A.R. 153, 155, 171 Mendelovici, A. 150 Mendlow, G. 131–2, 144, 149 mentalizing 177–82, 184, 187, 189 meta-ethical pluralism 30–1, 37–8, 40, 42, 49, 56–7 Meyer, M. 210, 221, 226 Miller, D.T. 92, 97, 105, 127, 128, 211–2 Miller, G. 175, 205 Milner, A.D. 176, 202 mindreading 153, 157 mirror neuron 176–7, 179, 181, 187 Mitchell, D. 148, 150 Mitchell, J.P. 178, 205 Moore, M.S. 106, 128



moral and conventional violations 133 moral concept 261 Moral Deprivation–Excuse Thesis (MDE) 9–11, 16, 22, 26 moral domain 31–3, 36–8, 40–57 moral ignorance 11–27 moral objectivity 33, 36, 43–4, 48–9, 55 moral reasoning 132–3, 137–8, 140, 143, 146–7 moral relativism 35 n. 2, 42 moral valence 131–2, 138–40, 147 Moreci, P. 149 Morris, M. 149 Mueller, C. 221 Mueller-Isberner, R. 149 Mullen, E. 28–9 Multi-dimensional Scaling (MDS) 240–2, 244, 250 Munro, R. 241 Murphy, R. 129 Nadelhoffer, T. 153, 155, 171 Nagel, T. 174, 189, 195, 205 Nanay, B. 153, 155–6, 171 negative psychological states 272, 275, 277 Nelkin, D. 17 n. 16, 23 n. 19 Nelson, S. 131, 134, 143, 150 Newman, J. 149, 151 Ng, P. 148 Nichols, S. 30, 93, 128, 153, 155, 171, 238–9 normative evaluation 263, 265, 267, 272–3, 275, 285–6 Nucci, L. 30 objective list theories 255 objectively grounded 30–1, 35–56 Ochsner, K.N. 201, 205 Ongur, D. 182, 205 opposing domains 186 Ortony, A. 143, 150 Osterheider, M. 149 Owen, A.M. 179, 202 Ozonoff, S. 167, 171 Parkinson, C. 147, 150 Patrick, C. 142, 149–51 Paulhus, D.L. 193, 205 Pearl, J. 107, 109–11, 126, 128 Pellizzoni, S. 137, 150



INDEX

Pelphrey, K.A. 176, 205 Pennington, J. 129, 167, 169, 171 Percy, E.J. 128 Petrocelli, J.V. 93, 100–1, 104, 106, 124, 128 Pettit, D. 135, 141, 150, 262 Phelan, M. 144, 150, 153, 155–6, 171, 238 n. 2 phenomenal stance 189–94, 198–200 Phillips, J. 254 Phillips, W. 153, 159–60, 171 physical stance 180, 186, 189–95, 198–200 physicalism 188, 194, 200 Piaget, J. 30 Pine, D. 148, 150–1 pivotality 110–11, 113–16, 118, 121–4 Plakias, A. 286 Plato 255–6 Pope, K. 151 positive psychological states 262, 284–5 positive versus negative programs 295–7, 300 possible world 92, 105, 109–11, 124–5 potency 100–1 praise 105 praiseworthiness 9, 15–18, 22, 24, 26–7 prejudice 209, 211–12 Prentice, D. 211–12 pretheoretic intuitions 11, 25 Price, J.L. 182, 205 Prinz, J.J. 193, 204, 238 n. 2 probability 96–101, 103, 106, 118 prospective 92, 116 psychological essentialism 211–12, 215, 219 psychopathy 131–47, 193, 195 Quinn, P. 63, 69 Qunaibi, M. 149 Rachels, J. 29 Rachels, S. 29 Raine, A. 149 Rakic, P. 182, 205 Ramsay, J. 250 Rao, U. 149 Rapoport, A. 117–18, 128 real self, see deep self Reeder, G.D. 102, 128 Reid, M. 148, 150

responsibility 91–125 scalar 23, 25–6 Rest, J. 30 retrospective 92, 96, 116, 120 Rhodes, M. 211, 213–14, 216–17, 222 n. 5 Rickless, S. 65 Rips, L. 240 Robbennolt, J.K. 92, 113, 128 Robbins, P. 173, 189–91, 193, 195, 203–5 Rodriguez, J. 213 Roedder, E. 262 Roese, N.J. 92, 129 Rogers, C.R. 190, 205 Rose, D. 134–5, 142, 147, 153, 171 Roseman, I. 143, 150 Rothbart, M. 211 Rothemund, Y. 142, 150 Roxborough, C. 261 Rozin, P. 143–4, 151 Russell, J. 153, 158, 167, 170–1 Ryan, N. 149 Ryan, R. 258 n. 8 Ryff, C. 258 n. 8 sanity (normative competence) 11 Sarkissian, H. 30, 53, 144, 150, 153, 155–6, 171 Sartorio, C. 124, 129 Sass, H. 149 Saul, J. 224 Saxe, R. 178, 182, 205 Schaffer, J. 108, 129 Schechter, J. 150–1 Scheines, R. 129 Schilbach, L. 178, 205 Schleifer, M. 129 Schlenker, B.R. 117, 120, 129 Schoenemann, P.T. 182, 205 Schooler, J.W. 93, 129, 194, 206 Schroeder, T. 25 n. 22 Schug, R. 149 Searle, J. 233, 249–50 Seligman, M. 258 Semendeferi, K. 182, 206 Shafer-Landau, R. 29 Shapley, L. 118, 129 shared intention 232–6, 241, 250 Shaver, K.G. 107, 113, 129 Shepard, R. 240

INDEX

Sherman, S.J. 128 Shih, M. 224 Shoben, E. 240 Shoemaker, D. 10 n. 5, 12 n. 8 Shubik, M. 118, 129 Shulman, G.L. 184, 201, 206 Shultz, T.R. 107, 126, 129 side-effect effect (Knobe effect) 17, 132–47, 155 Sidgwick, H. 260 Siegal, M. 150 Sims, C. 148, 150 Sinclair, S. 151 sine qua non 106 Singer, B. 258 n. 8 Sinnott-Armstrong, W. 150 Skitka, L. 28–9 Sloman, S. 67 Smetana, J. 30 Smith, D.L. 194, 206 Smith, E. 240 Smith, I.M. 158, 171 Smith, J. 149 Smith, M. 28, 148 Snare, F. 29 Snow, C.P. 196, 206 Snowden, R. 149 social cognition 176–8, 182, 189, 191–2 solo action 241–8 Song, M. 30 Spellman, B.A. 97–101, 103, 106–7, 124, 129 Spirtes, P. 107, 129 spontaneous cognition 187, 191 Sripada, C. 19 Steele, C. 223 Steiner, I.D. 112, 129 stereotype threat 223–5 stereotyping 212–13, 223–5 Stodart, T. 30 Stoit, A.M.B. 158, 171 Strawson, P.F. 93, 129 Stuhlmüller, A. 128 Subramaniam, K. 181, 191, 206 Suikkanen, J. 257, 286 Sumner, L. 258 Suri, S. 95, 128 Surian, L. 150 Sutton, R.M. 127 Swartz, T. 150 symmetry thesis 9–10, 15–22, 25–7

Takeuchi, H. 191, 206 Talmy, L. 65 Tardif, T. 209–10, 214 n. 3, 216 n. 4, 226 Tarr, M.J. 179, 206 task positive network (TPN) 179, 181–2, 184–8, 190, 192–4 Taylor, C. 25 n. 22 Taylor, M. 211 Tenenbaum, J.B. 294, 300 theory of mind 153, 161, 166–7 Thomson, J. 69 n. 9 Tiberius, V. 286 togetherness 233–4 Tomasello, M. 159, 172 Tonnaer, K. 148 Tormala, Z.L. 128 Towbin, K. 148, 150 trade-off hypothesis 156 Tranel, D. 149, 151 Trzesniewski, K. 220 Turiel, E. 30, 133, 151 Ulatowski, J. 153, 155, 171 Ullman, T. 128 unhappiness 268–75, 282 Uttich, K. 133, 135, 151 Vaidyanathan, U. 142, 151 van Dijk, E. 118, 126 Van Essen, D.C. 182, 201–2, 206 Van Overwalle, F. 178–9, 181 Vincent, J.L. 202, 206 Vivhelin, K. 69 n. 8 Vohs, K.D. 93, 129, 194, 206 Vonasch, A. 260 n. 13 Wainryb, C. 29, 30, 53 Ward, W.C. 97, 127 Waytz, A. 177, 206 we-intention 232, 234, 241, 249–50 Weisberg, J. 178, 205 well-being 255–6, 258, 263, 278, 282, 285–6 Wells, G.L. 94–5, 129 Werth, U. 149 Wheatley, T. 150 White, S. 137, 151 Widerker, D. 93, 129 Wiest, C. 150 Wiggett, A.J. 176, 207





INDEX

Wilkes, K.V. 187, 207 Williams, M. 212 Williamson, D. 149 Wolf, S. 8–11, 17 n. 6, 20, 22 Wolff, P. 65–6 Wong, D. 29 Woodward, J. 104, 109, 129, 303, 308–12, 318, 321, 324 Wright, J.C. 28–32, 35–7, 153, 155–6, 172 Yablo, S. 109, 129 Yazbek, A. 159, 169

Young, L. 132, 134, 141, 143, 147, 149, 151, 262 Yu, C. 118, 129 Yu, H. 150 Yzerbyt, V. 212 Zalla, T. 132, 151, 153, 158, 160, 166–7, 169, 172 Zeier, J. 149 Ziegler, S. 151 Zultan, R. 92, 114–16, 119, 121, 123–4, 127, 130 Zyzniewski, L.E. 126